Donate
Independent, objective, nonpartisan research

R 304HJR

Authors

R 304HJR

Tagged with:

Publication PDFs

Database

This is the content currently stored in the post and postmeta tables.

View live version

object(Timber\Post)#3711 (44) { ["ImageClass"]=> string(12) "Timber\Image" ["PostClass"]=> string(11) "Timber\Post" ["TermClass"]=> string(11) "Timber\Term" ["object_type"]=> string(4) "post" ["custom"]=> array(5) { ["_wp_attached_file"]=> string(12) "R_304HJR.pdf" ["wpmf_size"]=> string(6) "339031" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(211585) "In Short Supply? Cycles and Trends in California Housing ••• Hans P. Johnson Rosa M. Moller Michael Dardia 2004 Library of Congress Cataloging-in-Publication Data Johnson, Hans P. In short supply? : cycles and trends in California housing / Hans P. Johnson, Rosa M. Moller, and Michael Dardia. p. cm. ISBN: 1-58213-079-5 1. Housing—California. 2. California—Population. 3. California—Economic conditions. I. Title: Cycles and trends in California housing. II. Moller, Rosa Maria. III. Dardia, Michael. IV. Title. HD7303.C2J64 2004 333.33'8'09794—dc22 2004001896 Copyright © 2004 by Public Policy Institute of California All rights reserved San Francisco, CA Short sections of text, not to exceed three paragraphs, may be quoted without written permission provided that full attribution is given to the source and the above copyright notice is included. PPIC does not take or support positions on any ballot measure or on any local, state, or federal legislation, nor does it endorse, support, or oppose any political parties or candidates for public office. Research publications reflect the views of the authors and do not necessarily reflect the views of the staff, officers, or Board of Directors of the Public Policy Institute of California. Foreword During the 1990s, there was a disturbing and widely noted decline in the construction of new housing units in California. In the 1980s, almost 2.1 million units were built, averaging 207,000 a year. The following decade, new production fell to 1.1 million for an average of 110,000 per year. This decline has been attributed to a myriad of causes, from demanding housing codes and construction requirements to tighter growth controls and planning guidelines. Taken together, the story goes, these factors have made new construction ever more difficult. This difficulty, in turn, has led to the dramatic falloff in production and its predictable consequence: high and increasing housing costs. After reviewing the data, Hans Johnson, Rosa Moller, and Michael Dardia have concluded that a housing shortage does exist, but that it is much smaller than implied by the reduction of new construction in the 1990s. Slower population growth, unprecedented outflows of domestic migrants, greater numbers of immigrants and children (who consume less housing than other demographic groups), and business cycle swings all contributed to a much lower demand for housing in the 1990s. The authors assign nearly 80 percent of the downturn in housing production to these demand-side variables and argue that the actual housing shortfall statewide is closer to 138,000 units—not the 1,000,000 units implied by a simple comparison of production figures over the two decades. Almost the entire shortage has occurred in the Bay Area, Los Angeles, and San Diego—the state’s most populous regions. Pressure on the housing stock is nothing new for these areas. At least since the end of the Second World War, some California homebuyers have been willing to pay huge premiums to live where the amenities are, and the authors note that large macroeconomic trends may explain more of this willingness than levels of new construction do. This part of the analysis once again points to the role demand has played in shaping the state’s housing market over the last two decades. Acceptable in theory, this iii finding is unlikely to console first-time homebuyers in the state’s largest metropolitan areas. Here the study acknowledges the supply side of the equation by noting other research that indicates these regions were the least receptive to new production during this period. Although the authors find little or no evidence of shortage in the rest of the state, they note that residents of many of those regions face affordability challenges. To the untrained eye, it may be difficult to distinguish between housing shortages and affordability challenges, but the authors correctly attend to the substantial variation in the state’s regional housing markets. From a public policy perspective, this variation is especially important insofar as it points policymakers away from a single, statewide response. Once again, PPIC research has underscored the regional diversity of California’s markets, and if the policy lesson sounds familiar, it has proven difficult to master. David W. Lyon President and CEO Public Policy Institute of California iv Summary By many measures, California appears to be in the midst of a severe housing crisis. New construction is anemic compared to past levels. Prices are high in many regions and in some areas, such as the Bay Area, exorbitant. Vacancy rates, especially for rental units, are low. And the number of persons per housing unit has continued to increase in California even as it has declined in the rest of the nation. Other measures, however, suggest that the crisis is not so severe, or even that there is no crisis in much of the state. The median value of homes in the state was lower in 2000 than in 1990, once inflation is taken into account. In addition, incomes rose faster than rents or housing prices in much of the state. Housing prices and rents remain higher in California’s largest metropolitan areas than in the rest of the nation, but this could be a long-standing phenomenon related to incomes and amenities. Nationally, new housing construction was markedly slower in this business cycle as well. Finally, the source of California’s population growth—children and immigrants—could explain the higher household sizes. In this report, we explore California’s housing markets and attempt to explain why housing construction has lagged in the 1990s compared to previous business cycles. Can the sluggish pace of construction in the 1990s be explained by unique macroeconomic developments and demographic changes that took place in that period? To address this question, we analyze the effects of macroeconomic and demographic factors on California’s housing market. We find that the slowdown in new construction in the 1990s can largely be explained by these factors, but we also find evidence of housing shortages in the state’s largest metropolitan areas. The slowdown of housing construction in the 1990s was experienced nationally as well as in California, suggesting that the sluggish pace of new construction in the 1990s is at least partly explained by the national macroeconomic v events of that period. The more pronounced slowdown in California can be attributed to the severity of the 1990 recession, the slow pace of the recovery, and the demographic changes that took place in the state. In particular, we found that • There is evidence of a shortage in the supply of new housing units in California in 2000, but the magnitude of the shortage is much lower than has been estimated by others. Our best estimates are that the shortage of housing units in the state is around 138,000 units, far lower than the estimates often cited in the media. This suggests that most (over 80 percent) of the apparent shortage in California’s occupied housing units can be attributed to demographic differences between California and the rest of the nation. • Regional variations in California’s housing markets are large, with almost all of the shortage in new housing units originating in the Bay Area, Los Angeles, and San Diego, the state’s most populous regions. • The slowdown in population growth during the 1990s can explain much of the slowdown in new housing construction. Adjusting for census undercounts, California’s population grew by 3.8 million people in the 1990s, far lower than the 6.2 million increase of the 1980s. Slower population growth in the 1990s can be attributed to tremendous and unprecedented flows of domestic migrants out of California. • In addition to the slowdown in population growth, the composition of California’s population growth in the 1990s explains much of the apparent lack of new housing in the state, with immigrants and children constituting a much larger share of the state’s growth in the 1990s than in the 1980s. Immigrants and children tend to consume less housing than other demographic groups. Children do not form their own households, and immigrants often live in larger, extended-family households (even where housing prices are relatively low). Once we control for these and other demographic differences between vi California and the nation, we observe similar trends in the creation of new households. • More than 80 percent of annual changes in new construction are explained by changes in macroeconomic and demographic factors. This explained portion varies by region, ranging from 67 percent for the North Coast region to as much as 90 percent for the Central Eastern area. • Business cycle effects are an important determinant of annual fluctuations in housing construction. In particular, changes in new construction are closely related to changes in the business cycle (as measured by changes in unemployment). Construction slowed considerably during the recession of the early 1990s. • Changes in macroeconomic factors (such as changes in interest rates, income, prices, and expectations of inflation) driven by changes in economic growth (in the United States and the rest of the world) and by U.S. economic policies are closely related to changes in new construction. • Developments in the financial markets partly determined the slower pace of new construction in the 1990s. Investors see real estate and stocks as substitutes. Investments in stocks in the late 1990s had a dampening effect on new home construction, as expected returns on housing decreased relative to investments in stocks. Furthermore, we suspect that builders’ and developers’ expectations of profits may have been influenced by lower expectations of inflation resulting from the way that the Federal Reserve conducted monetary policies and the unusual low inflation in the period. Profit expectations were also influenced by demographic changes, specifically lower rates of household formation that accompanied reduced population growth in key age groups and with domestic out-migration. These demographic changes led to decreased investment in new construction as builders and developers expected slower growth in the demand for new housing. In some markets, expectations may have fallen too far, leading to an underproduction of new housing units. This underproduction could vii have been exacerbated by restrictive land-use policies, especially in the Bay Area and Los Angeles markets. These findings do not mean that there are no hardships with respect to housing supply and new construction in California. There may be serious problems in markets for low-income housing, and there is evidence of a housing shortage in the state’s largest metropolitan areas. However, they do suggest that the supply crisis may be overstated, and that our position in 2000 was perhaps better, and certainly not much worse, than in 1990. viii Contents Foreword ..................................... Summary..................................... Figures ...................................... Tables ....................................... Acknowledgments ............................... iii v xi xiii xv 1. INTRODUCTION ........................... What Are the Assertions Regarding a Housing Shortage? .... How Can a Housing Shortage Be Measured? ........... Outline of This Report ......................... 2. TRENDS AND PATTERNS IN CALIFORNIA HOUSING ................................ Prices and Rents ............................. Affordability ................................ New Construction ............................ How Is Population Growth Accommodated? ........... New Housing Units.......................... Increases in Household Size ..................... Declines in Vacancy Rates...................... Jobs and Housing ............................ Summary .................................. 3. CAN MACROECONOMIC FACTORS EXPLAIN CALIFORNIA’S HOUSING MARKETS? ............ Business Cycles, Macroeconomic Events, and Construction .. A Simple Housing Market Model .................. Factors Affecting Housing Demand ................. Supply Determinants .......................... Ten Regional Housing Markets in California ........... Results of the Model of California Housing Markets ...... Summary .................................. 1 1 3 4 7 7 16 21 23 23 25 28 30 31 35 36 42 43 44 45 46 51 ix 4. CAN DEMOGRAPHIC CHANGES EXPLAIN CALIFORNIA’S HOUSING MARKETS? ............ Demographic Determinants of Demand .............. Population Growth .......................... Age .................................... Gender .................................. Nativity and Ethnicity ........................ Marital Status ............................. Householder Rate Model: Temporal Trends in the Probability of Forming a Household ............. Temporal Trends in Householder Rates ............. Differences in Householder Rates Between California and the Rest of the United States .................. Householder Rates in California Metropolitan Areas ..... Estimating the Shortage of Occupied Housing Units..... Summary .................................. 53 55 55 56 56 58 60 61 61 65 67 70 75 5. CONCLUSION AND POLICY CONSIDERATIONS.... 79 Appendix A. A Housing Market Model ....................... 89 B. Householder Rate Models ....................... 103 C. Panel Data Analysis of Housing Supply ............... 107 References .................................... 115 About the Authors ............................... 121 Related PPIC Publications .......................... 123 x Figures 2.1. Median Home Prices in California, 1975–2002, Nominal and Real (2000 $) ................... 2.2. Annual Inflation-Adjusted (Real 2000 $) Median Home Prices in Selected Counties in California, 1989–2000 ... 2.3. Median Home Prices in the United States and California, 1998–2000, in Current Dollars .......... 2.4. Share of Households That Can Afford the MedianPriced Home in Major Counties in California, 1990 and 2000 .................................. 2.5. New Multifamily and Single-Family Housing Construction in California, 1967–2002 ............ 2.6. New Housing Construction in the United States and California, Indexed to 1.0 in 1986 (Peak Year of Production), 1973–2002 ..................... 3.1. California Construction and Employment Cycles ...... 3.2. Expectations of Inflation, 1972–2001 ............. 3.3. Term Structure (Difference Between the Long-Term and Short-Term Interest Rates), 1971–2001 ............ 3.4. Comparison of Actual Changes in California New Housing Construction and Model Fitted Values, 1967– 2000 .................................. 4.1. Growth in Child and Adult Populations in California, by Decade................................. 4.2. Probability of Being a Householder in California, by Age, 2001 ............................... 4.3. Change in California Population, by Age Group and Decade................................. 4.4. Change in Foreign-Born and U.S.-Born Populations in California, by Decade ....................... 4.5. Decennial Change in Population in California, by Marital Status ............................ 8 9 10 17 21 22 36 40 41 50 55 57 58 59 60 xi 4.6. Probability of Being a Householder in California and the Rest of the United States, 1976–2001 ............. 62 4.7. The Effect of Demographic Changes on the Probability of Being a Householder in the Rest of the United States, 1976–2001 .............................. 63 4.8. The Effect of Demographic Changes on the Probability of Being a Householder in California, 1976–2001 ..... 64 4.9. Difference in Householder Probabilities Between California and the Rest of the United States ......... 65 4.10. Age-Standardized Householder Rates in California and the Rest of the United States, by Ethnicity and Nativity, 2000 .................................. 67 4.11a. Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, No Demographic Controls, 2000 ................ 68 4.11b. Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, with Controls for Age, Gender, Ethnicity, and Nativity, 2000 ....... 69 4.12. Estimates of the Undersupply of Housing in California Based on Householder Rates in California and the Rest of the United States, by Decade ................. 72 4.13. New Occupied Housing Units, by Decade: Actual Versus Necessary for California (Controlling for Demographic Differences) .................... 73 4.14. Decennial Shortage of New Occupied Housing Units ... 74 4.15a. Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of the United States, No Demographic Controls, 2000 ...... 76 4.15b. Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of the United States, Controls for Age, Gender, Ethnicity, and Nativity, 2000 ............................ 77 C.1. Comparison of Actual and Predicted Annual Housing Production in California, 1973–1999 ............. 111 xii Tables 2.1. Median Value of Owner-Occupied Housing Units in California’s Counties, 1980, 1990, and 2000 ........ 12 2.2. Median Gross Rent in California’s Counties, 1980, 1990, and 2000 ........................... 14 2.3. Rent and Homeowner Costs as a Percentage of Income................................. 19 2.4. Ratio of Household Population Change to Total Housing Units Change in California’s Counties: 1970s, 1980s, and 1990s .......................... 24 2.5. Number of Persons per Occupied Household, 1970– 2000 .................................. 26 2.6. Vacancy Rates in California’s Counties, 1990 and 2000 .................................. 29 2.7. Nonfarm Jobs per Housing Unit in California’s Counties, 1990 and 2000 ..................... 32 3.1. Effect of Explanatory Factors on Changes in New Housing Construction: Results of the Macroeconomic Model, by Region .......................... 47 A.1. Analysis of Changes in New Construction in California, by Region (Time-Series Results, No Lags) .......... 97 A.2. Time-Series Results Allowing for Lagged Responses .... 99 C.1. Panel Regression Results ..................... 108 C.2. Results of County-Level Analysis of Annual Housing Production .............................. 110 C.3. Housing, Developable Land, and Resistance to Growth (1990–1999) ............................. 113 xiii Acknowledgments We have benefited from the comments, suggestions, and assistance of a number of people. We thank Anthony Downs, Bill Fulton, Paul Lewis, Dowell Myers, Dean Misczynski, and John Quigley for their reviews of an earlier draft of this report. Michael Teitz has provided encouragement for this project from its inception and has contributed many helpful comments. Joyce Peterson and Peter Richardson have provided important structural and editorial assistance. The authors are responsible for any errors of fact or interpretation. xv 1. Introduction Despite a strong economic performance overall during the late 1990s, California’s housing production appears to have failed to keep pace with either population or job growth. Indeed, the state’s most recent economic boom seems to have created a shortage of housing— particularly in highly populated coastal areas—as evidenced by a rapid rise in home prices. Some observers have pointed to the trend toward larger household sizes as further evidence of a housing shortage. This report attempts to explain why housing construction has lagged in the 1990s compared to previous business cycles. The study addresses the question of whether the sluggish pace of construction in the 1990s can be explained by unique macroeconomic developments and demographic changes that took place in that period. To address this question, we analyze the effects of macroeconomic and demographic factors on California’s housing market. Our macroeconomic model examines factors that economists have found to be the most important predictors of changes in new housing construction. Our demographic model considers whether the composition of the state’s population growth has had an effect on household formation and therefore on housing demand. Our results provide important information on the responsiveness of California housing markets to changes in demographics and economic growth. What Are the Assertions Regarding a Housing Shortage? The popular perception—shared by policymakers and the public alike—is that California is in the midst of a housing crisis that has been driven by the failure of new housing construction to keep up with the pace of growth in the state. It is easy to see why this is the general impression. 1 • From 1990 to 1999, 1.1 million housing units were built in California, or an average of 110,000 per year; from 1980 to 1989, almost 2.1 million were built (207,000 per year). The ratio of new residents to new housing units rose from 3.1 over the 1980s to 3.9 over the 1990s. • The median home price in California increased from $193,334 in 1990 to $241,779 in 2000, and housing affordability1 in 2000 was only 34 percent statewide and below 20 percent in most of the San Francisco Bay Area. • The share of homeowners who spend more than 30 percent of their income on housing rose from 29.7 percent in 1989 to 31.5 percent in 1999. These statistics are buttressed by any number of anecdotes about people having to leave their hometowns because of unaffordable housing and of employers unable to attract or retain workers because of exorbitant housing prices. All this evidence implies that there is a serious housing shortage in California. Arrayed against these facts, however, are other figures that either fail to document the outcomes one would expect in the midst of a housing crisis or put the dire statistics into historical context. • Although it is true that housing prices increased significantly over the 1990s, once we correct for inflation, the median home price in California was 3.5 percent lower in 2000 than it was in 1990. • Likewise, home prices are strongly related to personal income, which increased at a noticeably faster pace than inflation from 1990 to 2000. Although housing affordability in California at the peak of the recent business cycle was only 34 percent, it was 20 percent at the previous peak in 1990. Similarly, in only three of the counties where affordability is tracked back to 1990 was affordability lower in 2000 than it was in 1990; in most ____________ 1Housing affordability is defined as the percentage of households that can afford the median priced single-family home. Annual figures are prepared by the California Association of Realtors. 2 counties, housing is more affordable now than it was in the last cycle. Outside the San Francisco Bay Area, housing affordability improved dramatically over the 1990s. • Household size (the number of persons per household) has increased since 1990, but demographic changes explain most of the increases we observe (see Moller et al., 2002). • Despite the fact that homeownership costs rose as a share of income in the 1990s, the percentage of renters who spent more than 30 percent of their income on rent fell from 47.7 percent to 44.7 percent. • Although the homeownership rate in California lags the U.S. average (56.9 percent versus 66.2 percent), it nevertheless rose slightly, from 55.6 percent to 56.9 percent, from 1990 to 2000. It is the contradiction between these two sets of facts that we address in this study. How Can a Housing Shortage Be Measured? We use several approaches and datasets to evaluate California’s housing markets. Housing markets are better understood as regional rather than state phenomena. Some regions might be more supplyconstrained than others—and macroeconomic and demographic forces could operate differentially between regions. Identifying a housing shortage is difficult, depending in part on the measures chosen as evidence of a shortage. The simplest method, for example, would consider only rents and prices—increases could be taken as evidence of demand outstripping supply. This limited approach would be misleading, however, as it would fail to take into account many other determinants of housing prices, such as income and interest rates. Because of the complexities of housing supply and demand, we have approached the issue of whether there is a shortage in the supply of new housing in two complementary ways. Our first approach is a traditional econometric model that analyzes the effects of factors commonly used to explain housing market patterns. These factors include interest rates, unemployment rates, income, and prices. We analyze the California housing market as a whole as well as ten regional markets. This 3 econometric approach allows us to identify the extent to which macroeconomic factors influence changes in new housing construction. These annual fluctuations can be quite dramatic and are clearly related to macroeconomic factors. Details of the econometric model are in Appendix A. Our second method investigates the demographic determinants of housing demand by building a householder rate model. This approach is particularly important given the large demographic changes in many of the state’s metropolitan areas. The demographic model provides us with important information on how the nature of California’s population change influences the demand for housing. Because the population of the state does not vary dramatically from one year to the next, the effects are seen gradually over a 25-year period (1977 through 2001). Details of the demographic model are in Appendix B. By developing multiple methods and measures of California’s housing markets, we are able to assess the importance of very different inputs. In the end, we observe important changes in long-term demographic determinants of housing demand and identify important cyclical variations associated with business cycles. Outline of This Report Before proceeding to our detailed analyses of the state and regional housing markets, we summarize the major trends and indicators in housing and population in California. In Chapter 2, we provide descriptive measures of California’s housing market, for both the state and its regions. These descriptive measures allow us to examine the most accessible and well-known indicators of housing markets in California, including new construction, prices, vacancy rates, and household size. We also provide comparisons with national trends when available and of interest. Chapter 3 presents our econometric model of California’s housing markets, discusses the macroeconomic factors that help determine annual fluctuations in housing construction, and provides results of the macroeconomic models. Chapter 4 analyzes the effect of changes in the state’s population on the consumption of housing in the state. A householder rate model is constructed to determine differences in household formation between demographic subgroups in California 4 (and its largest metropolitan areas) and the rest of the nation. Chapter 5 summarizes our findings and discusses their implications for policy. Finally, it is important to note that this report does not assess all aspects of the California housing market. In particular, we do not focus on low-income housing in the state. Instead, we focus on the housing market as a whole. There is quite possibly a significant mismatch between the type of new housing units being built and the type of new households forming in the state. That and other topics are subjects of future research. 5 2. Trends and Patterns in California Housing This chapter provides summary data on key measures of California’s housing. We discuss prices and rents, affordability, new construction, population change, persons per household, jobs and housing, and vacancy rates. We use annual data series from the California Association of Realtors and the Department of Finance, as well as recently released decennial census data. Annual measures of change provide important evidence of business cycle events, whereas long-term demographic changes are assessed across decades. Many of the measures suggest that California’s position in 2000 was not appreciably different from its position in 1990. Nonetheless, these measures show considerably more housing stress in some of the state’s largest metropolitan areas than in the rest of the nation Prices and Rents Between 1975 and 2002, the median home price in California rose over seven-fold, from $42,267 to $315,870.1 These dollar figures are not adjusted for inflation and therefore give the impression of a larger rise in prices than is warranted. Figure 2.1 shows median home prices in California from 1975 to 2002 in both nominal and real (i.e., inflationadjusted) terms. Once corrected for inflation, the median home price in California hit a peak of $266,846 in 1989 and then declined substantially to 1996. By 2000, the median home price was still lower than the 1989 peak, but subsequent increases led to a new high of almost $300,000 by 2002. ____________ 1Annual prices reported are averages of monthly prices compiled by the California Association of Realtors. These are the prices of sold houses; below, we present information on median values of all owner-occupied houses as estimated by their owners in decennial censuses. 7 Median home price ($) 350,000 300,000 250,000 200,000 150,000 100,000 50,000 0 Real (in 2000 dollars) Nominal 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 Figure 2.1—Median Home Prices in California, 1975–2002, Nominal and Real (2000 $) The statewide median price is drawn from regions with quite different price levels. The importance of regional economic fortunes can be seen when the real and nominal median prices are compared across several of the largest counties in California. Figure 2.2 compares the median home price in Santa Clara, Los Angeles, and Orange Counties from 1989 to 2000 in real terms (adjusting for inflation). The early 1990s recession clearly took a toll on the Los Angeles housing market, less so in Orange County. From Figure 2.2, it is clear that the median statewide home price does not reflect the state of regional housing markets: The median real home price in Los Angeles County in 2000 was still 26 percent below its previous peak in 1989; Orange County prices were basically unchanged; and prices in Santa Clara County had increased 45 percent faster than inflation. In fact, of the counties that the California Association of Realtors has tracked since 1990, the only counties outside the Bay Area that had real price gains during the 1990s were Monterey and San Diego. In most other markets, prices in 2000 were still lower in real terms and some were down 20 percent or more from 1990. 8 600,000 500,000 400,000 Santa Clara County Orange County Los Angeles County Median home price ($) 300,000 200,000 100,000 0 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Figure 2.2—Annual Inflation-Adjusted (Real 2000 $) Median Home Prices in Selected Counties in California, 1989–2000 In 2000, the median home price (among houses sold) in California was $241,779—74 percent higher than the national median price of $139,000. Even though California is the largest state and is itself composed of several major metropolitan areas, it is small in comparison to the nation as a whole, and its metropolitan areas are more tightly linked to each other than to other areas in the nation. This proximity has made housing prices in the state significantly more volatile than national prices. Figure 2.3 shows how median home prices compare at the state and national levels (the national figure includes California).2 The median price in California rose to 2.2 times the national median price in 1989, fell to 1.5 times in 1996 and 1997, and then rose back to 1.7 times the national median in 2000. However, not all regional markets in California are expensive by national standards. Median home prices in Fresno, Riverside, and San Bernardino Counties were less ____________ 2Throughout this report, we provide both California and national data. Our preferred comparison is California versus the rest of the nation, but where such information is not available—as in this case—we provide data for California and the nation (including California) and provide a note to that effect. 9 300,000 250,000 California United States Median home price ($) 200,000 150,000 100,000 50,000 0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 SOURCE: National Association of Realtors, data for the United States include California. Figure 2.3—Median Home Prices in the United States and California, 1988–2000, in Current Dollars expensive in 2000 than the national average, and Sacramento was only 5 percent higher. Data from the decennial censuses allow us to identify long-term changes in rents and values in all of California’s counties for rented housing units and for all owner-occupied housing units (not just those sold).3 Patterns of change in rents and values of owner-occupied housing units were similar, with California and many of its counties experiencing much higher rents and housing values than the rest of the country, but with most if not all of the increase occurring during the 1980s rather than the 1990s. Indeed, in real terms, median values of owner-occupied housing units declined in California from 1980 to 1990 ____________ 3We consider several measures of housing to identify local market conditions, comparing indicators from the most recent census with those of past censuses. Although these ten-year changes do not allow for the consideration of annual patterns of change, the 1980, 1990, and 2000 censuses all coincided with strong economic conditions. Thus, the decennial patterns are not affected by short-term business cycles and should provide us with an accurate picture of long-term trends at the local level in California. 10 (in nominal terms, values increased 9 percent). Regional variations are substantial (see Table 2.1). Were it not for declines in Los Angeles County, median values would have increased in the state from 1990 to 2000. As with rents, counties with the greatest declines in housing values were in the Los Angeles metropolitan area (five of the six counties with the greatest declines were Los Angeles, Ventura, San Bernardino, Riverside, and Orange). Still, a majority of counties (36) experienced an increase in real values between 1990 and 2000. And from 1980 to 2000, eight California counties experienced real increases in values of more than 50 percent. Increases were focused in the Bay Area (including Santa Clara, San Francisco, San Mateo, Alameda, Marin, and Napa Counties) and the Central Coast (Santa Cruz and San Benito Counties). By 2000, California’s most expensive county, Marin, had home values over four times higher than the national average. Nonetheless, 15 of the state’s 58 counties still had median home values lower than the national average of $119,600 in 2000, including two of the state’s medium-sized metropolitan areas (Fresno and Bakersfield). As shown in Table 2.2, 33 of the state’s counties experienced declines in rents from 1990 to 2000, with the greatest declines in the Los Angeles metropolitan area (including Los Angeles, Orange, Riverside, San Bernardino, and Ventura Counties). In the Los Angeles area, the declines of the 1990s followed the spectacular increase in rents that occurred during the 1980s. These declines can be attributed to the strength and depth of the recession of the early 1990s—a recession that was especially pronounced in the Los Angeles area—and to a speculative bubble in prices around 1990. Only five counties experienced real increases in rents of more than 10 percent for the decade from 1990 to 2000, and only three were counties with large populations (San Mateo, Santa Clara, and San Francisco). Still, rents in much of California remain far higher than in the rest of the nation. By 2000, rents in 35 of California’s 58 counties were higher than the national average; only two of the state’s larger metropolitan areas, Fresno and Bakersfield, had average rents lower than those of the nation as a whole. From these trends, it does not appear that prices in this business cycle are much different from prices in the previous one. As of 2000, most regional markets had not yet regained their 1989–1990 peak prices 11 Table 2.1 Median Value of Owner-Occupied Housing Units in California’s Counties, 1980, 1990, and 2000 Median Value (Real 2000 $) % Change 1980– 1990– 1980– 1980 1990 2000 1990 2000 2000 United States 98,848 103,425 119,600 5 16 21 United States excluding California 99,266 95,391 112,500 –4 18 13 California 180,316 251,583 225,000 40 –16 18 Far North Del Norte 105,644 111,484 121,100 6 9 15 Humboldt 120,917 114,591 133,500 –5 17 10 Lake 124,312 123,525 122,600 –1 –1 –1 Lassen 106,704 89,731 106,700 –16 19 0 Mendocino 148,071 159,910 170,200 8 6 15 Modoc 85,067 62,281 69,100 –27 11 –19 Nevada 169,072 199,919 205,700 18 3 22 Plumas 139,798 116,145 137,900 –17 19 –1 Sierra 123,675 105,398 128,600 –15 22 4 Siskiyou 111,796 87,659 100,300 –22 14 –10 Trinity 116,463 106,434 112,000 –9 5 –4 Upper Sacramento Valley Butte 125,372 121,713 129,800 –3 74 Colusa 100,340 88,177 107,500 –12 22 7 Glenn 101,825 86,882 94,900 –15 9 –7 Shasta 129,403 117,828 120,800 –9 3 –7 Sutter 126,857 118,735 120,700 –6 2 –5 Tehama 102,250 89,083 103,100 –13 16 1 Yuba 94,400 86,882 89,700 –8 3 –5 Sacramento Metro El Dorado 177,558 201,862 194,400 14 –4 9 Placer 165,254 218,176 213,900 32 –2 29 Sacramento 137,040 167,031 144,200 22 –14 5 Yolo 141,283 177,908 169,800 26 –5 20 San Joaquin Valley Fresno 131,949 107,211 104,900 –19 –2 –20 Kern 116,038 106,693 93,300 –8 –13 –20 Kings 99,067 91,155 97,600 –8 7 –1 Madera 127,494 110,966 118,800 –13 7 –7 Merced 112,644 116,663 111,100 4 –5 –1 San Joaquin 119,645 156,025 142,400 30 –9 19 Stanislaus 127,494 159,262 125,300 25 –21 –2 Tulare 103,734 95,039 97,800 –8 3 –6 12 Table 2.1 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Median Value (Real 2000 $) % Change 1980– 1990– 1980– 1980 1990 2000 1990 2000 2000 180,952 200,681 320,325 165,890 221,894 263,897 232,077 143,192 187,528 291,722 281,104 459,530 236,433 381,711 441,273 372,518 190,597 259,740 303,100 267,800 514,600 251,300 396,400 469,200 446,400 178,300 273,200 61 40 43 43 72 67 61 33 39 4 68 –5 33 12 61 6 51 4 79 6 78 20 92 –6 25 5 46 183,498 155,283 164,617 220,621 199,620 254,172 265,567 276,055 322,668 330,696 265,800 284,000 230,000 293,000 377,500 39 71 68 46 66 5 7 –17 –9 14 45 83 40 33 89 184,346 140,858 138,737 169,709 134,070 224,864 144,040 146,573 155,378 148,127 153,436 128,057 203,415 158,485 184,200 –20 153,600 10 156,900 7 161,300 –10 141,900 –4 236,300 –10 149,800 10 26 0 –1 9 6 13 5 –5 11 6 16 5 –5 4 143,616 179,720 146,500 25 134,494 166,384 131,500 24 –18 –21 2 –2 186,680 229,319 197,923 289,780 324,092 315,287 209,300 55 270,000 41 248,700 59 –28 –17 –21 12 18 26 101,401 93,356 100,000 –8 193,044 241,095 227,200 25 7 –1 –6 18 SOURCE: Authors’ calculations of decennial census data. NOTES: Values are in 2000 dollars. Adjustments for inflation were made separately for California and the United States. Median values for counties in 1980 do not include owner-occupied condominiums. In 1990, the value of all owner-occupied dwelling units was 1.0 percent lower than the value of non-condominium owneroccupied units. In California, the difference was only 0.6 percent. 13 Table 2.2 Median Gross Rent in California’s Counties, 1980, 1990, and 2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Median Gross Rent ($) 1980 508 1990 589 2000 602 560 581 600 803 747 465 546 519 496 530 537 513 596 567 490 533 561 532 610 600 378 425 429 605 774 746 458 474 525 403 548 513 443 474 471 435 475 487 484 568 563 409 458 494 460 460 458 509 559 563 443 501 506 443 474 486 435 496 488 717 737 702 560 745 780 520 682 659 496 660 687 501 562 534 509 570 518 458 532 533 465 548 562 488 557 518 473 633 617 507 624 611 479 522 516 % Change 1980– 1990– 1990 2000 16 2 4 34 –7 18 –5 71 16 –5 95 15 –2 12 1 28 –4 3 11 36 –6 7 –1 92 18 –1 12 8 00 10 1 13 1 73 14 –2 3 –5 33 5 31 –3 33 4 12 –5 12 –9 16 0 18 3 14 –7 34 –3 23 –2 9 –1 14 Table 2.2 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Median Gross Rent ($) 1980 1990 2000 564 811 852 624 874 898 791 1,067 1,162 596 818 818 605 846 928 713 996 1,144 709 1,001 1,185 522 764 797 613 835 864 613 809 776 484 708 765 592 742 719 636 847 830 643 923 924 740 535 659 509 640 685 513 616 599 484 533 516 426 508 502 656 712 682 530 647 611 573 741 660 552 720 648 588 811 704 759 1,023 923 672 976 892 465 510 504 596 791 761 % Change 1980– 1990– 1990 2000 44 5 40 3 35 9 37 0 40 10 40 15 41 18 46 4 36 3 32 –4 46 8 25 –3 33 –2 44 0 –28 23 26 7 20 –3 10 –3 19 –1 9 –4 22 –6 29 –11 31 –10 38 –13 35 –10 45 –9 10 –1 33 –4 SOURCE: Authors’ calculations of decennial census data. NOTES: Rents are in 2000 dollars. Adjustments for inflation were made separately for California and the United States. 15 once inflation is accounted for, and only the San Francisco Bay Area had seen significant real price increases. What matters from a policy perspective, however, is how these price changes affect residents in these markets. An indicator that helps us to see these effects is housing affordability. Hence, we next examine trends in affordability in the largest counties in the state. Affordability Affordability makes the state’s housing supply a compelling policy issue. Housing is the most reliable indicator of a region’s cost of living. It is also the largest budget item for most households and represents the bulk of most homeowners’ net worth. There are many potentially useful ways to characterize affordability, but the definition used by most analysts, and one of the measures we consider, is the share of an area’s households that can afford the median-priced home at current interest rates and loan standards.4 This measure has two potential flaws in the context of this report, and they work in opposite directions. First, not everyone is in the market for a single-family home. Some households are not likely to be in the market at a given time and, more important, many households already own homes at lower-than-current prices, regardless of whether they could afford to buy their home at current prices. Second, the affordability index evaluates only the ability of households to purchase the median-priced home. Even in a market characterized by a plentiful supply of homes, we would expect only about half of the households to be able to purchase the median-priced home (assuming similar distributions of household income and home prices). Using decennial census data, we consider two additional measures of affordability: the share of renters who face high rent burdens (defined as paying more than 30 percent of their household income on rent), and the share of homeowners who face high housing cost burdens (defined as ____________ 4This assumes a 20 percent down payment, with mortgage payments not to exceed 30 percent of monthly pre-tax income. Affordability figures are calculated by the California Association of Realtors for this group of counties from 1990 on. 16 paying more than 30 percent of their income on housing costs, including mortgage payments). In addition to being affected by home prices, affordability is driven by personal income and interest rates. Over the last ten years, interest rates have fallen and personal incomes have risen relative to inflation, and these trends have mitigated the effects of rising home prices (and in fact contributed to those rising prices). Figure 2.4 shows the affordability in 1990 and 2000 for 17 of the largest counties in California. Counties in the greater San Francisco Bay Area (including Monterey and Santa Cruz) are grouped together on the left side of the graph, and they show the dramatic difference in affordability between that region and the rest of the state. It is also clear that this gap existed in the previous business cycle, although the more affluent parts of the greater Los Angeles area were in the same range as some Bay Area counties. Only Contra Costa and Sonoma Counties were noticeably less affordable in 2000 than they were in 1990. The rest of the Bay Area and San Diego faced the same affordability situation as they did at the previous peak, and the other 70 1990 60 2000 50 40 30 20 10 0 Figure 2.4—Share of Households That Can Afford the Median-Priced Home in Major Counties in California, 1990 and 2000 17 Percentage SCaoSnSSnataFanrrMAtnnlStaaoaaoannMtnCcmCMilaCeooteasrrasrtcredmiueaoazyoaan SanLSoBaSsceaRrriAnVnavOFneraremrDrgniaetedseiienslunntdggnreooesoeoa counties were much more affordable than previously. This level of affordability is obviously aided by the fact that interest rates in the latter half of the 1990s were substantially lower than they were in the 1980s, but it does seem to conflict with the notion of a statewide housing shortage that arose in the 1990s. Still, California has far more homeowners and slightly more renters under financial stress because of high housing costs than the rest of the nation (see Table 2.3).5 As bad as the situation is in California, it did not appear to appreciably worsen in the 1990s. In 14 counties, including some with the most expensive housing in the state (Ventura, Santa Clara, Orange, Contra Costa, San Mateo, Marin, Alameda, and Santa Cruz), the proportion of homeowners paying more than 30 percent of their income for housing remained unchanged or even declined, although the levels remained high. In those counties, rising incomes at least kept pace with rising housing costs. Substantial increases in homeowner costs relative to income occurred primarily in the Upper Sacramento Valley, San Joaquin Valley, and in some mostly rural mountainous counties. In many of those areas, housing costs are among the lowest in the state. Relatively low wages and incomes would seem to be more of a culprit than high housing costs for homeowners in those areas. For renters in California, the situation actually improved slightly from 1990 to 2000. Statewide, in 48 counties the proportion of renters paying more than 30 percent of their income in rent either did not change or declined. Increases in the proportion occurred primarily in less populated counties outside the state’s metropolitan areas. San Francisco had one of the largest declines and one of the lowest proportions of renters paying more than 30 percent of household income for rent—a consequence of rapidly increasing incomes and, probably, rent control and out-migration of those with lower incomes. Places with high rents relative to incomes tended to be counties with relatively high poverty rates. Among the state’s larger metropolitan areas, Fresno had ____________ 5This section presents data from the decennial censuses for all renters and homeowners. Calculations of income as a percentage of rent or homeowner costs are based on incomes for the year preceding the census, whereas rents and homeowner costs were those experienced at the time of the census. 18 Table 2.3 Rent and Homeowner Costs as a Percentage of Income United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare % of Homeowners Paying More Than 30% of Household Income 1990 2000 20 22 % of Renters Paying More Than 30% of Household Income 1990 2000 41 40 18 21 40 39 30 31 48 45 13 23 45 51 18 25 50 54 27 31 51 47 18 22 39 45 23 26 46 43 16 19 32 38 27 32 50 48 20 24 48 41 25 24 29 36 19 23 44 46 21 27 49 46 20 26 55 53 18 25 29 39 18 23 42 36 20 30 48 48 19 27 43 40 21 26 45 42 20 28 46 44 30 32 48 43 28 29 47 41 24 27 48 43 22 27 52 53 22 29 48 48 22 27 44 46 21 25 39 42 23 31 45 45 21 30 46 43 24 29 45 46 24 29 47 45 21 29 48 44 19 Table 2.3 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego % of Homeowners Paying More Than 30% of Household Income 1990 2000 % of Renters Paying More Than 30% of Household Income 1990 2000 31 31 46 42 31 30 46 42 32 32 50 46 26 29 48 40 26 30 45 37 32 31 43 42 31 29 43 40 31 30 43 42 31 32 49 44 29 33 46 42 32 37 35 40 31 32 53 51 29 32 53 51 34 34 53 48 37 25 29 36 21 25 39 43 23 32 48 41 19 19 40 37 22 28 39 37 30 39 32 41 25 27 44 48 33 32 50 47 30 31 49 46 31 35 49 46 33 32 47 44 35 32 48 42 21 29 47 47 32 32 49 45 SOURCE: Authors’ calculations of decennial census data. NOTE: Incomes are from the year preceding the census, whereas rents and homeowner costs were those experienced at the time of the census. the highest proportion of renters paying more than 30 percent of their income in rents, yet Fresno had among the lowest rents in the state and experienced a decline (in real terms) in median rents from 1990 to 2000. 20 Poverty and low incomes seem to be a more important factor in the location of financial stress with respect to housing than does the cost of the housing itself. New Construction One of the most compelling pieces of evidence regarding a possible housing shortage is the 47 percent decline in construction between the 1980s and the 1990s. From 1980 to 1989, almost 2.1 million housing units were built in California, for an average of 207,000 per year; from 1990 to 1999, only 1.1 million were built, for an average of 111,000 per year. The drop in construction was most severe in the case of multifamily units, where construction in the decade 1990–1999 fell by 70 percent from the pace in the 1980s. Multifamily units made up 53 percent of the total units in the previous peak year of 1986 but only 29 percent in 2000. In contrast, single-family construction fell by only 27 percent over the same period. Figure 2.5 shows annual housing construction from 1967 to 2002 for both single- and multifamily units. The break between the pattern of new construction in the last three 350,000 300,000 Multifamily Single-family 250,000 Number of new units 200,000 150,000 100,000 50,000 0 11996697 1971 1973 111999777975 1981 1983 111999888975 1991 1993 111999999975 2001 Figure 2.5—New Multifamily and Single-Family Housing Construction in California, 1967–2002 21 housing cycles and the current one is often “Exhibit A” in the argument for a housing supply crisis in the state. This change in the cyclical pattern of new construction also took place nationally. Although the acceleration in new construction after 1982 was higher in California than in the rest of the United States, the situation reversed after the second half of 1989, when the reduction in the number of new units was much more dramatic in California. From its trough in 1991, annual production in the rest of the United States increased by 88 percent by 2002. In California, the increase in annual production from 1991 to 2002 was only 51 percent, but the increase from its lowest production year (1993) to 2002 was 88 percent. Figure 2.6 shows the ratio of annual production to the peak-year production for the United States and California, with 1986 production indexed to 1.0. Housing production in California trailed that of the nation at least in part because the state had something of a double-dip construction recession. Not only did construction in the state take two years longer to 1.20 Rest of United States 1.00 California 0.80 0.60 0.40 0.20 0.00 SOURCE: Authors’ calculations of U.S. Census Bureau data. Figure 2.6—New Housing Consruction in the United States and California, Indexed to 1.0 in 1986 (Peak Year of Production), 1973–2002 22 Index 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 rebound than it did at the national level, but it stumbled again a year later (in 1995). Since 1995, new housing construction has increased markedly but remains far below the 1986 peak. How Is Population Growth Accommodated? New Housing Units Housing growth has not kept pace with population growth in California’s counties. Both for the state as a whole and in 49 of California’s 58 counties—including all the large urban counties in the state—the ratio of population change to the change in housing units was much greater in the 1980s and 1990s than in the 1970s (see Table 2.4).6 Statewide, this ratio doubled between the 1970s and 1980s. The ratio increased again from the 1980s to the 1990s, but much less dramatically. Counties with extremely high gains in household population relative to the number of net new housing units in the 1990s include places with very high housing prices, such as San Mateo, Santa Barbara, Santa Clara, and Alameda Counties. For those counties, the increase in the ratio could be an indication that high housing costs are leading people to live with more roommates or family members than they otherwise would. For other counties with large population growth relative to housing growth, such as Los Angeles and Colusa Counties, the nature of the population growth at least partly explains the high ratio; with immigrants and children forming a large share of population growth in those counties, housing demand is lower than if the population growth was primarily among U.S.-born adults.7 Perhaps most noteworthy, though, is that the dramatic increase in the ratio between population growth and new occupied households occurred in the 1980s as well as in the 1990s. In almost half of California’s counties (28 of 58), the ratio either remained unchanged or declined in the 1990s as compared to the 1980s. Thus, population growth outpacing housing ____________ 6The only exceptions were the relatively lightly populated mountainous counties of Lassen, Siskiyou, Modoc, Mono, Inyo, Plumas, Trinity, Sierra, and Alpine. 7Chapter 4 considers how the nature of California’s population growth in the 1990s might have affected the demand for new housing. 23 Table 2.4 Ratio of Household Population Change to Total Housing Units Change in California’s Counties: 1970s, 1980s, and 1990s United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin 1970–1980 1.19 1.13 1.63 1.59 0.85 1.54 1.90 1.67 1.14 1.96 1.63 2.04 1.46 1.65 1.79 0.90 1.85 1.88 1.61 1.60 1.09 1.98 1.74 1.29 1.47 1.71 1.56 1.81 2.20 1.76 1.46 1.91 2.15 0.60 1.32 0.80 1980–1990 1.53 1.28 3.10 2.06 1.71 2.35 1.86 2.75 0.94 2.08 0.95 0.79 1.42 0.84 2.54 3.42 3.61 2.28 3.15 2.66 4.00 2.40 2.30 2.67 2.88 3.58 3.15 3.52 3.86 5.09 4.19 3.45 3.98 2.77 2.24 0.88 1990–2000 2.32 2.19 3.92 1.97 1.42 2.06 1.35 1.79 –1.53 2.01 0.71 6.72 0.40 –0.09 2.12 5.09 2.48 1.89 3.37 2.01 1.87 3.11 2.55 3.13 3.13 3.60 3.09 3.47 2.99 3.26 3.51 3.99 3.84 4.82 3.75 2.72 24 Table 2.4 (continued) Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1970–1980 1.52 –5.87 0.74 1.65 2.18 2.07 1980–1990 2.83 3.78 3.07 2.94 2.83 2.36 1990–2000 3.21 3.19 6.72 4.85 3.47 3.03 1.83 3.52 4.41 2.34 3.28 3.92 1.72 2.46 2.30 1.37 2.88 6.11 2.16 3.63 3.49 1.49 0.04 0.49 1.76 1.99 2.33 1.28 1.77 2.23 0.99 0.47 0.00 1.70 1.71 1.89 0.84 0.63 2.44 1.47 1.96 1.73 1.64 2.61 3.62 1.78 2.96 4.80 1.43 4.40 6.07 1.95 3.04 4.56 2.17 3.02 3.64 1.97 3.53 3.33 1.90 2.79 3.48 SOURCE: Authors’ calculations of decennial census data. growth is as much a 1980s phenomenon as a 1990s phenomenon in much of California. Increases in Household Size If much of California’s population growth was not accommodated through new housing, how was it accommodated? Some population growth can be accommodated through previously unoccupied housing, and some can be accommodated by increases in household sizes. Table 2.5 shows changes in the number of persons per occupied household from 1970 through 2000. Since 1980, household sizes have been 25 Table 2.5 Number of Persons per Occupied Household, 1970–2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo 1970 3.11 3.13 2.95 3.05 3.00 2.50 2.93 2.94 2.93 2.70 2.81 2.67 2.90 2.89 2.82 2.95 3.02 3.04 3.16 3.04 3.27 2.89 3.04 3.08 3.07 3.20 3.19 3.35 3.20 3.35 3.03 3.09 3.25 2.84 3.19 2.94 2.92 2.34 2.97 1980 2.75 2.76 2.68 2.64 2.55 2.36 2.70 2.61 2.62 2.55 2.60 2.40 2.59 2.63 2.46 2.69 2.75 2.66 2.75 2.65 2.76 2.62 2.71 2.56 2.59 2.83 2.82 3.04 2.98 2.97 2.71 2.78 2.99 2.53 2.69 2.43 2.55 2.19 2.58 1990 2.63 2.61 2.79 2.63 2.49 2.38 2.66 2.57 2.49 2.51 2.41 2.45 2.48 2.49 2.48 2.84 2.77 2.58 2.75 2.60 2.85 2.66 2.66 2.58 2.63 2.96 2.92 3.08 3.05 3.17 2.94 2.91 3.12 2.59 2.64 2.33 2.54 2.29 2.64 2000 2.59 2.56 2.87 2.58 2.39 2.39 2.59 2.53 2.39 2.47 2.29 2.32 2.35 2.29 2.48 3.01 2.84 2.52 2.87 2.62 2.87 2.63 2.63 2.64 2.71 3.09 3.03 3.18 3.18 3.25 3.00 3.03 3.28 2.71 2.72 2.34 2.62 2.30 2.74 26 Table 2.5 (continued) Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1970 3.23 3.14 2.92 3.11 3.29 2.82 2.99 2.70 2.72 2.74 2.76 2.78 2.60 2.83 2.78 2.97 3.11 2.83 3.21 3.43 3.52 2.94 1980 2.76 2.82 2.56 2.85 3.12 2.51 2.62 2.54 2.84 2.49 2.54 2.45 2.48 2.43 2.55 2.69 2.82 2.69 2.78 3.00 3.24 2.62 1990 2.81 2.88 2.55 2.96 3.15 2.53 2.73 2.66 2.47 2.41 2.50 2.35 2.42 2.48 2.46 2.85 2.97 2.91 2.87 3.02 3.26 2.69 2000 2.92 2.90 2.60 3.14 3.32 2.49 2.80 2.71 2.50 2.39 2.44 2.31 2.37 2.43 2.36 2.98 3.15 2.98 3.00 3.04 3.33 2.73 SOURCE: Authors’ calculations of decennial census data. increasing in California while they have been declining in the rest of the United States. Increases in household sizes in California appear to be driven at least partly, perhaps solely, by changes in the demographic composition of the state’s population. In particular, immigrants, especially Latino immigrants, tend to have larger families and more often live with extended family members than do U.S.-born residents of the state. Large flows of immigrants accounted for much of California’s population growth in the 1980s and 1990s. In 2000, household sizes were highest in counties with large Latino populations, including Imperial, San Benito, and the San Joaquin Valley counties. Counties with declining and relatively low household sizes were primarily rural, mostly white non-Hispanic, counties in the northern and mountainous 27 areas of the state. Areas of the state with the most expensive housing are not necessarily those with the greatest household sizes. For example, San Francisco, Marin, Alameda, and San Mateo Counties all had average household sizes lower than the state average in 2000. Higher household sizes and crowding seem to be more driven by demographic effects than prices (see Moller et al., 2002). Declines in Vacancy Rates Population growth can also be accommodated as formerly unoccupied housing units become occupied. Vacancy rates declined substantially in California and in almost all of the state’s counties between 1990 and 2000 (Table 2.6).8 Low vacancy rates are an indication of housing demand outstripping supply. The percentage of homes for sale is only slightly lower in California than in the entire United States, but the percentage of rental units available for rent is much lower. Homeowner vacancy rates were exceptionally low (0.8 percent or less) in most Bay Area counties (including San Mateo, Santa Clara, Alameda, Marin, Contra Costa, San Francisco, and Sonoma) and in the Central Coast (Santa Barbara and Santa Cruz Counties). The tight supply of houses in these areas at the time of the 2000 census was reflected in rising home prices. The number and percentage of homes for sale was relatively high in the northern mountain counties of the state but also in the Inland Empire and Kern County. Rental vacancy rates show similar geographic patterns, with exceptionally low rates—less than 3 percent—in most of the Bay Area (particularly San Mateo, Santa Clara, Marin, Sonoma, Alameda, San Francisco, Contra Costa, and Napa Counties) and all of the Central Coast (including Santa Cruz, San Benito, Santa Barbara, and Monterey Counties). In Southern California, Ventura County had a vacancy rate of only 2.6 percent. However, in 20 of California’s counties, rental vacancy rates actually rose between 1990 and 2000. In much of the San ____________ 8The vacancy rate is the percentage of housing units that are either for rent or for sale. The homeowner vacancy rate is the number of housing units for sale divided by the number of owner-occupied housing units plus the number for sale. The rental vacancy rate is the number of housing units for rent divided by the total number of housing units that are rented or available for rent. 28 Table 2.6 Vacancy Rates in California’s Counties, 1990 and 2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa Homeowner Vacancy Rate 1990 2000 2.1 1.7 2.1 1.8 2.0 1.4 3.1 3.0 1.3 1.7 3.9 4.1 2.2 4.1 1.1 1.4 3.6 5.1 1.8 1.3 2.2 2.9 1.9 0.7 2.1 3.0 2.4 3.8 1.4 2.1 1.0 2.3 0.9 1.5 1.5 2.2 1.0 1.5 1.4 2.3 1.0 1.8 2.0 1.2 1.6 1.2 1.5 1.4 1.0 0.9 1.5 1.6 2.1 2.6 1.3 1.8 1.4 1.7 1.0 1.4 1.8 1.2 2.2 1.3 1.3 1.8 1.4 0.7 1.7 0.8 1.7 0.7 2.0 1.3 Rental Vacancy Rate 1990 2000 8.5 6.8 8.9 7.3 5.9 3.7 8.7 10.6 4.8 4.7 6.7 10.3 9.6 13.4 5.1 3.3 7.8 9.3 5.2 3.1 7.6 9.9 12.0 11.3 7.8 9.2 10.2 8.5 4.0 5.2 4.7 3.0 3.2 8.2 4.0 5.9 4.4 4.8 5.4 8.6 4.7 6.7 5.3 5.8 7.5 6.4 6.8 4.8 3.6 3.4 5.5 5.5 6.4 8.2 5.9 5.6 3.3 4.5 3.3 4.2 4.5 3.8 4.8 3.2 4.1 5.8 5.5 2.5 6.4 2.7 4.0 2.2 4.8 2.8 29 Table 2.6 (continued) San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Homeowner Vacancy Rate 1990 2000 1.7 0.8 1.7 0.5 1.4 0.5 2.1 0.9 1.7 0.8 2.2 1.4 3.1 1.0 2.8 1.1 2.0 0.8 2.4 0.8 1.5 0.9 2.1 1.9 3.6 2.1 3.0 1.8 2.1 2.4 5.1 2.1 2.3 2.2 4.9 2.5 3.2 3.1 1.9 1.6 1.8 0.9 2.0 0.9 1.6 1.4 2.0 1.0 Rental Vacancy Rate 1990 2000 5.7 2.5 4.3 1.8 4.5 1.8 6.0 3.7 5.1 2.4 3.8 2.9 4.3 2.7 5.8 3.2 5.0 2.8 4.3 2.5 55.3 8.4 5.3 4.4 6.2 6.2 4.9 6.9 14.5 7.7 37.6 20.9 5.9 6.9 9.9 7.2 8.8 7.3 5.9 3.3 6.6 3.0 4.9 2.6 5.0 4.9 6.2 3.1 SOURCE: Authors’ calculations of decennial census data. Joaquin Valley, the Upper Sacramento Valley, and the northern mountain counties, rental vacancy rates increased. In 2000, rental vacancy rates were relatively high in the Inland Empire, but not as high as in 1990. In the Bay Area, rental vacancy rates have risen substantially since 2000. Jobs and Housing Another way to consider whether counties are building a sufficient number of new housing units is to examine changes in jobs relative to 30 changes in housing units. Areas that add a significant number of jobs but not a significant amount of housing could be said to be creating a jobs-housing imbalance. Table 2.7 provides ratios of jobs per housing unit in 1990 and 2000. Only three counties had both a jobs-housing ratio greater than the state’s in 1990 and experienced substantial increases in the ratio in the 1990s (San Mateo, Santa Clara, and Yolo). Another three counties had a jobs-housing ratio greater than the state’s in 1990 and experienced moderate increases in the 1990s (Santa Barbara, Monterey, and Orange). Together, these six counties could be said to have accommodated substantial numbers of new jobs, but not substantial numbers of new housing units. Only eight of California’s counties experienced a decline in the ratio, and all but one are relatively small counties. The one exception is Los Angeles County. Notably, Los Angeles County experienced a decline in its jobs-housing ratio, so that by 2000 its ratio was similar to that of the state as a whole. Los Angeles County’s role as a job center in Southern California waned during the 1990s, with that county being the locus of California’s severe recession of the early 1990s. Summary Substantial variation in housing conditions exists across the state, although some common patterns emerge. Most notably, we find some indicators of housing demand outstripping supply: New housing production did not keep pace with population growth; already low vacancy rates in many counties declined even further in the 1990s, especially for rental units; real prices increased in most counties in the 1990s; and households in the state became more crowded in most counties. However, for many indicators, the trends of the 1990s were much less notable than those of the 1980s. For example, price increases were much more substantial in the 1980s than in the 1990s. Within the state, the Bay Area often stood out, with other housing markets in other regions of the state showing few if any signs of a lack of overall supply. The general picture is one of a tight housing market, but with the 1980s exhibiting more remarkable changes than the 1990s. 31 Table 2.7 Nonfarm Jobs per Housing Unit in California’s Counties, 1990 and 2000 California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo 1990 1.15 0.77 0.88 0.38 0.78 0.83 0.54 0.56 0.54 0.42 0.72 0.42 0.80 1.07 0.84 0.84 0.64 0.59 0.72 0.49 0.78 1.11 1.18 0.95 1.01 0.97 0.81 0.93 1.01 1.00 1.07 0.56 0.89 0.94 0.95 1.70 1.17 2000 1.22 0.75 0.90 0.42 0.81 0.89 0.59 0.64 0.54 0.46 0.66 0.41 0.86 1.12 0.78 0.88 0.73 0.71 0.71 0.62 1.01 1.18 1.42 1.00 1.04 1.03 0.97 0.93 1.08 1.07 1.12 0.63 0.95 1.08 1.18 1.73 1.44 Change 0.07 –0.01 0.03 0.04 0.03 0.06 0.06 0.08 0.00 0.03 –0.06 –0.02 0.06 0.05 –0.06 0.04 0.09 0.11 –0.01 0.13 0.23 0.06 0.23 0.05 0.03 0.06 0.15 0.00 0.07 0.07 0.05 0.06 0.06 0.14 0.23 0.03 0.27 32 Table 2.7 (continued) Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1990 1.52 0.80 0.90 1.15 0.91 0.86 1.16 1.03 0.52 0.65 0.36 0.83 0.62 0.48 0.56 0.66 0.76 1.31 1.35 1.08 1.23 1.03 2000 1.78 0.85 1.06 1.25 0.92 0.97 1.26 1.07 0.65 0.74 0.35 0.85 0.55 0.54 0.56 0.80 0.91 1.25 1.44 1.17 1.13 1.16 Change 0.26 0.05 0.16 0.11 0.01 0.11 0.11 0.04 0.12 0.10 0.00 0.02 –0.07 0.06 0.00 0.14 0.14 –0.06 0.10 0.09 –0.09 0.13 SOURCE: Authors’ calculations of decennial census data and California Employment Development Department data. 33 3. Can Macroeconomic Factors Explain California’s Housing Markets? In this chapter, we seek to determine whether the apparent lack of new housing construction in the late 1990s can be mostly explained by changes in macroeconomic factors, such as interest rates, unemployment rates, income, and prices. These economic factors are determined by market changes in the U.S. economy and in the rest of the world, and changes in the U.S. economic policies. To analyze these influences, we use a simple economic model of the housing market. Housing production in the United States and California is characterized by its cyclical behavior (periods of rapid increases in construction followed by periods of contraction), with large annual fluctuations in new housing construction. These features are captured in our model, which considers the influence of macroeconomic factors on construction in the short and longer term. Our analysis of macroeconomic influences on housing reveals that much of the puzzling trend observed in new construction in California in the 1990s (i.e., slow growth in housing production in a context of rapid economic growth) is a result of the unusual economic environment driven by major increases in productivity, technological breakthroughs, macroeconomic policies, and financial events of that period along with the demographic determinants of demand discussed in the next chapter. The portion of the changes in new construction that are not explained by these factors might be related to specific characteristics of local markets, such as local land-use policies and zoning decisions regarding new housing. We discuss those factors in the final chapter. In this chapter, we first describe how new construction is affected by macroeconomic events (including fiscal and monetary policies) and examine some major 35 differences in the economic environment between the period before the recession of 1990–1991 and the rest of the 1990s. We then turn to our macroeconomic model. Business Cycles, Macroeconomic Events, and Construction Historically, increases in unemployment are closely related to downturns in construction activity. This point is illustrated by Figure 3.1, which compares new residential construction and unemployment rates in California. The figure shows that in 1974–1975, 1981–1982, and 1990–1993, years of economic recession, construction activity decreased sharply. Conversely, upturns in construction have coincided with economic booms. The figure also shows that from 1983 through 1986, construction grew rapidly, but the recovery in the California housing markets from the U.S. recession that started in 1990 was very slow compared to previous recoveries (see Figure 3.1). 12.0 350 10.0 300 250 8.0 200 6.0 150 4.0 100 2.0 Unemployment rate New construction 50 0.0 0 111211111111111111909999999999999999608899999888777776713197531975975319 SOURCE: Center for Continuing Study of the California Economy (1993). Figure 3.1—California Construction and Employment Cycles Unemployment rate (%) Number of permits (1000s) 36 The robust performance of construction activity during the 1970s and early 1980s coincided with important changes in the demography and the economic environment of the state, which significantly increased the demand for housing units: 1. High population growth with a large increase in the number of people in the prime household formation ages (discussed in the next chapter). 2. Significant increases in per capita income, expectations of future high inflation, and changes in tax policy that increased the attractiveness of housing assets. Rising inflation expectations caused the price of long-lived, tax-favored assets such as housing to jump significantly. Furthermore, in 1986, a tax policy change phased out interest deductions on consumer debt except for that secured by home mortgages. This made ownership of housing even more attractive by providing an avenue for acquiring taxfavored debt. 3. Low real interest rates (which were even negative at times before 1981), making the cost of loans very low. A decrease in the real interest rate decreases the user costs of housing and increases the attractiveness of new investment in housing. Beginning in 1986, construction activity started to decrease when economic growth temporarily slowed down. The economy gained strength in 1987 and 1988 but turned sluggish again in 1989 and 1990, with a consequent decline in construction (see Figure 3.1). Macroeconomic forces that led to the slowdown in construction included a new oil shock, restrictive monetary policy, large federal budget deficits, and concerns about inflation. These factors put upward pressure on interest rates and decreased residential investment. This was a period of tighter credit with lower availability of loans—a problem that was compounded by higher standards of capital requirements imposed by bank regulators. Lending to businesses was considered risky and the collateral value on residential and commercial real estate loans fell with declining real estate values (resulting from less demand for housing). The combination of high interest rates and loss of deposits to money market funds that started in the early 1980s created liquidity problems 37 for many depository institutions that found it increasingly difficult to meet withdrawal demands. For thrift institutions—the main source of mortgage lending—the problem arose principally from borrowing on a short-term basis to make longer-term loans. As rates rose, the thrifts had to pay higher rates to retain deposits, but they could earn the higher yields only as their long-term assets gradually matured and the funds were invested in higher-yielding assets. For commercial banks, the main problem was losses from defaults on international loans, energy development loans, and agricultural loans. Between 1980 and 1984, 189 banks failed—an average of 38 per year—and the number of institutions insured by the Federal Savings and Loan Insurance Corporation fell by 20 percent. Furthermore, in mid-1984 one of the nation’s largest banks had to be rescued by a multibillion-dollar package arranged by the federal regulatory agencies. During the same year, one of the nation’s largest savings and loan associations ran into trouble. Starting in early 1988, the Federal Reserve began to raise short-term interest rates. This together with weak conditions in the credit market strongly affected construction activity in California, which declined sharply by the second half of 1989. Long-term investments became less attractive, the economy slowed down considerably, and domestic outmigration sharply reduced the demand for housing. Housing prices started to fall. In 1989, policymakers responded to sluggish growth (and lower inflation) with a more expansionary monetary policy. The Federal Reserve began to cut short-term interest rates in mid-1989 and then cut rates more aggressively in late 1990, for a cumulative decline of about 4 percentage points by December 1990. The U.S. economy started growing modestly and unevenly during 1992, when core inflation and interest rates were at their lowest levels in a generation. Since 1992, the Federal Reserve has been successful in pursuing lower interest rates and keeping inflation under control. By the end of 1993, inflation was low and the public’s expectations of higher prices declined significantly. Congress passed a significant long-term deficit reduction package, raising taxes and lowering spending for the next five years. A strong deficit reduction allowed the Federal Reserve to continue keeping interest rates low. 38 Macroeconomic policies in the mid-1990s helped both the U.S. and California economies by curbing inflationary expectations and influencing financial markets. For example, starting in mid-1996, monetary authorities signaled that they would control potential inflationary pressures from the tight labor markets (as low unemployment obligates firms to offer higher wages). In 1997, Federal Reserve policymakers publicly agreed that there was still room for economic growth driven by increases in productivity. The federal budget deficit decreased sharply and capital investment grew significantly as did labor productivity. By the mid-1990s the performance of the U.S. economy was excellent. The economy was growing rapidly, the unemployment rate declined to its lowest level in 25 years, and inflation was not a problem. This was a period of low interest rates and easy credit. The stock markets were booming, fostered by optimism stemming from a series of technological breakthroughs in electronics, biotechnology, and communications. However, construction activity in both the United States and California did not pick up as fast as in previous recoveries (see Figures 2.6 and 3.1). Why, despite record-low mortgage rates, were the housing markets, particularly in the single-family sector, not keeping pace with the accelerated economic growth? The trend observed after 1990 is largely the result of the unusual events that took place in three areas: (1) financial events and macroeconomic policies related to balancing the budget and interest rate adjustments targeted by the Federal Reserve, (2) economic changes that deeply affected the California economy, and (3) demographic changes in the population. The weak recovery in new construction of the 1990s is at least partly explained by the public’s expectations of significantly lower inflation resulting from increases in productivity, the way monetary policy was conducted, and the climate of optimism in the financial markets that drove the attention of investors to alternative markets. These expectations of low inflation and strong stock market returns made investments in real estate much less attractive relative to other investments. Not only did these macroeconomic factors curb expectations of housing appreciation, the demographic composition of 39 Expected inflation rate (%) the state (as discussed in the next chapter) also reduced expectations of future returns on housing construction. Builders and developers integrated a demographically driven slowdown in the demand for housing in their projections of future real estate values. Figure 3.2 illustrates the point that expectations of future inflation were decreasing.1 Figure 3.3. shows the difference between long-term and short-term interest rates. Decreases in this difference are associated with lower levels of construction, as investors prefer shorter-term investments when the yields of longer-term investments decrease.2 This difference shrank significantly after 1994 and remained unusually low until 2000, given the economic performance of the economy of these years. The severity and length of the California recession together with the demographic composition of the state also help explain why construction in California took longer to recover. The U.S. recession that started by 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 SOURCE: Livingston Survey (described in Appendix A). Figure 3.2—Expectations of Inflation, 1972–2001 ____________ 1The expectation variable is calculated as the median one-year-ahead expected inflation rates as reported by the Livingston Survey (see Appendix A) 2See Appendix A. 40 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Interest rate (%) 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 4.00 3.00 2.00 1.00 0.00 –1.00 –2.00 –3.00 SOURCE: Federal Reserve. Figure 3.3—Term Structure (Difference Between the Long-Term and Short-Term Interest Rates), 1971–2001 mid-1990 and finished in March 1991 had a deep and long-term effect on the state. For California, the 1990 recession was the longest recession since the Great Depression of the 1930s, whereas the U.S. contraction lasted less than a year. By the end of 1993, the U.S. economic recovery was well established, but the California economy was still weak. Consequently, California had a much greater downturn in construction activity than other states. Home prices dropped dramatically in highpriced areas including Los Angeles, San Diego, and Orange Counties. The California recovery was slower than that in the rest of the country because it was hampered by several factors, including (1) large household and business debt burdens, (2) high vacancy rates in commercial real estate, (3) tight credit practices by many lenders, (4) stagnant growth in much of the rest of the world, reducing the demand for California exports, (5) a dramatic outflow of domestic migrants, and (6) a decrease in federal purchases, especially of military goods and services. 41 Because of California’s high share of U.S. aerospace employment, the state was also particularly affected by the federal defense-spending cuts that were initiated in 1986 and intensified by 1989. Between the second quarter of 1990 and the second quarter of 1993, California lost 528,000 jobs—142,000 in the aerospace industry alone. Most of the job losses took place in Southern California. This sharp slowdown in economic activity also negatively affected residential construction, particularly in Southern California. A Simple Housing Market Model To systematically evaluate the effect of macroeconomic factors on housing markets in California, we develop a model of the state’s housing markets.3 Two specific features in our analysis make it different from previous research. First, we address the effects of macroeconomic factors on both sides of the market for new housing (demand and supply). Most of the studies dealing with the housing sector tend to concentrate on the demand side (Hanushek and Quigley, 1982; Mayer and Somerville, 1996). Second, unlike cross-sectional analyses of data for specific regions or cities, we use a time-series analysis because we seek to understand why new housing construction in California was more sluggish in the 1990s than in other periods of economic growth. This approach allows us to evaluate the influence of macroeconomic factors in explaining the annual variations in new housing construction in California. To control other influences on the housing market, such as quality of life characteristics and the influence of local policies, we divided California into ten regions sharing similar quality-of-life and local policy traits. The determination of each region is explained below. We focus on changes in new construction of total housing units, as measured by the number of permits issued for housing units. Some authors separate single-family construction from multifamily data in an effort to account for potential differences between the two. However, others do not separate these sectors because the substitution between single-family and multifamily units is very ambiguous (Boldin, 1993). We followed the latter approach. Finally, we developed an alternative econometric model using ____________ 3Our housing market model and the methodology are presented in Appendix B. 42 a panel analysis. The results of that model are consistent with the model presented here and are discussed in Appendix C. Factors Affecting Housing Demand We included the following macroeconomic factors affecting the aggregate demand for housing: changes in income, employment, and interest rates (cost of mortgages), actual and expected appreciation of real estate (housing price increases), and the price of alternative investments (stock prices). We also included controls for the growth and demographic composition of the population, although changes in these variables tend to occur more slowly. Measures of income and unemployment levels are usually incorporated in typical housing demand equations. As income and employment increase, the demand for new housing is expected to increase. Higher current housing prices decrease the current demand for housing, whereas expectations of higher future housing prices increase demand. Our model describes expected appreciation or expected housing valuation as a function of expected inflation (see Appendix A). Furthermore, because housing is both a commodity and an investment, real estate can be seen as a substitute asset for stocks, particularly when inflation is high. In periods of high economic growth, housing and stock prices tend to move together; however, in inflationary environments, houses have done better than stocks. Historically, stocks have better returns when inflation is lower (Wasserman, 1998). Thus, returns on stocks are also a determinant of housing demand because they are an alternative investment. Typically, housing demand or supply equations use either mortgage rates or the prime rate as a proxy for the cost of credit. The higher the interest rate, the lower the demand for housing. We looked at the difference between the long-term and short-term interest rate because it measures the tightness of credit better than the level of either interest rate separately. It is also a better measure because a large spread between the long- and short-term interest rates indicates expectations of rising inflation. The relationship between this spread and the demand of housing is expected to be positive. (See Appendix A for a more complete explanation.) 43 Changes in population growth, the age structure of the population, and immigration also affect the demand for housing. As the population grows, more housing is needed, but a high proportion of children in the population leads to smaller increases in housing demand. These changes are analyzed directly and in more detail in the householder rate models of Chapter 4. Demographic changes also affect expectations of future housing prices—a key determinant of new construction. In addition to macroeconomic and demographic forces, other factors influence the demand for housing in specific areas, such as the availability of amenities and the quality of life in a region. For example, housing is more expensive in areas with good weather, art centers, parks, cinemas, and low crime rates. However, measuring the effect of quality of life in the demand for housing is very difficult, and economists have studied these effects from different angles. With the exception of Gabriel, Mattey, and Wascher (1996 and 1999), who constructed a time-series of quality-of-life rankings, virtually all the research on quality-of-life estimates has focused on the relative ranking of localities at a single point in time. Most economists have entered the quality-of-life debate arguing that price differences across locations in wages or land rents should compensate for the differences in the quality characteristics of locations. Rosen (1979) and Roback (1982) inferred quality levels from differences in the price of housing across locations. Other authors following similar approaches include Hoehn et al. (1987), Blomquist et al. (1988), Gyourko and Tracy (1992), and Stover and Leven (1992). We control for differences in quality of life by dividing the state into regions with similar geographic and socioeconomic characteristics. The determination of these regions is explained later in this chapter. Supply Determinants The supply side of the housing market is also affected by changes in expected housing appreciation, changes in economic activity (unemployment), construction costs, interest rates, the availability of land, and local policies. Housing appreciation leads to higher levels of new construction. Because the cost of capital is the main cost to developers and builders, interest rates and new construction are negatively related. The 44 higher construction costs are, the lower the level of new construction. High economic growth, lower stock prices relative to real estate prices, and expectation of higher returns on long-lived assets have a positive influence on the decision to invest in real estate. The amount of available land for new residential developments also determines the amount of new construction in an area. Because we model changes in housing construction from one year to the next, we assume zero changes in available land (see the model description in Appendix A). Local policies also influence the supply of housing. These policies include slow-growth ordinances and the preference of local governments for commercial rather than residential development because of public finance pressures. Using two surveys of 490 Californian cities and counties, Levine (1999) examined the effects of local growth-control enactment between 1979 and 1988 on net housing construction. He found that local growth-management measures displaced new construction, particularly rental housing. He also found that measures that limited available land or downsized existing zoning had stronger effects than other measures. Ten Regional Housing Markets in California To control differences attributable to quality of life and local policies, we grouped counties into ten regional markets and tested our model for each of these regions. These regions have common characteristics related to quality of life as well as the type of regulations (measures of growth control) enacted in their localities. Our regions are basically the same as those used by Glickfeld and Levine in their 1990 study on land use regulation. These regions and their constituent counties are 1. North Coast: Del Norte, Humboldt, Mendocino 2. North Central: Butte, Colusa, Glenn, Lake, Shasta, Siskiyou, Sutter, Tehama, Trinity, Yuba 3. North Eastern: Lassen, Modoc, Nevada, Plumas, Sierra 4. San Francisco Bay Area: Alameda, Contra Costa, Marin, Napa, San Mateo, Santa Clara, Santa Cruz, Solano, Sonoma, San Francisco 45 5. Stockton-Sacramento: El Dorado, Placer, Sacramento, San Joaquin, Stanislaus, Yolo 6. Central Coast: Monterey, San Luis Obispo, Santa Barbara, El Dorado 7. Central Inland: Fresno, Kern, Kings, Merced, San Benito, Tulare 8. Central Eastern: Alpine, Amador, Calaveras, Inyo, Madera, Mariposa, Mono, Tuolumne 9. Los Angeles–San Bernardino: Ventura, Los Angeles, Orange, San Bernardino, Riverside 10. San Diego Following the Glickfeld and Levine regions allows us to qualitatively consider the role of land-use regulations as elaborated in their work. Results of the Model of California Housing Markets Our simple housing market model indicates that macroeconomic policies and global events explain a large part of the annual changes in new construction in California and in specific regions. Most of the explanatory factors showed the expected signs, but not all factors turned out to be significant in each region.4 Table 3.1 summarizes the results of the model, showing the relationship (positive or negative) between demographic and macroeconomic factors and changes in new home construction for each of the ten regions and the state.5 Results are shown for the lagged effects model.6 Because not all determinant factors affect demand and supply of housing in all California regions at the same speed or intensity, lags are necessary to understand those effects. For ____________ 4Our results are solid, considering that (1) we are dealing with first differences rather than levels, (2) the short historical period for our time-series analysis, and (3) the limitations imposed by data quality (such as county unemployment rates and costs). 5Appendix Table A.2 shows the coefficients of each determinant and the level of significance of each factor included in that estimating equation. 6The results obtained when we used lags were significantly better than the ones obtained when we related current changes in the determinants of housing to current changes in new housing construction (shown in Appendix Table A.1). 46 Table 3.1 Effect of Explanatory Factors on Changes in New Housing Construction: Results of the Macroeconomic Model, by Region 47 Explanatory Factor Los Angeles– North North North Stockton- Central Central Central San San Coast Central Eastern Bay Area Sacramento Coast Eastern Inland Bernardino Diego California Change in population ++ ++ ++ Change in income ++ (a) + Lagged new construction – – – – – – Change in unemployment – –– – Change in difference of long-term/short-term interest rate + + ++ Change in appreciation (expected inflation) + + ++ Change in real stock prices – –– –– Change in real cost – – –– R squared (variation explained) 0.67 0.78 0.78 + + –– –– ++ –– – – 0.76 ++ ++ ++ ++ ++ – + ++ –– –– –– –– –– –– – –– ++ ++ – ++ + ++ – –– –– –– –– –– – –– –– 0.85 0.77 0.90 0.72 +– + + + ++ ––– ––– ++ ++ ++ ––– – – –– –+– 0.77 0.76 0.83 SOURCE: Appendix Table A.1. NOTES: ++ means that the relationship is positive and statistically significant. + means that the relationship is positive but not statistically significant. – means that the relationship is negative but not statistically significant. – – means that the relationship is negative and statistically significant. aVariable not included because of colinearity with real cost. example, some regions may respond faster than others to changes in interest rates, income, or expected appreciation. Some factors may affect some regional housing markets more rapidly than others. Increases in the difference between the long- and short-term interest rates—a measure of credit availability—may stimulate new construction in the following year if, for example, the decisions to build are made one year before construction, or perhaps 18 months or more before the date of the permit issue. Thus, changes in the price of alternative investments, expected appreciation, and even the level of income may not affect construction inmediately but one year or more later. Moreover, population changes may affect the market more rapidly by changing market prices. First, we discuss our findings for each of the factors in the model, and then discuss the results more generally. Population growth did not turn out to be a significant factor in the general equation for California, but it was a statistically significant factor in various regions. Results did not change significantly when we introduced the number of non-Hispanic adults rather than total population. (The demographic model presented in the next chapter evaluates the effects of California population on housing demand in more detail.) Income appeared to be significant in the explanation of changes in new construction in California and various regions. The Los Angeles– San Bernardino area was one of the exceptions. Income is expected to be a short-run determinant of residential construction. Many authors omit this factor completely from their models on investment on residential construction (see Grebler and Maisel, 1963). However, authors such as Klein (1966) suggest that the correct role for income is through its shortterm effect on the value of housing. If builders start their plans during a recession and income begins to rise rapidly, builders may alter their original plans and could expand or speed their existing construction. On the demand side, higher levels of income are positively related to housing purchases. Lagged new construction (or the change in the housing stock) was a significant factor in the explanation of changes in new construction. 48 Higher increases in the housing stock in the previous year are associated with less current new construction. Changes in new construction were closely related to changes in the business cycle, as measured by changes in unemployment. Increases in unemployment lead to less construction. Costs played a significant role in explaining changes in new construction in only four regions. Our cost measurement showed the expected negative association with new construction (negative signs) in most of our equations. In other words, everything else constant, higher building costs lead to less new construction. Our results corroborate the hypothesis that developments in financial markets determined, in part, the slower pace of new construction in the 1990s. All regional equations showed a negative association between stock prices and changes in new construction. In both California and most regions, stock prices appear to be statistically significant in their association with changes in new construction. Results indicate that investors see real estate and stocks as substitutes. A more refined analysis of this relationship could provide a deeper insight on this issue. The association between annual changes in credit conditions (the difference between long-term and short-term interest rates) and annual changes in construction was statistically significant in most of the equations. The influence of expected appreciation (measured by expected inflation) appeared to be less significant than we would have expected. One possible explanation is that the effects of the difference between long-term and short-term interest rates is already accounting for the effect of changes in expected appreciation. Another possibility is that our proxy for expected inflation is a better measure for the United States as a whole than for California (see Appendix A). Results of our model for the state of California as a whole and for ten California regions suggest that more than 80 percent of annual changes in new construction in California is explained by the influence of macroeconomic policies and demographic changes (see the last row of Table 3.1). This means that there is a large proportion of the annual variation in construction beyond the control of state policy. The 49 Changes in new housing construction (1000s) 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 explained portion varies by region, ranging from 67 percent for the North Coast, 76 percent for the Bay Area, 77 percent for the Los Angeles and San Bernardino area, 85 percent for the Sacramento-Stockton area, and 90 percent for the Central Eastern area. Figure 3.4 shows actual changes in residential new construction and projected values from our California housing model. Generally, our housing market model does a good job of predicting annual changes in new construction, including the low levels recorded in the 1990s. The model performed well particularly in central regions and areas that did not appear to have major shortages. For the most populated regions, such as the Bay Area, San Diego, and the Los Angeles–San Bernardino regions, where the state has the highest cumulative housing shortages, the results are slightly weaker. Perhaps this indicates that housing markets in these regions are heavily determined by local characteristics as well as the influence of local policies and growth controls. 150 Change in new construction Predicted change in new construction 100 50 0 –50 –100 –150 SOURCE: CRIB and time-series model. Figure 3.4—Comparison of Actual Changes in California New Housing Construction and Model Fitted Values, 1967–2000 50 Summary In this chapter, we discussed macroeconomic forces that help determine housing construction and developed an econometric model of new housing construction in California. In short, the results suggest that macroeconomic events of the 1990s can explain much of the slowdown in new housing construction in the state. Our results were consistent statewide and for ten regions of the state, although they were not as robust in some of the state’s largest metropolitan areas. 51 4. Can Demographic Changes Explain California’s Housing Markets? Population growth and housing growth go hand in hand: The more people added to the state, the greater the housing demand. However, not all population growth is equal when it comes to housing demand. Some groups require more housing than other groups. Large increases in the child population and small increases in the adult population (i.e., more children per adult) will lead to less housing demand than will small increases in the child population and large increases in the adult population. Adults form new households, children do not. Similarly, immigrants are more likely to live in extended families. Thus, the amount of housing consumed by immigrants is less, on average, than the amount of housing consumed by U.S. natives.1 In this chapter, we explore the demographic determinants of California’s housing demand. First, we examine demographic trends that shape California’s housing demand. We discuss trends in the most important of those demographic trends, including population change itself, age structure, gender, nativity, and marital status. Next, we construct a householder rate model. The model estimates the probability that a person with a certain set of characteristics will be a householder.2 High probabilities translate into high housing demand, and low probabilities imply low housing demand. The demographic characteristics that determine householder rates include age, race and ____________ 1Average household sizes are higher for immigrants even where housing costs are lower than U.S. averages. 2Each occupied housing unit has one householder. The householder is the person in whose name the housing unit is owned or rented. If more than one person owns or rents the unit, only one member of the household is chosen as the householder. A householder is also referred to as a head of household. 53 ethnicity, nativity, gender, and marital status. (The complete model and data are described in Appendix B.) Finally, to quantify the demographic component of housing demand in California, we develop simulations of housing demand using the householder rate model. In contrast to business cycle effects, demographic changes tend to be felt more strongly in the long run, as annual fluctuations are not large for most demographic measures. Thus, our focus here is on long-term changes. The householder rate model does have limitations. First, changes in vacancy rates are not explicitly considered in the model. In California, some of the increase in the number of occupied households between 1990 and 2000 occurred as formerly vacant housing units became occupied. Second, a restricted supply of housing may prevent some people from moving to California. These potential migrants cannot be measured by this model. The results suggest that the nature of California’s population growth explains much of the apparent lack of new housing in the state. In particular, in the 1990s a relatively fast growth in child and immigrant populations did not translate into the same level of housing demand as did past population growth, which was driven by domestic migrants and baby boomers aging into prime household formation years. Without controlling for demographic characteristics, householder rates have risen in the United States but declined in California. Once we control for demographic determinants of household formation, however, differences between California and the rest of the United States are substantially reduced. Indeed, once we control for demographic characteristics—age being the most important—the direction of change is the same in both the United States and California; in both places, householder rates declined in the 1980s. Moreover, with demographic controls, temporal patterns of change in the 1990s are virtually identical between California and the rest of the United States. These findings indicate that California’s household formation rates were not particularly different from those of the rest of the nation during the 1990s. Thus, although the increase in new housing units might not have kept pace with overall population growth, the nature of the state’s population growth can explain a large part of the apparent discrepancy. From a demographic 54 standpoint, California’s housing markets do not appear to be functioning differently from those in the rest of the United States. Demographic Determinants of Demand Population Growth California’s population growth slowed considerably in the 1990s. On a percentage basis, the 1990s were the state’s slowest-growing decade since records have been kept (see Figure 4.1). Indeed, official growth in the state was only slightly higher than that in the rest of the United States. Absolute gains in the state’s population were large—over 4 million people from 1990 to 2000. In the 1980s, however, the state added over 6 million new residents, and thus absolute gains in the 1990s were over 30 percent lower than those of the 1980s. If we adjust official census tabulations of the population for the respective undercounts in 1980, 1990, and 2000, population growth during the 1990s appears even lower relative to that of the 1980s: Total population change was 7,000 6,000 1980–1990 1990–2000 6,093 5,000 4,000 4,731 4,115 Population change (1000s) 3,000 2,621 2,000 1,000 1,362 1,494 0 Children Adults Total SOURCE: Authors’ calculations of decennial census data. Figure 4.1—Growth in Child and Adult Populations in California, by Decade 55 3.8 million in the 1990s, compared to 6.2 million for the 1980s. (Official census counts place the decennial change at 4.1 million in the 1990s, and 6.1 million in the 1980s.) Slower population growth relative to that of previous decades led to less housing demand in the 1990s than in the 1980s. Age The number of children in California continued to grow substantially during the 1990s, but increases in the number of adults were much lower than during the 1980s. During the 1990s, California’s child population increased 19 percent, compared to only 12 percent for adults in the state. This pattern is in contrast to the 1980s, when the state’s adult population grew faster than the state’s child population. Remarkably, absolute population growth for children was greater in the 1990s than in the 1980s, whereas population growth for adults from one decade to the next declined by almost half (Figure 4.1). Thus, not only did lower population growth during the 1990s lead to less demand for new housing than in the 1980s, but the relatively large share of children in that growth also led to less demographic-based demand. Among adults, the likelihood of forming a household is dependent on age (Figure 4.2). Household formation increases dramatically as people age from young adulthood (people in their 20s) to middle ages (adults between the ages of 35 and 44). Population increases in these age groups will lead to greater housing demand than increases in other age groups. In the 1980s, population growth in California was strong in these age groups, as baby boomers entered these prime householder years (Figure 4.3). During the 1990s, however, the much smaller baby bust cohorts replaced the baby boom cohorts, and the population of 20 to 35 year olds actually declined. Gender Women are less likely to be householders than men—primarily because most married couples list the husband as the householder on census forms rather than the wife. However, over time women have become more likely to form their own households, both because the share of never-married mothers has increased and because of past (before 56 Probability (%) 50.0 45.0 40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 17.5 22.5 27.5 32.5 37.5 42.5 47.5 55 65 75 Age SOURCE: Authors’ calculations of 2001 CPS data. Figure 4.2—Probability of Being a Householder in California, by Age, 2001 the 1990s) increases in divorce rates (Myers, 1992).3 Still, by 2000 only 30 percent of women age 15 and over were householders compared to over 60 percent of men. Population growth in California in the 1990s consisted of more women than men: Among people age 15 and over, population growth totaled 1.6 million for females, compared to only 1.4 million for males. This is in direct contrast to the 1980s, when the population of males age 15 and over increased 2.5 million, compared to only 2.1 million women. More research is necessary to understand why the gender ratio of California’s population growth switched from primarily male to primarily female, but it could be related to the Immigration Reform and Control Act of 1986. This act provided legal status for formerly undocumented immigrants, an overwhelmingly male group. After gaining legal status in the late 1980s, many formerly unauthorized ____________ 3Among married couples, women have been increasingly listed as the householder, although the percentages are still low. 57 1,000 800 1980–1990 1990–2000 Population change (1000s) 600 400 200 0 –200 –400 <5 3124134255667750505050050505––––––––––––––531241342556677–949494944949499 80+ Age Figure 4.3—Change in California Population, by Age Group and Decade immigrants were joined in the United States by their family members, including wives. Whatever the cause, this shift to a greater increase in female than male population growth in the 1990s led to less housing demand than if the increase were more composed of men. Finally, increases in the ages at which men and women marry have implications for housing demand. These factors are taken up below. Nativity and Ethnicity Population growth attributable to immigration leads to less housing consumption than population growth among U.S. natives. Foreign-born residents of California tend to live in larger families and households than U.S. natives. In 2000, the average number of people in households headed by foreign-born persons was 3.5 in California, compared to 2.5 for households headed by U.S. natives.4 In the 1980s, California ____________ 4Authors’ calculations based on 2000 Census Supplementary Survey data. 58 experienced large increases in both foreign-born and U.S.-born populations, as natural increase, domestic migration, and international migration to the state were large. In the 1990s, international migration to and natural increase in California remained strong, even as the state lost many domestic migrants to other states. The result, as shown in Figure 4.4, was a much greater relative increase in the foreign-born population than the U.S.-born population in the 1990s. This change contributed to a decline in the demand for new housing. Changes in the ethnic composition of the state’s population, which are strongly related to migration flows, also dampened new housing demand. For example, non-Hispanic whites are more likely to live alone than other groups and thus tend to consume more housing. Relatively few international migrants to California are non-Hispanic whites, whereas large proportions of California’s domestic out-migrants during the 1990s were white. As a consequence, California’s non-Hispanic white population declined during the 1990s after having increased during the 1980s. This decline had a dampening effect on the demand 3,500 3,000 2,868 3,224 1980–1990 1990–2000 2,500 2,405 Population change (1000s) 2,000 1,500 1,706 1,000 500 0 Foreign-born SOURCE: Authors’ calculations of decennial census data. U.S.-born Figure 4.4—Change in Foreign-Born and U.S.-Born Populations in California, by Decade 59 for new housing in the 1990s. California’s largest population increase in the 1990s occurred among Hispanics, the ethnic group with the lowest householder rates. Marital Status Marital status is also an important determinant of household formation and thus housing demand. Divorced and widowed adults are much more likely to form their own households than never-married adults. Married couples have high household formation rates, but married individuals tend to have lower household formation rates than divorced or widowed adults; it takes two to make a couple, and only one is considered the householder. During the 1990s, the number of married adults increased almost as much as in the 1980s (Figure 4.5). In contrast, increases in the number of adults of other marital statuses were much lower in the 1990s than in the 1980s. The sharp drop in number of divorced adults is partly a reflection of lower divorce rates in the 2,500 2,000 1,500 1,984 1,647 1,790 1980–1990 1990–2000 Population change (1000s) 1,000 871 700 500 326 0 Married Never married Divorced/Separated SOURCE: Authors’ calculations of decennial census data. NOTE: Children younger than age 15 are excluded. 174 72 Widowed Figure 4.5—Decennial Change in Population in California, by Marital Status and Decade 60 1990s. These changes in marital status in the 1990s tend to dampen housing demand. Householder Rate Model: Temporal Trends in the Probability of Forming a Household The demographic changes discussed above have consequences for the demand for housing in California. To quantify these consequences altogether and evaluate temporal changes in household formation, we develop statistical models to estimate the probability that an individual has formed his or her own household. The probability of being a householder, also referred to as the householder rate, depends on an individual’s demographic characteristics. Higher probabilities are associated with more housing, lower probabilities with less housing. Because householder rates, and thus housing consumption, vary with demographic characteristics, California’s unique demographic changes have important implications for housing demand in the state. We compare temporal trends in California with those in the rest of the country. To consider the role of various demographic factors, we adjust those trends first by controlling for temporal differences in the demographic composition of the population and second by controlling for differences between California’s population and that of the rest of the country. Details of the data and methods can be found in Appendix B. Temporal Trends in Householder Rates Across time and without controlling for demographic factors, householder rates have been moving in opposite directions in California and the rest of the nation. In California, rates have been decreasing, whereas in the rest of the United States they have been increasing. By the start of the 21st century, householder rates in California were lower than at any time in the past quarter century. At the same time, householder rates in the United States were higher than at anytime in the past 25 years (Figure 4.6). In other words, persons age 15 and over are less likely to be the head of a household in California in 2001 than in 1976, whereas that probability has risen in the rest of the United States. The declines in California and increases in the United States occurred 61 Probability 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 0.56 0.54 California Rest of United States 0.52 0.50 0.48 0.46 0.44 0.42 0.40 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Children younger than age 15 are excluded. Figure 4.6—Probability of Being a Householder in California and the Rest of the United States, 1976–2001 almost entirely in the late 1980s. In both the United States and California, householder rates changed little in the 1990s. This difference in overall householder rates translates directly into more people per household in California than in the rest of the nation. Many have concluded that this difference reflects California’s unique housing crisis. However, the pattern of change in householder rates in the United States is somewhat different when we control for demographic changes in the population over time. These demographic controls allow us to evaluate the effect of various demographic determinants on the probability of being a householder. Controlling for gender, age, and ethnicity, we observe much less temporal change in householder rates in the rest of the United States than in the uncontrolled model (Figure 4.7). Among the demographic shifts that have led to higher householder rates in the nation, the most important factor is age. In 1975, all baby boomers were children or young adults (younger than age 30), ages at which householder rates are relatively low. By 2001, baby boomers were ages 37 to 55, ages at which householder rates are high. Simply put, in most of the nation, housing consumption outpaced growth in the 62 Probability 0.56 0.54 No controls Controls for age and gender 0.52 Controls for age, gender, and ethnicity 0.50 0.48 0.46 0.44 0.42 0.40 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Controls maintain the demographic composition of the rest of the United States in 1976. Figure 4.7—The Effect of Demographic Changes on the Probability of Being a Householder in the Rest of the United States, 1976–2001 nation’s population because baby boomers entered their prime years of household formation. If we control for changes in age structure over time, householder rates in most of the United States would have changed very little over the past 25 years. In contrast, changes in the ethnic composition of the nation’s population have not played a very important role. This is at least partly because those changes have not been very pronounced in most of the nation. Thus, much of the increase in householder rates from the late 1970s to the late 1990s in most of the United States is simply an artifact of demographic changes in the nation’s population, specifically the aging of the baby boomers. Controlling for demographic factors in California also changes temporal patterns in overall householder rates. Controlling for age and gender, householder rates in the state declined tremendously from the late 1970s to the early 1990s (Figure 4.8). Were it not for the aging of the baby boomers, householder rates would have declined even more than they actually did. That is, once we control for age, we see even more dramatic declines in the probability of being a householder over 63 Probability 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 0.56 0.54 No controls Controls for age, gender, and ethnicity 0.52 Controls for age and gender 0.50 0.48 0.46 0.44 0.42 0.40 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Controls maintain the demographic composition of California in 1976. Figure 4.8—The Effect of Demographic Changes on the Probability of Being a Householder in California, 1976–2001 time than in the model with no controls. This strong effect of the baby boom is similar to that observed for the rest of the United States. Unlike the case in the rest of the nation, however, the racial and ethnic composition of the state’s population also affected the temporal pattern of change in householder rates. In California, the state’s changing racial and ethnic population mattered almost as much as age. Declines in the probability of being a householder are less pronounced once we control for race and ethnicity. This means that part of the decline in householder rates in California could be attributed to a demographic shift in the state’s population from racial and ethnic groups that have relatively high householder rates to groups that have relatively low householder rates. Thus, declines in householder rates in California would have been even more severe were it not for the baby boomers aging into prime householder years. However, those declines would have been less substantial if the racial and ethnic composition of the state had not changed so much. 64 Differences in Householder Rates Between California and the Rest of the United States Differences in householder rates between California and the rest of the country provide an indicator of housing consumption in California and the rest of the country. When the difference is positive, Californians could be said to be consuming more housing in comparison to people in the rest of the country, and when the difference is negative, Californians are consuming less. In the late 1970s, California had substantially higher overall householder rates than the rest of the nation (Figure 4.9, no controls). By the mid-1990s and continuing to 2001, the state had householder rates that were substantially lower than in the rest of the United States. The dramatic decline in householder rates in California relative to the rest of the nation is still evident when we control for demographic differences between California and the rest of the United States (Figure 4.9, controls). In the late 1970s, Californians were much more likely to be householders than people of similar age, gender, and ethnicity in the 0.08 Controls for age, gender, and ethnicity 0.06 Controls for age, gender, ethnicity, and nativity No controls 0.04 Difference in probabilities (%) 0.02 0 –0.02 –0.04 1975 1980 1985 1990 1995 2000 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: These lines represent householder rates in California minus those in the rest of the United States. Figure 4.9—Difference in Householder Probabilities Between California and the Rest of the United States 65 rest of the nation. By 1993, householder rates were lower in California, even controlling for demographic differences. However, householder rates increased in California relative to rates in the nation in the late 1990s, so that by 1999 householder rates in California were similar to those in the rest of the nation once we control for demographic differences. In other words, when we compare Californians to their demographic counterparts in the rest of the nation, the probability of being a householder is about the same in both places. Thus, California has changed from a place with unusually high householder rates—given its population composition—to a place with householder rates that are what we would expect given the state’s population composition in comparison with the rest of the country. Today, California’s demography explains much of the apparent lack of housing consumption in the state relative to the nation. The importance of the state’s unique demography is evident when we examine householder rates for population subgroups. Overall, householder rates are almost 3 percentage points lower in California than in the rest of the United States (Figure 4.10).5 However, for U.S.-born adults, householder rates are similar in California and the rest of the United States (50.7 percent in California compared to 50.6 percent in the rest of the United States). Indeed, among U.S.-born groups, whites and Asians in California actually have higher householder rates than their counterparts in the rest of the nation. California’s lower overall householder rates are due to lower rates among U.S.-born Latinos and the foreign-born. In particular, recent Mexican immigrants have extremely low householder rates throughout the United States. Those rates are even lower in California than in the rest of the United States (25.6 percent versus 30.0 percent, not shown in Figure 4.10). Combined with California’s relatively high share of Mexican immigrants, these low rates explain much of California’s apparent shortfall in householder rates. ____________ 5The Census Supplementary Survey has higher householder rates in the rest of the United States and lower ones in California than the 1999 or 2001 CPS. Differences in age-standardized rates are 3.1 percent in the Census Supplementary Survey, versus 1.7 and 2.3 in the 1999 and 2001 Current Population Surveys. 66 0% Total 10% 20% U.S.-born Foreign-born White U.S.-born White foreign-born Latino U.S.-born Latino foreign-born Asian and Pacific Islander U.S.-born Asian and Pacific Islander foreign-born African American Rest of United States 30% California 40% 50% 60% 50.0% 47.4% 50.7% 50.4% 44.4% 40.6% 50.3% 51.4% 46.8% 46.3% 49.1% 45.2% 41.7% 38.9% 47.8% 48.0% 40.5% 37.7% 53.8% 52.6% SOURCE: Authors’ estimates based on 1999–2001 CPS and the 2000 Census Supplementary Survey. Figure 4.10—Age-Standardized Householder Rates in California and the Rest of the United States, by Ethnicity and Nativity, 2000 Householder Rates in California Metropolitan Areas Householder rates vary across California’s regions as well. Thus, although we find that demographic factors explain much if not all of the difference in householder rates between California and the rest of the United States, the picture for specific metropolitan areas within the state might also differ. Using the Current Population Surveys from 1999 to 2001, we compare householder rates for primary metropolitan statistical areas in California with householder rates in the rest of the country. Appendix B describes the methods used to develop the estimates. Without controlling for demographic differences, the probability of being a householder is lower in most of California’s metropolitan areas than in the rest of the nation (Figure 4.11a). Among the state’s larger metropolitan areas, only Sacramento and Fresno appear to have householder rates that are similar to those in the rest of the country. 67 Lower rates in California Higher rates in California –0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 San Luis Obispo Chico Stockton Fresno Sacramento Rest of California Merced Modesto Yuba City–Marysville Visalia Santa Rosa Riverside–San Bernardino Oakland Los Angeles–Long Beach Bakersfield Orange County Ventura Vallejo–Napa San Jose San Diego Salinas–Monterey Davis–Woodland Santa Barbara San Francisco NOTE: Figure shows point estimates and 90 percent confidence intervals. Figure 4.11a—Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, No Demographic Controls, 2000 68 Lower rates in California Higher rates in California –0.20 –0.15 –0.10 –0.05 0 0.05 0.10 0.15 0.20 San Luis Obispo Chico Stockton Fresno Sacramento Rest of California Merced Modesto Yuba City–Marysville Visalia Santa Rosa Riverside–San Bernardino Oakland Los Angeles–Long Beach Bakersfield Orange County Ventura Vallejo–Napa San Jose San Diego Salinas–Monterey Davis–Woodland Santa Barbara San Francisco NOTE: Figure shows point estimates and 90 percent confidence intervals. Figure 4.11b—Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, with Controls for Age, Gender, Ethnicity, and Nativity, 2000 69 Thirteen metropolitan areas had householder rates that were significantly lower than rates in the rest of the country, including Los Angeles–Long Beach, San Francisco, San Diego, Orange County, Riverside–San Bernardino, and San Jose. Controlling for demographic factors, however, a different picture emerges. In 11 metropolitan areas, householder rates are either higher than or no different from rates in the rest of the country, and in another six metropolitan areas, householder rates are not significantly different from those in the rest of the United States (Figure 4.11b). Only six metropolitan areas had lower estimated householder rates than in the rest of the country when we control for demographic composition. Nonetheless, those six include some of the state’s largest metropolitan areas: Los Angeles–Long Beach, San Francisco, Orange County, San Diego, and San Jose. In those metropolitan areas, adults are less likely to be householders than their demographic counterparts in the rest of the country. The effect of the demographic controls varies by metropolitan area. In Los Angeles–Long Beach, the controls erase most of the difference in householder rates between the metropolitan area and the rest of the nation, whereas in San Francisco the demographic controls make little difference. This, of course, is due to the nature of the population in these two metropolitan areas. Los Angeles–Long Beach has a much larger share of demographic subgroups that tend to have low householder rates, especially foreign-born Latinos, than does the rest of the country or San Francisco. Estimating the Shortage of Occupied Housing Units We use the results of our statistical analyses to estimate the state’s shortage of occupied housing units by comparing householder rates in California to those in the rest of the United States. We do so once without controlling for demographic differences and again controlling for demographic differences. In essence, we answer the following questions: • How many occupied housing units would California have if it had the same overall householder rates as those in the rest of the country? 70 • How many occupied housing units would California have if it had the same householder rates by demographic subgroup as in the rest of the country? The difference between the number of actual occupied housing units in California and the number that the state would have if it had the same householder rates as in the rest of the nation, controlling for demographic subgroup, is taken as an estimate of California’s housing shortage. We develop statewide estimates and estimates for specific metropolitan areas in the state. To ensure greater precision with large sample sizes, we restrict the estimates to decennial census years. Because such data were not yet available for 2000 when this report was written, we use the Census 2000 Supplementary Survey in conjunction with Current Population Survey data to develop estimates for 2000. Details of the approach are in Appendix B. It is important to keep in mind that these are estimates of occupied housing units rather than total housing units. The two differ by the number of unoccupied housing units, including those for seasonal use as well as those for rent or sale. To the extent that California’s vacancy rates are low and have declined over the past decade, these estimates understate the shortage of total housing units; that is, householder rates are higher in California than they would be if the vacancy rate had not declined during the 1990s. This could be remedied by assuming some ideal occupancy rate—something we do not pursue here.6 It is also important to keep in mind that householder rates in the rest of the United States are taken as the standard for comparison. One could argue that those rates are problematically low, especially for certain groups, and embody a national housing shortage. The results, as shown in Figure 4.12, indicate that California has gone from a position of housing excess in 1970 and 1980 to increasingly ____________ 6Vacancy rates for rental units in California declined from 5.9 percent in 1990 to 3.7 percent in 2000. Statewide, these vacancy rates are at “normal” levels, typically regarded as being in the range of 3 to 5 percent (Landis et al., 2000). In the Bay Area and the counties of Santa Cruz, Monterey, Santa Barbara, and Ventura, rental vacancy rates in 2000 were lower than 3 percent. Since 2000, however, vacancy rates have risen substantially in Bay Area counties (U.S. Department of Housing and Urban Development, 2003). 71 Supply (1000s) 1,000 800 600 400 200 0 –200 –400 –600 –800 –1,000 1970 1980 No controls Controls 1990 2000 Figure 4.12—Estimates of the Undersupply of Housing in California Based on Householder Rates in California and the Rest of the United States, by Decade large shortages in 1990 and 2000. The extent of the shortage depends very much on whether we control for differences between California’s population and that of the rest of the country. With no controls for the demographic composition of the state’s population, California’s housing shortage in 2000 is estimated to be 870,000 units. Most of that shortage was generated in the 1980s. During that decade, California went from an excess of 147,000 housing units to a shortage of 610,000 units, a difference of 757,000 units. In comparison, during the 1990s the shortage worsened by another 260,000 units. However, when we control for demographic differences between California’s population and that of the rest of the country, the shortage appears much smaller—about 138,000 units. This suggests that most (84 percent) of the apparent shortage in California’s occupied housing units can be attributed to demographic differences between California and the rest of the nation. We also find that most of this shortage was generated in the 1980s rather than the 1990s. The increase in the shortage of occupied housing units, controlling for demography, was 55,000 units during the 1990s. In contrast, during the 1980s the state went from an excess of 476,000 occupied housing units to a shortage of 72 83,000 based on the model with demographic controls. The increase in number of occupied housing units in California during the 1990s was very close to the number that would be necessary to maintain the differential that the state started with in 1990 (Figure 4.13). Even without demographic controls, California’s housing shortage is much more a creation of the 1980s than the 1990s (Figure 4.14). The shortage of occupied housing units in California is concentrated in the state’s largest metropolitan areas. Still, many of the state’s metropolitan areas, including some larger metropolitan areas such as Sacramento and Fresno, exhibit little or no shortage in housing units (Figure 4.15a). When we take into account the demographic composition of each metropolitan area, the shortages are not so severe as when we do not adjust for age, gender, ethnicity, and nativity (Figure 4.15b). Demographic controls are especially important in Los Angeles. For example, the shortage in the Los Angeles metropolitan area decreases from a point estimate of 314,000 units to 101,000 units when we control for Los Angeles’ unique demographic composition. In contrast, adjusting for the demographic composition in the San Francisco metropolitan area makes little difference, as that metropolitan area’s 2,500 2,000 1,500 Actual Necessary to match rest of U.S. householder rates Necessary to maintain start of decade differential Number of units (1000s) 1,000 500 0 1970–1980 1980–1990 1990–2000 Figure 4.13—New Occupied Housing Units, by Decade: Actual Versus Necessary for California (Controlling for Demographic Differences) 73 Number of units (1000s) 1,000 800 600 1970–1980 1980–1990 1990–2000 400 200 0 –200 –400 –600 Necessary to match rest of U.S. householder rates Necessary to maintatin start of decade differential No controls Necessary to match rest of U.S. householder rates Necessary to maintain start of decade differential Controls for gender, age, nativity, and ethnicity Figure 4.14—Decennial Shortage of New Occupied Housing Units population does not have large shares of subgroups with low householder rates. These estimates of California’s housing shortage appear conservative when compared with oft-cited numbers. For example, the California Building Industry Association estimates that the state needs 220,000 new housing units each year to keep pace with population growth but that the state is producing only a little over half that amount (Weintraub, 2002). The California Department of Housing and Community Development projects that California will need at least 200,000 new housing units each year to 2020 (Landis et al., 2000), but the state has averaged less than 117,000 new permits per year between 1990 and 2001 (California Budget Project, 2002). The California Budget Project puts the shortage of affordable units at 651,000 statewide and 289,000 in Los Angeles County (California Budget Project, 2002). A Fannie Mae Foundation report places the shortfall in new housing during the 1990s at 548,000 units—an estimate that was derived from an approach similar to the householder rate approach developed in this report except that it does 74 not adjust for demographic composition (Myers and Park, 2002).7 Indeed, none of the estimates of the shortage of new construction in California adequately consider the demographic composition of the state’s growth. Our estimates also suggest that the creation of the shortage primarily occurred in the 1980s rather than the 1990s. Many analysts note that the increase in new housing units during the 1980s was almost twice as great as in the 1990s, arguing that the shortage in new housing units is primarily a 1990s phenomenon (e.g., Myers and Park, 2002). We find that the increase in occupied housing units in California during the 1980s did not keep pace with the state’s strong population growth in general and the changing age structure of the state in particular. Our models suggest that, given the large numbers of baby boomers aging into the prime householder ages during the 1980s and with tremendous population growth, the number of new occupied housing units necessary to match householder rates in the rest of the United States was a great deal higher than actually achieved in the state. Although the shortage did indeed worsen in the 1990s, the relative contribution to the shortfall was much smaller in the 1990s than in the 1980s. Summary In this chapter, we have shown that California’s relatively anemic increase in new housing units during the 1990s is at least partly due to the nature of the state’s population growth. In particular, population growth slowed in the 1990s compared to the 1980s and growth became more concentrated among groups—children and immigrants, for example—that consume less housing than other groups. We find evidence of housing shortages in key areas of the state, but these shortages are lower than commonly cited figures. ____________ 7Also, even our unadjusted householder rate model restricts the analysis to people age 15 and older, whereas the Fannie Mae Foundation report uses an average household size approach that includes all members of the household regardless of age. To the extent that much of California’s population growth was among children who do not form their own households, the Fannie Mae approach will overstate housing needs even relative to the householder rate model without controls. 75 –200,000 –100,000 Number of Housing Units 0 100,000 200,000 300,000 400,000 Sacramento Rest of California Fresno Stockton San Luis Obispo Chico Modesto Merced Santa Rosa Yuba City–Marysville Visalia Riverside–San Bernardino Davis–Woodland Vallejo–Napa Bakersfield Oakland Ventura Salinas–Monterey Santa Barbara San Jose Orange County San Diego Los Angeles–Long Beach San Francisco NOTE: Negative values reflect an oversupply; positive values reflect an undersupply. Figure 4.15a—Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of United States, No Demographic Controls, 2000 76 –200,000 –100,000 Sacramento Number of Housing Units 0 100,000 200,000 300,000 400,000 Rest of California Fresno Stockton San Luis Obispo Chico Modesto Merced Santa Rosa Yuba City–Marysville Visalia Riverside–San Bernardino Davis–Woodland Vallejo–Napa Bakersfield Oakland Ventura Salinas–Monterey Santa Barbara San Jose Orange County San Diego Los Angeles–Long Beach San Francisco NOTE: Negative values reflect an oversupply; positive values reflect an undersupply. Figure 4.15b—Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of United States, Controls for Age, Gender, Ethnicity, and Nativity, 2000 77 5. Conclusion and Policy Considerations This report evaluates the effect of business cycles and demographic factors on the demand and supply of housing in California. We found that more than 80 percent of annual changes in construction in California can be explained by the influence of macroeconomic effects and demographic changes, leaving small room for local and state policies. Our results suggest that the sluggish response of construction to the latest economic recovery could be explained by: • Slower population growth in the 1990s compared to the 1980s, which slowed the demand for housing. The nature of that growth, too, explains much of the apparent lack of new housing in the state, with immigrants and children accounting for most of this growth, but consuming proportionately fewer housing units than other demographic groups. • The severity and duration of the economic recession that took place in 1990 in California, leading to low housing valuation. • The Federal Reserve’s monetary policy, which decreased the public’s expectations of future inflation and caused the price of long-lived assets to decline. • High returns on alternative investments (such as stocks and mutual funds) that led to a decrease in new housing investment. • Decrease in builders’ and developers’ expectations of profits on new housing construction, because of both changes in future inflation and interest rates and a slowdown in housing demand resulting from demographic changes. Although our analysis runs through the 1990s, there is little reason yet to believe that the results would be altered by very recent events. 79 Since 2000, housing supply has continued to increase while job growth has stopped. Declining interest rates have fueled further price gains in most regions and have countered the severe job and income losses in the Bay Area. Most analysts expect housing prices to decline (in real terms) once interest rates rise from 40-year lows, and some have argued that the Bay Area is especially prone to declines (Leamer, 2002).1 Our results suggest that the housing supply crisis may be overstated—at least in the context of the entire state—and that our position today is perhaps better, and certainly not much worse, than where it was in 1990. The housing cycle in the 1990s is quite different from the three previous cycles but this is also true at the national level— although the falloff in production is not nearly as pronounced as it is in California. Indeed, some of the same arguments over whether there is a housing shortage or a housing bubble are also taking place at the national level. Housing prices have not fallen despite the recession that began in 2001, raising the specter of a housing bubble akin to the stock market bubble. Other observers are much more concerned about the slower rate of housing production in this cycle and warn of an ongoing shortage nationwide. A recent analysis, however, suggests that even if housing prices nationally have outpaced their rental equivalents, a few years of flat home prices and normal rent increases would bring the two markets back into alignment.2 Any policy responses to problems in the state’s housing market need to be mindful of both the important connection between underlying macroeconomic and demographic forces and real estate investment, as well as the regional nature of housing markets. Our findings suggest that there is not a large statewide crisis in housing production in California. This is not to say that there are no legitimate concerns in the state with regards to housing. In this report, ____________ 1Of the Bay Area, Leamer (2003) states: “Expect a long agonizing decline in prices like we had in LA in the early 1990s.” 2This analysis (Krainer, 2003) also demonstrated the importance of the after-tax mortgage rate in the owner cost of housing. Not only do declining mortgage rates reduce the new owner’s cost but, through refinancing, reduces the ongoing cost of ownership for existing homeowners. Both effects serve to prop up housing prices even in a weak economy. 80 we have not considered other issues such as housing affordability for lowincome families, the condition of rental housing, or overcrowding.3 Previous research suggests that the state does have problems in these areas. For example, in our earlier report we found that overcrowding occurs more frequently in California than in the rest of the United States but also that poverty and demographics were greater determinants of overcrowding than were high housing prices (Moller et al., 2002). Overcrowding is much more common in poor areas of the state with relatively low-priced housing than it is in expensive places such as the Bay Area.4 The mismatch between the type of new housing that is constructed and the needs of the newest Californians is a particularly important topic for future research. Given that houses are long-lived assets, rapid demographic change can increase demand for a particular type of housing unit that easily outstrips new supply. Much of the population increase in the 1990s came from an increase in family size, primarily among low-income immigrant families. Such families are typically renters, and the rental housing stock in many areas is not sufficiently oriented toward large households. In addition, as shown in Chapter 2, production of rental (i.e., multifamily) housing fell much more dramatically than did production of single-family housing. Although the amount of new housing units produced in California is largely explained by macroeconomic and demographic factors, housing shortages do exist in the state’s largest metropolitan areas: the Bay Area, coastal Los Angeles (Los Angeles, Orange, and Ventura Counties), and San Diego County. The magnitude of the shortages is much lower by our estimates than those put forth by others, and even some of these shortages may be worked off after several years of job losses (most dramatically in the Bay Area) and the completion of new housing. Prices in these regions are higher partly because of their strong desirability as ____________ 3We addressed this issue in a previous report (Moller et al., 2002). 4Using 1990 census data, Myers et al. (1996) also found that demographic factors and income were more important than housing affordability in explaining differences in overcrowding between metropolitan areas. 81 places to live—their amenity value—and partly because of obstacles to new supply. These obstacles can be regulatory or financial, or both. There are ongoing battles, even in the midst of a widely reported housing crisis, by groups who oppose most new greenfield development (often termed “sprawl”) and by those who oppose most infill development in their communities (often referred to as the NIMBY5 syndrome. These actions serve to reduce the supply of new housing in areas where developers have found demand for it. Two of the regions of California with evidence of housing shortages are also the least receptive to new housing construction. According to a PPIC survey of city managers in 1998, respondents in the Los Angeles region and the Bay Area were much more likely to state that residential growth issues were “often” or “always” controversial (Lewis and Neiman, 2000). They were also more likely than respondents in other regions to state that there was little or no vacant land available for new development (Lewis and Barbour, 1998). Using GIS and mapping techniques, researchers in a study for the California Department of Housing and Community Development (HCD) concluded that sufficient vacant land exists statewide for 20 years of expected population growth, but restrictions in some key urban areas—including some Bay Area and South Coast counties—are likely to present serious obstacles in the future (Landis et al., 2000). According to the authors of the HCD study The key constraint on the State’s development capacity is political, not environmental. As this analysis reveals, the careless adoption of even moderately-limiting urban growth boundaries by high-growth counties in the Bay Area and Central Valley would constrain land supplies below the levels required to meet future housing production needs. In many cases, voters have directly restricted new housing growth. Dozens of local jurisdictions in California have passed measures to limit growth (Glickfeld and Levine, 1992; Lewis and Neiman, 2000; Landis et al., 2000). Solimar Research Group has documented 389 slow-growth ballot measures in California from 1986 through 2000, with 59 percent of them passing. Of course, ballot measures are not the only way citizens ____________ 5The acronym stands for Not In My Back Yard. 82 express their displeasure with new developments. City councils often reflect the lack of support for new housing of their constituents and have simply failed to allow new housing developments or delayed their approval long enough for developers to drop their proposed projects. A recent study of housing affordability in the nation’s metropolitan areas found that in those areas with the least affordable housing, land-use controls were the most compelling explanation for high relative prices (Glaeser, Gyourko, and Saks, 2003). It also demonstrated that there was little change in affordability by metropolitan area between 1989 and 1999—supporting our conclusion that much of what is driving concerns about new housing supply is cyclical. There can be financial obstacles to new housing supply as well. In response to charges that litigation costs were contributing to the sharp decline in multifamily construction, the state legislature recently proposed several bills to restrict construction defect litigation and encourage remediation efforts over litigation. Even without opposition from local residents, infill development can often be more expensive to construct than the market will bear. A recent study undertaken for the California Department of Housing and Community Development (Landis, 2003) categorizes the barriers to infill production. Although a number are political in nature, several are purely economic: a shortage of adequate lot sizes, greater need for infrastructure upgrades, high land prices, lack of economies of scale, and greater design costs to work within existing land uses. Even for greenfield development, there can be sizable expenses in addition to direct construction costs. Most local jurisdictions in California have resorted to the use of development fees and special districts to recoup the cost of new housing (new housing leads to increased demand for schools, parks, and basic infrastructure). A study of these fees in Contra Costa County reported typical costs of $20,000 to $30,000 per new housing unit (Dresch and Sheffrin, 1997) and news reports pegged fees in Santa Clara County at over $30,000 per unit. Even jurisdictions not facing pressure from residents to restrict growth can be reluctant to allow large-scale development that might trigger more demand for public services. Most city managers believe that the additional tax revenues generated from new housing development do not 83 cover the full cost of the public services they require (Lewis and Barbour, 1999). The state is active in California’s housing markets through a number of funding programs. These programs are generally targeted to meet specific housing needs and are not necessarily designed to increase the supply of new housing as a whole. Although it is beyond the scope of this research to evaluate the efficacy of such programs, their sheer number and size (in terms of expenditures) suggest that they do affect California’s housing markets, and at least some lead to greater housing construction than would otherwise occur in the state. The undersupply of new housing in the Bay Area, coastal Los Angeles, and San Diego County is cause for concern. Increasing employment in those areas in the late 1990s exacerbated the jobs-housing imbalance and led many workers to live in areas far removed from their jobs. High and increasing housing prices reflect both existing amenities and the shortage in those areas, and these prices can create affordability problems for middle- and low-income households that do not already own their own homes. Any policy intended to increase the supply of new housing in the state’s largest metropolitan areas must address the inherent conflict between existing homeowners and renters (as well as new entrants to the region). Current homeowners reap direct financial benefits from any restriction on new housing supply that raises the price of existing homes. But their opposition to new development does not necessarily reflect personal financial interest alone. Many homeowners have made their largest financial investment in a community with certain characteristics, and their local government is expected to reflect their concerns. New infill development can increase traffic congestion, overwhelm the capacity of local public facilities, and require an increased tax burden on current residents. Although often opposed by statewide and even national groups, new greenfield development can sometimes conflict with the priorities of current residents as well. Part of the character of their local community may be proximity to open spaces and scenic views that would be lost to new housing development. In light of these forces, the problems facing potential new residents are often less apparent and urgent to most policymakers. 84 Policies could be implemented to make new residential development a more financially attractive option to local officials. These policies could come in the form of both rewards and punishments. For example, the state could impose financial penalties on jurisdictions that do not provide their “fair share” of new housing.6 These penalties could include withholding of funds for missing housing goals as well as rewards for meeting them (Lewis, 2003). Jurisdictions that meet their statedetermined share of new housing could be rewarded financially by having the state offset some of the costs of providing new infrastructure. Reducing local reliance on sales tax dollars and increasing local access to property tax revenue might also increase the attractiveness of new housing to local officials. Of course, there have been no shortage of proposals and recommendations to reallocate property and sales tax revenues to reduce the fiscalization of land use, and the specific details of any legislation will determine the support and opposition it faces. If implemented, such a policy will help illuminate just how much of an influence the fiscalization of land use has really had on housing supply. Citizen opposition to new development might be an even greater impediment to new housing construction in the Bay Area and coastal Los Angeles. Citizens can express their opposition to new development in myriad ways, from passing initiatives that limit growth to electing city councils that restrict growth to intervening in the planning and approval process for new developments. When combined with the fiscal concerns regarding housing versus other land uses, such citizen opposition often squelches new housing. Policies to address citizen opposition must consider the source of that opposition. Opposition to new housing is not necessarily based on housing per se but originates from concerns about the effect of the new development on the environment and existing infrastructure. Specifically, concerns are often raised about increasing traffic congestion, loss of open space, and overcrowding of schools and other public facilities. Policies that address these concerns could lessen opposition to new growth. In some cases, educating the public about the ____________ 6Compliance with the state’s housing element does not seem to result in an increased supply of housing (Lewis, 2003). 85 consequences of not planning and allowing for new housing could change local opinion. Finally, policies could be enacted to address the jobs-housing imbalance by focusing on jobs rather than housing. For example, the state could encourage job growth in areas with relatively abundant housing (especially in inland areas adjacent to the state’s most undersupplied regions). A common response across the country to population pressures has been the growth of once outlying regions as economic centers in their own right. As a natural outgrowth of job growth in the inner Bay Area, satellite job centers have bloomed in Pleasanton and Livermore; further job growth in these locations could provide shorter commutes for residents in neighboring counties. For cities and counties with housing shortages, the state could reward jurisdictions that meet some standard of zoning for adequate housing when approving new employment-generating land uses and penalize those jurisdictions that do not meet such goals. The Jobs/Housing Balance Incentive Program is an example of this kind of effort, providing financial rewards for new housing construction in high job growth areas where housing has not kept pace with construction.7 All of these policy options are controversial and would create winners and losers. Our primary finding, that the state’s housing shortage is smaller than previously thought and concentrated in three regions— albeit the state’s most populous regions—suggests that any policies designed to address the housing shortage should focus on the situations in those regions. How the regional housing markets fare when interest rates rise from their historic lows—and how high the Bay Area’s housing prices remain if job growth is delayed much longer—will clarify how much new intervention may be required. An important, related policy question is whether the state should provide incentives for more housing construction in the built-out and environmentally protected coastal areas or instead increase housing by supporting infrastructure development in less built-out areas adjacent to existing job centers. In the final analysis, ____________ 7In “high employment demand areas,” the program provides incentives of $1,300 per unit for new units above some baseline standard (California Department of Housing and Community Development, 2003). In the context of developer fees of $30,000 or more per unit, this may not provide a large incentive in the tightest markets. 86 the conflict between local control of land-use decisions and increasing new housing construction is not one that is likely to be resolved any time soon. 87 Appendix A A Housing Market Model Our housing model used to assess the influence of macroeconomic factors on California housing was summarized in a single equation that explains residential new construction. This equation is derived from a model of residential housing markets in equilibrium and relates changes in new construction to changes in the factors that determine the levels of housing demand and supply. The effects of governmental economic policies and other worldwide influences are measured through changes in interest rates, unemployment, income, expected inflation, and stock prices. These are factors that affect both the demand and supply of new housing. Our housing model takes into account the fact that the adjustment of new housing construction to changes in housing demand is not instantaneous. A Simple Model for the Housing Market The demand for housing (Dh) is determined by Dh = Dh(Ph,Ps,E(p),Pop,I, Y, A) where: (A.1) Ph = Ps = E(p) = Pop = I= Y= A= price of housing price of stock expected price of housing, or expected inflation population variables (structure of the population, immigration, population growth) interest rates income amenities. The supply of housing (Sh) is determined by 89 where Sh = Sh(Ph,Ps,E(p),I,C,Pol) (A.2) Ph = Ps = E(p) = I= C= Pol = price of housing price of stock expected price of housing, or expected inflation interest rates (mortgage rates) construction costs policy variables. The optimal amount of housing is found when supply and demand are in equilibrium: Sh* = Dh *. (A.3) However, equilibrium is not instantaneous. It takes time for changes in supply to meet changes in demand. We introduce a simple stock adjustment model, where the equilibrium stock of housing is Sh*(Ph, Ps, E(p), I, C, Pol). We represent the adjustment process as Sht – Sht –1 = g(Sht * – Sht –1) 0 < g <1. (A.4) Equation (A.4) specifies that the change in Sh will respond only partially to the difference between the desired stock of housing and the past value of stock of housing. The adjustment coefficient g determines the rate of response. Substituting for Sh* in Equation (A.4), and solving for Sht yields Sht – Sht –1 = g(Sh * (Ph,Ps,E(Ph),I,C,Pol) – Sht –1) . (A.5) Using the fact that Sh* = Dh*, and that, in equilibrium, quantities determine prices, the price of housing disappears from the equilibrium equation. Using the fact that Sht- – Sht–1 = NC = new construction, and taking first differences of these equations, we obtain NCt – NCt –1 = g *[f * t(E(Ph)t – E(Ph )t –1,Pst – Pst –1, (A.6) It – It –1,CtCt –1, Yt – Yt –1,Popt – Popt –1,It – It –1) – NCt –1] 90 where St – St–1 is equal to new construction if we disregard the demolition of housing stock. E(Ph) was estimated independently. A lagged variable of new construction (NCt–1 – NCt–1) was also incorporated to include the effect of previous changes in new construction on current new construction, when it was estimated to be necessary to control persistent effects on the disturbances. Measurement of Variables NC = new construction Number of permits issued for total family units as by the Construction Industry Research Board (CIRB) Ps = price of stock Estimated from Standard and Poor’s 500 (S&P) at http://chart.yahoo.com/t?a=10&b= 01&c=55&d=10&e=23&f=01&g=m&s=%5 Espc&y=0&z=5. An alternative way to measure the return on stocks is the S&P earnings yield (E/P) (Kaiser, 1999). We used changes in the S&P, since the historical series is already adjusted by dividends and stock splits. E(p) = Expected price of housing, or expected inflation To measure expected appreciation we use various alternative approaches. We worked with our own projections of expected prices, based on an adaptive expectation model using two and three lags of California changes in the Consumer Price Index (CPI) and the CPI housing component in California and the United States. In a similar way, we also modeled changes in 91 California median home prices. Finally, we used different expected inflation rates as reported in the Livingston Surveys. The Livingston Survey data are discussed below. Since we are working with flows, we have included change of expected appreciation (or prices). The advantages of working with price changes rather than levels are discussed in Fratantoni and Schuh (2000). Pop = population variables (population growth) California Department of Finance annual estimates I = interest rates We estimated the effect of both mortgage rates and the prime rate. Builders indicate that their costs of funds depend on the prime rate rather than the Treasury bills or mortgage rates. We found that statistically it was better to work with the difference between the long-term and the short-term rate. We used the difference between Treasury securities maturing in ten years and commercial paper maturing in three months. Y = personal income California Department of Finance. A = amenities This variable was controlled by using a regional approach grouping counties of similar characteristics. 92 C = construction costs In the estimation of the model, we tried to include various construction cost indexes published by the Census Bureau, but we did not get good results. Then we used the construction regional cost index compiled by CIRB, which estimates costs for nine regions. Because the board’s classification of counties in regions differed from ours, we built our own regional cost indexes by assigning to each county the cost estimated by the CIRB for the region where the CIRB assigned that county. Then, we used the number of total permits to estimate a new weighted average cost for our ten regions. Some Methodological Remarks on Our Time-Series Approach We estimated our reduced-equation model for each region (see above).1 We also estimated an aggregate equation for new construction in California. To avoid the problem of measuring the housing stock for various regions, we worked with first differences. Furthermore, we estimated the model using a Prais-Winsten regression. This method takes into account previous effects that persist over time affecting the error terms of subsequent specifications.2 We also tried ARIMA specifications, but this type of approach did not provide better results than the ones obtained with the Prais-Winsten regression. The ARIMA ____________ 1The model was reduced to one equation. 2We also used this method for cross-pooled time-series analysis since, in addition to the effects that often persist over time affecting the error terms, there are also effects if the random shock affecting economic activity in one region causes economic activity in an adjacent region because of close economic ties between the regions. The CochraneOrcutt method is a procedure to control the correlation between errors from one period to another. The Prais-Winsten method is the Cochrane-Orcutt method when it incorporates its special transformation to estimate the correlation coefficient between the error term of current and previous periods to the first observations in the time-series. 93 method controls for both the effect of past values and the effect of current and lagged random disturbances (unmeasured effects). Modeling Expected Appreciation Since we do not have many observation points in our time series, we calculated expected appreciation, E(p), independently to include it in our estimating equation as an exogenous factor. First we calculated our own expected prices series. Since all prices tend to move together and are highly correlated, we used projected inflation rates and median housing prices as measures of expected appreciation. There are various theories on expectation formations. Adaptive expectations assume that expectations are based only on past knowledge. Rational expectations believe that expectations are based on all information available and an understanding of how the market works. Extrapolative expectations are based on the past but follow along a trend line into the future. To calculate our own inflationary expectations, we used two approaches. First we used an adaptive expectation model that included distributed lags of price changes. We chose this model because various studies have rejected the notion that people form their expectations rationally. For example, Hamilton and Schwab (1985) raised doubts as to the general applicability of the rational expectations model. Collins, Lipman, and Groeneman (1992) also reject the applicability of rational expectations to the real estate market. Our second approach was to use expected inflation rates as reported by the public at each year included in our time-series. This was accomplished by using various measures of expected inflation one and two years ahead for each year, as reported by the Livingston Survey.3 The Livingston Survey data performed better in our estimating equations and provided more alternatives for modeling expected prices, since reports include the public perception for one year and two years ahead. Although we also explored alternative sources for expected prices such as one-year-ahead expected prices as calculated by the DRI model, and Green Book forecasts (a briefing document with macroeconomic ____________ 3Described below. 94 forecasts—including inflation—prepared by staff economists at the Board of Governors about three days before each Federal Open Market Committee meeting), we used the Livingston Survey. The Green Book forecasts become available to the public only five years after they are made; DRI forecasts were highly correlated with some other determinants of new construction. None of the various measures of inflation yielded very good and consistent results for all our ten regional equations. However, most equations showed better results when we used expected inflation rates as reported by the Livingston Survey (median rates of projected inflation) instead of our own projections. The Livingston Survey The Livingston Survey dataset contains a forecast of economic variables from a survey of forecasters. Joseph Livingston, a columnist for the Philadelphia Inquirer, began the survey in June 1946. Livingston continued the survey until 1990, when the Federal Reserve Bank of Philadelphia took it over. The survey is conducted twice a year, in June and December. The survey forecasts 18 macroeconomic variables describing national output, prices, unemployment, and other macroeconomic data. Each variable in the survey asks for forecasts of the level of the variable in the current quarter, in the next quarter, and in the same quarter in the following year. It also asks for the annual average of the variable for the current year (six months ahead), and one year ahead. For example, for the Consumer Price Index in the survey of June 1992, it asks for forecasts of the monthly level of the CPI in June 1992, December 1992, and June 1993, and the annual average of the CPI in 1992 and 1993. The survey is administered to affiliates of non-financial businesses, academic institutions, commercial banking, government, Federal Reserve Bank, insurance companies, and labor unions. One problem is that the affiliations for each participant change over time. In this study, we used the one-year-ahead forecasted CPI from the current month of the survey. This variable is reported from the December surveys and predicts the average level of the CPI for the next calendar year. We used the median values of this variable because they 95 gave better results than the average values. We also tried forecasts for two years ahead. However, the one-year-ahead forecast provided better statistical results. The CPI data reported by the survey have different base years. We worked with the CPI series adjusted to the base 1982 = 100. For more information on the Livingston Survey, see http://www. phil.frb.org/files/liv/document.html. Modeling the Influence of Interest Rates on Housing Markets Interest rates and new construction are negatively related. Typically, housing demand or supply equations estimate the effect of either mortgage rates or the prime rate. Builders indicate that their cost of funds depends on the prime rate rather than the Treasury bills or mortgage rates. We looked at the difference between the long-term and short-term interest rate as measured by the difference between two types of securities: Treasury securities maturing in ten years and commercial paper maturing in three months. This differential is expected to be positively related to new residential construction. The difference between the long-term and short-term interest rate indicates that the expectations of future inflation are rising, driving shortterm rates up. When short-term rates are above some long-run average, investors may expect them to fall and receive capital gains. On the other hand, if investors expect capital losses, the differential between long- and short-term rates will widen. If they expect capital gains, the differential will narrow. At the moment of the change in expectations, long-term rates will rise relatively faster than short-term rates. The spread between the long-term and short-term interest rates also represents a better measure of the tightness of credit than does the level of either interest rate. Easy monetary policy through its effects on expectations about capital gains will lead to a large spread between longterm and short-term interest rates, and tight money to a small spread. During periods of tight money, lending institutions will give higher priority to loans for business investment. Other sources such as bond and stock markets are effectively closed to all but the largest residential builders. 96 Table A.1 Analysis of Changes in New Construction in California, by Region (Time-Series Results, No Lags) 97 North Coast R2 0.67 F test 3.57 Probability 0.02 Metropolitan Statistical Area 223 North Central 0.6 3.44 0.02 810 North Eastern 0.82 7.77 0 226 Bay Stockton- Area Sacramento 0.64 0.7 3.13 4.11 0.03 0.01 5498 3922 Central Coast 0.6 2.59 0.06 1067 Central Eastern 0.89 14.34 0 298 Los Angeles– Central San Inland Bernardino 0.76 0.57 5.62 2.27 0 0.09 2408 20594 San Diego 0.72 4.41 0.01 5649 California 0.64 4.24 0 35891 Change in population Coefficient t Probability 0.077 0.123 0.12 0.091 0.126 0.071 0.134 0.134 2.1 2.78 2.1 1.4 1.14 1.43 3.91 1.57 0.05 0.01 0.05 0.18 0.28 0.18 0 0.14 0.157 1.14 0.27 –0.126 –1.47 0.16 0.072 0.68 0.51 Change in income Coefficient t Probability 0.231 2.32 0.04 (a) 0.153 0.77 0.45 –0.013 –0.22 0.83 0.066 1.46 0.17 0.098 0.9 0.38 0.461 3.41 0 0.191 1.6 0.13 0.146 0.87 0.4 0.316 0.93 0.37 74.049 0.65 0.53 Lagged new construction Coefficient t Probability –0.665 –1.74 0.1 –0.579 –2.97 0.01 –0.681 –4.46 0 –0.822 –3.64 0 –1.094 –4.57 0 –0.47 –2.81 0.01 –0.593 –5.02 0 –1.343 –5.34 0 –0.5 –1.87 0.08 –0.125 –0.08 0.94 –0.32 –1.48 0.16 Change in unemployment Coefficient –21.285 –423.648 t –0.31 –2.67 Probability 0.76 0.02 –80.57 –721.119 –1018.914 –1.53 –0.35 –2.27 0.15 0.74 0.04 –955.54 –2.14 0.05 –33.921 –614.251 –0.42 –1 0.68 0.33 –13335.94 –3497.584 –39508.48 –1.35 –1.31 –2.83 0.2 0.21 0.01 Table A.1 (continued) 98 Change in difference of long-term minus shortterm interest rate Coefficient t Probability North North North Coast Central Eastern Los Angeles– Bay Stockton- Central Central Central San San Area Sacramento Coast Eastern Inland Bernardino Diego California 10.54 0.19 0.85 –1.845 –63.259 –1385.153 –1625.548 –0.01 –1.26 –1.28 –2 0.99 0.23 0.22 0.07 43.925 –201.784 –731.818 0.17 –2.92 –1.57 0.87 0.01 0.14 466.545 3251.318 16318.46 0.07 2.36 2.12 0.94 0.03 0.05 Change in appreciation (expected inflation) Coefficient t Probability 11.814 –46.034 0.09 –0.17 0.93 0.87 21.873 2721.669 0.31 1.93 0.76 0.07 996.303 0.94 0.37 233.906 –266.904 0.72 –2.37 0.49 0.03 733.462 1.01 0.33 –0.305 –945.606 –11197.75 0 –0.53 –1.15 1 0.6 0.26 Change in real stock prices Coefficient –325.812 –909.265 –407.326 t –1.55 –1.67 –2.81 Probability 0.14 0.12 0.01 3183.985 –2462.197 –1254.795 –456.879 2313.619 0.68 –0.8 –1.68 –2.62 1.15 0.51 0.44 0.12 0.02 0.27 –16880.39 –1723.627 –14505.81 –1.27 –0.4 –0.67 0.23 0.7 0.51 Change in real cost Coefficient t Probability –31416 –91444 –1.47 –0.79 0.16 0.44 –84595 –1496151 –1572724 –2.38 –1.08 –1.56 0.03 0.3 0.14 37061 –110850.1 –110850 0.34 –3.37 –1.07 0.74 0.01 0.3 –115914.3 –0.05 0.96 401967 –2481687 0.81 –0.85 0.43 0.4 DW (Durbin Watson stat.) 2.02 1.95 2.13 1.33 1.61 1.96 2.13 1.24 aIncome highly correlated to real cost. 2 2.03 2.1 NOTES: Cell entries in boldface are statistically significant. Cell entries in italics are of the wrong sign but not statistically significant. Cell entries that are shaded are of the wrong sign and are statistically significant. All other cell entries are statistically significant with the right sign. Table A.2 Time-Series Results Allowing for Lagged Responses 99 R2 Adjusted R2 F Test Probability Metropolitan Statistical Area North Coasta 0.67 0.48 3.57 0.02 North Central 0.78 0.69 8.31 0.00 North Eastern 0.78 0.66 6.24 0.00 Bay Area 0.76 0.64 6.10 0.00 Stockton– Saramento 0.85 0.76 9.59 0.00 Central Coast 0.77 0.63 5.77 0.00 Central Eastern 0.90 0.85 16.07 0.00 Los Angeles– Central San Inland Bernardino 0.72 0.77 0.56 0.63 4.57 5.78 0.01 0.00 San Diego California 0.76 0.83 0.62 0.76 5.42 11.48 0.00 0.00 223 670 229 4385 3675 845 311 2009 16110 5195 28256 Change in population Coefficient t Probability 0.077 2.10 0.05 0.207 5.53 0.00 0.225 2.51 0.03 0.018 0.33 0.75 0.205 2.34 0.04 0.072 2.05 0.06 0.226 5.44 0.00 0.108 1.87 0.08 0.081 0.64 0.53 –0.119 –1.50 0.16 0.035 0.45 0.66 Change in income Coefficient t Probability 0.231 2.32 0.04 (b) 0.227 1.15 0.27 0.730 1.50 0.16 0.123 2.63 0.02 –0.030 –0.29 0.78 0.225 1.72 0.11 0.372 2.68 0.02 0.005 0.03 0.98 0.390 1.37 0.19 229.561 2.31 0.03 Lagged new construction Coefficient t Probability –0.665 –1.74 0.10 –0.803 –4.67 0.00 –0.876 –3.70 0.00 –0.201 –2.02 0.06 –0.910 –3.68 0.00 –0.388 –3.37 0.01 –0.959 –5.64 0.00 –0.969 –4.63 0.00 –0.210 –0.98 0.34 –0.025 –0.180 0.86 –0.235 –1.51 0.15 Change in unemployment Coefficient t Probability –21.285 –0.31 0.76 –472.151 –4.24 0.00 –78.797 –4257.78 –2942.576 –1250.212 –1.56 –3.05 –4.85 –3.51 0.14 0.01 0.00 0.00 –98.087 –1167.395 –16303.760 –983.431 –16178.720 –1.10 –1.76 –1.68 –0.45 –1.41 0.29 0.10 0.12 0.66 0.18 Table A.2 (continued) 100 North Coasta Change in difference between long–term and short–term interest rate Coefficient 10.540 t 0.19 Probability 0.85 North Central L.1 133.363 1.14 0.27 North Stockton– Eastern Bay Area Saramento L.1 103.594 1.80 0.09 L.1 2364.337 2.55 0.02 L.1 3161.962 4.52 0.00 Central Coast L.1 475.901 2.34 0.03 Central Eastern L.1 –60.133 –0.93 0.37 Central Inland Los Angeles– San Bernardino San Diego California L.1 1269.820 3.21 0.01 L.1 7604.326 1.82 0.09 L.1 3026.018 3.20 0.01 L.1 16271.780 3.96 0.00 Change in appreciation (expected inflation) Coefficient 11.814 t 0.09 Probability 0.93 111.535 0.51 0.62 L.1 L.1 145.153 –3116.065 1.95 –2.61 0.07 0.02 1186.058 1.18 0.26 180.400 0.74 0.47 L.1 44.106 0.44 0.67 –198.192 –0.26 0.80 L.1 –7361.258 –1044.524 –1.12 –0.66 0.28 0.52 –8959.499 –1.20 0.25 Change in real stock prices Coefficient t Probability –325.812 –1.55 0.14 L.1 1744.118 –3.78 0.00 L.1 L.1 –531.920 –3566.201 –2.06 –1.18 0.06 0.26 L.1 –4676.496 –1.800 0.09 L.1 –1217.889 –2.11 0.05 L.1 –1129.649 –4.48 0.00 L.1 –2638.260 –2.10 0.06 L.1 –4665.539 –0.43 0.68 L.1 L.1 –81.903 –39150.930 –0.02 –2.36 0.98 0.03 Change in real cost Coefficient –31416.4 –64667.98 –89322.48 –146346.1 –1584431.0 –15447.310 –126930.90 –623250.00 –786015.400 t –1.47 –0.890 –2.36 –0.37 –3.84 –0.20 –4.08 –2.140 –0.46 Probability 0.16 0.38 0.03 0.72 0.00 0.85 0.00 0.05 0.65 63737.95 –758840.40 0.14 –0.36 0.89 0.73 C 473.449 1262.604 378.205 605.766 –3400.848 1047.274 813.804 6334.513 –9865.349 1421.356 –6027.783 0.99 1.9 1.41 0.14 –1.16 2.06 2.43 2.460 –0.59 0.54 –0.28 0.34 0.08 0.18 0.89 0.27 0.06 0.03 0.03 0.56 0.60 0.78 Table A.2 (continued) 101 DW (Durbin Watson stat.) North North North Stockton– Central Coasta Central Eastern Bay Area Saramento Coast Central Eastern Central Inland Los Angeles– San Bernardino San Diego California 2.02 2.26 1.97 2.05 2.15 2.00 2.37 2.05 2.10 2.08 2.08 aOriginal equation (lags did not improve the results). bIncome highly correlated to real cost. NOTES: Cell entries in boldface are statistically significant. Cell entries that are shaded are of the wrong sign and are statistically significant. All other cell entries are statistically significant with the right sign. Appendix B Householder Rate Models We use several sources of data to estimate householder rates for two types of models: the annual models and the decennial models. The annual models are based on the annual March supplements of the Current Population Survey from 1976 through 2001; the decennial models rely on the censuses of 1970, 1980, and 1990, and the Census Supplementary Survey of 2000 combined with CPS data from 1999 and 2001. The annual models allow us to examine changes that occur each year and thus to observe the effects of business cycles, but they suffer from the relatively small sample size of the CPS. Because the sample size is not large, our ability to identify trends in local housing markets and for specific demographic groups is limited. The decennial models allow for an examination of rates for specific demographic groups and metropolitan areas, as sample sizes are much larger, but do not allow for an examination of annual changes. The annual models are logistic regression models of the probability that an individual age 15 and older is a householder. We develop three types of annual models. In one type, the key independent variable is whether the individual lives in California. The model is run separately for each year 1976 through 2001. Coefficients for the intercept and the California variable are used to estimate the probability of being a householder in California and the rest of the United States. Demographic variables are added sequentially. These variables are gender, age, ethnicity, and, starting in 1994, nativity.1 The model with all demographic variables is called the model with demographic controls. The coefficient for the California variable in the model with demographic controls provides an overall measure of how different ____________ 1Before 1994, the March supplement of the CPS did not contain information on nativity. 103 California’s householder rates are from those in the rest of the country separate from the effects of the demographic factors. In a second type of annual model, data are restricted to California and the key independent variable is the year of the survey. Coefficients for the intercept and the year variable are used to estimate the probability of being a householder in California over time. Demographic variables are added sequentially. The coefficient for the year variable in the model with no demographic controls provides an estimate of temporal changes in householder rates in California. The coefficient for the year variable in the model with demographic controls provides an overall measure of how much of the change in householder rates in California can be attributed to temporal changes that are not caused by changes in the demographic composition of the state’s population. The third type of annual model is similar to the second type, except the data are restricted to the rest of the United States. The decennial models allow for an examination of patterns of change and differences between subgroups (for U.S.-born versus foreign-born Latinos) as well as for lower levels of geography (metropolitan areas). In the models comparing California to the rest of the United States, the probability of being a householder is first estimated with no controls (the only independent variable being an indicator for California residence), and then with a series of demographic controls (age, gender, ethnicity, and nativity). The models are run separately for each census year. As in the annual models, the parameter estimate for the California indicator variable provides a measure of the difference between householder rates in California and those in the rest of the nation. The California metropolitan area models are similar, with the exception of replacing the California indicator variable with a series of metropolitan area indicator variables for the 23 metropolitan areas (and an additional indicator for “rest of California”) identifiable in the CPS. Because the Census Supplementary Survey of 2000 does not provide substate geographic identifiers, only the 1999 and 2001 CPS data were used in the metropolitan area model. The CPS shows smaller differences in householder rates between California and the rest of the country than does the Census Supplementary Survey (which shows that California’s householder rates are 3.1 percentage points lower than the nation’s but 104 only 1.7 percentage points lower in the 1999 CPS and 2.3 percentage points lower in the 2001 CPS). Throughout the report, we have converted the logistic regression results to rates or probabilities. All of the parameter estimates are based on weighted regressions, although standard errors and confidence intervals are based on unweighted regressions. We use weighted regressions because our goal is to estimate householder rates for the entire population. For example, the results of the California annual model (a weighted logistic regression with the dependent variable being an indicator of whether the adult is a householder and the only independent variables being year dummies for 1976 through 2000) are the same as a simple tabulation of the proportion of California’s adults who are householders for each year from 1976 through 2001. 105 Appendix C Panel Data Analysis of Housing Supply In this appendix we follow the same basic economic model of housing supply used in Chapter 3 (and also described in more detail in Appendix B). We take advantage of the number of separate counties in the state to analyze housing supply in California from 1972 to 2001 in a way that permits us to compare the number of housing units produced1 in each county to the number predicted by our model. We use the same variables as the time-series analysis but predict the number of new housing units based on a common response across regions to changes in the explanatory variables. As before, the variables are population, personal income, previous construction, unemployment, credit market conditions, inflation expectations, stock prices, and construction costs. Factors specific to each county—such as the amount of developable land remaining and the presence of growth controls and other regulatory constraints on housing— are controlled for with a variable representing each county. This is the same approach taken in many studies of the national housing market, so it is not unreasonable to assume that the underlying relationship between, for example, construction costs and housing production is the same regardless of which county we are examining. This approach is more restrictive than the approach taken in the time-series analysis but it produces estimates of supply shortfalls that are comparable to the housing unit figures estimated in the householder analysis in Chapter 3. Table C.1 reports the results of two panel regressions. The first column shows the log of new housing units regressed on the log of total population and the second column shows the log of new units regressed on the log of adult population. The first column specification is more ____________ 1We actually use data on the number of housing permits issued in each year, but for most counties, in most years, the share of permits that are produced exceeds 90 percent. 107 Table C.1 Panel Regression Results (t-values in parentheses) Log (county population) Log (county adult population) Real county income per capita County unemployment rate Difference in long- and short-term interest rates Expected inflation Real stock prices Real county construction costs County fixed effects Constant R2 No. of observations (58 counties) Model 1 .0249 (0.18) — –3.98 e–07 (–0.86) –0.0367 (–6.45) Model 2 — 0.2887 (2.41) –5.46 e–07 (–1.18) –0.0361 (–6.41) 0.0262 (1.83) –0.0069 (–0.39) –0.0464 (–3.79) –0.0109 (–6.87) Suppressed 9.0286 (5.96) 0.0731 1,392 0.0308 (2.14) –0.0005 (–0.03) –0.0461 (–3.93) –0.0123 (–8.27) Suppressed 6.4946 (5.42) 0.7683 1,392 consistent with that from the time-series models and the second column better reflects the demographic drivers of housing demand as discussed in the householder analysis. The model incorporating the counties’ adult, rather than total, population changes is a much better fit and gives results that are more consistent with those found both in the time-series and householder analyses.2 Total population change is not statistically significant whereas change in the adult population is, as we expected. Although it is not significant, the sign on per capita income is consistent with those hypotheses that predict more resistance to new housing as a community’s income levels increase. When unemployment rates rise in a county, new housing supply falls. Easier credit conditions, measured as the difference between long and short rates, are associated with an increase in housing production. Increases in either inflation expectations or stock market returns have negative signs—implying lower production—but only stock market returns were significant. Last, increased construction costs are strongly associated with lower levels of housing production. Because we are using lagged terms, and because data for some variables are missing in the earliest years of the period we are interested ____________ 2Note that because of the structure of the fixed-effects approach, the lagged value for units produced is omitted from the models. 108 in, we effectively predict housing supply at the county level for the period from 1973 to 1999. The total number of new housing units produced is explained fairly well by the model incorporating adult population rather than total population. Our major result in this analysis is that slightly more housing was produced in the 1980s business cycle than was predicted in the model, and significantly less housing was produced in the 1990s than predicted. Of the explanatory variables included in the model, only inflation expectations were not significant in at least some of the alternative forms that were examined.3 The only issue about interpreting the results in a fixed-effect model is that the net over- and undersupply over the entire period is constrained to be zero, which may be too binding a constraint even though the analysis covers almost three complete housing cycles. The housing oversupply at the state level in the 1980s cycle was approximately 4 percent of the total number of units predicted; the undersupply in the 1990s cycle was 15–17 percent of the predicted total (Table C.2)—at least through 1999. (The numeric comparisons are made from cycle trough to cycle trough to capture all of the new construction within each building cycle, but the results are similar when we simply compare the number of units predicted by decade.) The estimated shortfall of 15–17 percent of the predicted new supply over the decade works out to be a deficit of approximately 17,000 housing units per year for California as a whole. Most studies of housing dynamics use the change in total population as the measure of population change, and we saw in Chapter 3 that this will overpredict the amount of new housing required. The detailed demographic data needed for the analyses in Chapter 3 are available only in census years, but we can use counties’ adult population on an annual basis for these models to better handle the issue of family size. It should also be noted that most of the discussions of the housing situation in California have described an undersupply on the order of 40–50 percent ____________ 3In addition to several panel regressions, we estimated the model as a cross-section with all years’ data being estimated as a function of first differences to obtain estimates divorced from any time dimension. We also examined the sensitivity of the results to changing the starting or ending dates by one year. In all of these cases, most variables showed the same pattern of coefficient magnitudes and statistical significance. 109 Table C.2 Results of County-Level Analysis of Annual Housing Production (All Units) Cross-section analysis Panel analysis Cross-section analysis Panel analysis Cross-section analysis Panel analysis Actual Predicted 1973–1982 1,759,855 1,738,761 1,542,991 1,523,094 1983–1993 2,179,996 2,093,717 2,179,996 2,095,823 1994–1999 659,543 768,944 659,543 758,639 Over-/ Undersupply (%) 1.2 1.3 4.0 3.9 –16.6 –15.0 of the required level. Our estimate is less than a third that size—even without the additional demographic characteristics that were used in the householder analysis. Using another standard, a comparison of national and state housing production from 1991 to 2000 yields an aggregate shortfall of approximately 30 percent—assuming that California should have produced housing at the same rate of growth that the nation did. If we allow for the fact that California emerged from recession two years later than the nation did, the implied shortfall shrinks to only 10 percent. In Figure C.1, we show the difference between the number of new housing units predicted by our panel regressions and the actual production. There seems to have been at least some oversupply at the 1980s peak, but the shortfall in the 1990s is more evident and more consistent from year to year. Even the predicted peak is at an annual production figure that is 46 percent below the 1980s peak and 38 percent below the 1970s peak (although it is not yet clear whether this cycle will eventually have the kind of clearly defined peak that has characterized previous cycles). It should not be surprising that the San Francisco Bay Area has the largest shortfall in new housing, or that the shortfall seems to have accelerated during the Internet boom. It was mentioned above that the percentage changes in real per capita income and real home prices in the Bay Area matched quite closely, but even when income is accounted for 110 Number of units (1000s) 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 350 300 Actual Predicted 250 200 150 100 50 0 Figure C.1—Comparison of Actual and Predicted Annual Housing Production in California, 1973–1999 there is a significant and growing gap. For the current cycle from 1994 to 1999, the shortfall is estimated to be 21 percent, compared to an overproduction of 3 percent in the 1980s. San Diego also shows a consistent gap in housing production in the 1990s but it also had a sizable overproduction during the 1980s peak. The shortfall from 1994 onward is 20 percent, although production rose sharply after 1996; overproduction in the 1980s was 6 percent. These numbers are consistent with the notion that the 1980s was characterized by at least some level of overbuilding in response to the tax incentives for real estate development as well as similar patterns in construction employment relative to that in other sectors. The Sacramento area grew rapidly in the 1990s and, in terms of employment and output change, it suffered only a mild recession in the early 1990s. It is the only major metropolitan area to have both actual and predicted housing production that is approaching the peak levels seen in the previous cycles. The region did not see overbuilding in either of the last two cycles and—despite the rapid rate of housing production since 1995—it still had an estimated shortfall of almost 5 percent in the last five years of our forecast period. So far in this business cycle, it still 111 seems to be in relatively strong shape in terms of both employment and new construction. It is the last, and largest, metropolitan area that has the greatest potential effect on the statewide housing balance. By some measures, such as construction employment, Los Angeles experienced something of a construction bubble in the 1980s and its deflation was compounded by the collapse of the aerospace industry. According to our analysis, there was a slight overproduction of housing in the 1970s and a corresponding shortfall in the early 1980s recession. In the growth phase of the 1980s business cycle, the region once again produced more housing than the model predicted—almost 69,000 units from 1983 to 1990—and undersupplied housing during the deep recession and during the current growth period. Despite the persistent shortfall of housing in the greater Los Angeles region throughout the 1990s, the region essentially balanced the prior excess supply with the current shortfall. By examining the actual and predicted values for each county in the region, it appears that the largest imbalance comes from Los Angeles County itself. Overproduction there was a relatively small 10 percent in the 1980s and underproduction in the 1990s amounted to almost 30 percent of the predicted new supply. New 1980s construction in the other counties in the region outpaced the predicted values by 3 percent in Orange County, 10 percent in Riverside and San Bernardino, and 53 percent in Ventura. In the 1990s, the shortfalls amounted to 15 percent in Riverside and San Bernardino and 10 percent in Orange and Ventura Counties. Some of the “oversupply” in Ventura County in the 1980s can probably be explained by the evolving commuting connections with its much larger neighbor county—if there is more developable land in an adjacent county, that is where construction is more likely to occur in the absence of regulatory barriers. The integrated labor markets of these counties imply that any short-run imbalances in one county may reflect the relative conditions of the individual county labor and land markets (in addition to receptivity to development). However, it is difficult to square sizable shortfalls during the 1990s with the fact that median house prices in Los Angeles County declined by 26 percent from 1989 to 2000—even after correcting for inflation. 112 Finally, the results of the model without fixed effects are compared to two important factors—a region’s ability and willingness to accommodate growth. The first factor is the amount of developable land within the county; the second is community resistance to growth. Table C.3 displays these three numbers for the regions that had sufficient responses on a city manager survey regarding attitudes about growth: deviation between predicted and actual housing production from 1990 to 1999; remaining developable land as a share of total (from Landis et al., 2000), and a population-weighted index of city managers’ appraisals of their city’s resistance to new development (from a comprehensive survey conducted by Lewis and Neiman, 2000, with 1 = least resistance and 5 = most resistance). It is hard to divine a strong pattern among these five observations, but it is interesting to note that the region with the largest shortfall of new housing (San Francisco Bay Area) is the one with the greatest relative share of available land but the strongest political resistance to growth. This observation is consistent with a number of studies that have shown a relationship between growth barriers and home supply and prices.4 The region with the next largest shortfall (San Diego) is more accommodative of growth but has a smaller supply of surplus land. Table C.3 Housing, Developable Land, and Resistance to Growth (1990–1999) Actual–predicted new units, % Developable land as a share of total, % Resistance to growth San Francisco Bay Area Los Angeles, Orange, Ventura Counties –21 –7 44.9 22.3 2.58 2.34 Riverside, San Bernardino Counties –15 20.8 2.30 San Diego County –20 28.9 2.18 Sacramento Region –5 59.8 2.09 ____________ 4See Malpezzi (2001) for an overview of the major findings from the literature about the effects of growth control and zoning on housing. 113 References Blomquist, Glenn, Mark Berger, and John Hoehn, “New Estimates of the Quality of Life in Urban Areas,” American Economic Review, Vol. 78, 1988, pp. 89–107. Boldin, Michael D., “Econometric Analysis of the Recent Downturn in Housing Construction: Was It a Credit-Crunch? Federal Reserve Bank of New York, Research Paper # 9332, December 1993. Browne, Lynn Elaine, “National and Regional Housing Patterns,” Federal Reserve Bank, New England Economic Review, July/August 2000. California Budget Project, Locked Out 2002: California’s Affordable Housing Crisis Continues, Sacramento, California, 2002. California Department of Housing and Community Development, “Guidelines for the Jobs Housing Balance Incentive Grant Program,” 2003, available at http://www.hcd.ca.gov/ca/jhbig/ jhbig20030123guide.pdf. Carlstrom, Charles, and Christy D. Rollow, “Regional Variations in White-Black Earnings,” Federal Reserve Bank of Cleveland, Economic Review, Quarter 2, 1998. Center for Continuing Study of the California Economy, California Economic Growth 1993, 1993, available at http://www.ccsce.com/ about_ccsce.html. Collins, Susan, Barbara Lipman, and Sid Groeneman, “Experience and Expectations: The Home Price Predictions of Owners and Renters,” National Association of Realtors, Washington, D.C., August 1992. Dresch, Marla, and Steven M. Sheffrin, Who Pays for Development Fees and Exactions? Public Policy Institute of California, San Francisco, California, 1997. 115 Fratantoni, Michael, and Scott Schuh, “Monetary Policy, Housing Investment, and Heterogeneous Regional Markets,” May 2000. Fulton, William, Rolf Pendall, Mai Nguyen, and Alicia Harrison, “Who Sprawls Most? How Growth Patterns Differ Across the U.S.,” Brookings Institution, Washington, D.C., 2001. Fulton, William, Paul Shigley, Alicia Harrison, and Peter Sezzi, Trends in Local Land Use Ballot Measure, 1986–2000, Solimar Research Group, Ventura, California, 2000. Gabriel, Stuart A., Joe P. Mattey, and William L. Wascher, “Compensating Differentials and Evolution of the Quality-of-Life Among U.S. States,” Working Papers in Applied Economic Theory 96-07, Federal Reserve Bank of San Francisco, 1996. Gabriel, Stuart A., Joe P. Mattey, and William L. Wascher, “House Price Differentials and Dynamics: Evidence from the Los Angeles and San Francisco Metropolitan Areas, FRBSF Economic Review, No. 1, 1999. Gavin, William T., and Rachel J. Mandal, “Forecasting Inflation and Growth: Do Private Forecasts Match Those of Policymakers? Federal Reserve Bank of St. Louis, September 28, 2000. Glaeser, Edward L., Joseph Gyourko, and Raven Saks, Why Is Manhattan So Expensive? Regulation and the Rise in House Prices, Harvard Institute of Economic Research Discussion Paper 2020, November 2003. Glickfeld, Madelyn, and Ned Levine, “The New Land Use Regulation ‘Revolution’: Why California’s Local Jurisdictions Enact Growth Control and Management Measures,” Graduate School of Architecture and Urban Planning, University of California, Los Angeles, California, June 1990. Glickfeld, Madelyn, and Ned Levine, “Regional Growth . . . Local Reaction: The Enactment and Effects of Local Growth Control and Management Measures in California,” Lincoln Institute of Land Policy, Cambridge, Massachusetts, 1992. 116 Grebler, L., and S. J. Maisel, “Determinants of Residential Construction: A Review of Present Knowledge,” in Commission on Money and Credit, Impacts of Monetary Policy, Prentice-Hall, Englewood Cliffs, New Jersey, 1963, pp. 496–505. Gyourko, Joseph, and Joseph Tracy, “The Structure of Local Public Finance and the Quality of Life,” Journal of Political Economy, Vol. 99, No. 4, 1992, pp. 774–806. Hamilton, Bruce W., and Robert M. Schwab, “Expected Appreciation in Urban Housing Markets,” Journal of Urban Economics, Vol. 18, 1985, pp. 103–118. Hanushek, Eric A., and John M. Quigley, “The Determinants of Housing Demand,” Research in Urban Economics, Vol. 2, 1982, pp. 221–242. Hoehn, John, Mark Berger, and Glenn Blomquist, “A Hedonic Model of Interregional Wages, Rents, and Amenity Values,” Journal of Regional Science, Vol. 27, 1987, pp. 605–620. Kaiser, Ronald W., “Using Capital Markets’ Value Cycles in Allocating to Real Estate vs. Stocks or Bonds,” Journal of Real Estate Portfolio Management, Vol. 5, No. 1, 1999. Klein, L. R., The Keynesian Revolution, 2nd ed., Macmillan, New York, 1966, p. 233. Krainer, John, “House Price Bubbles,” Federal Reserve Bank of San Francisco Economic Letter, No. 2003-06, 2003. Landis, John D., “Opening the Doors to Infill Housing,” PowerPoint presentation, Department of City and Regional Planning, UC Berkeley, 2003. Landis, John D., Michael Smith-Heimer, Michael Larice, Michael Reilly, Mary Corley, and Oliver Jerchow, Raising the Roof: California Housing Development Projections and Constraints 1997–2020, Statewide Housing Plan, Department of Housing and Community Development, Sacramento, California, 2000. 117 Leamer, Edward E., “Bubble Trouble? Your Home Has a P/E Ratio Too,” UCLA Anderson Forecast Report, 2002. Leamer, Edward E., “Update June 2, 2003; Bubble Trouble? Your Home Has a P/E Ratio Too,” UCLA Anderson Forecast Report, 2003. Levine, Ned, “The Effects of Local Growth Controls on Regional Housing Production and Population Redistribution in California,” Urban Studies, Vol. 36, No. 12, 1999, pp. 2047– 2068. Lewis, Paul, California’s Housing Element Law: The Issue of Local Noncompliance, Public Policy Institute of California, San Francisco, California, 2003. Lewis, Paul, and Elisa Barbour, California Cities and the Local Sales Tax, Public Policy Institute of California, San Francisco, California, 1999. Lewis, Paul, and Elisa Barbour, “Development Priorities in California Cities: Results from a PPIC Survey,” Occasional Paper, Public Policy Institute of California, San Francisco, California, 1998. Lewis, Paul, and Max Neiman, “Residential Development and Growth Control Policies: Survey Results from Cities in Three California Regions,” Occasional Paper, Public Policy Institute of California, San Francisco, California, 2000. Maisel, S. J., “A Theory of Fluctuations in Residential Construction Starts,” American Economic Review, Vol. 53, No. 3, June 1963. Malpezzi, Stephen, “NIMBYs and Knowledge: Urban Regulation and the ‘New Economy,’” working paper, Center for Urban Land Economic Research, University of Wisconsin, WisconsinMadison, 2001. Mayer, Christopher J., and C. Tsuriel Somerville, “Using Empirical and Theoretical Models of Housing Supply,” Working Paper No. 9612, Federal Reserve Bank of Boston, December 1996. Moller, Rosa, Hans Johnson, and Michael Dardia, What Explains Crowding in California? California Research Bureau, Sacramento, California, February 2002. 118 Myers, Dowell, Analysis with Local Census Data, Academic Press, San Diego, California, 1992. Myers, Dowell, and Julie Park, “The Great Housing Collapse in California,” FannieMae Foundation, 2002. Myers, Dowell, William C. Baer, and Seong-Youn Choi, “The Changing Problem of Overcrowded Housing,” Journal of the American Planning Association, Vol. 62, 1996, pp. 66–84. Roback, Jennifer, “Wages, Rents, and the Quality of Life,” Journal of Political Economy, Vol. 90, 1982, pp. 1257–1278. Rosen, Sherwin, “Wage-Based Indexes of Urban Quality of Life,” in P. Mieszkowski and M. Strazheim (eds.), Current Issues in Urban Economics, John Hopkins University Press, Baltimore, Maryland, 1979, pp. 74–104. Stover, Mark Edward, and Charles Leven, “Methodological Issues in the Determination of the Quality of Life in Urban Areas,” Urban Studies, Vol. 29, No. 5, 1992, pp. 737–754. U.S. Department of Housing and Urban Development, U.S. Housing Market Conditions, 3rd Quarter 2003, 2003, available at http:// www.huduser.org/periodicals/ushmc/fall03/ushmc_fall03.pdf. Wasserman, Miriam, “Appreciating the House. Housing as an Investment,” Federal Reserve Bank of Boston, Regional Review, 1998. Weintraub, Daniel, “The Prices for Housing Soar—Along with Demand,” Sacramento Bee, December 1, 2002. 119 About the Authors HANS JOHNSON Hans Johnson is a demographer at the Public Policy Institute of California. His research interests include international and domestic migration, population estimates and projections, and state and local demography. Before joining PPIC as a research fellow, he was the senior demographer at the California Research Bureau, where he conducted research for the state legislature and the governor’s office on population issues. He has also worked as a demographer at the California Department of Finance, specializing in population projections. He holds a Ph.D. in demography from the University of California, Berkeley. ROSA MOLLER Rosa M. Moller is a senior economist at the California Research Bureau, where she has written and co-authored reports and technical briefs on a wide range of issues for the legislature and state government. Before joining the California Research Bureau in 1991, she worked on various socioeconomic and demographic studies for the United Nations, including the Economic Commission for Latin America. She received her Ph.D. in economics from the University of California, Davis, and a B.A. and M.A. in sociology from the Catholic University of Chile in Santiago. MICHAEL DARDIA Michael Dardia is vice president at The SPHERE Institute. His research interests cover workforce development, redevelopment, regional economics, and public finance and taxation issues. Before joining SPHERE, he was a research fellow at the Public Policy Institute of California and a consultant to the RAND Corporation. He is a member of the California Workforce Investment Board’s Economic Advisory Council and has participated in numerous other state and local government advisory groups. He edits the growth and employment series of the California Policy Review and writes the regional highlights for the Bay Area Economic Pulse. He earned his Ph.D. in policy analysis from the RAND Graduate School. 121 Related PPIC Publications Who Pays for Development Fees and Exactions? Marla Dresch and Steven M. Sheffrin Planned Developments in California: Private Communities and Public Life Tracy M. Gordon California’s Housing Element Law: The Issue of Local Noncompliance Paul G. Lewis Cities Under Pressure: Local Growth Controls and Residential Development Policy Paul G. Lewis and Max Neiman PPIC publications may be ordered by phone or from our website (800) 232-5343 [mainland U.S.] (415) 291-4400 [Canada, Hawaii, overseas] www.ppic.org 123" } ["___content":protected]=> string(102) "

R 304HJR

" ["_permalink":protected]=> string(98) "https://www.ppic.org/publication/in-short-supply-cycles-and-trends-in-california-housing/r_304hjr/" ["_next":protected]=> array(0) { } ["_prev":protected]=> array(0) { } ["_css_class":protected]=> NULL ["id"]=> int(8292) ["ID"]=> int(8292) ["post_author"]=> string(1) "1" ["post_content"]=> string(0) "" ["post_date"]=> string(19) "2017-05-20 02:36:19" ["post_excerpt"]=> string(0) "" ["post_parent"]=> int(3457) ["post_status"]=> string(7) "inherit" ["post_title"]=> string(8) "R 304HJR" ["post_type"]=> string(10) "attachment" ["slug"]=> string(8) "r_304hjr" ["__type":protected]=> NULL ["_wp_attached_file"]=> string(12) "R_304HJR.pdf" ["wpmf_size"]=> string(6) "339031" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(211585) "In Short Supply? Cycles and Trends in California Housing ••• Hans P. Johnson Rosa M. Moller Michael Dardia 2004 Library of Congress Cataloging-in-Publication Data Johnson, Hans P. In short supply? : cycles and trends in California housing / Hans P. Johnson, Rosa M. Moller, and Michael Dardia. p. cm. ISBN: 1-58213-079-5 1. Housing—California. 2. California—Population. 3. California—Economic conditions. I. Title: Cycles and trends in California housing. II. Moller, Rosa Maria. III. Dardia, Michael. IV. Title. HD7303.C2J64 2004 333.33'8'09794—dc22 2004001896 Copyright © 2004 by Public Policy Institute of California All rights reserved San Francisco, CA Short sections of text, not to exceed three paragraphs, may be quoted without written permission provided that full attribution is given to the source and the above copyright notice is included. PPIC does not take or support positions on any ballot measure or on any local, state, or federal legislation, nor does it endorse, support, or oppose any political parties or candidates for public office. Research publications reflect the views of the authors and do not necessarily reflect the views of the staff, officers, or Board of Directors of the Public Policy Institute of California. Foreword During the 1990s, there was a disturbing and widely noted decline in the construction of new housing units in California. In the 1980s, almost 2.1 million units were built, averaging 207,000 a year. The following decade, new production fell to 1.1 million for an average of 110,000 per year. This decline has been attributed to a myriad of causes, from demanding housing codes and construction requirements to tighter growth controls and planning guidelines. Taken together, the story goes, these factors have made new construction ever more difficult. This difficulty, in turn, has led to the dramatic falloff in production and its predictable consequence: high and increasing housing costs. After reviewing the data, Hans Johnson, Rosa Moller, and Michael Dardia have concluded that a housing shortage does exist, but that it is much smaller than implied by the reduction of new construction in the 1990s. Slower population growth, unprecedented outflows of domestic migrants, greater numbers of immigrants and children (who consume less housing than other demographic groups), and business cycle swings all contributed to a much lower demand for housing in the 1990s. The authors assign nearly 80 percent of the downturn in housing production to these demand-side variables and argue that the actual housing shortfall statewide is closer to 138,000 units—not the 1,000,000 units implied by a simple comparison of production figures over the two decades. Almost the entire shortage has occurred in the Bay Area, Los Angeles, and San Diego—the state’s most populous regions. Pressure on the housing stock is nothing new for these areas. At least since the end of the Second World War, some California homebuyers have been willing to pay huge premiums to live where the amenities are, and the authors note that large macroeconomic trends may explain more of this willingness than levels of new construction do. This part of the analysis once again points to the role demand has played in shaping the state’s housing market over the last two decades. Acceptable in theory, this iii finding is unlikely to console first-time homebuyers in the state’s largest metropolitan areas. Here the study acknowledges the supply side of the equation by noting other research that indicates these regions were the least receptive to new production during this period. Although the authors find little or no evidence of shortage in the rest of the state, they note that residents of many of those regions face affordability challenges. To the untrained eye, it may be difficult to distinguish between housing shortages and affordability challenges, but the authors correctly attend to the substantial variation in the state’s regional housing markets. From a public policy perspective, this variation is especially important insofar as it points policymakers away from a single, statewide response. Once again, PPIC research has underscored the regional diversity of California’s markets, and if the policy lesson sounds familiar, it has proven difficult to master. David W. Lyon President and CEO Public Policy Institute of California iv Summary By many measures, California appears to be in the midst of a severe housing crisis. New construction is anemic compared to past levels. Prices are high in many regions and in some areas, such as the Bay Area, exorbitant. Vacancy rates, especially for rental units, are low. And the number of persons per housing unit has continued to increase in California even as it has declined in the rest of the nation. Other measures, however, suggest that the crisis is not so severe, or even that there is no crisis in much of the state. The median value of homes in the state was lower in 2000 than in 1990, once inflation is taken into account. In addition, incomes rose faster than rents or housing prices in much of the state. Housing prices and rents remain higher in California’s largest metropolitan areas than in the rest of the nation, but this could be a long-standing phenomenon related to incomes and amenities. Nationally, new housing construction was markedly slower in this business cycle as well. Finally, the source of California’s population growth—children and immigrants—could explain the higher household sizes. In this report, we explore California’s housing markets and attempt to explain why housing construction has lagged in the 1990s compared to previous business cycles. Can the sluggish pace of construction in the 1990s be explained by unique macroeconomic developments and demographic changes that took place in that period? To address this question, we analyze the effects of macroeconomic and demographic factors on California’s housing market. We find that the slowdown in new construction in the 1990s can largely be explained by these factors, but we also find evidence of housing shortages in the state’s largest metropolitan areas. The slowdown of housing construction in the 1990s was experienced nationally as well as in California, suggesting that the sluggish pace of new construction in the 1990s is at least partly explained by the national macroeconomic v events of that period. The more pronounced slowdown in California can be attributed to the severity of the 1990 recession, the slow pace of the recovery, and the demographic changes that took place in the state. In particular, we found that • There is evidence of a shortage in the supply of new housing units in California in 2000, but the magnitude of the shortage is much lower than has been estimated by others. Our best estimates are that the shortage of housing units in the state is around 138,000 units, far lower than the estimates often cited in the media. This suggests that most (over 80 percent) of the apparent shortage in California’s occupied housing units can be attributed to demographic differences between California and the rest of the nation. • Regional variations in California’s housing markets are large, with almost all of the shortage in new housing units originating in the Bay Area, Los Angeles, and San Diego, the state’s most populous regions. • The slowdown in population growth during the 1990s can explain much of the slowdown in new housing construction. Adjusting for census undercounts, California’s population grew by 3.8 million people in the 1990s, far lower than the 6.2 million increase of the 1980s. Slower population growth in the 1990s can be attributed to tremendous and unprecedented flows of domestic migrants out of California. • In addition to the slowdown in population growth, the composition of California’s population growth in the 1990s explains much of the apparent lack of new housing in the state, with immigrants and children constituting a much larger share of the state’s growth in the 1990s than in the 1980s. Immigrants and children tend to consume less housing than other demographic groups. Children do not form their own households, and immigrants often live in larger, extended-family households (even where housing prices are relatively low). Once we control for these and other demographic differences between vi California and the nation, we observe similar trends in the creation of new households. • More than 80 percent of annual changes in new construction are explained by changes in macroeconomic and demographic factors. This explained portion varies by region, ranging from 67 percent for the North Coast region to as much as 90 percent for the Central Eastern area. • Business cycle effects are an important determinant of annual fluctuations in housing construction. In particular, changes in new construction are closely related to changes in the business cycle (as measured by changes in unemployment). Construction slowed considerably during the recession of the early 1990s. • Changes in macroeconomic factors (such as changes in interest rates, income, prices, and expectations of inflation) driven by changes in economic growth (in the United States and the rest of the world) and by U.S. economic policies are closely related to changes in new construction. • Developments in the financial markets partly determined the slower pace of new construction in the 1990s. Investors see real estate and stocks as substitutes. Investments in stocks in the late 1990s had a dampening effect on new home construction, as expected returns on housing decreased relative to investments in stocks. Furthermore, we suspect that builders’ and developers’ expectations of profits may have been influenced by lower expectations of inflation resulting from the way that the Federal Reserve conducted monetary policies and the unusual low inflation in the period. Profit expectations were also influenced by demographic changes, specifically lower rates of household formation that accompanied reduced population growth in key age groups and with domestic out-migration. These demographic changes led to decreased investment in new construction as builders and developers expected slower growth in the demand for new housing. In some markets, expectations may have fallen too far, leading to an underproduction of new housing units. This underproduction could vii have been exacerbated by restrictive land-use policies, especially in the Bay Area and Los Angeles markets. These findings do not mean that there are no hardships with respect to housing supply and new construction in California. There may be serious problems in markets for low-income housing, and there is evidence of a housing shortage in the state’s largest metropolitan areas. However, they do suggest that the supply crisis may be overstated, and that our position in 2000 was perhaps better, and certainly not much worse, than in 1990. viii Contents Foreword ..................................... Summary..................................... Figures ...................................... Tables ....................................... Acknowledgments ............................... iii v xi xiii xv 1. INTRODUCTION ........................... What Are the Assertions Regarding a Housing Shortage? .... How Can a Housing Shortage Be Measured? ........... Outline of This Report ......................... 2. TRENDS AND PATTERNS IN CALIFORNIA HOUSING ................................ Prices and Rents ............................. Affordability ................................ New Construction ............................ How Is Population Growth Accommodated? ........... New Housing Units.......................... Increases in Household Size ..................... Declines in Vacancy Rates...................... Jobs and Housing ............................ Summary .................................. 3. CAN MACROECONOMIC FACTORS EXPLAIN CALIFORNIA’S HOUSING MARKETS? ............ Business Cycles, Macroeconomic Events, and Construction .. A Simple Housing Market Model .................. Factors Affecting Housing Demand ................. Supply Determinants .......................... Ten Regional Housing Markets in California ........... Results of the Model of California Housing Markets ...... Summary .................................. 1 1 3 4 7 7 16 21 23 23 25 28 30 31 35 36 42 43 44 45 46 51 ix 4. CAN DEMOGRAPHIC CHANGES EXPLAIN CALIFORNIA’S HOUSING MARKETS? ............ Demographic Determinants of Demand .............. Population Growth .......................... Age .................................... Gender .................................. Nativity and Ethnicity ........................ Marital Status ............................. Householder Rate Model: Temporal Trends in the Probability of Forming a Household ............. Temporal Trends in Householder Rates ............. Differences in Householder Rates Between California and the Rest of the United States .................. Householder Rates in California Metropolitan Areas ..... Estimating the Shortage of Occupied Housing Units..... Summary .................................. 53 55 55 56 56 58 60 61 61 65 67 70 75 5. CONCLUSION AND POLICY CONSIDERATIONS.... 79 Appendix A. A Housing Market Model ....................... 89 B. Householder Rate Models ....................... 103 C. Panel Data Analysis of Housing Supply ............... 107 References .................................... 115 About the Authors ............................... 121 Related PPIC Publications .......................... 123 x Figures 2.1. Median Home Prices in California, 1975–2002, Nominal and Real (2000 $) ................... 2.2. Annual Inflation-Adjusted (Real 2000 $) Median Home Prices in Selected Counties in California, 1989–2000 ... 2.3. Median Home Prices in the United States and California, 1998–2000, in Current Dollars .......... 2.4. Share of Households That Can Afford the MedianPriced Home in Major Counties in California, 1990 and 2000 .................................. 2.5. New Multifamily and Single-Family Housing Construction in California, 1967–2002 ............ 2.6. New Housing Construction in the United States and California, Indexed to 1.0 in 1986 (Peak Year of Production), 1973–2002 ..................... 3.1. California Construction and Employment Cycles ...... 3.2. Expectations of Inflation, 1972–2001 ............. 3.3. Term Structure (Difference Between the Long-Term and Short-Term Interest Rates), 1971–2001 ............ 3.4. Comparison of Actual Changes in California New Housing Construction and Model Fitted Values, 1967– 2000 .................................. 4.1. Growth in Child and Adult Populations in California, by Decade................................. 4.2. Probability of Being a Householder in California, by Age, 2001 ............................... 4.3. Change in California Population, by Age Group and Decade................................. 4.4. Change in Foreign-Born and U.S.-Born Populations in California, by Decade ....................... 4.5. Decennial Change in Population in California, by Marital Status ............................ 8 9 10 17 21 22 36 40 41 50 55 57 58 59 60 xi 4.6. Probability of Being a Householder in California and the Rest of the United States, 1976–2001 ............. 62 4.7. The Effect of Demographic Changes on the Probability of Being a Householder in the Rest of the United States, 1976–2001 .............................. 63 4.8. The Effect of Demographic Changes on the Probability of Being a Householder in California, 1976–2001 ..... 64 4.9. Difference in Householder Probabilities Between California and the Rest of the United States ......... 65 4.10. Age-Standardized Householder Rates in California and the Rest of the United States, by Ethnicity and Nativity, 2000 .................................. 67 4.11a. Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, No Demographic Controls, 2000 ................ 68 4.11b. Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, with Controls for Age, Gender, Ethnicity, and Nativity, 2000 ....... 69 4.12. Estimates of the Undersupply of Housing in California Based on Householder Rates in California and the Rest of the United States, by Decade ................. 72 4.13. New Occupied Housing Units, by Decade: Actual Versus Necessary for California (Controlling for Demographic Differences) .................... 73 4.14. Decennial Shortage of New Occupied Housing Units ... 74 4.15a. Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of the United States, No Demographic Controls, 2000 ...... 76 4.15b. Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of the United States, Controls for Age, Gender, Ethnicity, and Nativity, 2000 ............................ 77 C.1. Comparison of Actual and Predicted Annual Housing Production in California, 1973–1999 ............. 111 xii Tables 2.1. Median Value of Owner-Occupied Housing Units in California’s Counties, 1980, 1990, and 2000 ........ 12 2.2. Median Gross Rent in California’s Counties, 1980, 1990, and 2000 ........................... 14 2.3. Rent and Homeowner Costs as a Percentage of Income................................. 19 2.4. Ratio of Household Population Change to Total Housing Units Change in California’s Counties: 1970s, 1980s, and 1990s .......................... 24 2.5. Number of Persons per Occupied Household, 1970– 2000 .................................. 26 2.6. Vacancy Rates in California’s Counties, 1990 and 2000 .................................. 29 2.7. Nonfarm Jobs per Housing Unit in California’s Counties, 1990 and 2000 ..................... 32 3.1. Effect of Explanatory Factors on Changes in New Housing Construction: Results of the Macroeconomic Model, by Region .......................... 47 A.1. Analysis of Changes in New Construction in California, by Region (Time-Series Results, No Lags) .......... 97 A.2. Time-Series Results Allowing for Lagged Responses .... 99 C.1. Panel Regression Results ..................... 108 C.2. Results of County-Level Analysis of Annual Housing Production .............................. 110 C.3. Housing, Developable Land, and Resistance to Growth (1990–1999) ............................. 113 xiii Acknowledgments We have benefited from the comments, suggestions, and assistance of a number of people. We thank Anthony Downs, Bill Fulton, Paul Lewis, Dowell Myers, Dean Misczynski, and John Quigley for their reviews of an earlier draft of this report. Michael Teitz has provided encouragement for this project from its inception and has contributed many helpful comments. Joyce Peterson and Peter Richardson have provided important structural and editorial assistance. The authors are responsible for any errors of fact or interpretation. xv 1. Introduction Despite a strong economic performance overall during the late 1990s, California’s housing production appears to have failed to keep pace with either population or job growth. Indeed, the state’s most recent economic boom seems to have created a shortage of housing— particularly in highly populated coastal areas—as evidenced by a rapid rise in home prices. Some observers have pointed to the trend toward larger household sizes as further evidence of a housing shortage. This report attempts to explain why housing construction has lagged in the 1990s compared to previous business cycles. The study addresses the question of whether the sluggish pace of construction in the 1990s can be explained by unique macroeconomic developments and demographic changes that took place in that period. To address this question, we analyze the effects of macroeconomic and demographic factors on California’s housing market. Our macroeconomic model examines factors that economists have found to be the most important predictors of changes in new housing construction. Our demographic model considers whether the composition of the state’s population growth has had an effect on household formation and therefore on housing demand. Our results provide important information on the responsiveness of California housing markets to changes in demographics and economic growth. What Are the Assertions Regarding a Housing Shortage? The popular perception—shared by policymakers and the public alike—is that California is in the midst of a housing crisis that has been driven by the failure of new housing construction to keep up with the pace of growth in the state. It is easy to see why this is the general impression. 1 • From 1990 to 1999, 1.1 million housing units were built in California, or an average of 110,000 per year; from 1980 to 1989, almost 2.1 million were built (207,000 per year). The ratio of new residents to new housing units rose from 3.1 over the 1980s to 3.9 over the 1990s. • The median home price in California increased from $193,334 in 1990 to $241,779 in 2000, and housing affordability1 in 2000 was only 34 percent statewide and below 20 percent in most of the San Francisco Bay Area. • The share of homeowners who spend more than 30 percent of their income on housing rose from 29.7 percent in 1989 to 31.5 percent in 1999. These statistics are buttressed by any number of anecdotes about people having to leave their hometowns because of unaffordable housing and of employers unable to attract or retain workers because of exorbitant housing prices. All this evidence implies that there is a serious housing shortage in California. Arrayed against these facts, however, are other figures that either fail to document the outcomes one would expect in the midst of a housing crisis or put the dire statistics into historical context. • Although it is true that housing prices increased significantly over the 1990s, once we correct for inflation, the median home price in California was 3.5 percent lower in 2000 than it was in 1990. • Likewise, home prices are strongly related to personal income, which increased at a noticeably faster pace than inflation from 1990 to 2000. Although housing affordability in California at the peak of the recent business cycle was only 34 percent, it was 20 percent at the previous peak in 1990. Similarly, in only three of the counties where affordability is tracked back to 1990 was affordability lower in 2000 than it was in 1990; in most ____________ 1Housing affordability is defined as the percentage of households that can afford the median priced single-family home. Annual figures are prepared by the California Association of Realtors. 2 counties, housing is more affordable now than it was in the last cycle. Outside the San Francisco Bay Area, housing affordability improved dramatically over the 1990s. • Household size (the number of persons per household) has increased since 1990, but demographic changes explain most of the increases we observe (see Moller et al., 2002). • Despite the fact that homeownership costs rose as a share of income in the 1990s, the percentage of renters who spent more than 30 percent of their income on rent fell from 47.7 percent to 44.7 percent. • Although the homeownership rate in California lags the U.S. average (56.9 percent versus 66.2 percent), it nevertheless rose slightly, from 55.6 percent to 56.9 percent, from 1990 to 2000. It is the contradiction between these two sets of facts that we address in this study. How Can a Housing Shortage Be Measured? We use several approaches and datasets to evaluate California’s housing markets. Housing markets are better understood as regional rather than state phenomena. Some regions might be more supplyconstrained than others—and macroeconomic and demographic forces could operate differentially between regions. Identifying a housing shortage is difficult, depending in part on the measures chosen as evidence of a shortage. The simplest method, for example, would consider only rents and prices—increases could be taken as evidence of demand outstripping supply. This limited approach would be misleading, however, as it would fail to take into account many other determinants of housing prices, such as income and interest rates. Because of the complexities of housing supply and demand, we have approached the issue of whether there is a shortage in the supply of new housing in two complementary ways. Our first approach is a traditional econometric model that analyzes the effects of factors commonly used to explain housing market patterns. These factors include interest rates, unemployment rates, income, and prices. We analyze the California housing market as a whole as well as ten regional markets. This 3 econometric approach allows us to identify the extent to which macroeconomic factors influence changes in new housing construction. These annual fluctuations can be quite dramatic and are clearly related to macroeconomic factors. Details of the econometric model are in Appendix A. Our second method investigates the demographic determinants of housing demand by building a householder rate model. This approach is particularly important given the large demographic changes in many of the state’s metropolitan areas. The demographic model provides us with important information on how the nature of California’s population change influences the demand for housing. Because the population of the state does not vary dramatically from one year to the next, the effects are seen gradually over a 25-year period (1977 through 2001). Details of the demographic model are in Appendix B. By developing multiple methods and measures of California’s housing markets, we are able to assess the importance of very different inputs. In the end, we observe important changes in long-term demographic determinants of housing demand and identify important cyclical variations associated with business cycles. Outline of This Report Before proceeding to our detailed analyses of the state and regional housing markets, we summarize the major trends and indicators in housing and population in California. In Chapter 2, we provide descriptive measures of California’s housing market, for both the state and its regions. These descriptive measures allow us to examine the most accessible and well-known indicators of housing markets in California, including new construction, prices, vacancy rates, and household size. We also provide comparisons with national trends when available and of interest. Chapter 3 presents our econometric model of California’s housing markets, discusses the macroeconomic factors that help determine annual fluctuations in housing construction, and provides results of the macroeconomic models. Chapter 4 analyzes the effect of changes in the state’s population on the consumption of housing in the state. A householder rate model is constructed to determine differences in household formation between demographic subgroups in California 4 (and its largest metropolitan areas) and the rest of the nation. Chapter 5 summarizes our findings and discusses their implications for policy. Finally, it is important to note that this report does not assess all aspects of the California housing market. In particular, we do not focus on low-income housing in the state. Instead, we focus on the housing market as a whole. There is quite possibly a significant mismatch between the type of new housing units being built and the type of new households forming in the state. That and other topics are subjects of future research. 5 2. Trends and Patterns in California Housing This chapter provides summary data on key measures of California’s housing. We discuss prices and rents, affordability, new construction, population change, persons per household, jobs and housing, and vacancy rates. We use annual data series from the California Association of Realtors and the Department of Finance, as well as recently released decennial census data. Annual measures of change provide important evidence of business cycle events, whereas long-term demographic changes are assessed across decades. Many of the measures suggest that California’s position in 2000 was not appreciably different from its position in 1990. Nonetheless, these measures show considerably more housing stress in some of the state’s largest metropolitan areas than in the rest of the nation Prices and Rents Between 1975 and 2002, the median home price in California rose over seven-fold, from $42,267 to $315,870.1 These dollar figures are not adjusted for inflation and therefore give the impression of a larger rise in prices than is warranted. Figure 2.1 shows median home prices in California from 1975 to 2002 in both nominal and real (i.e., inflationadjusted) terms. Once corrected for inflation, the median home price in California hit a peak of $266,846 in 1989 and then declined substantially to 1996. By 2000, the median home price was still lower than the 1989 peak, but subsequent increases led to a new high of almost $300,000 by 2002. ____________ 1Annual prices reported are averages of monthly prices compiled by the California Association of Realtors. These are the prices of sold houses; below, we present information on median values of all owner-occupied houses as estimated by their owners in decennial censuses. 7 Median home price ($) 350,000 300,000 250,000 200,000 150,000 100,000 50,000 0 Real (in 2000 dollars) Nominal 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 Figure 2.1—Median Home Prices in California, 1975–2002, Nominal and Real (2000 $) The statewide median price is drawn from regions with quite different price levels. The importance of regional economic fortunes can be seen when the real and nominal median prices are compared across several of the largest counties in California. Figure 2.2 compares the median home price in Santa Clara, Los Angeles, and Orange Counties from 1989 to 2000 in real terms (adjusting for inflation). The early 1990s recession clearly took a toll on the Los Angeles housing market, less so in Orange County. From Figure 2.2, it is clear that the median statewide home price does not reflect the state of regional housing markets: The median real home price in Los Angeles County in 2000 was still 26 percent below its previous peak in 1989; Orange County prices were basically unchanged; and prices in Santa Clara County had increased 45 percent faster than inflation. In fact, of the counties that the California Association of Realtors has tracked since 1990, the only counties outside the Bay Area that had real price gains during the 1990s were Monterey and San Diego. In most other markets, prices in 2000 were still lower in real terms and some were down 20 percent or more from 1990. 8 600,000 500,000 400,000 Santa Clara County Orange County Los Angeles County Median home price ($) 300,000 200,000 100,000 0 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Figure 2.2—Annual Inflation-Adjusted (Real 2000 $) Median Home Prices in Selected Counties in California, 1989–2000 In 2000, the median home price (among houses sold) in California was $241,779—74 percent higher than the national median price of $139,000. Even though California is the largest state and is itself composed of several major metropolitan areas, it is small in comparison to the nation as a whole, and its metropolitan areas are more tightly linked to each other than to other areas in the nation. This proximity has made housing prices in the state significantly more volatile than national prices. Figure 2.3 shows how median home prices compare at the state and national levels (the national figure includes California).2 The median price in California rose to 2.2 times the national median price in 1989, fell to 1.5 times in 1996 and 1997, and then rose back to 1.7 times the national median in 2000. However, not all regional markets in California are expensive by national standards. Median home prices in Fresno, Riverside, and San Bernardino Counties were less ____________ 2Throughout this report, we provide both California and national data. Our preferred comparison is California versus the rest of the nation, but where such information is not available—as in this case—we provide data for California and the nation (including California) and provide a note to that effect. 9 300,000 250,000 California United States Median home price ($) 200,000 150,000 100,000 50,000 0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 SOURCE: National Association of Realtors, data for the United States include California. Figure 2.3—Median Home Prices in the United States and California, 1988–2000, in Current Dollars expensive in 2000 than the national average, and Sacramento was only 5 percent higher. Data from the decennial censuses allow us to identify long-term changes in rents and values in all of California’s counties for rented housing units and for all owner-occupied housing units (not just those sold).3 Patterns of change in rents and values of owner-occupied housing units were similar, with California and many of its counties experiencing much higher rents and housing values than the rest of the country, but with most if not all of the increase occurring during the 1980s rather than the 1990s. Indeed, in real terms, median values of owner-occupied housing units declined in California from 1980 to 1990 ____________ 3We consider several measures of housing to identify local market conditions, comparing indicators from the most recent census with those of past censuses. Although these ten-year changes do not allow for the consideration of annual patterns of change, the 1980, 1990, and 2000 censuses all coincided with strong economic conditions. Thus, the decennial patterns are not affected by short-term business cycles and should provide us with an accurate picture of long-term trends at the local level in California. 10 (in nominal terms, values increased 9 percent). Regional variations are substantial (see Table 2.1). Were it not for declines in Los Angeles County, median values would have increased in the state from 1990 to 2000. As with rents, counties with the greatest declines in housing values were in the Los Angeles metropolitan area (five of the six counties with the greatest declines were Los Angeles, Ventura, San Bernardino, Riverside, and Orange). Still, a majority of counties (36) experienced an increase in real values between 1990 and 2000. And from 1980 to 2000, eight California counties experienced real increases in values of more than 50 percent. Increases were focused in the Bay Area (including Santa Clara, San Francisco, San Mateo, Alameda, Marin, and Napa Counties) and the Central Coast (Santa Cruz and San Benito Counties). By 2000, California’s most expensive county, Marin, had home values over four times higher than the national average. Nonetheless, 15 of the state’s 58 counties still had median home values lower than the national average of $119,600 in 2000, including two of the state’s medium-sized metropolitan areas (Fresno and Bakersfield). As shown in Table 2.2, 33 of the state’s counties experienced declines in rents from 1990 to 2000, with the greatest declines in the Los Angeles metropolitan area (including Los Angeles, Orange, Riverside, San Bernardino, and Ventura Counties). In the Los Angeles area, the declines of the 1990s followed the spectacular increase in rents that occurred during the 1980s. These declines can be attributed to the strength and depth of the recession of the early 1990s—a recession that was especially pronounced in the Los Angeles area—and to a speculative bubble in prices around 1990. Only five counties experienced real increases in rents of more than 10 percent for the decade from 1990 to 2000, and only three were counties with large populations (San Mateo, Santa Clara, and San Francisco). Still, rents in much of California remain far higher than in the rest of the nation. By 2000, rents in 35 of California’s 58 counties were higher than the national average; only two of the state’s larger metropolitan areas, Fresno and Bakersfield, had average rents lower than those of the nation as a whole. From these trends, it does not appear that prices in this business cycle are much different from prices in the previous one. As of 2000, most regional markets had not yet regained their 1989–1990 peak prices 11 Table 2.1 Median Value of Owner-Occupied Housing Units in California’s Counties, 1980, 1990, and 2000 Median Value (Real 2000 $) % Change 1980– 1990– 1980– 1980 1990 2000 1990 2000 2000 United States 98,848 103,425 119,600 5 16 21 United States excluding California 99,266 95,391 112,500 –4 18 13 California 180,316 251,583 225,000 40 –16 18 Far North Del Norte 105,644 111,484 121,100 6 9 15 Humboldt 120,917 114,591 133,500 –5 17 10 Lake 124,312 123,525 122,600 –1 –1 –1 Lassen 106,704 89,731 106,700 –16 19 0 Mendocino 148,071 159,910 170,200 8 6 15 Modoc 85,067 62,281 69,100 –27 11 –19 Nevada 169,072 199,919 205,700 18 3 22 Plumas 139,798 116,145 137,900 –17 19 –1 Sierra 123,675 105,398 128,600 –15 22 4 Siskiyou 111,796 87,659 100,300 –22 14 –10 Trinity 116,463 106,434 112,000 –9 5 –4 Upper Sacramento Valley Butte 125,372 121,713 129,800 –3 74 Colusa 100,340 88,177 107,500 –12 22 7 Glenn 101,825 86,882 94,900 –15 9 –7 Shasta 129,403 117,828 120,800 –9 3 –7 Sutter 126,857 118,735 120,700 –6 2 –5 Tehama 102,250 89,083 103,100 –13 16 1 Yuba 94,400 86,882 89,700 –8 3 –5 Sacramento Metro El Dorado 177,558 201,862 194,400 14 –4 9 Placer 165,254 218,176 213,900 32 –2 29 Sacramento 137,040 167,031 144,200 22 –14 5 Yolo 141,283 177,908 169,800 26 –5 20 San Joaquin Valley Fresno 131,949 107,211 104,900 –19 –2 –20 Kern 116,038 106,693 93,300 –8 –13 –20 Kings 99,067 91,155 97,600 –8 7 –1 Madera 127,494 110,966 118,800 –13 7 –7 Merced 112,644 116,663 111,100 4 –5 –1 San Joaquin 119,645 156,025 142,400 30 –9 19 Stanislaus 127,494 159,262 125,300 25 –21 –2 Tulare 103,734 95,039 97,800 –8 3 –6 12 Table 2.1 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Median Value (Real 2000 $) % Change 1980– 1990– 1980– 1980 1990 2000 1990 2000 2000 180,952 200,681 320,325 165,890 221,894 263,897 232,077 143,192 187,528 291,722 281,104 459,530 236,433 381,711 441,273 372,518 190,597 259,740 303,100 267,800 514,600 251,300 396,400 469,200 446,400 178,300 273,200 61 40 43 43 72 67 61 33 39 4 68 –5 33 12 61 6 51 4 79 6 78 20 92 –6 25 5 46 183,498 155,283 164,617 220,621 199,620 254,172 265,567 276,055 322,668 330,696 265,800 284,000 230,000 293,000 377,500 39 71 68 46 66 5 7 –17 –9 14 45 83 40 33 89 184,346 140,858 138,737 169,709 134,070 224,864 144,040 146,573 155,378 148,127 153,436 128,057 203,415 158,485 184,200 –20 153,600 10 156,900 7 161,300 –10 141,900 –4 236,300 –10 149,800 10 26 0 –1 9 6 13 5 –5 11 6 16 5 –5 4 143,616 179,720 146,500 25 134,494 166,384 131,500 24 –18 –21 2 –2 186,680 229,319 197,923 289,780 324,092 315,287 209,300 55 270,000 41 248,700 59 –28 –17 –21 12 18 26 101,401 93,356 100,000 –8 193,044 241,095 227,200 25 7 –1 –6 18 SOURCE: Authors’ calculations of decennial census data. NOTES: Values are in 2000 dollars. Adjustments for inflation were made separately for California and the United States. Median values for counties in 1980 do not include owner-occupied condominiums. In 1990, the value of all owner-occupied dwelling units was 1.0 percent lower than the value of non-condominium owneroccupied units. In California, the difference was only 0.6 percent. 13 Table 2.2 Median Gross Rent in California’s Counties, 1980, 1990, and 2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Median Gross Rent ($) 1980 508 1990 589 2000 602 560 581 600 803 747 465 546 519 496 530 537 513 596 567 490 533 561 532 610 600 378 425 429 605 774 746 458 474 525 403 548 513 443 474 471 435 475 487 484 568 563 409 458 494 460 460 458 509 559 563 443 501 506 443 474 486 435 496 488 717 737 702 560 745 780 520 682 659 496 660 687 501 562 534 509 570 518 458 532 533 465 548 562 488 557 518 473 633 617 507 624 611 479 522 516 % Change 1980– 1990– 1990 2000 16 2 4 34 –7 18 –5 71 16 –5 95 15 –2 12 1 28 –4 3 11 36 –6 7 –1 92 18 –1 12 8 00 10 1 13 1 73 14 –2 3 –5 33 5 31 –3 33 4 12 –5 12 –9 16 0 18 3 14 –7 34 –3 23 –2 9 –1 14 Table 2.2 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Median Gross Rent ($) 1980 1990 2000 564 811 852 624 874 898 791 1,067 1,162 596 818 818 605 846 928 713 996 1,144 709 1,001 1,185 522 764 797 613 835 864 613 809 776 484 708 765 592 742 719 636 847 830 643 923 924 740 535 659 509 640 685 513 616 599 484 533 516 426 508 502 656 712 682 530 647 611 573 741 660 552 720 648 588 811 704 759 1,023 923 672 976 892 465 510 504 596 791 761 % Change 1980– 1990– 1990 2000 44 5 40 3 35 9 37 0 40 10 40 15 41 18 46 4 36 3 32 –4 46 8 25 –3 33 –2 44 0 –28 23 26 7 20 –3 10 –3 19 –1 9 –4 22 –6 29 –11 31 –10 38 –13 35 –10 45 –9 10 –1 33 –4 SOURCE: Authors’ calculations of decennial census data. NOTES: Rents are in 2000 dollars. Adjustments for inflation were made separately for California and the United States. 15 once inflation is accounted for, and only the San Francisco Bay Area had seen significant real price increases. What matters from a policy perspective, however, is how these price changes affect residents in these markets. An indicator that helps us to see these effects is housing affordability. Hence, we next examine trends in affordability in the largest counties in the state. Affordability Affordability makes the state’s housing supply a compelling policy issue. Housing is the most reliable indicator of a region’s cost of living. It is also the largest budget item for most households and represents the bulk of most homeowners’ net worth. There are many potentially useful ways to characterize affordability, but the definition used by most analysts, and one of the measures we consider, is the share of an area’s households that can afford the median-priced home at current interest rates and loan standards.4 This measure has two potential flaws in the context of this report, and they work in opposite directions. First, not everyone is in the market for a single-family home. Some households are not likely to be in the market at a given time and, more important, many households already own homes at lower-than-current prices, regardless of whether they could afford to buy their home at current prices. Second, the affordability index evaluates only the ability of households to purchase the median-priced home. Even in a market characterized by a plentiful supply of homes, we would expect only about half of the households to be able to purchase the median-priced home (assuming similar distributions of household income and home prices). Using decennial census data, we consider two additional measures of affordability: the share of renters who face high rent burdens (defined as paying more than 30 percent of their household income on rent), and the share of homeowners who face high housing cost burdens (defined as ____________ 4This assumes a 20 percent down payment, with mortgage payments not to exceed 30 percent of monthly pre-tax income. Affordability figures are calculated by the California Association of Realtors for this group of counties from 1990 on. 16 paying more than 30 percent of their income on housing costs, including mortgage payments). In addition to being affected by home prices, affordability is driven by personal income and interest rates. Over the last ten years, interest rates have fallen and personal incomes have risen relative to inflation, and these trends have mitigated the effects of rising home prices (and in fact contributed to those rising prices). Figure 2.4 shows the affordability in 1990 and 2000 for 17 of the largest counties in California. Counties in the greater San Francisco Bay Area (including Monterey and Santa Cruz) are grouped together on the left side of the graph, and they show the dramatic difference in affordability between that region and the rest of the state. It is also clear that this gap existed in the previous business cycle, although the more affluent parts of the greater Los Angeles area were in the same range as some Bay Area counties. Only Contra Costa and Sonoma Counties were noticeably less affordable in 2000 than they were in 1990. The rest of the Bay Area and San Diego faced the same affordability situation as they did at the previous peak, and the other 70 1990 60 2000 50 40 30 20 10 0 Figure 2.4—Share of Households That Can Afford the Median-Priced Home in Major Counties in California, 1990 and 2000 17 Percentage SCaoSnSSnataFanrrMAtnnlStaaoaaoannMtnCcmCMilaCeooteasrrasrtcredmiueaoazyoaan SanLSoBaSsceaRrriAnVnavOFneraremrDrgniaetedseiienslunntdggnreooesoeoa counties were much more affordable than previously. This level of affordability is obviously aided by the fact that interest rates in the latter half of the 1990s were substantially lower than they were in the 1980s, but it does seem to conflict with the notion of a statewide housing shortage that arose in the 1990s. Still, California has far more homeowners and slightly more renters under financial stress because of high housing costs than the rest of the nation (see Table 2.3).5 As bad as the situation is in California, it did not appear to appreciably worsen in the 1990s. In 14 counties, including some with the most expensive housing in the state (Ventura, Santa Clara, Orange, Contra Costa, San Mateo, Marin, Alameda, and Santa Cruz), the proportion of homeowners paying more than 30 percent of their income for housing remained unchanged or even declined, although the levels remained high. In those counties, rising incomes at least kept pace with rising housing costs. Substantial increases in homeowner costs relative to income occurred primarily in the Upper Sacramento Valley, San Joaquin Valley, and in some mostly rural mountainous counties. In many of those areas, housing costs are among the lowest in the state. Relatively low wages and incomes would seem to be more of a culprit than high housing costs for homeowners in those areas. For renters in California, the situation actually improved slightly from 1990 to 2000. Statewide, in 48 counties the proportion of renters paying more than 30 percent of their income in rent either did not change or declined. Increases in the proportion occurred primarily in less populated counties outside the state’s metropolitan areas. San Francisco had one of the largest declines and one of the lowest proportions of renters paying more than 30 percent of household income for rent—a consequence of rapidly increasing incomes and, probably, rent control and out-migration of those with lower incomes. Places with high rents relative to incomes tended to be counties with relatively high poverty rates. Among the state’s larger metropolitan areas, Fresno had ____________ 5This section presents data from the decennial censuses for all renters and homeowners. Calculations of income as a percentage of rent or homeowner costs are based on incomes for the year preceding the census, whereas rents and homeowner costs were those experienced at the time of the census. 18 Table 2.3 Rent and Homeowner Costs as a Percentage of Income United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare % of Homeowners Paying More Than 30% of Household Income 1990 2000 20 22 % of Renters Paying More Than 30% of Household Income 1990 2000 41 40 18 21 40 39 30 31 48 45 13 23 45 51 18 25 50 54 27 31 51 47 18 22 39 45 23 26 46 43 16 19 32 38 27 32 50 48 20 24 48 41 25 24 29 36 19 23 44 46 21 27 49 46 20 26 55 53 18 25 29 39 18 23 42 36 20 30 48 48 19 27 43 40 21 26 45 42 20 28 46 44 30 32 48 43 28 29 47 41 24 27 48 43 22 27 52 53 22 29 48 48 22 27 44 46 21 25 39 42 23 31 45 45 21 30 46 43 24 29 45 46 24 29 47 45 21 29 48 44 19 Table 2.3 (continued) Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego % of Homeowners Paying More Than 30% of Household Income 1990 2000 % of Renters Paying More Than 30% of Household Income 1990 2000 31 31 46 42 31 30 46 42 32 32 50 46 26 29 48 40 26 30 45 37 32 31 43 42 31 29 43 40 31 30 43 42 31 32 49 44 29 33 46 42 32 37 35 40 31 32 53 51 29 32 53 51 34 34 53 48 37 25 29 36 21 25 39 43 23 32 48 41 19 19 40 37 22 28 39 37 30 39 32 41 25 27 44 48 33 32 50 47 30 31 49 46 31 35 49 46 33 32 47 44 35 32 48 42 21 29 47 47 32 32 49 45 SOURCE: Authors’ calculations of decennial census data. NOTE: Incomes are from the year preceding the census, whereas rents and homeowner costs were those experienced at the time of the census. the highest proportion of renters paying more than 30 percent of their income in rents, yet Fresno had among the lowest rents in the state and experienced a decline (in real terms) in median rents from 1990 to 2000. 20 Poverty and low incomes seem to be a more important factor in the location of financial stress with respect to housing than does the cost of the housing itself. New Construction One of the most compelling pieces of evidence regarding a possible housing shortage is the 47 percent decline in construction between the 1980s and the 1990s. From 1980 to 1989, almost 2.1 million housing units were built in California, for an average of 207,000 per year; from 1990 to 1999, only 1.1 million were built, for an average of 111,000 per year. The drop in construction was most severe in the case of multifamily units, where construction in the decade 1990–1999 fell by 70 percent from the pace in the 1980s. Multifamily units made up 53 percent of the total units in the previous peak year of 1986 but only 29 percent in 2000. In contrast, single-family construction fell by only 27 percent over the same period. Figure 2.5 shows annual housing construction from 1967 to 2002 for both single- and multifamily units. The break between the pattern of new construction in the last three 350,000 300,000 Multifamily Single-family 250,000 Number of new units 200,000 150,000 100,000 50,000 0 11996697 1971 1973 111999777975 1981 1983 111999888975 1991 1993 111999999975 2001 Figure 2.5—New Multifamily and Single-Family Housing Construction in California, 1967–2002 21 housing cycles and the current one is often “Exhibit A” in the argument for a housing supply crisis in the state. This change in the cyclical pattern of new construction also took place nationally. Although the acceleration in new construction after 1982 was higher in California than in the rest of the United States, the situation reversed after the second half of 1989, when the reduction in the number of new units was much more dramatic in California. From its trough in 1991, annual production in the rest of the United States increased by 88 percent by 2002. In California, the increase in annual production from 1991 to 2002 was only 51 percent, but the increase from its lowest production year (1993) to 2002 was 88 percent. Figure 2.6 shows the ratio of annual production to the peak-year production for the United States and California, with 1986 production indexed to 1.0. Housing production in California trailed that of the nation at least in part because the state had something of a double-dip construction recession. Not only did construction in the state take two years longer to 1.20 Rest of United States 1.00 California 0.80 0.60 0.40 0.20 0.00 SOURCE: Authors’ calculations of U.S. Census Bureau data. Figure 2.6—New Housing Consruction in the United States and California, Indexed to 1.0 in 1986 (Peak Year of Production), 1973–2002 22 Index 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 rebound than it did at the national level, but it stumbled again a year later (in 1995). Since 1995, new housing construction has increased markedly but remains far below the 1986 peak. How Is Population Growth Accommodated? New Housing Units Housing growth has not kept pace with population growth in California’s counties. Both for the state as a whole and in 49 of California’s 58 counties—including all the large urban counties in the state—the ratio of population change to the change in housing units was much greater in the 1980s and 1990s than in the 1970s (see Table 2.4).6 Statewide, this ratio doubled between the 1970s and 1980s. The ratio increased again from the 1980s to the 1990s, but much less dramatically. Counties with extremely high gains in household population relative to the number of net new housing units in the 1990s include places with very high housing prices, such as San Mateo, Santa Barbara, Santa Clara, and Alameda Counties. For those counties, the increase in the ratio could be an indication that high housing costs are leading people to live with more roommates or family members than they otherwise would. For other counties with large population growth relative to housing growth, such as Los Angeles and Colusa Counties, the nature of the population growth at least partly explains the high ratio; with immigrants and children forming a large share of population growth in those counties, housing demand is lower than if the population growth was primarily among U.S.-born adults.7 Perhaps most noteworthy, though, is that the dramatic increase in the ratio between population growth and new occupied households occurred in the 1980s as well as in the 1990s. In almost half of California’s counties (28 of 58), the ratio either remained unchanged or declined in the 1990s as compared to the 1980s. Thus, population growth outpacing housing ____________ 6The only exceptions were the relatively lightly populated mountainous counties of Lassen, Siskiyou, Modoc, Mono, Inyo, Plumas, Trinity, Sierra, and Alpine. 7Chapter 4 considers how the nature of California’s population growth in the 1990s might have affected the demand for new housing. 23 Table 2.4 Ratio of Household Population Change to Total Housing Units Change in California’s Counties: 1970s, 1980s, and 1990s United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin 1970–1980 1.19 1.13 1.63 1.59 0.85 1.54 1.90 1.67 1.14 1.96 1.63 2.04 1.46 1.65 1.79 0.90 1.85 1.88 1.61 1.60 1.09 1.98 1.74 1.29 1.47 1.71 1.56 1.81 2.20 1.76 1.46 1.91 2.15 0.60 1.32 0.80 1980–1990 1.53 1.28 3.10 2.06 1.71 2.35 1.86 2.75 0.94 2.08 0.95 0.79 1.42 0.84 2.54 3.42 3.61 2.28 3.15 2.66 4.00 2.40 2.30 2.67 2.88 3.58 3.15 3.52 3.86 5.09 4.19 3.45 3.98 2.77 2.24 0.88 1990–2000 2.32 2.19 3.92 1.97 1.42 2.06 1.35 1.79 –1.53 2.01 0.71 6.72 0.40 –0.09 2.12 5.09 2.48 1.89 3.37 2.01 1.87 3.11 2.55 3.13 3.13 3.60 3.09 3.47 2.99 3.26 3.51 3.99 3.84 4.82 3.75 2.72 24 Table 2.4 (continued) Napa San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1970–1980 1.52 –5.87 0.74 1.65 2.18 2.07 1980–1990 2.83 3.78 3.07 2.94 2.83 2.36 1990–2000 3.21 3.19 6.72 4.85 3.47 3.03 1.83 3.52 4.41 2.34 3.28 3.92 1.72 2.46 2.30 1.37 2.88 6.11 2.16 3.63 3.49 1.49 0.04 0.49 1.76 1.99 2.33 1.28 1.77 2.23 0.99 0.47 0.00 1.70 1.71 1.89 0.84 0.63 2.44 1.47 1.96 1.73 1.64 2.61 3.62 1.78 2.96 4.80 1.43 4.40 6.07 1.95 3.04 4.56 2.17 3.02 3.64 1.97 3.53 3.33 1.90 2.79 3.48 SOURCE: Authors’ calculations of decennial census data. growth is as much a 1980s phenomenon as a 1990s phenomenon in much of California. Increases in Household Size If much of California’s population growth was not accommodated through new housing, how was it accommodated? Some population growth can be accommodated through previously unoccupied housing, and some can be accommodated by increases in household sizes. Table 2.5 shows changes in the number of persons per occupied household from 1970 through 2000. Since 1980, household sizes have been 25 Table 2.5 Number of Persons per Occupied Household, 1970–2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo 1970 3.11 3.13 2.95 3.05 3.00 2.50 2.93 2.94 2.93 2.70 2.81 2.67 2.90 2.89 2.82 2.95 3.02 3.04 3.16 3.04 3.27 2.89 3.04 3.08 3.07 3.20 3.19 3.35 3.20 3.35 3.03 3.09 3.25 2.84 3.19 2.94 2.92 2.34 2.97 1980 2.75 2.76 2.68 2.64 2.55 2.36 2.70 2.61 2.62 2.55 2.60 2.40 2.59 2.63 2.46 2.69 2.75 2.66 2.75 2.65 2.76 2.62 2.71 2.56 2.59 2.83 2.82 3.04 2.98 2.97 2.71 2.78 2.99 2.53 2.69 2.43 2.55 2.19 2.58 1990 2.63 2.61 2.79 2.63 2.49 2.38 2.66 2.57 2.49 2.51 2.41 2.45 2.48 2.49 2.48 2.84 2.77 2.58 2.75 2.60 2.85 2.66 2.66 2.58 2.63 2.96 2.92 3.08 3.05 3.17 2.94 2.91 3.12 2.59 2.64 2.33 2.54 2.29 2.64 2000 2.59 2.56 2.87 2.58 2.39 2.39 2.59 2.53 2.39 2.47 2.29 2.32 2.35 2.29 2.48 3.01 2.84 2.52 2.87 2.62 2.87 2.63 2.63 2.64 2.71 3.09 3.03 3.18 3.18 3.25 3.00 3.03 3.28 2.71 2.72 2.34 2.62 2.30 2.74 26 Table 2.5 (continued) Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1970 3.23 3.14 2.92 3.11 3.29 2.82 2.99 2.70 2.72 2.74 2.76 2.78 2.60 2.83 2.78 2.97 3.11 2.83 3.21 3.43 3.52 2.94 1980 2.76 2.82 2.56 2.85 3.12 2.51 2.62 2.54 2.84 2.49 2.54 2.45 2.48 2.43 2.55 2.69 2.82 2.69 2.78 3.00 3.24 2.62 1990 2.81 2.88 2.55 2.96 3.15 2.53 2.73 2.66 2.47 2.41 2.50 2.35 2.42 2.48 2.46 2.85 2.97 2.91 2.87 3.02 3.26 2.69 2000 2.92 2.90 2.60 3.14 3.32 2.49 2.80 2.71 2.50 2.39 2.44 2.31 2.37 2.43 2.36 2.98 3.15 2.98 3.00 3.04 3.33 2.73 SOURCE: Authors’ calculations of decennial census data. increasing in California while they have been declining in the rest of the United States. Increases in household sizes in California appear to be driven at least partly, perhaps solely, by changes in the demographic composition of the state’s population. In particular, immigrants, especially Latino immigrants, tend to have larger families and more often live with extended family members than do U.S.-born residents of the state. Large flows of immigrants accounted for much of California’s population growth in the 1980s and 1990s. In 2000, household sizes were highest in counties with large Latino populations, including Imperial, San Benito, and the San Joaquin Valley counties. Counties with declining and relatively low household sizes were primarily rural, mostly white non-Hispanic, counties in the northern and mountainous 27 areas of the state. Areas of the state with the most expensive housing are not necessarily those with the greatest household sizes. For example, San Francisco, Marin, Alameda, and San Mateo Counties all had average household sizes lower than the state average in 2000. Higher household sizes and crowding seem to be more driven by demographic effects than prices (see Moller et al., 2002). Declines in Vacancy Rates Population growth can also be accommodated as formerly unoccupied housing units become occupied. Vacancy rates declined substantially in California and in almost all of the state’s counties between 1990 and 2000 (Table 2.6).8 Low vacancy rates are an indication of housing demand outstripping supply. The percentage of homes for sale is only slightly lower in California than in the entire United States, but the percentage of rental units available for rent is much lower. Homeowner vacancy rates were exceptionally low (0.8 percent or less) in most Bay Area counties (including San Mateo, Santa Clara, Alameda, Marin, Contra Costa, San Francisco, and Sonoma) and in the Central Coast (Santa Barbara and Santa Cruz Counties). The tight supply of houses in these areas at the time of the 2000 census was reflected in rising home prices. The number and percentage of homes for sale was relatively high in the northern mountain counties of the state but also in the Inland Empire and Kern County. Rental vacancy rates show similar geographic patterns, with exceptionally low rates—less than 3 percent—in most of the Bay Area (particularly San Mateo, Santa Clara, Marin, Sonoma, Alameda, San Francisco, Contra Costa, and Napa Counties) and all of the Central Coast (including Santa Cruz, San Benito, Santa Barbara, and Monterey Counties). In Southern California, Ventura County had a vacancy rate of only 2.6 percent. However, in 20 of California’s counties, rental vacancy rates actually rose between 1990 and 2000. In much of the San ____________ 8The vacancy rate is the percentage of housing units that are either for rent or for sale. The homeowner vacancy rate is the number of housing units for sale divided by the number of owner-occupied housing units plus the number for sale. The rental vacancy rate is the number of housing units for rent divided by the total number of housing units that are rented or available for rent. 28 Table 2.6 Vacancy Rates in California’s Counties, 1990 and 2000 United States United States excluding California California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa Homeowner Vacancy Rate 1990 2000 2.1 1.7 2.1 1.8 2.0 1.4 3.1 3.0 1.3 1.7 3.9 4.1 2.2 4.1 1.1 1.4 3.6 5.1 1.8 1.3 2.2 2.9 1.9 0.7 2.1 3.0 2.4 3.8 1.4 2.1 1.0 2.3 0.9 1.5 1.5 2.2 1.0 1.5 1.4 2.3 1.0 1.8 2.0 1.2 1.6 1.2 1.5 1.4 1.0 0.9 1.5 1.6 2.1 2.6 1.3 1.8 1.4 1.7 1.0 1.4 1.8 1.2 2.2 1.3 1.3 1.8 1.4 0.7 1.7 0.8 1.7 0.7 2.0 1.3 Rental Vacancy Rate 1990 2000 8.5 6.8 8.9 7.3 5.9 3.7 8.7 10.6 4.8 4.7 6.7 10.3 9.6 13.4 5.1 3.3 7.8 9.3 5.2 3.1 7.6 9.9 12.0 11.3 7.8 9.2 10.2 8.5 4.0 5.2 4.7 3.0 3.2 8.2 4.0 5.9 4.4 4.8 5.4 8.6 4.7 6.7 5.3 5.8 7.5 6.4 6.8 4.8 3.6 3.4 5.5 5.5 6.4 8.2 5.9 5.6 3.3 4.5 3.3 4.2 4.5 3.8 4.8 3.2 4.1 5.8 5.5 2.5 6.4 2.7 4.0 2.2 4.8 2.8 29 Table 2.6 (continued) San Francisco San Mateo Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego Homeowner Vacancy Rate 1990 2000 1.7 0.8 1.7 0.5 1.4 0.5 2.1 0.9 1.7 0.8 2.2 1.4 3.1 1.0 2.8 1.1 2.0 0.8 2.4 0.8 1.5 0.9 2.1 1.9 3.6 2.1 3.0 1.8 2.1 2.4 5.1 2.1 2.3 2.2 4.9 2.5 3.2 3.1 1.9 1.6 1.8 0.9 2.0 0.9 1.6 1.4 2.0 1.0 Rental Vacancy Rate 1990 2000 5.7 2.5 4.3 1.8 4.5 1.8 6.0 3.7 5.1 2.4 3.8 2.9 4.3 2.7 5.8 3.2 5.0 2.8 4.3 2.5 55.3 8.4 5.3 4.4 6.2 6.2 4.9 6.9 14.5 7.7 37.6 20.9 5.9 6.9 9.9 7.2 8.8 7.3 5.9 3.3 6.6 3.0 4.9 2.6 5.0 4.9 6.2 3.1 SOURCE: Authors’ calculations of decennial census data. Joaquin Valley, the Upper Sacramento Valley, and the northern mountain counties, rental vacancy rates increased. In 2000, rental vacancy rates were relatively high in the Inland Empire, but not as high as in 1990. In the Bay Area, rental vacancy rates have risen substantially since 2000. Jobs and Housing Another way to consider whether counties are building a sufficient number of new housing units is to examine changes in jobs relative to 30 changes in housing units. Areas that add a significant number of jobs but not a significant amount of housing could be said to be creating a jobs-housing imbalance. Table 2.7 provides ratios of jobs per housing unit in 1990 and 2000. Only three counties had both a jobs-housing ratio greater than the state’s in 1990 and experienced substantial increases in the ratio in the 1990s (San Mateo, Santa Clara, and Yolo). Another three counties had a jobs-housing ratio greater than the state’s in 1990 and experienced moderate increases in the 1990s (Santa Barbara, Monterey, and Orange). Together, these six counties could be said to have accommodated substantial numbers of new jobs, but not substantial numbers of new housing units. Only eight of California’s counties experienced a decline in the ratio, and all but one are relatively small counties. The one exception is Los Angeles County. Notably, Los Angeles County experienced a decline in its jobs-housing ratio, so that by 2000 its ratio was similar to that of the state as a whole. Los Angeles County’s role as a job center in Southern California waned during the 1990s, with that county being the locus of California’s severe recession of the early 1990s. Summary Substantial variation in housing conditions exists across the state, although some common patterns emerge. Most notably, we find some indicators of housing demand outstripping supply: New housing production did not keep pace with population growth; already low vacancy rates in many counties declined even further in the 1990s, especially for rental units; real prices increased in most counties in the 1990s; and households in the state became more crowded in most counties. However, for many indicators, the trends of the 1990s were much less notable than those of the 1980s. For example, price increases were much more substantial in the 1980s than in the 1990s. Within the state, the Bay Area often stood out, with other housing markets in other regions of the state showing few if any signs of a lack of overall supply. The general picture is one of a tight housing market, but with the 1980s exhibiting more remarkable changes than the 1990s. 31 Table 2.7 Nonfarm Jobs per Housing Unit in California’s Counties, 1990 and 2000 California Far North Del Norte Humboldt Lake Lassen Mendocino Modoc Nevada Plumas Sierra Siskiyou Trinity Upper Sacramento Valley Butte Colusa Glenn Shasta Sutter Tehama Yuba Sacramento Metro El Dorado Placer Sacramento Yolo San Joaquin Valley Fresno Kern Kings Madera Merced San Joaquin Stanislaus Tulare Bay Area Alameda Contra Costa Marin Napa San Francisco San Mateo 1990 1.15 0.77 0.88 0.38 0.78 0.83 0.54 0.56 0.54 0.42 0.72 0.42 0.80 1.07 0.84 0.84 0.64 0.59 0.72 0.49 0.78 1.11 1.18 0.95 1.01 0.97 0.81 0.93 1.01 1.00 1.07 0.56 0.89 0.94 0.95 1.70 1.17 2000 1.22 0.75 0.90 0.42 0.81 0.89 0.59 0.64 0.54 0.46 0.66 0.41 0.86 1.12 0.78 0.88 0.73 0.71 0.71 0.62 1.01 1.18 1.42 1.00 1.04 1.03 0.97 0.93 1.08 1.07 1.12 0.63 0.95 1.08 1.18 1.73 1.44 Change 0.07 –0.01 0.03 0.04 0.03 0.06 0.06 0.08 0.00 0.03 –0.06 –0.02 0.06 0.05 –0.06 0.04 0.09 0.11 –0.01 0.13 0.23 0.06 0.23 0.05 0.03 0.06 0.15 0.00 0.07 0.07 0.05 0.06 0.06 0.14 0.23 0.03 0.27 32 Table 2.7 (continued) Santa Clara Solano Sonoma Central Coast Monterey San Benito San Luis Obispo Santa Barbara Santa Cruz Sierras Alpine Amador Calaveras Inyo Mariposa Mono Tuolumne Inland Empire Riverside San Bernardino South Coast Los Angeles Orange Ventura San Diego Imperial San Diego 1990 1.52 0.80 0.90 1.15 0.91 0.86 1.16 1.03 0.52 0.65 0.36 0.83 0.62 0.48 0.56 0.66 0.76 1.31 1.35 1.08 1.23 1.03 2000 1.78 0.85 1.06 1.25 0.92 0.97 1.26 1.07 0.65 0.74 0.35 0.85 0.55 0.54 0.56 0.80 0.91 1.25 1.44 1.17 1.13 1.16 Change 0.26 0.05 0.16 0.11 0.01 0.11 0.11 0.04 0.12 0.10 0.00 0.02 –0.07 0.06 0.00 0.14 0.14 –0.06 0.10 0.09 –0.09 0.13 SOURCE: Authors’ calculations of decennial census data and California Employment Development Department data. 33 3. Can Macroeconomic Factors Explain California’s Housing Markets? In this chapter, we seek to determine whether the apparent lack of new housing construction in the late 1990s can be mostly explained by changes in macroeconomic factors, such as interest rates, unemployment rates, income, and prices. These economic factors are determined by market changes in the U.S. economy and in the rest of the world, and changes in the U.S. economic policies. To analyze these influences, we use a simple economic model of the housing market. Housing production in the United States and California is characterized by its cyclical behavior (periods of rapid increases in construction followed by periods of contraction), with large annual fluctuations in new housing construction. These features are captured in our model, which considers the influence of macroeconomic factors on construction in the short and longer term. Our analysis of macroeconomic influences on housing reveals that much of the puzzling trend observed in new construction in California in the 1990s (i.e., slow growth in housing production in a context of rapid economic growth) is a result of the unusual economic environment driven by major increases in productivity, technological breakthroughs, macroeconomic policies, and financial events of that period along with the demographic determinants of demand discussed in the next chapter. The portion of the changes in new construction that are not explained by these factors might be related to specific characteristics of local markets, such as local land-use policies and zoning decisions regarding new housing. We discuss those factors in the final chapter. In this chapter, we first describe how new construction is affected by macroeconomic events (including fiscal and monetary policies) and examine some major 35 differences in the economic environment between the period before the recession of 1990–1991 and the rest of the 1990s. We then turn to our macroeconomic model. Business Cycles, Macroeconomic Events, and Construction Historically, increases in unemployment are closely related to downturns in construction activity. This point is illustrated by Figure 3.1, which compares new residential construction and unemployment rates in California. The figure shows that in 1974–1975, 1981–1982, and 1990–1993, years of economic recession, construction activity decreased sharply. Conversely, upturns in construction have coincided with economic booms. The figure also shows that from 1983 through 1986, construction grew rapidly, but the recovery in the California housing markets from the U.S. recession that started in 1990 was very slow compared to previous recoveries (see Figure 3.1). 12.0 350 10.0 300 250 8.0 200 6.0 150 4.0 100 2.0 Unemployment rate New construction 50 0.0 0 111211111111111111909999999999999999608899999888777776713197531975975319 SOURCE: Center for Continuing Study of the California Economy (1993). Figure 3.1—California Construction and Employment Cycles Unemployment rate (%) Number of permits (1000s) 36 The robust performance of construction activity during the 1970s and early 1980s coincided with important changes in the demography and the economic environment of the state, which significantly increased the demand for housing units: 1. High population growth with a large increase in the number of people in the prime household formation ages (discussed in the next chapter). 2. Significant increases in per capita income, expectations of future high inflation, and changes in tax policy that increased the attractiveness of housing assets. Rising inflation expectations caused the price of long-lived, tax-favored assets such as housing to jump significantly. Furthermore, in 1986, a tax policy change phased out interest deductions on consumer debt except for that secured by home mortgages. This made ownership of housing even more attractive by providing an avenue for acquiring taxfavored debt. 3. Low real interest rates (which were even negative at times before 1981), making the cost of loans very low. A decrease in the real interest rate decreases the user costs of housing and increases the attractiveness of new investment in housing. Beginning in 1986, construction activity started to decrease when economic growth temporarily slowed down. The economy gained strength in 1987 and 1988 but turned sluggish again in 1989 and 1990, with a consequent decline in construction (see Figure 3.1). Macroeconomic forces that led to the slowdown in construction included a new oil shock, restrictive monetary policy, large federal budget deficits, and concerns about inflation. These factors put upward pressure on interest rates and decreased residential investment. This was a period of tighter credit with lower availability of loans—a problem that was compounded by higher standards of capital requirements imposed by bank regulators. Lending to businesses was considered risky and the collateral value on residential and commercial real estate loans fell with declining real estate values (resulting from less demand for housing). The combination of high interest rates and loss of deposits to money market funds that started in the early 1980s created liquidity problems 37 for many depository institutions that found it increasingly difficult to meet withdrawal demands. For thrift institutions—the main source of mortgage lending—the problem arose principally from borrowing on a short-term basis to make longer-term loans. As rates rose, the thrifts had to pay higher rates to retain deposits, but they could earn the higher yields only as their long-term assets gradually matured and the funds were invested in higher-yielding assets. For commercial banks, the main problem was losses from defaults on international loans, energy development loans, and agricultural loans. Between 1980 and 1984, 189 banks failed—an average of 38 per year—and the number of institutions insured by the Federal Savings and Loan Insurance Corporation fell by 20 percent. Furthermore, in mid-1984 one of the nation’s largest banks had to be rescued by a multibillion-dollar package arranged by the federal regulatory agencies. During the same year, one of the nation’s largest savings and loan associations ran into trouble. Starting in early 1988, the Federal Reserve began to raise short-term interest rates. This together with weak conditions in the credit market strongly affected construction activity in California, which declined sharply by the second half of 1989. Long-term investments became less attractive, the economy slowed down considerably, and domestic outmigration sharply reduced the demand for housing. Housing prices started to fall. In 1989, policymakers responded to sluggish growth (and lower inflation) with a more expansionary monetary policy. The Federal Reserve began to cut short-term interest rates in mid-1989 and then cut rates more aggressively in late 1990, for a cumulative decline of about 4 percentage points by December 1990. The U.S. economy started growing modestly and unevenly during 1992, when core inflation and interest rates were at their lowest levels in a generation. Since 1992, the Federal Reserve has been successful in pursuing lower interest rates and keeping inflation under control. By the end of 1993, inflation was low and the public’s expectations of higher prices declined significantly. Congress passed a significant long-term deficit reduction package, raising taxes and lowering spending for the next five years. A strong deficit reduction allowed the Federal Reserve to continue keeping interest rates low. 38 Macroeconomic policies in the mid-1990s helped both the U.S. and California economies by curbing inflationary expectations and influencing financial markets. For example, starting in mid-1996, monetary authorities signaled that they would control potential inflationary pressures from the tight labor markets (as low unemployment obligates firms to offer higher wages). In 1997, Federal Reserve policymakers publicly agreed that there was still room for economic growth driven by increases in productivity. The federal budget deficit decreased sharply and capital investment grew significantly as did labor productivity. By the mid-1990s the performance of the U.S. economy was excellent. The economy was growing rapidly, the unemployment rate declined to its lowest level in 25 years, and inflation was not a problem. This was a period of low interest rates and easy credit. The stock markets were booming, fostered by optimism stemming from a series of technological breakthroughs in electronics, biotechnology, and communications. However, construction activity in both the United States and California did not pick up as fast as in previous recoveries (see Figures 2.6 and 3.1). Why, despite record-low mortgage rates, were the housing markets, particularly in the single-family sector, not keeping pace with the accelerated economic growth? The trend observed after 1990 is largely the result of the unusual events that took place in three areas: (1) financial events and macroeconomic policies related to balancing the budget and interest rate adjustments targeted by the Federal Reserve, (2) economic changes that deeply affected the California economy, and (3) demographic changes in the population. The weak recovery in new construction of the 1990s is at least partly explained by the public’s expectations of significantly lower inflation resulting from increases in productivity, the way monetary policy was conducted, and the climate of optimism in the financial markets that drove the attention of investors to alternative markets. These expectations of low inflation and strong stock market returns made investments in real estate much less attractive relative to other investments. Not only did these macroeconomic factors curb expectations of housing appreciation, the demographic composition of 39 Expected inflation rate (%) the state (as discussed in the next chapter) also reduced expectations of future returns on housing construction. Builders and developers integrated a demographically driven slowdown in the demand for housing in their projections of future real estate values. Figure 3.2 illustrates the point that expectations of future inflation were decreasing.1 Figure 3.3. shows the difference between long-term and short-term interest rates. Decreases in this difference are associated with lower levels of construction, as investors prefer shorter-term investments when the yields of longer-term investments decrease.2 This difference shrank significantly after 1994 and remained unusually low until 2000, given the economic performance of the economy of these years. The severity and length of the California recession together with the demographic composition of the state also help explain why construction in California took longer to recover. The U.S. recession that started by 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 SOURCE: Livingston Survey (described in Appendix A). Figure 3.2—Expectations of Inflation, 1972–2001 ____________ 1The expectation variable is calculated as the median one-year-ahead expected inflation rates as reported by the Livingston Survey (see Appendix A) 2See Appendix A. 40 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Interest rate (%) 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 4.00 3.00 2.00 1.00 0.00 –1.00 –2.00 –3.00 SOURCE: Federal Reserve. Figure 3.3—Term Structure (Difference Between the Long-Term and Short-Term Interest Rates), 1971–2001 mid-1990 and finished in March 1991 had a deep and long-term effect on the state. For California, the 1990 recession was the longest recession since the Great Depression of the 1930s, whereas the U.S. contraction lasted less than a year. By the end of 1993, the U.S. economic recovery was well established, but the California economy was still weak. Consequently, California had a much greater downturn in construction activity than other states. Home prices dropped dramatically in highpriced areas including Los Angeles, San Diego, and Orange Counties. The California recovery was slower than that in the rest of the country because it was hampered by several factors, including (1) large household and business debt burdens, (2) high vacancy rates in commercial real estate, (3) tight credit practices by many lenders, (4) stagnant growth in much of the rest of the world, reducing the demand for California exports, (5) a dramatic outflow of domestic migrants, and (6) a decrease in federal purchases, especially of military goods and services. 41 Because of California’s high share of U.S. aerospace employment, the state was also particularly affected by the federal defense-spending cuts that were initiated in 1986 and intensified by 1989. Between the second quarter of 1990 and the second quarter of 1993, California lost 528,000 jobs—142,000 in the aerospace industry alone. Most of the job losses took place in Southern California. This sharp slowdown in economic activity also negatively affected residential construction, particularly in Southern California. A Simple Housing Market Model To systematically evaluate the effect of macroeconomic factors on housing markets in California, we develop a model of the state’s housing markets.3 Two specific features in our analysis make it different from previous research. First, we address the effects of macroeconomic factors on both sides of the market for new housing (demand and supply). Most of the studies dealing with the housing sector tend to concentrate on the demand side (Hanushek and Quigley, 1982; Mayer and Somerville, 1996). Second, unlike cross-sectional analyses of data for specific regions or cities, we use a time-series analysis because we seek to understand why new housing construction in California was more sluggish in the 1990s than in other periods of economic growth. This approach allows us to evaluate the influence of macroeconomic factors in explaining the annual variations in new housing construction in California. To control other influences on the housing market, such as quality of life characteristics and the influence of local policies, we divided California into ten regions sharing similar quality-of-life and local policy traits. The determination of each region is explained below. We focus on changes in new construction of total housing units, as measured by the number of permits issued for housing units. Some authors separate single-family construction from multifamily data in an effort to account for potential differences between the two. However, others do not separate these sectors because the substitution between single-family and multifamily units is very ambiguous (Boldin, 1993). We followed the latter approach. Finally, we developed an alternative econometric model using ____________ 3Our housing market model and the methodology are presented in Appendix B. 42 a panel analysis. The results of that model are consistent with the model presented here and are discussed in Appendix C. Factors Affecting Housing Demand We included the following macroeconomic factors affecting the aggregate demand for housing: changes in income, employment, and interest rates (cost of mortgages), actual and expected appreciation of real estate (housing price increases), and the price of alternative investments (stock prices). We also included controls for the growth and demographic composition of the population, although changes in these variables tend to occur more slowly. Measures of income and unemployment levels are usually incorporated in typical housing demand equations. As income and employment increase, the demand for new housing is expected to increase. Higher current housing prices decrease the current demand for housing, whereas expectations of higher future housing prices increase demand. Our model describes expected appreciation or expected housing valuation as a function of expected inflation (see Appendix A). Furthermore, because housing is both a commodity and an investment, real estate can be seen as a substitute asset for stocks, particularly when inflation is high. In periods of high economic growth, housing and stock prices tend to move together; however, in inflationary environments, houses have done better than stocks. Historically, stocks have better returns when inflation is lower (Wasserman, 1998). Thus, returns on stocks are also a determinant of housing demand because they are an alternative investment. Typically, housing demand or supply equations use either mortgage rates or the prime rate as a proxy for the cost of credit. The higher the interest rate, the lower the demand for housing. We looked at the difference between the long-term and short-term interest rate because it measures the tightness of credit better than the level of either interest rate separately. It is also a better measure because a large spread between the long- and short-term interest rates indicates expectations of rising inflation. The relationship between this spread and the demand of housing is expected to be positive. (See Appendix A for a more complete explanation.) 43 Changes in population growth, the age structure of the population, and immigration also affect the demand for housing. As the population grows, more housing is needed, but a high proportion of children in the population leads to smaller increases in housing demand. These changes are analyzed directly and in more detail in the householder rate models of Chapter 4. Demographic changes also affect expectations of future housing prices—a key determinant of new construction. In addition to macroeconomic and demographic forces, other factors influence the demand for housing in specific areas, such as the availability of amenities and the quality of life in a region. For example, housing is more expensive in areas with good weather, art centers, parks, cinemas, and low crime rates. However, measuring the effect of quality of life in the demand for housing is very difficult, and economists have studied these effects from different angles. With the exception of Gabriel, Mattey, and Wascher (1996 and 1999), who constructed a time-series of quality-of-life rankings, virtually all the research on quality-of-life estimates has focused on the relative ranking of localities at a single point in time. Most economists have entered the quality-of-life debate arguing that price differences across locations in wages or land rents should compensate for the differences in the quality characteristics of locations. Rosen (1979) and Roback (1982) inferred quality levels from differences in the price of housing across locations. Other authors following similar approaches include Hoehn et al. (1987), Blomquist et al. (1988), Gyourko and Tracy (1992), and Stover and Leven (1992). We control for differences in quality of life by dividing the state into regions with similar geographic and socioeconomic characteristics. The determination of these regions is explained later in this chapter. Supply Determinants The supply side of the housing market is also affected by changes in expected housing appreciation, changes in economic activity (unemployment), construction costs, interest rates, the availability of land, and local policies. Housing appreciation leads to higher levels of new construction. Because the cost of capital is the main cost to developers and builders, interest rates and new construction are negatively related. The 44 higher construction costs are, the lower the level of new construction. High economic growth, lower stock prices relative to real estate prices, and expectation of higher returns on long-lived assets have a positive influence on the decision to invest in real estate. The amount of available land for new residential developments also determines the amount of new construction in an area. Because we model changes in housing construction from one year to the next, we assume zero changes in available land (see the model description in Appendix A). Local policies also influence the supply of housing. These policies include slow-growth ordinances and the preference of local governments for commercial rather than residential development because of public finance pressures. Using two surveys of 490 Californian cities and counties, Levine (1999) examined the effects of local growth-control enactment between 1979 and 1988 on net housing construction. He found that local growth-management measures displaced new construction, particularly rental housing. He also found that measures that limited available land or downsized existing zoning had stronger effects than other measures. Ten Regional Housing Markets in California To control differences attributable to quality of life and local policies, we grouped counties into ten regional markets and tested our model for each of these regions. These regions have common characteristics related to quality of life as well as the type of regulations (measures of growth control) enacted in their localities. Our regions are basically the same as those used by Glickfeld and Levine in their 1990 study on land use regulation. These regions and their constituent counties are 1. North Coast: Del Norte, Humboldt, Mendocino 2. North Central: Butte, Colusa, Glenn, Lake, Shasta, Siskiyou, Sutter, Tehama, Trinity, Yuba 3. North Eastern: Lassen, Modoc, Nevada, Plumas, Sierra 4. San Francisco Bay Area: Alameda, Contra Costa, Marin, Napa, San Mateo, Santa Clara, Santa Cruz, Solano, Sonoma, San Francisco 45 5. Stockton-Sacramento: El Dorado, Placer, Sacramento, San Joaquin, Stanislaus, Yolo 6. Central Coast: Monterey, San Luis Obispo, Santa Barbara, El Dorado 7. Central Inland: Fresno, Kern, Kings, Merced, San Benito, Tulare 8. Central Eastern: Alpine, Amador, Calaveras, Inyo, Madera, Mariposa, Mono, Tuolumne 9. Los Angeles–San Bernardino: Ventura, Los Angeles, Orange, San Bernardino, Riverside 10. San Diego Following the Glickfeld and Levine regions allows us to qualitatively consider the role of land-use regulations as elaborated in their work. Results of the Model of California Housing Markets Our simple housing market model indicates that macroeconomic policies and global events explain a large part of the annual changes in new construction in California and in specific regions. Most of the explanatory factors showed the expected signs, but not all factors turned out to be significant in each region.4 Table 3.1 summarizes the results of the model, showing the relationship (positive or negative) between demographic and macroeconomic factors and changes in new home construction for each of the ten regions and the state.5 Results are shown for the lagged effects model.6 Because not all determinant factors affect demand and supply of housing in all California regions at the same speed or intensity, lags are necessary to understand those effects. For ____________ 4Our results are solid, considering that (1) we are dealing with first differences rather than levels, (2) the short historical period for our time-series analysis, and (3) the limitations imposed by data quality (such as county unemployment rates and costs). 5Appendix Table A.2 shows the coefficients of each determinant and the level of significance of each factor included in that estimating equation. 6The results obtained when we used lags were significantly better than the ones obtained when we related current changes in the determinants of housing to current changes in new housing construction (shown in Appendix Table A.1). 46 Table 3.1 Effect of Explanatory Factors on Changes in New Housing Construction: Results of the Macroeconomic Model, by Region 47 Explanatory Factor Los Angeles– North North North Stockton- Central Central Central San San Coast Central Eastern Bay Area Sacramento Coast Eastern Inland Bernardino Diego California Change in population ++ ++ ++ Change in income ++ (a) + Lagged new construction – – – – – – Change in unemployment – –– – Change in difference of long-term/short-term interest rate + + ++ Change in appreciation (expected inflation) + + ++ Change in real stock prices – –– –– Change in real cost – – –– R squared (variation explained) 0.67 0.78 0.78 + + –– –– ++ –– – – 0.76 ++ ++ ++ ++ ++ – + ++ –– –– –– –– –– –– – –– ++ ++ – ++ + ++ – –– –– –– –– –– – –– –– 0.85 0.77 0.90 0.72 +– + + + ++ ––– ––– ++ ++ ++ ––– – – –– –+– 0.77 0.76 0.83 SOURCE: Appendix Table A.1. NOTES: ++ means that the relationship is positive and statistically significant. + means that the relationship is positive but not statistically significant. – means that the relationship is negative but not statistically significant. – – means that the relationship is negative and statistically significant. aVariable not included because of colinearity with real cost. example, some regions may respond faster than others to changes in interest rates, income, or expected appreciation. Some factors may affect some regional housing markets more rapidly than others. Increases in the difference between the long- and short-term interest rates—a measure of credit availability—may stimulate new construction in the following year if, for example, the decisions to build are made one year before construction, or perhaps 18 months or more before the date of the permit issue. Thus, changes in the price of alternative investments, expected appreciation, and even the level of income may not affect construction inmediately but one year or more later. Moreover, population changes may affect the market more rapidly by changing market prices. First, we discuss our findings for each of the factors in the model, and then discuss the results more generally. Population growth did not turn out to be a significant factor in the general equation for California, but it was a statistically significant factor in various regions. Results did not change significantly when we introduced the number of non-Hispanic adults rather than total population. (The demographic model presented in the next chapter evaluates the effects of California population on housing demand in more detail.) Income appeared to be significant in the explanation of changes in new construction in California and various regions. The Los Angeles– San Bernardino area was one of the exceptions. Income is expected to be a short-run determinant of residential construction. Many authors omit this factor completely from their models on investment on residential construction (see Grebler and Maisel, 1963). However, authors such as Klein (1966) suggest that the correct role for income is through its shortterm effect on the value of housing. If builders start their plans during a recession and income begins to rise rapidly, builders may alter their original plans and could expand or speed their existing construction. On the demand side, higher levels of income are positively related to housing purchases. Lagged new construction (or the change in the housing stock) was a significant factor in the explanation of changes in new construction. 48 Higher increases in the housing stock in the previous year are associated with less current new construction. Changes in new construction were closely related to changes in the business cycle, as measured by changes in unemployment. Increases in unemployment lead to less construction. Costs played a significant role in explaining changes in new construction in only four regions. Our cost measurement showed the expected negative association with new construction (negative signs) in most of our equations. In other words, everything else constant, higher building costs lead to less new construction. Our results corroborate the hypothesis that developments in financial markets determined, in part, the slower pace of new construction in the 1990s. All regional equations showed a negative association between stock prices and changes in new construction. In both California and most regions, stock prices appear to be statistically significant in their association with changes in new construction. Results indicate that investors see real estate and stocks as substitutes. A more refined analysis of this relationship could provide a deeper insight on this issue. The association between annual changes in credit conditions (the difference between long-term and short-term interest rates) and annual changes in construction was statistically significant in most of the equations. The influence of expected appreciation (measured by expected inflation) appeared to be less significant than we would have expected. One possible explanation is that the effects of the difference between long-term and short-term interest rates is already accounting for the effect of changes in expected appreciation. Another possibility is that our proxy for expected inflation is a better measure for the United States as a whole than for California (see Appendix A). Results of our model for the state of California as a whole and for ten California regions suggest that more than 80 percent of annual changes in new construction in California is explained by the influence of macroeconomic policies and demographic changes (see the last row of Table 3.1). This means that there is a large proportion of the annual variation in construction beyond the control of state policy. The 49 Changes in new housing construction (1000s) 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 explained portion varies by region, ranging from 67 percent for the North Coast, 76 percent for the Bay Area, 77 percent for the Los Angeles and San Bernardino area, 85 percent for the Sacramento-Stockton area, and 90 percent for the Central Eastern area. Figure 3.4 shows actual changes in residential new construction and projected values from our California housing model. Generally, our housing market model does a good job of predicting annual changes in new construction, including the low levels recorded in the 1990s. The model performed well particularly in central regions and areas that did not appear to have major shortages. For the most populated regions, such as the Bay Area, San Diego, and the Los Angeles–San Bernardino regions, where the state has the highest cumulative housing shortages, the results are slightly weaker. Perhaps this indicates that housing markets in these regions are heavily determined by local characteristics as well as the influence of local policies and growth controls. 150 Change in new construction Predicted change in new construction 100 50 0 –50 –100 –150 SOURCE: CRIB and time-series model. Figure 3.4—Comparison of Actual Changes in California New Housing Construction and Model Fitted Values, 1967–2000 50 Summary In this chapter, we discussed macroeconomic forces that help determine housing construction and developed an econometric model of new housing construction in California. In short, the results suggest that macroeconomic events of the 1990s can explain much of the slowdown in new housing construction in the state. Our results were consistent statewide and for ten regions of the state, although they were not as robust in some of the state’s largest metropolitan areas. 51 4. Can Demographic Changes Explain California’s Housing Markets? Population growth and housing growth go hand in hand: The more people added to the state, the greater the housing demand. However, not all population growth is equal when it comes to housing demand. Some groups require more housing than other groups. Large increases in the child population and small increases in the adult population (i.e., more children per adult) will lead to less housing demand than will small increases in the child population and large increases in the adult population. Adults form new households, children do not. Similarly, immigrants are more likely to live in extended families. Thus, the amount of housing consumed by immigrants is less, on average, than the amount of housing consumed by U.S. natives.1 In this chapter, we explore the demographic determinants of California’s housing demand. First, we examine demographic trends that shape California’s housing demand. We discuss trends in the most important of those demographic trends, including population change itself, age structure, gender, nativity, and marital status. Next, we construct a householder rate model. The model estimates the probability that a person with a certain set of characteristics will be a householder.2 High probabilities translate into high housing demand, and low probabilities imply low housing demand. The demographic characteristics that determine householder rates include age, race and ____________ 1Average household sizes are higher for immigrants even where housing costs are lower than U.S. averages. 2Each occupied housing unit has one householder. The householder is the person in whose name the housing unit is owned or rented. If more than one person owns or rents the unit, only one member of the household is chosen as the householder. A householder is also referred to as a head of household. 53 ethnicity, nativity, gender, and marital status. (The complete model and data are described in Appendix B.) Finally, to quantify the demographic component of housing demand in California, we develop simulations of housing demand using the householder rate model. In contrast to business cycle effects, demographic changes tend to be felt more strongly in the long run, as annual fluctuations are not large for most demographic measures. Thus, our focus here is on long-term changes. The householder rate model does have limitations. First, changes in vacancy rates are not explicitly considered in the model. In California, some of the increase in the number of occupied households between 1990 and 2000 occurred as formerly vacant housing units became occupied. Second, a restricted supply of housing may prevent some people from moving to California. These potential migrants cannot be measured by this model. The results suggest that the nature of California’s population growth explains much of the apparent lack of new housing in the state. In particular, in the 1990s a relatively fast growth in child and immigrant populations did not translate into the same level of housing demand as did past population growth, which was driven by domestic migrants and baby boomers aging into prime household formation years. Without controlling for demographic characteristics, householder rates have risen in the United States but declined in California. Once we control for demographic determinants of household formation, however, differences between California and the rest of the United States are substantially reduced. Indeed, once we control for demographic characteristics—age being the most important—the direction of change is the same in both the United States and California; in both places, householder rates declined in the 1980s. Moreover, with demographic controls, temporal patterns of change in the 1990s are virtually identical between California and the rest of the United States. These findings indicate that California’s household formation rates were not particularly different from those of the rest of the nation during the 1990s. Thus, although the increase in new housing units might not have kept pace with overall population growth, the nature of the state’s population growth can explain a large part of the apparent discrepancy. From a demographic 54 standpoint, California’s housing markets do not appear to be functioning differently from those in the rest of the United States. Demographic Determinants of Demand Population Growth California’s population growth slowed considerably in the 1990s. On a percentage basis, the 1990s were the state’s slowest-growing decade since records have been kept (see Figure 4.1). Indeed, official growth in the state was only slightly higher than that in the rest of the United States. Absolute gains in the state’s population were large—over 4 million people from 1990 to 2000. In the 1980s, however, the state added over 6 million new residents, and thus absolute gains in the 1990s were over 30 percent lower than those of the 1980s. If we adjust official census tabulations of the population for the respective undercounts in 1980, 1990, and 2000, population growth during the 1990s appears even lower relative to that of the 1980s: Total population change was 7,000 6,000 1980–1990 1990–2000 6,093 5,000 4,000 4,731 4,115 Population change (1000s) 3,000 2,621 2,000 1,000 1,362 1,494 0 Children Adults Total SOURCE: Authors’ calculations of decennial census data. Figure 4.1—Growth in Child and Adult Populations in California, by Decade 55 3.8 million in the 1990s, compared to 6.2 million for the 1980s. (Official census counts place the decennial change at 4.1 million in the 1990s, and 6.1 million in the 1980s.) Slower population growth relative to that of previous decades led to less housing demand in the 1990s than in the 1980s. Age The number of children in California continued to grow substantially during the 1990s, but increases in the number of adults were much lower than during the 1980s. During the 1990s, California’s child population increased 19 percent, compared to only 12 percent for adults in the state. This pattern is in contrast to the 1980s, when the state’s adult population grew faster than the state’s child population. Remarkably, absolute population growth for children was greater in the 1990s than in the 1980s, whereas population growth for adults from one decade to the next declined by almost half (Figure 4.1). Thus, not only did lower population growth during the 1990s lead to less demand for new housing than in the 1980s, but the relatively large share of children in that growth also led to less demographic-based demand. Among adults, the likelihood of forming a household is dependent on age (Figure 4.2). Household formation increases dramatically as people age from young adulthood (people in their 20s) to middle ages (adults between the ages of 35 and 44). Population increases in these age groups will lead to greater housing demand than increases in other age groups. In the 1980s, population growth in California was strong in these age groups, as baby boomers entered these prime householder years (Figure 4.3). During the 1990s, however, the much smaller baby bust cohorts replaced the baby boom cohorts, and the population of 20 to 35 year olds actually declined. Gender Women are less likely to be householders than men—primarily because most married couples list the husband as the householder on census forms rather than the wife. However, over time women have become more likely to form their own households, both because the share of never-married mothers has increased and because of past (before 56 Probability (%) 50.0 45.0 40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 17.5 22.5 27.5 32.5 37.5 42.5 47.5 55 65 75 Age SOURCE: Authors’ calculations of 2001 CPS data. Figure 4.2—Probability of Being a Householder in California, by Age, 2001 the 1990s) increases in divorce rates (Myers, 1992).3 Still, by 2000 only 30 percent of women age 15 and over were householders compared to over 60 percent of men. Population growth in California in the 1990s consisted of more women than men: Among people age 15 and over, population growth totaled 1.6 million for females, compared to only 1.4 million for males. This is in direct contrast to the 1980s, when the population of males age 15 and over increased 2.5 million, compared to only 2.1 million women. More research is necessary to understand why the gender ratio of California’s population growth switched from primarily male to primarily female, but it could be related to the Immigration Reform and Control Act of 1986. This act provided legal status for formerly undocumented immigrants, an overwhelmingly male group. After gaining legal status in the late 1980s, many formerly unauthorized ____________ 3Among married couples, women have been increasingly listed as the householder, although the percentages are still low. 57 1,000 800 1980–1990 1990–2000 Population change (1000s) 600 400 200 0 –200 –400 <5 3124134255667750505050050505––––––––––––––531241342556677–949494944949499 80+ Age Figure 4.3—Change in California Population, by Age Group and Decade immigrants were joined in the United States by their family members, including wives. Whatever the cause, this shift to a greater increase in female than male population growth in the 1990s led to less housing demand than if the increase were more composed of men. Finally, increases in the ages at which men and women marry have implications for housing demand. These factors are taken up below. Nativity and Ethnicity Population growth attributable to immigration leads to less housing consumption than population growth among U.S. natives. Foreign-born residents of California tend to live in larger families and households than U.S. natives. In 2000, the average number of people in households headed by foreign-born persons was 3.5 in California, compared to 2.5 for households headed by U.S. natives.4 In the 1980s, California ____________ 4Authors’ calculations based on 2000 Census Supplementary Survey data. 58 experienced large increases in both foreign-born and U.S.-born populations, as natural increase, domestic migration, and international migration to the state were large. In the 1990s, international migration to and natural increase in California remained strong, even as the state lost many domestic migrants to other states. The result, as shown in Figure 4.4, was a much greater relative increase in the foreign-born population than the U.S.-born population in the 1990s. This change contributed to a decline in the demand for new housing. Changes in the ethnic composition of the state’s population, which are strongly related to migration flows, also dampened new housing demand. For example, non-Hispanic whites are more likely to live alone than other groups and thus tend to consume more housing. Relatively few international migrants to California are non-Hispanic whites, whereas large proportions of California’s domestic out-migrants during the 1990s were white. As a consequence, California’s non-Hispanic white population declined during the 1990s after having increased during the 1980s. This decline had a dampening effect on the demand 3,500 3,000 2,868 3,224 1980–1990 1990–2000 2,500 2,405 Population change (1000s) 2,000 1,500 1,706 1,000 500 0 Foreign-born SOURCE: Authors’ calculations of decennial census data. U.S.-born Figure 4.4—Change in Foreign-Born and U.S.-Born Populations in California, by Decade 59 for new housing in the 1990s. California’s largest population increase in the 1990s occurred among Hispanics, the ethnic group with the lowest householder rates. Marital Status Marital status is also an important determinant of household formation and thus housing demand. Divorced and widowed adults are much more likely to form their own households than never-married adults. Married couples have high household formation rates, but married individuals tend to have lower household formation rates than divorced or widowed adults; it takes two to make a couple, and only one is considered the householder. During the 1990s, the number of married adults increased almost as much as in the 1980s (Figure 4.5). In contrast, increases in the number of adults of other marital statuses were much lower in the 1990s than in the 1980s. The sharp drop in number of divorced adults is partly a reflection of lower divorce rates in the 2,500 2,000 1,500 1,984 1,647 1,790 1980–1990 1990–2000 Population change (1000s) 1,000 871 700 500 326 0 Married Never married Divorced/Separated SOURCE: Authors’ calculations of decennial census data. NOTE: Children younger than age 15 are excluded. 174 72 Widowed Figure 4.5—Decennial Change in Population in California, by Marital Status and Decade 60 1990s. These changes in marital status in the 1990s tend to dampen housing demand. Householder Rate Model: Temporal Trends in the Probability of Forming a Household The demographic changes discussed above have consequences for the demand for housing in California. To quantify these consequences altogether and evaluate temporal changes in household formation, we develop statistical models to estimate the probability that an individual has formed his or her own household. The probability of being a householder, also referred to as the householder rate, depends on an individual’s demographic characteristics. Higher probabilities are associated with more housing, lower probabilities with less housing. Because householder rates, and thus housing consumption, vary with demographic characteristics, California’s unique demographic changes have important implications for housing demand in the state. We compare temporal trends in California with those in the rest of the country. To consider the role of various demographic factors, we adjust those trends first by controlling for temporal differences in the demographic composition of the population and second by controlling for differences between California’s population and that of the rest of the country. Details of the data and methods can be found in Appendix B. Temporal Trends in Householder Rates Across time and without controlling for demographic factors, householder rates have been moving in opposite directions in California and the rest of the nation. In California, rates have been decreasing, whereas in the rest of the United States they have been increasing. By the start of the 21st century, householder rates in California were lower than at any time in the past quarter century. At the same time, householder rates in the United States were higher than at anytime in the past 25 years (Figure 4.6). In other words, persons age 15 and over are less likely to be the head of a household in California in 2001 than in 1976, whereas that probability has risen in the rest of the United States. The declines in California and increases in the United States occurred 61 Probability 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 0.56 0.54 California Rest of United States 0.52 0.50 0.48 0.46 0.44 0.42 0.40 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Children younger than age 15 are excluded. Figure 4.6—Probability of Being a Householder in California and the Rest of the United States, 1976–2001 almost entirely in the late 1980s. In both the United States and California, householder rates changed little in the 1990s. This difference in overall householder rates translates directly into more people per household in California than in the rest of the nation. Many have concluded that this difference reflects California’s unique housing crisis. However, the pattern of change in householder rates in the United States is somewhat different when we control for demographic changes in the population over time. These demographic controls allow us to evaluate the effect of various demographic determinants on the probability of being a householder. Controlling for gender, age, and ethnicity, we observe much less temporal change in householder rates in the rest of the United States than in the uncontrolled model (Figure 4.7). Among the demographic shifts that have led to higher householder rates in the nation, the most important factor is age. In 1975, all baby boomers were children or young adults (younger than age 30), ages at which householder rates are relatively low. By 2001, baby boomers were ages 37 to 55, ages at which householder rates are high. Simply put, in most of the nation, housing consumption outpaced growth in the 62 Probability 0.56 0.54 No controls Controls for age and gender 0.52 Controls for age, gender, and ethnicity 0.50 0.48 0.46 0.44 0.42 0.40 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Controls maintain the demographic composition of the rest of the United States in 1976. Figure 4.7—The Effect of Demographic Changes on the Probability of Being a Householder in the Rest of the United States, 1976–2001 nation’s population because baby boomers entered their prime years of household formation. If we control for changes in age structure over time, householder rates in most of the United States would have changed very little over the past 25 years. In contrast, changes in the ethnic composition of the nation’s population have not played a very important role. This is at least partly because those changes have not been very pronounced in most of the nation. Thus, much of the increase in householder rates from the late 1970s to the late 1990s in most of the United States is simply an artifact of demographic changes in the nation’s population, specifically the aging of the baby boomers. Controlling for demographic factors in California also changes temporal patterns in overall householder rates. Controlling for age and gender, householder rates in the state declined tremendously from the late 1970s to the early 1990s (Figure 4.8). Were it not for the aging of the baby boomers, householder rates would have declined even more than they actually did. That is, once we control for age, we see even more dramatic declines in the probability of being a householder over 63 Probability 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 0.56 0.54 No controls Controls for age, gender, and ethnicity 0.52 Controls for age and gender 0.50 0.48 0.46 0.44 0.42 0.40 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: Controls maintain the demographic composition of California in 1976. Figure 4.8—The Effect of Demographic Changes on the Probability of Being a Householder in California, 1976–2001 time than in the model with no controls. This strong effect of the baby boom is similar to that observed for the rest of the United States. Unlike the case in the rest of the nation, however, the racial and ethnic composition of the state’s population also affected the temporal pattern of change in householder rates. In California, the state’s changing racial and ethnic population mattered almost as much as age. Declines in the probability of being a householder are less pronounced once we control for race and ethnicity. This means that part of the decline in householder rates in California could be attributed to a demographic shift in the state’s population from racial and ethnic groups that have relatively high householder rates to groups that have relatively low householder rates. Thus, declines in householder rates in California would have been even more severe were it not for the baby boomers aging into prime householder years. However, those declines would have been less substantial if the racial and ethnic composition of the state had not changed so much. 64 Differences in Householder Rates Between California and the Rest of the United States Differences in householder rates between California and the rest of the country provide an indicator of housing consumption in California and the rest of the country. When the difference is positive, Californians could be said to be consuming more housing in comparison to people in the rest of the country, and when the difference is negative, Californians are consuming less. In the late 1970s, California had substantially higher overall householder rates than the rest of the nation (Figure 4.9, no controls). By the mid-1990s and continuing to 2001, the state had householder rates that were substantially lower than in the rest of the United States. The dramatic decline in householder rates in California relative to the rest of the nation is still evident when we control for demographic differences between California and the rest of the United States (Figure 4.9, controls). In the late 1970s, Californians were much more likely to be householders than people of similar age, gender, and ethnicity in the 0.08 Controls for age, gender, and ethnicity 0.06 Controls for age, gender, ethnicity, and nativity No controls 0.04 Difference in probabilities (%) 0.02 0 –0.02 –0.04 1975 1980 1985 1990 1995 2000 SOURCE: Authors’ simulation from CPS data for 1976 to 2001. NOTE: These lines represent householder rates in California minus those in the rest of the United States. Figure 4.9—Difference in Householder Probabilities Between California and the Rest of the United States 65 rest of the nation. By 1993, householder rates were lower in California, even controlling for demographic differences. However, householder rates increased in California relative to rates in the nation in the late 1990s, so that by 1999 householder rates in California were similar to those in the rest of the nation once we control for demographic differences. In other words, when we compare Californians to their demographic counterparts in the rest of the nation, the probability of being a householder is about the same in both places. Thus, California has changed from a place with unusually high householder rates—given its population composition—to a place with householder rates that are what we would expect given the state’s population composition in comparison with the rest of the country. Today, California’s demography explains much of the apparent lack of housing consumption in the state relative to the nation. The importance of the state’s unique demography is evident when we examine householder rates for population subgroups. Overall, householder rates are almost 3 percentage points lower in California than in the rest of the United States (Figure 4.10).5 However, for U.S.-born adults, householder rates are similar in California and the rest of the United States (50.7 percent in California compared to 50.6 percent in the rest of the United States). Indeed, among U.S.-born groups, whites and Asians in California actually have higher householder rates than their counterparts in the rest of the nation. California’s lower overall householder rates are due to lower rates among U.S.-born Latinos and the foreign-born. In particular, recent Mexican immigrants have extremely low householder rates throughout the United States. Those rates are even lower in California than in the rest of the United States (25.6 percent versus 30.0 percent, not shown in Figure 4.10). Combined with California’s relatively high share of Mexican immigrants, these low rates explain much of California’s apparent shortfall in householder rates. ____________ 5The Census Supplementary Survey has higher householder rates in the rest of the United States and lower ones in California than the 1999 or 2001 CPS. Differences in age-standardized rates are 3.1 percent in the Census Supplementary Survey, versus 1.7 and 2.3 in the 1999 and 2001 Current Population Surveys. 66 0% Total 10% 20% U.S.-born Foreign-born White U.S.-born White foreign-born Latino U.S.-born Latino foreign-born Asian and Pacific Islander U.S.-born Asian and Pacific Islander foreign-born African American Rest of United States 30% California 40% 50% 60% 50.0% 47.4% 50.7% 50.4% 44.4% 40.6% 50.3% 51.4% 46.8% 46.3% 49.1% 45.2% 41.7% 38.9% 47.8% 48.0% 40.5% 37.7% 53.8% 52.6% SOURCE: Authors’ estimates based on 1999–2001 CPS and the 2000 Census Supplementary Survey. Figure 4.10—Age-Standardized Householder Rates in California and the Rest of the United States, by Ethnicity and Nativity, 2000 Householder Rates in California Metropolitan Areas Householder rates vary across California’s regions as well. Thus, although we find that demographic factors explain much if not all of the difference in householder rates between California and the rest of the United States, the picture for specific metropolitan areas within the state might also differ. Using the Current Population Surveys from 1999 to 2001, we compare householder rates for primary metropolitan statistical areas in California with householder rates in the rest of the country. Appendix B describes the methods used to develop the estimates. Without controlling for demographic differences, the probability of being a householder is lower in most of California’s metropolitan areas than in the rest of the nation (Figure 4.11a). Among the state’s larger metropolitan areas, only Sacramento and Fresno appear to have householder rates that are similar to those in the rest of the country. 67 Lower rates in California Higher rates in California –0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 San Luis Obispo Chico Stockton Fresno Sacramento Rest of California Merced Modesto Yuba City–Marysville Visalia Santa Rosa Riverside–San Bernardino Oakland Los Angeles–Long Beach Bakersfield Orange County Ventura Vallejo–Napa San Jose San Diego Salinas–Monterey Davis–Woodland Santa Barbara San Francisco NOTE: Figure shows point estimates and 90 percent confidence intervals. Figure 4.11a—Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, No Demographic Controls, 2000 68 Lower rates in California Higher rates in California –0.20 –0.15 –0.10 –0.05 0 0.05 0.10 0.15 0.20 San Luis Obispo Chico Stockton Fresno Sacramento Rest of California Merced Modesto Yuba City–Marysville Visalia Santa Rosa Riverside–San Bernardino Oakland Los Angeles–Long Beach Bakersfield Orange County Ventura Vallejo–Napa San Jose San Diego Salinas–Monterey Davis–Woodland Santa Barbara San Francisco NOTE: Figure shows point estimates and 90 percent confidence intervals. Figure 4.11b—Difference in Householder Rates: California Metropolitan Areas Less the Rest of the United States, with Controls for Age, Gender, Ethnicity, and Nativity, 2000 69 Thirteen metropolitan areas had householder rates that were significantly lower than rates in the rest of the country, including Los Angeles–Long Beach, San Francisco, San Diego, Orange County, Riverside–San Bernardino, and San Jose. Controlling for demographic factors, however, a different picture emerges. In 11 metropolitan areas, householder rates are either higher than or no different from rates in the rest of the country, and in another six metropolitan areas, householder rates are not significantly different from those in the rest of the United States (Figure 4.11b). Only six metropolitan areas had lower estimated householder rates than in the rest of the country when we control for demographic composition. Nonetheless, those six include some of the state’s largest metropolitan areas: Los Angeles–Long Beach, San Francisco, Orange County, San Diego, and San Jose. In those metropolitan areas, adults are less likely to be householders than their demographic counterparts in the rest of the country. The effect of the demographic controls varies by metropolitan area. In Los Angeles–Long Beach, the controls erase most of the difference in householder rates between the metropolitan area and the rest of the nation, whereas in San Francisco the demographic controls make little difference. This, of course, is due to the nature of the population in these two metropolitan areas. Los Angeles–Long Beach has a much larger share of demographic subgroups that tend to have low householder rates, especially foreign-born Latinos, than does the rest of the country or San Francisco. Estimating the Shortage of Occupied Housing Units We use the results of our statistical analyses to estimate the state’s shortage of occupied housing units by comparing householder rates in California to those in the rest of the United States. We do so once without controlling for demographic differences and again controlling for demographic differences. In essence, we answer the following questions: • How many occupied housing units would California have if it had the same overall householder rates as those in the rest of the country? 70 • How many occupied housing units would California have if it had the same householder rates by demographic subgroup as in the rest of the country? The difference between the number of actual occupied housing units in California and the number that the state would have if it had the same householder rates as in the rest of the nation, controlling for demographic subgroup, is taken as an estimate of California’s housing shortage. We develop statewide estimates and estimates for specific metropolitan areas in the state. To ensure greater precision with large sample sizes, we restrict the estimates to decennial census years. Because such data were not yet available for 2000 when this report was written, we use the Census 2000 Supplementary Survey in conjunction with Current Population Survey data to develop estimates for 2000. Details of the approach are in Appendix B. It is important to keep in mind that these are estimates of occupied housing units rather than total housing units. The two differ by the number of unoccupied housing units, including those for seasonal use as well as those for rent or sale. To the extent that California’s vacancy rates are low and have declined over the past decade, these estimates understate the shortage of total housing units; that is, householder rates are higher in California than they would be if the vacancy rate had not declined during the 1990s. This could be remedied by assuming some ideal occupancy rate—something we do not pursue here.6 It is also important to keep in mind that householder rates in the rest of the United States are taken as the standard for comparison. One could argue that those rates are problematically low, especially for certain groups, and embody a national housing shortage. The results, as shown in Figure 4.12, indicate that California has gone from a position of housing excess in 1970 and 1980 to increasingly ____________ 6Vacancy rates for rental units in California declined from 5.9 percent in 1990 to 3.7 percent in 2000. Statewide, these vacancy rates are at “normal” levels, typically regarded as being in the range of 3 to 5 percent (Landis et al., 2000). In the Bay Area and the counties of Santa Cruz, Monterey, Santa Barbara, and Ventura, rental vacancy rates in 2000 were lower than 3 percent. Since 2000, however, vacancy rates have risen substantially in Bay Area counties (U.S. Department of Housing and Urban Development, 2003). 71 Supply (1000s) 1,000 800 600 400 200 0 –200 –400 –600 –800 –1,000 1970 1980 No controls Controls 1990 2000 Figure 4.12—Estimates of the Undersupply of Housing in California Based on Householder Rates in California and the Rest of the United States, by Decade large shortages in 1990 and 2000. The extent of the shortage depends very much on whether we control for differences between California’s population and that of the rest of the country. With no controls for the demographic composition of the state’s population, California’s housing shortage in 2000 is estimated to be 870,000 units. Most of that shortage was generated in the 1980s. During that decade, California went from an excess of 147,000 housing units to a shortage of 610,000 units, a difference of 757,000 units. In comparison, during the 1990s the shortage worsened by another 260,000 units. However, when we control for demographic differences between California’s population and that of the rest of the country, the shortage appears much smaller—about 138,000 units. This suggests that most (84 percent) of the apparent shortage in California’s occupied housing units can be attributed to demographic differences between California and the rest of the nation. We also find that most of this shortage was generated in the 1980s rather than the 1990s. The increase in the shortage of occupied housing units, controlling for demography, was 55,000 units during the 1990s. In contrast, during the 1980s the state went from an excess of 476,000 occupied housing units to a shortage of 72 83,000 based on the model with demographic controls. The increase in number of occupied housing units in California during the 1990s was very close to the number that would be necessary to maintain the differential that the state started with in 1990 (Figure 4.13). Even without demographic controls, California’s housing shortage is much more a creation of the 1980s than the 1990s (Figure 4.14). The shortage of occupied housing units in California is concentrated in the state’s largest metropolitan areas. Still, many of the state’s metropolitan areas, including some larger metropolitan areas such as Sacramento and Fresno, exhibit little or no shortage in housing units (Figure 4.15a). When we take into account the demographic composition of each metropolitan area, the shortages are not so severe as when we do not adjust for age, gender, ethnicity, and nativity (Figure 4.15b). Demographic controls are especially important in Los Angeles. For example, the shortage in the Los Angeles metropolitan area decreases from a point estimate of 314,000 units to 101,000 units when we control for Los Angeles’ unique demographic composition. In contrast, adjusting for the demographic composition in the San Francisco metropolitan area makes little difference, as that metropolitan area’s 2,500 2,000 1,500 Actual Necessary to match rest of U.S. householder rates Necessary to maintain start of decade differential Number of units (1000s) 1,000 500 0 1970–1980 1980–1990 1990–2000 Figure 4.13—New Occupied Housing Units, by Decade: Actual Versus Necessary for California (Controlling for Demographic Differences) 73 Number of units (1000s) 1,000 800 600 1970–1980 1980–1990 1990–2000 400 200 0 –200 –400 –600 Necessary to match rest of U.S. householder rates Necessary to maintatin start of decade differential No controls Necessary to match rest of U.S. householder rates Necessary to maintain start of decade differential Controls for gender, age, nativity, and ethnicity Figure 4.14—Decennial Shortage of New Occupied Housing Units population does not have large shares of subgroups with low householder rates. These estimates of California’s housing shortage appear conservative when compared with oft-cited numbers. For example, the California Building Industry Association estimates that the state needs 220,000 new housing units each year to keep pace with population growth but that the state is producing only a little over half that amount (Weintraub, 2002). The California Department of Housing and Community Development projects that California will need at least 200,000 new housing units each year to 2020 (Landis et al., 2000), but the state has averaged less than 117,000 new permits per year between 1990 and 2001 (California Budget Project, 2002). The California Budget Project puts the shortage of affordable units at 651,000 statewide and 289,000 in Los Angeles County (California Budget Project, 2002). A Fannie Mae Foundation report places the shortfall in new housing during the 1990s at 548,000 units—an estimate that was derived from an approach similar to the householder rate approach developed in this report except that it does 74 not adjust for demographic composition (Myers and Park, 2002).7 Indeed, none of the estimates of the shortage of new construction in California adequately consider the demographic composition of the state’s growth. Our estimates also suggest that the creation of the shortage primarily occurred in the 1980s rather than the 1990s. Many analysts note that the increase in new housing units during the 1980s was almost twice as great as in the 1990s, arguing that the shortage in new housing units is primarily a 1990s phenomenon (e.g., Myers and Park, 2002). We find that the increase in occupied housing units in California during the 1980s did not keep pace with the state’s strong population growth in general and the changing age structure of the state in particular. Our models suggest that, given the large numbers of baby boomers aging into the prime householder ages during the 1980s and with tremendous population growth, the number of new occupied housing units necessary to match householder rates in the rest of the United States was a great deal higher than actually achieved in the state. Although the shortage did indeed worsen in the 1990s, the relative contribution to the shortfall was much smaller in the 1990s than in the 1980s. Summary In this chapter, we have shown that California’s relatively anemic increase in new housing units during the 1990s is at least partly due to the nature of the state’s population growth. In particular, population growth slowed in the 1990s compared to the 1980s and growth became more concentrated among groups—children and immigrants, for example—that consume less housing than other groups. We find evidence of housing shortages in key areas of the state, but these shortages are lower than commonly cited figures. ____________ 7Also, even our unadjusted householder rate model restricts the analysis to people age 15 and older, whereas the Fannie Mae Foundation report uses an average household size approach that includes all members of the household regardless of age. To the extent that much of California’s population growth was among children who do not form their own households, the Fannie Mae approach will overstate housing needs even relative to the householder rate model without controls. 75 –200,000 –100,000 Number of Housing Units 0 100,000 200,000 300,000 400,000 Sacramento Rest of California Fresno Stockton San Luis Obispo Chico Modesto Merced Santa Rosa Yuba City–Marysville Visalia Riverside–San Bernardino Davis–Woodland Vallejo–Napa Bakersfield Oakland Ventura Salinas–Monterey Santa Barbara San Jose Orange County San Diego Los Angeles–Long Beach San Francisco NOTE: Negative values reflect an oversupply; positive values reflect an undersupply. Figure 4.15a—Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of United States, No Demographic Controls, 2000 76 –200,000 –100,000 Sacramento Number of Housing Units 0 100,000 200,000 300,000 400,000 Rest of California Fresno Stockton San Luis Obispo Chico Modesto Merced Santa Rosa Yuba City–Marysville Visalia Riverside–San Bernardino Davis–Woodland Vallejo–Napa Bakersfield Oakland Ventura Salinas–Monterey Santa Barbara San Jose Orange County San Diego Los Angeles–Long Beach San Francisco NOTE: Negative values reflect an oversupply; positive values reflect an undersupply. Figure 4.15b—Estimates of the Undersupply of Occupied Housing Units in California Metropolitan Areas Based on Householder Rates in California and the Rest of United States, Controls for Age, Gender, Ethnicity, and Nativity, 2000 77 5. Conclusion and Policy Considerations This report evaluates the effect of business cycles and demographic factors on the demand and supply of housing in California. We found that more than 80 percent of annual changes in construction in California can be explained by the influence of macroeconomic effects and demographic changes, leaving small room for local and state policies. Our results suggest that the sluggish response of construction to the latest economic recovery could be explained by: • Slower population growth in the 1990s compared to the 1980s, which slowed the demand for housing. The nature of that growth, too, explains much of the apparent lack of new housing in the state, with immigrants and children accounting for most of this growth, but consuming proportionately fewer housing units than other demographic groups. • The severity and duration of the economic recession that took place in 1990 in California, leading to low housing valuation. • The Federal Reserve’s monetary policy, which decreased the public’s expectations of future inflation and caused the price of long-lived assets to decline. • High returns on alternative investments (such as stocks and mutual funds) that led to a decrease in new housing investment. • Decrease in builders’ and developers’ expectations of profits on new housing construction, because of both changes in future inflation and interest rates and a slowdown in housing demand resulting from demographic changes. Although our analysis runs through the 1990s, there is little reason yet to believe that the results would be altered by very recent events. 79 Since 2000, housing supply has continued to increase while job growth has stopped. Declining interest rates have fueled further price gains in most regions and have countered the severe job and income losses in the Bay Area. Most analysts expect housing prices to decline (in real terms) once interest rates rise from 40-year lows, and some have argued that the Bay Area is especially prone to declines (Leamer, 2002).1 Our results suggest that the housing supply crisis may be overstated—at least in the context of the entire state—and that our position today is perhaps better, and certainly not much worse, than where it was in 1990. The housing cycle in the 1990s is quite different from the three previous cycles but this is also true at the national level— although the falloff in production is not nearly as pronounced as it is in California. Indeed, some of the same arguments over whether there is a housing shortage or a housing bubble are also taking place at the national level. Housing prices have not fallen despite the recession that began in 2001, raising the specter of a housing bubble akin to the stock market bubble. Other observers are much more concerned about the slower rate of housing production in this cycle and warn of an ongoing shortage nationwide. A recent analysis, however, suggests that even if housing prices nationally have outpaced their rental equivalents, a few years of flat home prices and normal rent increases would bring the two markets back into alignment.2 Any policy responses to problems in the state’s housing market need to be mindful of both the important connection between underlying macroeconomic and demographic forces and real estate investment, as well as the regional nature of housing markets. Our findings suggest that there is not a large statewide crisis in housing production in California. This is not to say that there are no legitimate concerns in the state with regards to housing. In this report, ____________ 1Of the Bay Area, Leamer (2003) states: “Expect a long agonizing decline in prices like we had in LA in the early 1990s.” 2This analysis (Krainer, 2003) also demonstrated the importance of the after-tax mortgage rate in the owner cost of housing. Not only do declining mortgage rates reduce the new owner’s cost but, through refinancing, reduces the ongoing cost of ownership for existing homeowners. Both effects serve to prop up housing prices even in a weak economy. 80 we have not considered other issues such as housing affordability for lowincome families, the condition of rental housing, or overcrowding.3 Previous research suggests that the state does have problems in these areas. For example, in our earlier report we found that overcrowding occurs more frequently in California than in the rest of the United States but also that poverty and demographics were greater determinants of overcrowding than were high housing prices (Moller et al., 2002). Overcrowding is much more common in poor areas of the state with relatively low-priced housing than it is in expensive places such as the Bay Area.4 The mismatch between the type of new housing that is constructed and the needs of the newest Californians is a particularly important topic for future research. Given that houses are long-lived assets, rapid demographic change can increase demand for a particular type of housing unit that easily outstrips new supply. Much of the population increase in the 1990s came from an increase in family size, primarily among low-income immigrant families. Such families are typically renters, and the rental housing stock in many areas is not sufficiently oriented toward large households. In addition, as shown in Chapter 2, production of rental (i.e., multifamily) housing fell much more dramatically than did production of single-family housing. Although the amount of new housing units produced in California is largely explained by macroeconomic and demographic factors, housing shortages do exist in the state’s largest metropolitan areas: the Bay Area, coastal Los Angeles (Los Angeles, Orange, and Ventura Counties), and San Diego County. The magnitude of the shortages is much lower by our estimates than those put forth by others, and even some of these shortages may be worked off after several years of job losses (most dramatically in the Bay Area) and the completion of new housing. Prices in these regions are higher partly because of their strong desirability as ____________ 3We addressed this issue in a previous report (Moller et al., 2002). 4Using 1990 census data, Myers et al. (1996) also found that demographic factors and income were more important than housing affordability in explaining differences in overcrowding between metropolitan areas. 81 places to live—their amenity value—and partly because of obstacles to new supply. These obstacles can be regulatory or financial, or both. There are ongoing battles, even in the midst of a widely reported housing crisis, by groups who oppose most new greenfield development (often termed “sprawl”) and by those who oppose most infill development in their communities (often referred to as the NIMBY5 syndrome. These actions serve to reduce the supply of new housing in areas where developers have found demand for it. Two of the regions of California with evidence of housing shortages are also the least receptive to new housing construction. According to a PPIC survey of city managers in 1998, respondents in the Los Angeles region and the Bay Area were much more likely to state that residential growth issues were “often” or “always” controversial (Lewis and Neiman, 2000). They were also more likely than respondents in other regions to state that there was little or no vacant land available for new development (Lewis and Barbour, 1998). Using GIS and mapping techniques, researchers in a study for the California Department of Housing and Community Development (HCD) concluded that sufficient vacant land exists statewide for 20 years of expected population growth, but restrictions in some key urban areas—including some Bay Area and South Coast counties—are likely to present serious obstacles in the future (Landis et al., 2000). According to the authors of the HCD study The key constraint on the State’s development capacity is political, not environmental. As this analysis reveals, the careless adoption of even moderately-limiting urban growth boundaries by high-growth counties in the Bay Area and Central Valley would constrain land supplies below the levels required to meet future housing production needs. In many cases, voters have directly restricted new housing growth. Dozens of local jurisdictions in California have passed measures to limit growth (Glickfeld and Levine, 1992; Lewis and Neiman, 2000; Landis et al., 2000). Solimar Research Group has documented 389 slow-growth ballot measures in California from 1986 through 2000, with 59 percent of them passing. Of course, ballot measures are not the only way citizens ____________ 5The acronym stands for Not In My Back Yard. 82 express their displeasure with new developments. City councils often reflect the lack of support for new housing of their constituents and have simply failed to allow new housing developments or delayed their approval long enough for developers to drop their proposed projects. A recent study of housing affordability in the nation’s metropolitan areas found that in those areas with the least affordable housing, land-use controls were the most compelling explanation for high relative prices (Glaeser, Gyourko, and Saks, 2003). It also demonstrated that there was little change in affordability by metropolitan area between 1989 and 1999—supporting our conclusion that much of what is driving concerns about new housing supply is cyclical. There can be financial obstacles to new housing supply as well. In response to charges that litigation costs were contributing to the sharp decline in multifamily construction, the state legislature recently proposed several bills to restrict construction defect litigation and encourage remediation efforts over litigation. Even without opposition from local residents, infill development can often be more expensive to construct than the market will bear. A recent study undertaken for the California Department of Housing and Community Development (Landis, 2003) categorizes the barriers to infill production. Although a number are political in nature, several are purely economic: a shortage of adequate lot sizes, greater need for infrastructure upgrades, high land prices, lack of economies of scale, and greater design costs to work within existing land uses. Even for greenfield development, there can be sizable expenses in addition to direct construction costs. Most local jurisdictions in California have resorted to the use of development fees and special districts to recoup the cost of new housing (new housing leads to increased demand for schools, parks, and basic infrastructure). A study of these fees in Contra Costa County reported typical costs of $20,000 to $30,000 per new housing unit (Dresch and Sheffrin, 1997) and news reports pegged fees in Santa Clara County at over $30,000 per unit. Even jurisdictions not facing pressure from residents to restrict growth can be reluctant to allow large-scale development that might trigger more demand for public services. Most city managers believe that the additional tax revenues generated from new housing development do not 83 cover the full cost of the public services they require (Lewis and Barbour, 1999). The state is active in California’s housing markets through a number of funding programs. These programs are generally targeted to meet specific housing needs and are not necessarily designed to increase the supply of new housing as a whole. Although it is beyond the scope of this research to evaluate the efficacy of such programs, their sheer number and size (in terms of expenditures) suggest that they do affect California’s housing markets, and at least some lead to greater housing construction than would otherwise occur in the state. The undersupply of new housing in the Bay Area, coastal Los Angeles, and San Diego County is cause for concern. Increasing employment in those areas in the late 1990s exacerbated the jobs-housing imbalance and led many workers to live in areas far removed from their jobs. High and increasing housing prices reflect both existing amenities and the shortage in those areas, and these prices can create affordability problems for middle- and low-income households that do not already own their own homes. Any policy intended to increase the supply of new housing in the state’s largest metropolitan areas must address the inherent conflict between existing homeowners and renters (as well as new entrants to the region). Current homeowners reap direct financial benefits from any restriction on new housing supply that raises the price of existing homes. But their opposition to new development does not necessarily reflect personal financial interest alone. Many homeowners have made their largest financial investment in a community with certain characteristics, and their local government is expected to reflect their concerns. New infill development can increase traffic congestion, overwhelm the capacity of local public facilities, and require an increased tax burden on current residents. Although often opposed by statewide and even national groups, new greenfield development can sometimes conflict with the priorities of current residents as well. Part of the character of their local community may be proximity to open spaces and scenic views that would be lost to new housing development. In light of these forces, the problems facing potential new residents are often less apparent and urgent to most policymakers. 84 Policies could be implemented to make new residential development a more financially attractive option to local officials. These policies could come in the form of both rewards and punishments. For example, the state could impose financial penalties on jurisdictions that do not provide their “fair share” of new housing.6 These penalties could include withholding of funds for missing housing goals as well as rewards for meeting them (Lewis, 2003). Jurisdictions that meet their statedetermined share of new housing could be rewarded financially by having the state offset some of the costs of providing new infrastructure. Reducing local reliance on sales tax dollars and increasing local access to property tax revenue might also increase the attractiveness of new housing to local officials. Of course, there have been no shortage of proposals and recommendations to reallocate property and sales tax revenues to reduce the fiscalization of land use, and the specific details of any legislation will determine the support and opposition it faces. If implemented, such a policy will help illuminate just how much of an influence the fiscalization of land use has really had on housing supply. Citizen opposition to new development might be an even greater impediment to new housing construction in the Bay Area and coastal Los Angeles. Citizens can express their opposition to new development in myriad ways, from passing initiatives that limit growth to electing city councils that restrict growth to intervening in the planning and approval process for new developments. When combined with the fiscal concerns regarding housing versus other land uses, such citizen opposition often squelches new housing. Policies to address citizen opposition must consider the source of that opposition. Opposition to new housing is not necessarily based on housing per se but originates from concerns about the effect of the new development on the environment and existing infrastructure. Specifically, concerns are often raised about increasing traffic congestion, loss of open space, and overcrowding of schools and other public facilities. Policies that address these concerns could lessen opposition to new growth. In some cases, educating the public about the ____________ 6Compliance with the state’s housing element does not seem to result in an increased supply of housing (Lewis, 2003). 85 consequences of not planning and allowing for new housing could change local opinion. Finally, policies could be enacted to address the jobs-housing imbalance by focusing on jobs rather than housing. For example, the state could encourage job growth in areas with relatively abundant housing (especially in inland areas adjacent to the state’s most undersupplied regions). A common response across the country to population pressures has been the growth of once outlying regions as economic centers in their own right. As a natural outgrowth of job growth in the inner Bay Area, satellite job centers have bloomed in Pleasanton and Livermore; further job growth in these locations could provide shorter commutes for residents in neighboring counties. For cities and counties with housing shortages, the state could reward jurisdictions that meet some standard of zoning for adequate housing when approving new employment-generating land uses and penalize those jurisdictions that do not meet such goals. The Jobs/Housing Balance Incentive Program is an example of this kind of effort, providing financial rewards for new housing construction in high job growth areas where housing has not kept pace with construction.7 All of these policy options are controversial and would create winners and losers. Our primary finding, that the state’s housing shortage is smaller than previously thought and concentrated in three regions— albeit the state’s most populous regions—suggests that any policies designed to address the housing shortage should focus on the situations in those regions. How the regional housing markets fare when interest rates rise from their historic lows—and how high the Bay Area’s housing prices remain if job growth is delayed much longer—will clarify how much new intervention may be required. An important, related policy question is whether the state should provide incentives for more housing construction in the built-out and environmentally protected coastal areas or instead increase housing by supporting infrastructure development in less built-out areas adjacent to existing job centers. In the final analysis, ____________ 7In “high employment demand areas,” the program provides incentives of $1,300 per unit for new units above some baseline standard (California Department of Housing and Community Development, 2003). In the context of developer fees of $30,000 or more per unit, this may not provide a large incentive in the tightest markets. 86 the conflict between local control of land-use decisions and increasing new housing construction is not one that is likely to be resolved any time soon. 87 Appendix A A Housing Market Model Our housing model used to assess the influence of macroeconomic factors on California housing was summarized in a single equation that explains residential new construction. This equation is derived from a model of residential housing markets in equilibrium and relates changes in new construction to changes in the factors that determine the levels of housing demand and supply. The effects of governmental economic policies and other worldwide influences are measured through changes in interest rates, unemployment, income, expected inflation, and stock prices. These are factors that affect both the demand and supply of new housing. Our housing model takes into account the fact that the adjustment of new housing construction to changes in housing demand is not instantaneous. A Simple Model for the Housing Market The demand for housing (Dh) is determined by Dh = Dh(Ph,Ps,E(p),Pop,I, Y, A) where: (A.1) Ph = Ps = E(p) = Pop = I= Y= A= price of housing price of stock expected price of housing, or expected inflation population variables (structure of the population, immigration, population growth) interest rates income amenities. The supply of housing (Sh) is determined by 89 where Sh = Sh(Ph,Ps,E(p),I,C,Pol) (A.2) Ph = Ps = E(p) = I= C= Pol = price of housing price of stock expected price of housing, or expected inflation interest rates (mortgage rates) construction costs policy variables. The optimal amount of housing is found when supply and demand are in equilibrium: Sh* = Dh *. (A.3) However, equilibrium is not instantaneous. It takes time for changes in supply to meet changes in demand. We introduce a simple stock adjustment model, where the equilibrium stock of housing is Sh*(Ph, Ps, E(p), I, C, Pol). We represent the adjustment process as Sht – Sht –1 = g(Sht * – Sht –1) 0 < g <1. (A.4) Equation (A.4) specifies that the change in Sh will respond only partially to the difference between the desired stock of housing and the past value of stock of housing. The adjustment coefficient g determines the rate of response. Substituting for Sh* in Equation (A.4), and solving for Sht yields Sht – Sht –1 = g(Sh * (Ph,Ps,E(Ph),I,C,Pol) – Sht –1) . (A.5) Using the fact that Sh* = Dh*, and that, in equilibrium, quantities determine prices, the price of housing disappears from the equilibrium equation. Using the fact that Sht- – Sht–1 = NC = new construction, and taking first differences of these equations, we obtain NCt – NCt –1 = g *[f * t(E(Ph)t – E(Ph )t –1,Pst – Pst –1, (A.6) It – It –1,CtCt –1, Yt – Yt –1,Popt – Popt –1,It – It –1) – NCt –1] 90 where St – St–1 is equal to new construction if we disregard the demolition of housing stock. E(Ph) was estimated independently. A lagged variable of new construction (NCt–1 – NCt–1) was also incorporated to include the effect of previous changes in new construction on current new construction, when it was estimated to be necessary to control persistent effects on the disturbances. Measurement of Variables NC = new construction Number of permits issued for total family units as by the Construction Industry Research Board (CIRB) Ps = price of stock Estimated from Standard and Poor’s 500 (S&P) at http://chart.yahoo.com/t?a=10&b= 01&c=55&d=10&e=23&f=01&g=m&s=%5 Espc&y=0&z=5. An alternative way to measure the return on stocks is the S&P earnings yield (E/P) (Kaiser, 1999). We used changes in the S&P, since the historical series is already adjusted by dividends and stock splits. E(p) = Expected price of housing, or expected inflation To measure expected appreciation we use various alternative approaches. We worked with our own projections of expected prices, based on an adaptive expectation model using two and three lags of California changes in the Consumer Price Index (CPI) and the CPI housing component in California and the United States. In a similar way, we also modeled changes in 91 California median home prices. Finally, we used different expected inflation rates as reported in the Livingston Surveys. The Livingston Survey data are discussed below. Since we are working with flows, we have included change of expected appreciation (or prices). The advantages of working with price changes rather than levels are discussed in Fratantoni and Schuh (2000). Pop = population variables (population growth) California Department of Finance annual estimates I = interest rates We estimated the effect of both mortgage rates and the prime rate. Builders indicate that their costs of funds depend on the prime rate rather than the Treasury bills or mortgage rates. We found that statistically it was better to work with the difference between the long-term and the short-term rate. We used the difference between Treasury securities maturing in ten years and commercial paper maturing in three months. Y = personal income California Department of Finance. A = amenities This variable was controlled by using a regional approach grouping counties of similar characteristics. 92 C = construction costs In the estimation of the model, we tried to include various construction cost indexes published by the Census Bureau, but we did not get good results. Then we used the construction regional cost index compiled by CIRB, which estimates costs for nine regions. Because the board’s classification of counties in regions differed from ours, we built our own regional cost indexes by assigning to each county the cost estimated by the CIRB for the region where the CIRB assigned that county. Then, we used the number of total permits to estimate a new weighted average cost for our ten regions. Some Methodological Remarks on Our Time-Series Approach We estimated our reduced-equation model for each region (see above).1 We also estimated an aggregate equation for new construction in California. To avoid the problem of measuring the housing stock for various regions, we worked with first differences. Furthermore, we estimated the model using a Prais-Winsten regression. This method takes into account previous effects that persist over time affecting the error terms of subsequent specifications.2 We also tried ARIMA specifications, but this type of approach did not provide better results than the ones obtained with the Prais-Winsten regression. The ARIMA ____________ 1The model was reduced to one equation. 2We also used this method for cross-pooled time-series analysis since, in addition to the effects that often persist over time affecting the error terms, there are also effects if the random shock affecting economic activity in one region causes economic activity in an adjacent region because of close economic ties between the regions. The CochraneOrcutt method is a procedure to control the correlation between errors from one period to another. The Prais-Winsten method is the Cochrane-Orcutt method when it incorporates its special transformation to estimate the correlation coefficient between the error term of current and previous periods to the first observations in the time-series. 93 method controls for both the effect of past values and the effect of current and lagged random disturbances (unmeasured effects). Modeling Expected Appreciation Since we do not have many observation points in our time series, we calculated expected appreciation, E(p), independently to include it in our estimating equation as an exogenous factor. First we calculated our own expected prices series. Since all prices tend to move together and are highly correlated, we used projected inflation rates and median housing prices as measures of expected appreciation. There are various theories on expectation formations. Adaptive expectations assume that expectations are based only on past knowledge. Rational expectations believe that expectations are based on all information available and an understanding of how the market works. Extrapolative expectations are based on the past but follow along a trend line into the future. To calculate our own inflationary expectations, we used two approaches. First we used an adaptive expectation model that included distributed lags of price changes. We chose this model because various studies have rejected the notion that people form their expectations rationally. For example, Hamilton and Schwab (1985) raised doubts as to the general applicability of the rational expectations model. Collins, Lipman, and Groeneman (1992) also reject the applicability of rational expectations to the real estate market. Our second approach was to use expected inflation rates as reported by the public at each year included in our time-series. This was accomplished by using various measures of expected inflation one and two years ahead for each year, as reported by the Livingston Survey.3 The Livingston Survey data performed better in our estimating equations and provided more alternatives for modeling expected prices, since reports include the public perception for one year and two years ahead. Although we also explored alternative sources for expected prices such as one-year-ahead expected prices as calculated by the DRI model, and Green Book forecasts (a briefing document with macroeconomic ____________ 3Described below. 94 forecasts—including inflation—prepared by staff economists at the Board of Governors about three days before each Federal Open Market Committee meeting), we used the Livingston Survey. The Green Book forecasts become available to the public only five years after they are made; DRI forecasts were highly correlated with some other determinants of new construction. None of the various measures of inflation yielded very good and consistent results for all our ten regional equations. However, most equations showed better results when we used expected inflation rates as reported by the Livingston Survey (median rates of projected inflation) instead of our own projections. The Livingston Survey The Livingston Survey dataset contains a forecast of economic variables from a survey of forecasters. Joseph Livingston, a columnist for the Philadelphia Inquirer, began the survey in June 1946. Livingston continued the survey until 1990, when the Federal Reserve Bank of Philadelphia took it over. The survey is conducted twice a year, in June and December. The survey forecasts 18 macroeconomic variables describing national output, prices, unemployment, and other macroeconomic data. Each variable in the survey asks for forecasts of the level of the variable in the current quarter, in the next quarter, and in the same quarter in the following year. It also asks for the annual average of the variable for the current year (six months ahead), and one year ahead. For example, for the Consumer Price Index in the survey of June 1992, it asks for forecasts of the monthly level of the CPI in June 1992, December 1992, and June 1993, and the annual average of the CPI in 1992 and 1993. The survey is administered to affiliates of non-financial businesses, academic institutions, commercial banking, government, Federal Reserve Bank, insurance companies, and labor unions. One problem is that the affiliations for each participant change over time. In this study, we used the one-year-ahead forecasted CPI from the current month of the survey. This variable is reported from the December surveys and predicts the average level of the CPI for the next calendar year. We used the median values of this variable because they 95 gave better results than the average values. We also tried forecasts for two years ahead. However, the one-year-ahead forecast provided better statistical results. The CPI data reported by the survey have different base years. We worked with the CPI series adjusted to the base 1982 = 100. For more information on the Livingston Survey, see http://www. phil.frb.org/files/liv/document.html. Modeling the Influence of Interest Rates on Housing Markets Interest rates and new construction are negatively related. Typically, housing demand or supply equations estimate the effect of either mortgage rates or the prime rate. Builders indicate that their cost of funds depends on the prime rate rather than the Treasury bills or mortgage rates. We looked at the difference between the long-term and short-term interest rate as measured by the difference between two types of securities: Treasury securities maturing in ten years and commercial paper maturing in three months. This differential is expected to be positively related to new residential construction. The difference between the long-term and short-term interest rate indicates that the expectations of future inflation are rising, driving shortterm rates up. When short-term rates are above some long-run average, investors may expect them to fall and receive capital gains. On the other hand, if investors expect capital losses, the differential between long- and short-term rates will widen. If they expect capital gains, the differential will narrow. At the moment of the change in expectations, long-term rates will rise relatively faster than short-term rates. The spread between the long-term and short-term interest rates also represents a better measure of the tightness of credit than does the level of either interest rate. Easy monetary policy through its effects on expectations about capital gains will lead to a large spread between longterm and short-term interest rates, and tight money to a small spread. During periods of tight money, lending institutions will give higher priority to loans for business investment. Other sources such as bond and stock markets are effectively closed to all but the largest residential builders. 96 Table A.1 Analysis of Changes in New Construction in California, by Region (Time-Series Results, No Lags) 97 North Coast R2 0.67 F test 3.57 Probability 0.02 Metropolitan Statistical Area 223 North Central 0.6 3.44 0.02 810 North Eastern 0.82 7.77 0 226 Bay Stockton- Area Sacramento 0.64 0.7 3.13 4.11 0.03 0.01 5498 3922 Central Coast 0.6 2.59 0.06 1067 Central Eastern 0.89 14.34 0 298 Los Angeles– Central San Inland Bernardino 0.76 0.57 5.62 2.27 0 0.09 2408 20594 San Diego 0.72 4.41 0.01 5649 California 0.64 4.24 0 35891 Change in population Coefficient t Probability 0.077 0.123 0.12 0.091 0.126 0.071 0.134 0.134 2.1 2.78 2.1 1.4 1.14 1.43 3.91 1.57 0.05 0.01 0.05 0.18 0.28 0.18 0 0.14 0.157 1.14 0.27 –0.126 –1.47 0.16 0.072 0.68 0.51 Change in income Coefficient t Probability 0.231 2.32 0.04 (a) 0.153 0.77 0.45 –0.013 –0.22 0.83 0.066 1.46 0.17 0.098 0.9 0.38 0.461 3.41 0 0.191 1.6 0.13 0.146 0.87 0.4 0.316 0.93 0.37 74.049 0.65 0.53 Lagged new construction Coefficient t Probability –0.665 –1.74 0.1 –0.579 –2.97 0.01 –0.681 –4.46 0 –0.822 –3.64 0 –1.094 –4.57 0 –0.47 –2.81 0.01 –0.593 –5.02 0 –1.343 –5.34 0 –0.5 –1.87 0.08 –0.125 –0.08 0.94 –0.32 –1.48 0.16 Change in unemployment Coefficient –21.285 –423.648 t –0.31 –2.67 Probability 0.76 0.02 –80.57 –721.119 –1018.914 –1.53 –0.35 –2.27 0.15 0.74 0.04 –955.54 –2.14 0.05 –33.921 –614.251 –0.42 –1 0.68 0.33 –13335.94 –3497.584 –39508.48 –1.35 –1.31 –2.83 0.2 0.21 0.01 Table A.1 (continued) 98 Change in difference of long-term minus shortterm interest rate Coefficient t Probability North North North Coast Central Eastern Los Angeles– Bay Stockton- Central Central Central San San Area Sacramento Coast Eastern Inland Bernardino Diego California 10.54 0.19 0.85 –1.845 –63.259 –1385.153 –1625.548 –0.01 –1.26 –1.28 –2 0.99 0.23 0.22 0.07 43.925 –201.784 –731.818 0.17 –2.92 –1.57 0.87 0.01 0.14 466.545 3251.318 16318.46 0.07 2.36 2.12 0.94 0.03 0.05 Change in appreciation (expected inflation) Coefficient t Probability 11.814 –46.034 0.09 –0.17 0.93 0.87 21.873 2721.669 0.31 1.93 0.76 0.07 996.303 0.94 0.37 233.906 –266.904 0.72 –2.37 0.49 0.03 733.462 1.01 0.33 –0.305 –945.606 –11197.75 0 –0.53 –1.15 1 0.6 0.26 Change in real stock prices Coefficient –325.812 –909.265 –407.326 t –1.55 –1.67 –2.81 Probability 0.14 0.12 0.01 3183.985 –2462.197 –1254.795 –456.879 2313.619 0.68 –0.8 –1.68 –2.62 1.15 0.51 0.44 0.12 0.02 0.27 –16880.39 –1723.627 –14505.81 –1.27 –0.4 –0.67 0.23 0.7 0.51 Change in real cost Coefficient t Probability –31416 –91444 –1.47 –0.79 0.16 0.44 –84595 –1496151 –1572724 –2.38 –1.08 –1.56 0.03 0.3 0.14 37061 –110850.1 –110850 0.34 –3.37 –1.07 0.74 0.01 0.3 –115914.3 –0.05 0.96 401967 –2481687 0.81 –0.85 0.43 0.4 DW (Durbin Watson stat.) 2.02 1.95 2.13 1.33 1.61 1.96 2.13 1.24 aIncome highly correlated to real cost. 2 2.03 2.1 NOTES: Cell entries in boldface are statistically significant. Cell entries in italics are of the wrong sign but not statistically significant. Cell entries that are shaded are of the wrong sign and are statistically significant. All other cell entries are statistically significant with the right sign. Table A.2 Time-Series Results Allowing for Lagged Responses 99 R2 Adjusted R2 F Test Probability Metropolitan Statistical Area North Coasta 0.67 0.48 3.57 0.02 North Central 0.78 0.69 8.31 0.00 North Eastern 0.78 0.66 6.24 0.00 Bay Area 0.76 0.64 6.10 0.00 Stockton– Saramento 0.85 0.76 9.59 0.00 Central Coast 0.77 0.63 5.77 0.00 Central Eastern 0.90 0.85 16.07 0.00 Los Angeles– Central San Inland Bernardino 0.72 0.77 0.56 0.63 4.57 5.78 0.01 0.00 San Diego California 0.76 0.83 0.62 0.76 5.42 11.48 0.00 0.00 223 670 229 4385 3675 845 311 2009 16110 5195 28256 Change in population Coefficient t Probability 0.077 2.10 0.05 0.207 5.53 0.00 0.225 2.51 0.03 0.018 0.33 0.75 0.205 2.34 0.04 0.072 2.05 0.06 0.226 5.44 0.00 0.108 1.87 0.08 0.081 0.64 0.53 –0.119 –1.50 0.16 0.035 0.45 0.66 Change in income Coefficient t Probability 0.231 2.32 0.04 (b) 0.227 1.15 0.27 0.730 1.50 0.16 0.123 2.63 0.02 –0.030 –0.29 0.78 0.225 1.72 0.11 0.372 2.68 0.02 0.005 0.03 0.98 0.390 1.37 0.19 229.561 2.31 0.03 Lagged new construction Coefficient t Probability –0.665 –1.74 0.10 –0.803 –4.67 0.00 –0.876 –3.70 0.00 –0.201 –2.02 0.06 –0.910 –3.68 0.00 –0.388 –3.37 0.01 –0.959 –5.64 0.00 –0.969 –4.63 0.00 –0.210 –0.98 0.34 –0.025 –0.180 0.86 –0.235 –1.51 0.15 Change in unemployment Coefficient t Probability –21.285 –0.31 0.76 –472.151 –4.24 0.00 –78.797 –4257.78 –2942.576 –1250.212 –1.56 –3.05 –4.85 –3.51 0.14 0.01 0.00 0.00 –98.087 –1167.395 –16303.760 –983.431 –16178.720 –1.10 –1.76 –1.68 –0.45 –1.41 0.29 0.10 0.12 0.66 0.18 Table A.2 (continued) 100 North Coasta Change in difference between long–term and short–term interest rate Coefficient 10.540 t 0.19 Probability 0.85 North Central L.1 133.363 1.14 0.27 North Stockton– Eastern Bay Area Saramento L.1 103.594 1.80 0.09 L.1 2364.337 2.55 0.02 L.1 3161.962 4.52 0.00 Central Coast L.1 475.901 2.34 0.03 Central Eastern L.1 –60.133 –0.93 0.37 Central Inland Los Angeles– San Bernardino San Diego California L.1 1269.820 3.21 0.01 L.1 7604.326 1.82 0.09 L.1 3026.018 3.20 0.01 L.1 16271.780 3.96 0.00 Change in appreciation (expected inflation) Coefficient 11.814 t 0.09 Probability 0.93 111.535 0.51 0.62 L.1 L.1 145.153 –3116.065 1.95 –2.61 0.07 0.02 1186.058 1.18 0.26 180.400 0.74 0.47 L.1 44.106 0.44 0.67 –198.192 –0.26 0.80 L.1 –7361.258 –1044.524 –1.12 –0.66 0.28 0.52 –8959.499 –1.20 0.25 Change in real stock prices Coefficient t Probability –325.812 –1.55 0.14 L.1 1744.118 –3.78 0.00 L.1 L.1 –531.920 –3566.201 –2.06 –1.18 0.06 0.26 L.1 –4676.496 –1.800 0.09 L.1 –1217.889 –2.11 0.05 L.1 –1129.649 –4.48 0.00 L.1 –2638.260 –2.10 0.06 L.1 –4665.539 –0.43 0.68 L.1 L.1 –81.903 –39150.930 –0.02 –2.36 0.98 0.03 Change in real cost Coefficient –31416.4 –64667.98 –89322.48 –146346.1 –1584431.0 –15447.310 –126930.90 –623250.00 –786015.400 t –1.47 –0.890 –2.36 –0.37 –3.84 –0.20 –4.08 –2.140 –0.46 Probability 0.16 0.38 0.03 0.72 0.00 0.85 0.00 0.05 0.65 63737.95 –758840.40 0.14 –0.36 0.89 0.73 C 473.449 1262.604 378.205 605.766 –3400.848 1047.274 813.804 6334.513 –9865.349 1421.356 –6027.783 0.99 1.9 1.41 0.14 –1.16 2.06 2.43 2.460 –0.59 0.54 –0.28 0.34 0.08 0.18 0.89 0.27 0.06 0.03 0.03 0.56 0.60 0.78 Table A.2 (continued) 101 DW (Durbin Watson stat.) North North North Stockton– Central Coasta Central Eastern Bay Area Saramento Coast Central Eastern Central Inland Los Angeles– San Bernardino San Diego California 2.02 2.26 1.97 2.05 2.15 2.00 2.37 2.05 2.10 2.08 2.08 aOriginal equation (lags did not improve the results). bIncome highly correlated to real cost. NOTES: Cell entries in boldface are statistically significant. Cell entries that are shaded are of the wrong sign and are statistically significant. All other cell entries are statistically significant with the right sign. Appendix B Householder Rate Models We use several sources of data to estimate householder rates for two types of models: the annual models and the decennial models. The annual models are based on the annual March supplements of the Current Population Survey from 1976 through 2001; the decennial models rely on the censuses of 1970, 1980, and 1990, and the Census Supplementary Survey of 2000 combined with CPS data from 1999 and 2001. The annual models allow us to examine changes that occur each year and thus to observe the effects of business cycles, but they suffer from the relatively small sample size of the CPS. Because the sample size is not large, our ability to identify trends in local housing markets and for specific demographic groups is limited. The decennial models allow for an examination of rates for specific demographic groups and metropolitan areas, as sample sizes are much larger, but do not allow for an examination of annual changes. The annual models are logistic regression models of the probability that an individual age 15 and older is a householder. We develop three types of annual models. In one type, the key independent variable is whether the individual lives in California. The model is run separately for each year 1976 through 2001. Coefficients for the intercept and the California variable are used to estimate the probability of being a householder in California and the rest of the United States. Demographic variables are added sequentially. These variables are gender, age, ethnicity, and, starting in 1994, nativity.1 The model with all demographic variables is called the model with demographic controls. The coefficient for the California variable in the model with demographic controls provides an overall measure of how different ____________ 1Before 1994, the March supplement of the CPS did not contain information on nativity. 103 California’s householder rates are from those in the rest of the country separate from the effects of the demographic factors. In a second type of annual model, data are restricted to California and the key independent variable is the year of the survey. Coefficients for the intercept and the year variable are used to estimate the probability of being a householder in California over time. Demographic variables are added sequentially. The coefficient for the year variable in the model with no demographic controls provides an estimate of temporal changes in householder rates in California. The coefficient for the year variable in the model with demographic controls provides an overall measure of how much of the change in householder rates in California can be attributed to temporal changes that are not caused by changes in the demographic composition of the state’s population. The third type of annual model is similar to the second type, except the data are restricted to the rest of the United States. The decennial models allow for an examination of patterns of change and differences between subgroups (for U.S.-born versus foreign-born Latinos) as well as for lower levels of geography (metropolitan areas). In the models comparing California to the rest of the United States, the probability of being a householder is first estimated with no controls (the only independent variable being an indicator for California residence), and then with a series of demographic controls (age, gender, ethnicity, and nativity). The models are run separately for each census year. As in the annual models, the parameter estimate for the California indicator variable provides a measure of the difference between householder rates in California and those in the rest of the nation. The California metropolitan area models are similar, with the exception of replacing the California indicator variable with a series of metropolitan area indicator variables for the 23 metropolitan areas (and an additional indicator for “rest of California”) identifiable in the CPS. Because the Census Supplementary Survey of 2000 does not provide substate geographic identifiers, only the 1999 and 2001 CPS data were used in the metropolitan area model. The CPS shows smaller differences in householder rates between California and the rest of the country than does the Census Supplementary Survey (which shows that California’s householder rates are 3.1 percentage points lower than the nation’s but 104 only 1.7 percentage points lower in the 1999 CPS and 2.3 percentage points lower in the 2001 CPS). Throughout the report, we have converted the logistic regression results to rates or probabilities. All of the parameter estimates are based on weighted regressions, although standard errors and confidence intervals are based on unweighted regressions. We use weighted regressions because our goal is to estimate householder rates for the entire population. For example, the results of the California annual model (a weighted logistic regression with the dependent variable being an indicator of whether the adult is a householder and the only independent variables being year dummies for 1976 through 2000) are the same as a simple tabulation of the proportion of California’s adults who are householders for each year from 1976 through 2001. 105 Appendix C Panel Data Analysis of Housing Supply In this appendix we follow the same basic economic model of housing supply used in Chapter 3 (and also described in more detail in Appendix B). We take advantage of the number of separate counties in the state to analyze housing supply in California from 1972 to 2001 in a way that permits us to compare the number of housing units produced1 in each county to the number predicted by our model. We use the same variables as the time-series analysis but predict the number of new housing units based on a common response across regions to changes in the explanatory variables. As before, the variables are population, personal income, previous construction, unemployment, credit market conditions, inflation expectations, stock prices, and construction costs. Factors specific to each county—such as the amount of developable land remaining and the presence of growth controls and other regulatory constraints on housing— are controlled for with a variable representing each county. This is the same approach taken in many studies of the national housing market, so it is not unreasonable to assume that the underlying relationship between, for example, construction costs and housing production is the same regardless of which county we are examining. This approach is more restrictive than the approach taken in the time-series analysis but it produces estimates of supply shortfalls that are comparable to the housing unit figures estimated in the householder analysis in Chapter 3. Table C.1 reports the results of two panel regressions. The first column shows the log of new housing units regressed on the log of total population and the second column shows the log of new units regressed on the log of adult population. The first column specification is more ____________ 1We actually use data on the number of housing permits issued in each year, but for most counties, in most years, the share of permits that are produced exceeds 90 percent. 107 Table C.1 Panel Regression Results (t-values in parentheses) Log (county population) Log (county adult population) Real county income per capita County unemployment rate Difference in long- and short-term interest rates Expected inflation Real stock prices Real county construction costs County fixed effects Constant R2 No. of observations (58 counties) Model 1 .0249 (0.18) — –3.98 e–07 (–0.86) –0.0367 (–6.45) Model 2 — 0.2887 (2.41) –5.46 e–07 (–1.18) –0.0361 (–6.41) 0.0262 (1.83) –0.0069 (–0.39) –0.0464 (–3.79) –0.0109 (–6.87) Suppressed 9.0286 (5.96) 0.0731 1,392 0.0308 (2.14) –0.0005 (–0.03) –0.0461 (–3.93) –0.0123 (–8.27) Suppressed 6.4946 (5.42) 0.7683 1,392 consistent with that from the time-series models and the second column better reflects the demographic drivers of housing demand as discussed in the householder analysis. The model incorporating the counties’ adult, rather than total, population changes is a much better fit and gives results that are more consistent with those found both in the time-series and householder analyses.2 Total population change is not statistically significant whereas change in the adult population is, as we expected. Although it is not significant, the sign on per capita income is consistent with those hypotheses that predict more resistance to new housing as a community’s income levels increase. When unemployment rates rise in a county, new housing supply falls. Easier credit conditions, measured as the difference between long and short rates, are associated with an increase in housing production. Increases in either inflation expectations or stock market returns have negative signs—implying lower production—but only stock market returns were significant. Last, increased construction costs are strongly associated with lower levels of housing production. Because we are using lagged terms, and because data for some variables are missing in the earliest years of the period we are interested ____________ 2Note that because of the structure of the fixed-effects approach, the lagged value for units produced is omitted from the models. 108 in, we effectively predict housing supply at the county level for the period from 1973 to 1999. The total number of new housing units produced is explained fairly well by the model incorporating adult population rather than total population. Our major result in this analysis is that slightly more housing was produced in the 1980s business cycle than was predicted in the model, and significantly less housing was produced in the 1990s than predicted. Of the explanatory variables included in the model, only inflation expectations were not significant in at least some of the alternative forms that were examined.3 The only issue about interpreting the results in a fixed-effect model is that the net over- and undersupply over the entire period is constrained to be zero, which may be too binding a constraint even though the analysis covers almost three complete housing cycles. The housing oversupply at the state level in the 1980s cycle was approximately 4 percent of the total number of units predicted; the undersupply in the 1990s cycle was 15–17 percent of the predicted total (Table C.2)—at least through 1999. (The numeric comparisons are made from cycle trough to cycle trough to capture all of the new construction within each building cycle, but the results are similar when we simply compare the number of units predicted by decade.) The estimated shortfall of 15–17 percent of the predicted new supply over the decade works out to be a deficit of approximately 17,000 housing units per year for California as a whole. Most studies of housing dynamics use the change in total population as the measure of population change, and we saw in Chapter 3 that this will overpredict the amount of new housing required. The detailed demographic data needed for the analyses in Chapter 3 are available only in census years, but we can use counties’ adult population on an annual basis for these models to better handle the issue of family size. It should also be noted that most of the discussions of the housing situation in California have described an undersupply on the order of 40–50 percent ____________ 3In addition to several panel regressions, we estimated the model as a cross-section with all years’ data being estimated as a function of first differences to obtain estimates divorced from any time dimension. We also examined the sensitivity of the results to changing the starting or ending dates by one year. In all of these cases, most variables showed the same pattern of coefficient magnitudes and statistical significance. 109 Table C.2 Results of County-Level Analysis of Annual Housing Production (All Units) Cross-section analysis Panel analysis Cross-section analysis Panel analysis Cross-section analysis Panel analysis Actual Predicted 1973–1982 1,759,855 1,738,761 1,542,991 1,523,094 1983–1993 2,179,996 2,093,717 2,179,996 2,095,823 1994–1999 659,543 768,944 659,543 758,639 Over-/ Undersupply (%) 1.2 1.3 4.0 3.9 –16.6 –15.0 of the required level. Our estimate is less than a third that size—even without the additional demographic characteristics that were used in the householder analysis. Using another standard, a comparison of national and state housing production from 1991 to 2000 yields an aggregate shortfall of approximately 30 percent—assuming that California should have produced housing at the same rate of growth that the nation did. If we allow for the fact that California emerged from recession two years later than the nation did, the implied shortfall shrinks to only 10 percent. In Figure C.1, we show the difference between the number of new housing units predicted by our panel regressions and the actual production. There seems to have been at least some oversupply at the 1980s peak, but the shortfall in the 1990s is more evident and more consistent from year to year. Even the predicted peak is at an annual production figure that is 46 percent below the 1980s peak and 38 percent below the 1970s peak (although it is not yet clear whether this cycle will eventually have the kind of clearly defined peak that has characterized previous cycles). It should not be surprising that the San Francisco Bay Area has the largest shortfall in new housing, or that the shortfall seems to have accelerated during the Internet boom. It was mentioned above that the percentage changes in real per capita income and real home prices in the Bay Area matched quite closely, but even when income is accounted for 110 Number of units (1000s) 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 350 300 Actual Predicted 250 200 150 100 50 0 Figure C.1—Comparison of Actual and Predicted Annual Housing Production in California, 1973–1999 there is a significant and growing gap. For the current cycle from 1994 to 1999, the shortfall is estimated to be 21 percent, compared to an overproduction of 3 percent in the 1980s. San Diego also shows a consistent gap in housing production in the 1990s but it also had a sizable overproduction during the 1980s peak. The shortfall from 1994 onward is 20 percent, although production rose sharply after 1996; overproduction in the 1980s was 6 percent. These numbers are consistent with the notion that the 1980s was characterized by at least some level of overbuilding in response to the tax incentives for real estate development as well as similar patterns in construction employment relative to that in other sectors. The Sacramento area grew rapidly in the 1990s and, in terms of employment and output change, it suffered only a mild recession in the early 1990s. It is the only major metropolitan area to have both actual and predicted housing production that is approaching the peak levels seen in the previous cycles. The region did not see overbuilding in either of the last two cycles and—despite the rapid rate of housing production since 1995—it still had an estimated shortfall of almost 5 percent in the last five years of our forecast period. So far in this business cycle, it still 111 seems to be in relatively strong shape in terms of both employment and new construction. It is the last, and largest, metropolitan area that has the greatest potential effect on the statewide housing balance. By some measures, such as construction employment, Los Angeles experienced something of a construction bubble in the 1980s and its deflation was compounded by the collapse of the aerospace industry. According to our analysis, there was a slight overproduction of housing in the 1970s and a corresponding shortfall in the early 1980s recession. In the growth phase of the 1980s business cycle, the region once again produced more housing than the model predicted—almost 69,000 units from 1983 to 1990—and undersupplied housing during the deep recession and during the current growth period. Despite the persistent shortfall of housing in the greater Los Angeles region throughout the 1990s, the region essentially balanced the prior excess supply with the current shortfall. By examining the actual and predicted values for each county in the region, it appears that the largest imbalance comes from Los Angeles County itself. Overproduction there was a relatively small 10 percent in the 1980s and underproduction in the 1990s amounted to almost 30 percent of the predicted new supply. New 1980s construction in the other counties in the region outpaced the predicted values by 3 percent in Orange County, 10 percent in Riverside and San Bernardino, and 53 percent in Ventura. In the 1990s, the shortfalls amounted to 15 percent in Riverside and San Bernardino and 10 percent in Orange and Ventura Counties. Some of the “oversupply” in Ventura County in the 1980s can probably be explained by the evolving commuting connections with its much larger neighbor county—if there is more developable land in an adjacent county, that is where construction is more likely to occur in the absence of regulatory barriers. The integrated labor markets of these counties imply that any short-run imbalances in one county may reflect the relative conditions of the individual county labor and land markets (in addition to receptivity to development). However, it is difficult to square sizable shortfalls during the 1990s with the fact that median house prices in Los Angeles County declined by 26 percent from 1989 to 2000—even after correcting for inflation. 112 Finally, the results of the model without fixed effects are compared to two important factors—a region’s ability and willingness to accommodate growth. The first factor is the amount of developable land within the county; the second is community resistance to growth. Table C.3 displays these three numbers for the regions that had sufficient responses on a city manager survey regarding attitudes about growth: deviation between predicted and actual housing production from 1990 to 1999; remaining developable land as a share of total (from Landis et al., 2000), and a population-weighted index of city managers’ appraisals of their city’s resistance to new development (from a comprehensive survey conducted by Lewis and Neiman, 2000, with 1 = least resistance and 5 = most resistance). It is hard to divine a strong pattern among these five observations, but it is interesting to note that the region with the largest shortfall of new housing (San Francisco Bay Area) is the one with the greatest relative share of available land but the strongest political resistance to growth. This observation is consistent with a number of studies that have shown a relationship between growth barriers and home supply and prices.4 The region with the next largest shortfall (San Diego) is more accommodative of growth but has a smaller supply of surplus land. Table C.3 Housing, Developable Land, and Resistance to Growth (1990–1999) Actual–predicted new units, % Developable land as a share of total, % Resistance to growth San Francisco Bay Area Los Angeles, Orange, Ventura Counties –21 –7 44.9 22.3 2.58 2.34 Riverside, San Bernardino Counties –15 20.8 2.30 San Diego County –20 28.9 2.18 Sacramento Region –5 59.8 2.09 ____________ 4See Malpezzi (2001) for an overview of the major findings from the literature about the effects of growth control and zoning on housing. 113 References Blomquist, Glenn, Mark Berger, and John Hoehn, “New Estimates of the Quality of Life in Urban Areas,” American Economic Review, Vol. 78, 1988, pp. 89–107. Boldin, Michael D., “Econometric Analysis of the Recent Downturn in Housing Construction: Was It a Credit-Crunch? Federal Reserve Bank of New York, Research Paper # 9332, December 1993. Browne, Lynn Elaine, “National and Regional Housing Patterns,” Federal Reserve Bank, New England Economic Review, July/August 2000. California Budget Project, Locked Out 2002: California’s Affordable Housing Crisis Continues, Sacramento, California, 2002. California Department of Housing and Community Development, “Guidelines for the Jobs Housing Balance Incentive Grant Program,” 2003, available at http://www.hcd.ca.gov/ca/jhbig/ jhbig20030123guide.pdf. Carlstrom, Charles, and Christy D. Rollow, “Regional Variations in White-Black Earnings,” Federal Reserve Bank of Cleveland, Economic Review, Quarter 2, 1998. Center for Continuing Study of the California Economy, California Economic Growth 1993, 1993, available at http://www.ccsce.com/ about_ccsce.html. Collins, Susan, Barbara Lipman, and Sid Groeneman, “Experience and Expectations: The Home Price Predictions of Owners and Renters,” National Association of Realtors, Washington, D.C., August 1992. Dresch, Marla, and Steven M. Sheffrin, Who Pays for Development Fees and Exactions? Public Policy Institute of California, San Francisco, California, 1997. 115 Fratantoni, Michael, and Scott Schuh, “Monetary Policy, Housing Investment, and Heterogeneous Regional Markets,” May 2000. Fulton, William, Rolf Pendall, Mai Nguyen, and Alicia Harrison, “Who Sprawls Most? How Growth Patterns Differ Across the U.S.,” Brookings Institution, Washington, D.C., 2001. Fulton, William, Paul Shigley, Alicia Harrison, and Peter Sezzi, Trends in Local Land Use Ballot Measure, 1986–2000, Solimar Research Group, Ventura, California, 2000. Gabriel, Stuart A., Joe P. Mattey, and William L. Wascher, “Compensating Differentials and Evolution of the Quality-of-Life Among U.S. States,” Working Papers in Applied Economic Theory 96-07, Federal Reserve Bank of San Francisco, 1996. Gabriel, Stuart A., Joe P. Mattey, and William L. Wascher, “House Price Differentials and Dynamics: Evidence from the Los Angeles and San Francisco Metropolitan Areas, FRBSF Economic Review, No. 1, 1999. Gavin, William T., and Rachel J. Mandal, “Forecasting Inflation and Growth: Do Private Forecasts Match Those of Policymakers? Federal Reserve Bank of St. Louis, September 28, 2000. Glaeser, Edward L., Joseph Gyourko, and Raven Saks, Why Is Manhattan So Expensive? Regulation and the Rise in House Prices, Harvard Institute of Economic Research Discussion Paper 2020, November 2003. Glickfeld, Madelyn, and Ned Levine, “The New Land Use Regulation ‘Revolution’: Why California’s Local Jurisdictions Enact Growth Control and Management Measures,” Graduate School of Architecture and Urban Planning, University of California, Los Angeles, California, June 1990. Glickfeld, Madelyn, and Ned Levine, “Regional Growth . . . Local Reaction: The Enactment and Effects of Local Growth Control and Management Measures in California,” Lincoln Institute of Land Policy, Cambridge, Massachusetts, 1992. 116 Grebler, L., and S. J. Maisel, “Determinants of Residential Construction: A Review of Present Knowledge,” in Commission on Money and Credit, Impacts of Monetary Policy, Prentice-Hall, Englewood Cliffs, New Jersey, 1963, pp. 496–505. Gyourko, Joseph, and Joseph Tracy, “The Structure of Local Public Finance and the Quality of Life,” Journal of Political Economy, Vol. 99, No. 4, 1992, pp. 774–806. Hamilton, Bruce W., and Robert M. Schwab, “Expected Appreciation in Urban Housing Markets,” Journal of Urban Economics, Vol. 18, 1985, pp. 103–118. Hanushek, Eric A., and John M. Quigley, “The Determinants of Housing Demand,” Research in Urban Economics, Vol. 2, 1982, pp. 221–242. Hoehn, John, Mark Berger, and Glenn Blomquist, “A Hedonic Model of Interregional Wages, Rents, and Amenity Values,” Journal of Regional Science, Vol. 27, 1987, pp. 605–620. Kaiser, Ronald W., “Using Capital Markets’ Value Cycles in Allocating to Real Estate vs. Stocks or Bonds,” Journal of Real Estate Portfolio Management, Vol. 5, No. 1, 1999. Klein, L. R., The Keynesian Revolution, 2nd ed., Macmillan, New York, 1966, p. 233. Krainer, John, “House Price Bubbles,” Federal Reserve Bank of San Francisco Economic Letter, No. 2003-06, 2003. Landis, John D., “Opening the Doors to Infill Housing,” PowerPoint presentation, Department of City and Regional Planning, UC Berkeley, 2003. Landis, John D., Michael Smith-Heimer, Michael Larice, Michael Reilly, Mary Corley, and Oliver Jerchow, Raising the Roof: California Housing Development Projections and Constraints 1997–2020, Statewide Housing Plan, Department of Housing and Community Development, Sacramento, California, 2000. 117 Leamer, Edward E., “Bubble Trouble? Your Home Has a P/E Ratio Too,” UCLA Anderson Forecast Report, 2002. Leamer, Edward E., “Update June 2, 2003; Bubble Trouble? Your Home Has a P/E Ratio Too,” UCLA Anderson Forecast Report, 2003. Levine, Ned, “The Effects of Local Growth Controls on Regional Housing Production and Population Redistribution in California,” Urban Studies, Vol. 36, No. 12, 1999, pp. 2047– 2068. Lewis, Paul, California’s Housing Element Law: The Issue of Local Noncompliance, Public Policy Institute of California, San Francisco, California, 2003. Lewis, Paul, and Elisa Barbour, California Cities and the Local Sales Tax, Public Policy Institute of California, San Francisco, California, 1999. Lewis, Paul, and Elisa Barbour, “Development Priorities in California Cities: Results from a PPIC Survey,” Occasional Paper, Public Policy Institute of California, San Francisco, California, 1998. Lewis, Paul, and Max Neiman, “Residential Development and Growth Control Policies: Survey Results from Cities in Three California Regions,” Occasional Paper, Public Policy Institute of California, San Francisco, California, 2000. Maisel, S. J., “A Theory of Fluctuations in Residential Construction Starts,” American Economic Review, Vol. 53, No. 3, June 1963. Malpezzi, Stephen, “NIMBYs and Knowledge: Urban Regulation and the ‘New Economy,’” working paper, Center for Urban Land Economic Research, University of Wisconsin, WisconsinMadison, 2001. Mayer, Christopher J., and C. Tsuriel Somerville, “Using Empirical and Theoretical Models of Housing Supply,” Working Paper No. 9612, Federal Reserve Bank of Boston, December 1996. Moller, Rosa, Hans Johnson, and Michael Dardia, What Explains Crowding in California? California Research Bureau, Sacramento, California, February 2002. 118 Myers, Dowell, Analysis with Local Census Data, Academic Press, San Diego, California, 1992. Myers, Dowell, and Julie Park, “The Great Housing Collapse in California,” FannieMae Foundation, 2002. Myers, Dowell, William C. Baer, and Seong-Youn Choi, “The Changing Problem of Overcrowded Housing,” Journal of the American Planning Association, Vol. 62, 1996, pp. 66–84. Roback, Jennifer, “Wages, Rents, and the Quality of Life,” Journal of Political Economy, Vol. 90, 1982, pp. 1257–1278. Rosen, Sherwin, “Wage-Based Indexes of Urban Quality of Life,” in P. Mieszkowski and M. Strazheim (eds.), Current Issues in Urban Economics, John Hopkins University Press, Baltimore, Maryland, 1979, pp. 74–104. Stover, Mark Edward, and Charles Leven, “Methodological Issues in the Determination of the Quality of Life in Urban Areas,” Urban Studies, Vol. 29, No. 5, 1992, pp. 737–754. U.S. Department of Housing and Urban Development, U.S. Housing Market Conditions, 3rd Quarter 2003, 2003, available at http:// www.huduser.org/periodicals/ushmc/fall03/ushmc_fall03.pdf. Wasserman, Miriam, “Appreciating the House. Housing as an Investment,” Federal Reserve Bank of Boston, Regional Review, 1998. Weintraub, Daniel, “The Prices for Housing Soar—Along with Demand,” Sacramento Bee, December 1, 2002. 119 About the Authors HANS JOHNSON Hans Johnson is a demographer at the Public Policy Institute of California. His research interests include international and domestic migration, population estimates and projections, and state and local demography. Before joining PPIC as a research fellow, he was the senior demographer at the California Research Bureau, where he conducted research for the state legislature and the governor’s office on population issues. He has also worked as a demographer at the California Department of Finance, specializing in population projections. He holds a Ph.D. in demography from the University of California, Berkeley. ROSA MOLLER Rosa M. Moller is a senior economist at the California Research Bureau, where she has written and co-authored reports and technical briefs on a wide range of issues for the legislature and state government. Before joining the California Research Bureau in 1991, she worked on various socioeconomic and demographic studies for the United Nations, including the Economic Commission for Latin America. She received her Ph.D. in economics from the University of California, Davis, and a B.A. and M.A. in sociology from the Catholic University of Chile in Santiago. MICHAEL DARDIA Michael Dardia is vice president at The SPHERE Institute. His research interests cover workforce development, redevelopment, regional economics, and public finance and taxation issues. Before joining SPHERE, he was a research fellow at the Public Policy Institute of California and a consultant to the RAND Corporation. He is a member of the California Workforce Investment Board’s Economic Advisory Council and has participated in numerous other state and local government advisory groups. He edits the growth and employment series of the California Policy Review and writes the regional highlights for the Bay Area Economic Pulse. He earned his Ph.D. in policy analysis from the RAND Graduate School. 121 Related PPIC Publications Who Pays for Development Fees and Exactions? Marla Dresch and Steven M. Sheffrin Planned Developments in California: Private Communities and Public Life Tracy M. Gordon California’s Housing Element Law: The Issue of Local Noncompliance Paul G. Lewis Cities Under Pressure: Local Growth Controls and Residential Development Policy Paul G. Lewis and Max Neiman PPIC publications may be ordered by phone or from our website (800) 232-5343 [mainland U.S.] (415) 291-4400 [Canada, Hawaii, overseas] www.ppic.org 123" ["post_date_gmt"]=> string(19) "2017-05-20 09:36:19" ["comment_status"]=> string(4) "open" ["ping_status"]=> string(6) "closed" ["post_password"]=> string(0) "" ["post_name"]=> string(8) "r_304hjr" ["to_ping"]=> string(0) "" ["pinged"]=> string(0) "" ["post_modified"]=> string(19) "2017-05-20 02:36:19" ["post_modified_gmt"]=> string(19) "2017-05-20 09:36:19" ["post_content_filtered"]=> string(0) "" ["guid"]=> string(50) "http://148.62.4.17/wp-content/uploads/R_304HJR.pdf" ["menu_order"]=> int(0) ["post_mime_type"]=> string(15) "application/pdf" ["comment_count"]=> string(1) "0" ["filter"]=> string(3) "raw" ["status"]=> string(7) "inherit" ["attachment_authors"]=> bool(false) }