Donate
Independent, objective, nonpartisan research

R 499HJR

Authors

R 499HJR

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_499HJR.pdf" ["wpmf_size"]=> string(6) "161736" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(117980) "The Basic Skills of Welfare Recipients: Implications for Welfare Reform Hans P. Johnson Sonya M. Tafoya 1999 Copyright © 1999 Public Policy Institute of California, San Francisco, CA. All rights reserved. PPIC permits short sections of text, not to exceed three paragraphs, to be quoted without written permission, provided that full attribution is given to the source and the above copyright notice is included. Foreword This report is the fourth in a series of studies undertaken by PPIC to understand the consequences of the Personal Responsibility and Work Opportunity and Reconciliation Act of 1996. The authors, Hans P. Johnson and Sonya M. Tafoya, analyze the National Adult Literacy Survey to assess the basic skills of adults on welfare and the likelihood that welfare recipients will be able to find and hold full-time jobs, given their educational background and skill level. In spite of the remarkable reduction in welfare rolls since the reform legislation of 1996, and the sustained growth of the California economy, the findings do not augur well for the poor still on the rolls. Welfare recipients in California are found to have substantially lower basic skills than other adults in the state and the nation, even when compared to other adults with the same level of education. Why, then, are the rolls shrinking and applications for assistance continuing to decline? The authors do not have a direct answer, but they do find that over 50 percent of the adults in California who have basic skills and iii demographic characteristics similar to welfare recipients, but who are not receiving welfare, work at least part time. Most hold jobs intermittently, and the jobs are low-paying. These findings suggest that some welfare recipients could be similarly employed. The authors wave a flag of caution, however, and note that any softening of the economy for a sustained period could hit these workers—the ones with the lowest levels of basic skills—the hardest. There is no major reform of public policy that has come under closer scrutiny than welfare reform. For those who cheer the strong economy and the declining caseload, there are others who see a grim tale of poorly educated and undernourished children whose parents will return to the rolls with the first downturn in the California job market. The authors suggest that improving the basic skills of welfare recipients, although difficult, merits public policy attention; some contact with the job market, however unsteady, is a realistic option for some, if not all, of those currently receiving assistance. Future publications by PPIC will explore this welfare/work relationship in further detail. David W. Lyon President and CEO Public Policy Institute of California iv Summary Large reductions in welfare caseloads have led many to conclude that welfare reform initiated by the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 has been a success. In California, for example, the number of families receiving welfare declined 22 percent from January 1997 to September 1998. Although impressive and certainly one indicator of success, this decline has occurred during a period of strong economic growth. The ability of welfare recipients to transition from welfare to work during a recession is less certain. Even the large decline in welfare caseloads during this current period of economic growth is not necessarily due to welfare recipients’ success in finding work. Some of the decline is due to a drop in the number of applications for welfare. Additionally, some may have left the welfare rolls but not to work, relying instead on friends and family for financial support. Many questions remain about the ability of welfare recipients to find work and the quality of jobs that they do find. v This report addresses the prospects of the nation’s and particularly California’s welfare population as it faces a new welfare system of work requirements, sanctions, and time limits. The report describes the basic skills of welfare recipients and evaluates the employment implications of such skills. In this report, we seek to answer three questions: • How do the basic skills of welfare recipients differ from those of other adults in general and workers in particular? • How much of the gap in skills between workers and welfare recipients can be explained by educational attainment? • What are the labor force status and characteristics of jobs held by persons with skills and characteristics similar to the skills and characteristics of welfare recipients? To answer these questions, we use data from the National Adult Literacy Survey. This nationally representative survey, conducted in 1992, includes a test of basic skills. It assessed the ability of respondents to perform tasks commonly encountered in daily living (e.g., understanding the argument in a newspaper editorial) and tasks that could be encountered in the workplace (e.g., completing a job application). We use several methods to answer the questions posited above, from simple descriptive statistics to logistic regression. These are our major findings: • Welfare recipients have substantially lower basic skills than other adults. In California, for example, almost 80 percent of welfare recipients have either low or very low basic skills, compared to 34 percent of full-time workers in the state. With such poor basic skills, most welfare recipients have difficulty successfully completing tasks commonly encountered in daily living. For example, the average welfare recipient in California has difficulty vi following simple written directions to perform a single mathematical operation (such as addition) using numbers easily located in the text. • Differences in educational attainment between welfare recipients and other adults explain some of the skills gap but not the majority of the gap. About 40 percent of the difference in basic skills scores between welfare recipients and other adults can be attributed to lower educational attainment levels of welfare recipients. However, welfare recipients have substantially lower basic skills than other adults with the same level of education. • We have some cause for optimism: In California, a substantial proportion (58 percent) of adults with basic skills and demographic characteristics similar to welfare recipients are working at least part time. • We also have cause for concern: The jobs held by people whose basic skills are similar to those of welfare recipients are characterized by low wages, intermittent employment, and less than full-time hours. In California, only one-third of adults with basic skills similar to welfare recipients were employed full time year-round. Although the ultimate success of welfare reform will be determined as recipients encounter time limits, the social and individual costs of failure require that we anticipate and respond to potential impediments to success before that time. Our findings suggest that although many welfare recipients can and will find work, a substantial proportion lack the skills for successful integration into the labor force. California faces a greater challenge than most other states: The basic skills of welfare recipients in California are lower than those of welfare recipients in the rest of the nation, and the skills gap between workers and welfare recipients is greater in California than in the rest of the nation. vii The low skills of welfare recipients are not easily amenable to change. Many welfare recipients have graduated from high school, yet even after a dozen years of schooling they are unable to perform simple tasks commonly encountered in the workplace. The track record of training programs is not especially promising. We are also skeptical that on-thejob training will provide these skills—especially considering the types of jobs that welfare recipients might hold. The difficulty in improving the basic skills of welfare recipients does not mean that we should not try. It does mean that we need to be realistic about the costs of providing meaningful training and of improving basic skills. Training programs for improving basic skills need to be critically assessed, with their costs weighed against their benefits. The most promising programs seem to be those that focus on employment and that integrate real job situations into the vocational and basic skills training. Ultimately, we might need to accept that a substantial portion of welfare recipients will continue to need some form of income support, either because their very low skills make them virtually unemployable or because the work they find is of such low quality (and quantity) that they are still living in poverty. viii Contents Foreword ..................................... Summary..................................... Figure and Tables ............................... Acknowledgments ............................... Acronyms .................................... iii v xi xiii xv 1. INTRODUCTION ........................... Historical Welfare Context ....................... California Welfare Context....................... Scope of This Research ......................... 2. DATA AND METHODOLOGY .................. The National Adult Literacy Survey ................. Study Methods .............................. 3. FINDINGS ................................ General Findings ............................. The Skills Gap: Basic Skills of Welfare Recipients and Workers ............................... Educational Attainment and the Basic Skills Gap......... Estimating the Employment Prospects of Welfare Recipients .............................. Labor Force Status .......................... Earnings ................................. 1 3 5 8 11 11 13 17 17 19 26 31 36 38 ix Occupations .............................. Industry ................................. Summary of Findings on Employment Prospects ....... 4. POLICY IMPLICATIONS ...................... Appendix A. The National Adult Literacy Survey: Examples of Tasks and Difficulty Levels ............................. B. The National Adult Literacy Survey: Sampling Design and Scoring ................................... C. NALS and Labor Force Outcomes .................. D. Regressions on Quantitative Basic Skills Score .......... E. Determination of Welfare Counterparts .............. References .................................... About the Authors ............................... 39 41 42 45 49 53 57 61 67 71 77 x Figure 3.1. Percentage Change in AFDC Families, by State, 1993– 1997 .................................. 25 Tables 3.1. Average Basic Skills of U.S. Adults ............... 3.2. Average Basic Skills Scores, by State and for the Nation .. 3.3. Distribution of Basic Skills Scores, by State and for the Nation ................................. 3.4. Average Basic Skills, by Welfare Status: United States ... 3.5. Distribution of Basic Skills, by Welfare Status: United States .................................. 3.6. Average Basic Skills, by Welfare Status: California Compared to the Rest of the Nation .............. 3.7. Distribution of Basic Skills, by Welfare Status: California and the Rest of the Nation ............. 3.8. The Skills Gap and Educational Attainment Levels in the Nation ................................ 18 19 20 21 22 22 23 27 xi 3.9. The Skills Gap and Population Composition Effects in the Nation .............................. 3.10. The Skills Gap and Educational Attainment Levels in California ............................... 3.11. The Skills Gap and Population Composition Effects in California ............................... 3.12. Characteristics of Welfare Recipients, by Work Status ... 3.13. Characteristics of Welfare Recipients and Welfare Counterparts ............................. 3.14. Labor Force Status of Welfare Counterparts and Other Adults ................................. 3.15. Earnings of Welfare Workers, Welfare Counterparts, and Other Adults ............................. 3.16. Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Currently Working Full Time ......... 3.17. Earnings of Welfare Workers with Very Low Basic Skills, Welfare Counterparts with Very Low Basic Skills, and Other Non-Welfare Adults .................... 3.18. Occupational Profile of Welfare Workers, Welfare Counterparts, and Other Adults ................. 3.19. Industrial Profile of Welfare Workers, Welfare Counterparts, and Other Adults ................. B.1. NALS Response Rates ....................... C.1. Wage Equations Using Quantitative Basic Skills and Education as Dependent Variables ............... D.1. Variables Used in Regressions to Evaluate Basic Skills Gap................................... D.2. Descriptive Statistics and Regression Results ......... E.1. Variables Used in Logit Regressions to Identify Welfare Counterparts ............................. E.2. Logistic Regressions Used to Identify Welfare Counterparts ............................. 29 30 30 33 35 36 39 40 41 42 43 54 58 62 63 68 69 xii Acknowledgments We are grateful to Hans Bos, Steve Reder, Margaret O’Brien-Strain, Michael Teitz, and Kim Rueben for their thoughtful reviews of an earlier version of this report. Andrew Kolstad was generous in sharing his time and expertise on the National Adult Literacy Survey. The California Research Bureau of the California State Library generously provided us with the California State Adult Literacy Survey. David Illig and others at the California Research Bureau provided helpful comments when this research was at various stages of development. Once again, Gary Bjork and Joyce Peterson proved to be the best of stylistic reviewers and Patricia Bedrosian an excellent editor. Although this report reflects the contributions of many people, the authors are solely responsible for its content. xiii Acronyms ADC AFDC AFDC-UP CalWORKs DHHS ETS FSA GAIN GED JOBS NALS NCES PRWORA SALS Aid to Dependent Children Aid to Families with Dependent Children Aid to Families with Dependent Children— Unemployed Parent California Work Opportunity and Responsibility to Kids Department of Health and Human Services Educational Testing Service Family Support Act Greater Avenues for Independence Program General Equivalency Diploma Job Opportunities and Basic Skills Training Program National Adult Literacy Survey National Center for Education Statistics Personal Responsibility and Work Opportunity Reconciliation Act of 1996 State Adult Literacy Survey xv SSI TANF WIN Supplemental Security Income Temporary Assistance to Needy Families Work Incentive Program xvi 1. Introduction Since the inception of welfare programs in the United States, one primary goal of policymakers has been to reduce the number of welfare recipients. Particularly over the past few decades, numerous programs have been devised to improve the employment prospects of welfare recipients and lead them forward to self-sufficiency. These welfare-towork programs have focused variously on job searches, unpaid work experience, monetary incentives (e.g., earnings disregards),1 classroom training, and remedial education. The latest and most dramatic incarnation of welfare reform, operating partly under the assumption that welfare recipients lack the proper motivation to work, requires welfare recipients to work after a certain amount of time on aid and limits the total amount of time an individual can receive assistance. The success of welfare reform largely depends on moving people from welfare ____________ 1Earnings disregards provide monetary work incentives for welfare recipients. Rather than reducing welfare benefits by the full amount of earnings, under earnings disregard programs, some welfare recipients who work are able to continue to receive full or only partially reduced benefits. 1 to work. However, it also depends on the duration and wages of that work. Income support programs might still be necessary if one goal of welfare reform is to lift welfare recipients out of poverty. Ascertaining the ability of welfare recipients to find work and determining the quality of the jobs they find are essential to assessing the effectiveness of welfare reform. However, projecting such labor force outcomes is difficult. It is well known that welfare recipients are less educated and less skilled than other adults in the labor force (see, for example, Burtless, 1995; Barton and Jenkins, 1995; MaCurdy and O’Brien-Strain, 1997; Pavetti, 1997; and Reder and Wikelund, 1994). However, it is not clear to what extent these low levels of skills are an impediment to employment. In this report, we use data on basic skills from a national survey of adults to determine the basic skills gap between welfare recipients and other adults and to estimate the employment implications for welfare recipients. We conduct our analyses for both California and the United States. California is an important state to single out for the study of basic skills as they relate to welfare recipients. As of September 1997, California was home to 23 percent of the nation’s welfare (Temporary Assistance to Needy Families—TANF) recipients (U.S. DHHS, 1998). This group, totaling 2,225,893 people, consisted of 663,396 adults and 1,559,497 children. Eighteen percent of families in this group were twoparent families. California spent $4.8 billion on its welfare program in fiscal year 1996–97 (California Department of Social Services, 1998). If the nation is to successfully reform welfare, California, with its large and diverse population, must be considered as a crucial factor in the equation. 2 Historical Welfare Context Aid to Dependent Children (ADC)2 was created in 1935 to ensure income security for mothers who had lost the income of a spouse as a result of death or disability (O’Neill and O’Neill, 1997). Work requirements and work skills, topics now central to welfare policy debates, were not among initial policy concerns, as mothers were not expected to work. However, in the 1960s, when women from every social class began to enter the labor force in large numbers, support for policies that allowed parents to receive public assistance rather than working to support their children began to decline (Jansson, 1997). In 1962, the first federally sponsored work requirement was instituted (Brock, Butler, and Long, 1993). Though small, it was followed by larger federal programs that stressed training and work requirements, thereby introducing the basic skills of welfare recipients as a factor in formulating welfare policy. In 1967, Congress created the Work Incentive Program (WIN), which introduced mandatory training programs for some welfare recipients.3 The program was intended to “reorient welfare toward work” (Gueron and Pauly, 1991). Supervised job searches and unpaid work experience were the main activities of the programs, and earnings disregards were instituted to encourage recipients to work their way off welfare. In practice, however, low enrollment and lack of adequate funding meant that the hope of “reorienting welfare to work” went unfulfilled (O’Neill and O’Neill, 1997; Friedlander, Greenberg, and ____________ 2ADC was the precursor to Aid to Families with Dependent Children (AFDC). 3Those recipients were heads of single-parent AFDC families without preschoolaged children, and heads of two-parent AFDC-UP (-unemployed parent) families (Friedlander, Greenberg, and Robins, 1997). 3 Robins, 1997). AFDC caseloads did not decline and welfare rolls swelled in the early 1970s (O’Neill and O’Neill, 1997; Majority Staff of the Committee on Ways and Means, 1996). In an effort to encourage innovative and cost-effective programs, the Omnibus Budget Reconciliation Act (1981) granted states the flexibility to design their own WIN demonstration projects (O’Neill and O’Neill, 1997; Friedlander, Greenberg, and Robins, 1997; also, see Gueron and Pauly, 1991, for a review of these projects). Based on the most promising of these demonstration projects, the Family Support Act (FSA) was passed in 1988 and the Job Opportunities and Basic Skills Training Program (JOBS) was established to replace WIN in providing federal funds for welfare-to-work program services (Gueron and Pauly, 1991). The FSA stressed the primary responsibility of parents to financially support their children, without changing the entitlement nature of AFDC. JOBS broadened the population of recipients mandated to participate in training and work, increased sanctions for nonparticipation, and committed federal funds to remedial and basic education in welfare-to-work programs (Friedlander, Greenberg, and Robins, 1997; Brock, Butler, and Long, 1993). The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) replaced both the AFDC and JOBS programs with TANF, effectively ending the entitlement nature of AFDC. PRWORA was based on the notion that welfare benefits have had the perverse effect of encouraging a cycle of dependency. The reform emphasizes time limits and work requirements (rather than education and job training), imposing a lifetime limit of 60 months of benefits and a work requirement after a maximum of two years of assistance. States that are unable to move welfare recipients to work face penalties. 4 California Welfare Context Under TANF, each state is given a block grant and some flexibility to design its own welfare program. California’s program is entitled California Work Opportunity and Responsibility to Kids (CalWORKs). New applicants to CalWORKs may receive aid for 18 continuous months, although counties may extend aid for an additional six months.4 As required by federal law, there is a five-year cumulative lifetime limit on aid, although children of adults who reach the lifetime limit will continue to receive aid (California Department of Social Services, 1997). Most welfare applicants are required to first engage in a job search.5 If the job search is unsuccessful, a county employee will conduct an assessment interview with the applicant, during which the applicant and the county will enter an agreement written up as a welfare-to-work plan. Applicants will then participate in welfare-to-work activities for the period specified in the plan. If the time period expires and the applicant has not found unsubsidized work, the county may extend the plan by six months. For new adult applicants, Adult Basic Education, vocational education, and education directly related to employment will qualify as work activities but only in cases where the education is needed to become employed (California Welfare and Institutions Code). Thus, CalWORKs supports the development of basic skills only to the extent that it is necessary to qualify an applicant to enter the workforce. The structure of CalWORKs’ welfare-to-work component has its roots in California’s Greater Avenues for Independence Program (GAIN), which was instituted in 1985, and began operating under the ____________ 4Current recipients may receive aid for 24 consecutive months. 5See the California Welfare and Institutions Code for a list of individuals exempt from this sequence of activities. 5 federal JOBS program in 1989. GAIN has been evaluated closely in six California counties—Alameda, Butte, Los Angeles, Riverside, San Diego, and Tulare. The successes of GAIN, particularly in Riverside County, were cited as examples for the state to follow in establishing welfare-towork programs under PRWORA (Legislative Analyst’s Office, 1997). Initially, the GAIN program differed from previous programs in that it made basic education mandatory for the subset of welfare recipients deemed “in need of basic education.”6 This group of registrants could either attend basic education classes or elect a job search activity. If they chose a job search and failed to obtain employment, they were required to attend basic education classes. The six counties evaluated varied in their emphasis on basic skills’ development versus quick employment (Riccio, Friedlander, and Freedman, 1994). The Riverside program, with its emphasis on “quick employment,” is one of the most successful welfare-to-work programs to date. The Riverside program increased the five-year average of those ever employed by 16 percent, increased average total earnings over five years by 42 percent, and reduced five-year average total AFDC payments by 15 percent (Freedman et al., 1996). The Riverside GAIN “quick employment” strategy has been adopted by CalWORKs. Yet the GAIN results also raise concerns about time limits set by PRWORA, especially for recipients with relatively low basic skills and numerous obstacles to employment. Although federal law allows a state to exempt up to 20 percent of its caseload from the five-year time limit, ____________ 6AB 1371 in 1996 repealed the mandate for basic education and required job search activities as the first activity, except for individuals who lack the education to succeed in even the most unskilled employment (California Assembly Bill 1371 at www.leginfo.ca. gov). 6 results from the GAIN program indicate that a larger exemption may be necessary. For example, by the last quarter of the fifth year, nearly onethird of those in the experimental group in Riverside were collecting AFDC payments—about the same number as in the control group. Similarly, in Los Angeles County,7 approximately half of both the control and experimental groups were collecting AFDC payments in the last quarter of the fifth year (Freedman et al., 1996). This is a discouraging finding, given that about one-third of the state welfare caseload is located in Los Angeles County (Riccio, Friedlander, and Freedman, 1994). When all six counties in the GAIN evaluation were taken into account, and the experimental group was compared with the control group, the five-year average of those ever employed increased by only 7 percent. The average total earnings over five years increased by 23 percent, and the five-year average total AFDC payments fell by 7 percent. The proportion of the experimental group collecting welfare at the end of the fifth year was 39 percent (Riccio, Friedlander, and Freedman, 1994). These results demonstrate that the success of GAIN in California was not universal. Strawn (1998) asserts that low earnings and lack of steady employment account for high levels of AFDC receipt in the fifth year of GAIN. She suggests that these outcomes are the result of quick employment programs, which increase average earnings mostly by helping recipients to work more, rather than helping them to find better jobs. Prior evaluations of GAIN and like programs have yielded similar conclusions, adding that even though these programs have shown ____________ 7In Los Angeles County, GAIN focused exclusively on long-term welfare recipients. 7 success, they have not lifted large numbers of their participants above poverty (U.S. Department of Labor, 1995). It is well known that the earnings capacity of both men and women at risk of need for public assistance has been declining since the 1970s and that the decline has been especially steep since the late 1980s (Brady and Wiseman, 1997). The earlier results of the GAIN program, combined with the realities of the low-skill labor market, highlight the importance of earnings disregards and the Earned Income Tax Credit in alleviating poverty. They also highlight the importance of job skills in successfully making the transition from welfare to work. Scope of This Research Although a primary tenet of the TANF legislation is that able-bodied welfare recipients should work, we do not know much about the labor force skills of welfare recipients. Educational attainment levels of welfare recipients are well known but might not be an adequate measure of a welfare recipient’s employability. In particular, educational attainment levels probably overstate the skills of welfare recipients. For example, high school graduates who are welfare recipients can be expected to be less skilled than high school graduates who are in the labor force. Some past research has compared the skills of welfare recipients to those in the rest of the population (see O’Neill and O’Neill, 1997, for a summary). However, such research has been limited because of relatively small sample sizes, and the skills test, the Armed Forces Qualifying Test, was administered years before entry into the labor force or receipt of welfare. To assess the employment prospects of welfare recipients, a measure of basic skills is necessary, preferably one that is contemporaneous with 8 labor force experience and that also captures the types of skills employers might value. In 1992, the National Center for Education Statistics (NCES) and the Educational Testing Service (ETS) conducted the National Adult Literacy Survey (NALS), administering the survey to a nationally representative group of adults, including welfare recipients. It was the first national scale survey to measure the basic skills of working-age persons contemporaneously with their labor force experience. Twelve states, including California, sponsored increased sample sizes for their states to obtain reliable information at the state level. The goal of NCES and ETS was to assess people’s ability to succeed in dealing with practical analytical problems involving reading, writing, and calculating— problems that they could be expected to encounter in their work, home, and civic lives. For example, the exam included such tasks as completing a job application, calculating the total cost of a purchase from an order form, totaling a bank deposit entry, using a bus schedule, and writing a brief letter explaining an error on a credit card bill (see Appendix A for examples of tasks and levels of difficulty). Because the NALS included a questionnaire rich in demographic and socioeconomic information, we have a source of information that is well suited to the study of the basic skills of the employed, the working poor, and welfare recipients. In this report, we address the prospects of the nation’s, and particularly California’s, welfare population as it faces a new welfare system of work requirements, sanctions, and time limits. Using the NALS database, we examine the characteristics of several groups of respondents, including welfare recipients, heavily dependent welfare recipients, workers not receiving public aid, and other adults. We analyze how the basic skills of welfare recipients differ from those of 9 other adults in general and from workers in particular. Although it is known that welfare recipients are, in general, less educated than workers, we determine how much of the gap in skills between workers and welfare recipients can be explained by educational attainment. Additionally, we identify the types of jobs held by persons with skills and characteristics similar to the skills and characteristics of welfare recipients. Finally, we discuss some of the policy implications of this research. 10 2. Data and Methodology The National Adult Literacy Survey The NALS was conducted in 1992 and included both a national household sample and supplemental household samples for 12 states, including California. The NALS gathered descriptive information and examined proficiency in basic skills for 26,091 respondents aged 16 and older.1 In California, the total sample size was 2,665. All respondents completed a background questionnaire, which provides demographic, linguistic, educational, and socioeconomic information including data on income, work, and public aid. This information was used to characterize the adult population of the United States, to understand factors related to the distribution of basic skills ____________ 1The total number includes the additional samples of approximately 1,000 people per state for each of 12 states that chose to fund additional sampling in their respective states (state samples are referred to as State Adult Literacy Surveys or SALS). These supplemental state samples allow for state-level analyses. See Appendix B for a more complete discussion of the samples and the data. 11 scores, and to compare the NALS results with previous studies. It was also used to summarize the data by various demographic groups and to increase the accuracy of the basic skills estimates for various subpopulations (see Appendix B). Respondents spent approximately 20 minutes completing the background questionnaire and 45 minutes completing a booklet of tasks measuring their prose, document, and quantitative skills. These groups of tasks were scored separately, so that each individual received scores along a prose scale, a document scale, and a quantitative scale. The tasks were designed to measure an individual’s ability to succeed in common, practical, analytical problems. Examples of the tasks are presented in Appendix A. In previous literacy surveys, adult skills were measured by grade-level criteria, such as understanding a sixth-grade vocabulary list, or correctly completing an eighth-grade mathematical exercise. Because such tasks do not reflect the kinds of tasks that adults must routinely perform, they are neither appropriate nor adequate for assessing adult basic skills in the context of assessing employment prospects (Kirsch et al., 1993). Thus, our analysis and discussion of basic skills in this report are based on the scores derived from the NALS.2 Our analyses indicate that basic skills scores are a good predictor of labor force outcomes (see Appendix C). Indeed, basic skills scores are at least as good a predictor of labor force ____________ 2NCES and ETS use the term literacy rather than basic skills. Our experience has been that many people understand literacy as a dichotomous skill (the ability to read and write). NCES and ETS, however, consider literacy to be much less discrete, noting that the NALS shows a wide range of literacy proficiencies. We chose the term basic skills rather than literacy because it is more readily understood as consisting of a range of abilities, and to make explicit that, more than simply testing for the ability to read and write, the exam contained practical reading, writing, calculating, analyzing, and reasoning tasks that adults face in their everyday lives. 12 outcomes as education. This suggests that the NALS exam measures skills that employers value. One potential problem of the survey is its age; although we would not expect a substantial change in the literacy proficiencies of adults since 1992, we might expect changes in the population receiving welfare. In 1992, the nation, especially California, was experiencing a recession. Welfare caseloads were substantially higher in 1992 than they are today. It is reasonable to expect that those most likely to leave welfare in the intervening years were those most skilled. Thus, we would expect our findings to understate the current difference in skills between welfare recipients and other adults. On the other hand, the labor force characteristics of persons not on welfare might have been depressed, in terms of both wages and employment rates during the recession. Because we analyze the employment prospects of welfare recipients by looking at the labor force status and characteristics of jobs of certain adults not on welfare, the survey’s timing might lead to an overstatement of the problems welfare recipients might face. The net effect is uncertain. Study Methods We use several methods to accomplish the various goals of our analyses. In almost all of our analyses, we use the quantitative literacy score as our measure of basic skills and restrict the sample to adults between the ages of 16 and 55 who are not enrolled in high school. We identify welfare recipients as persons who report living in a household that received AFDC, public assistance, or public welfare in the past 12 months.3 Because one of our primary goals is to evaluate the ability of ____________ 3The survey asked separate questions for Supplemental Security Income (SSI) and for Food Stamps. 13 welfare recipients to move off aid and into employment, we chose not to consider persons over the age of 55. Welfare recipients beyond 55 years of age will soon be, if they are not already, eligible for other forms of public assistance. We chose to exclude students still in high school, because those students are not in the labor force, generally not on welfare, and their basic skills are subject to substantial change as they complete more schooling. We might have chosen to exclude college students for the same reasons; however, we did not want to exclude a group of such substantial size and in the same age groups as many welfare recipients. Generally, we focus on quantitative basic skills because they are slightly better predictors of labor force outcomes than either document or prose skills (see Appendix C). In any event, the three types of basic skills are highly correlated and our results did not change in any substantial way when we considered one of the other types of skills. In this report, we first present general findings of the NALS. In describing and comparing the basic skills scores of adults and certain subgroups, we provide simple statistics such as means and distributions. These statistics are weighted to reflect state and national adult populations. Population means are calculated as the weighted mean of individual scores, and standard errors were adjusted to take into account the sampling design and the NALS scoring procedure (see Appendix B). We then develop one set of regression models to evaluate the difference in basic skills between welfare recipients and other adults. Our goal in these regression models is to determine if differences in basic skills can be ascribed to population composition differences between welfare recipients and other adults. Thus, the models we consider attempt to predict an individual’s basic skills score using a prescribed and limited set of variables that identify certain demographic and social characteristics. 14 We conduct separate regressions by welfare status and educational attainment level. The substantive results of this set of regressions are discussed in Chapter 3; Appendix D contains the regression results themselves. Finally, we develop a logistic regression model to predict the receipt of welfare. Our goal in this model is to identify persons similar to welfare recipients in terms of basic skills (and other characteristics) but who did not receive welfare. We seek to characterize the employment status and types of jobs of persons who are similar to welfare recipients in terms of education, basic skills, and some demographic characteristics. The model is described in Appendix E and the findings from the model are discussed in Chapter 3. 15 3. Findings General Findings The results of the NALS suggest that a substantial number of Americans lack fundamental basic skills.1 As shown in Table 3.1, almost one in four American adults has very low basic skills. People at this lowest level can be expected to fail at tasks that are often encountered in an increasingly technical workplace that demands mental rather than physical skills. For example, people in the lowest basic skills level are generally unable to follow simple directions for performing a single mathematical operation using numbers that are easily located in a text (see Appendix A for examples of tasks in the survey). About half of Americans have either very low basic skills or low basic skills. The large number of Americans with low basic skills is disturbing. In their review of the results of the NALS, the authors of the study state: ____________ 1Detailed tables of descriptive statistics for the general adult population are available in Kirsch et al. (1993). 17 If large percentages of adults had to do little more than be able to sign their name on a form or locate a single fact in a newspaper or table, then the levels of literacy seen in this survey might not warrant concern. We live in a nation, however, where both the volume and variety of written information are growing and where increasing numbers of citizens are expected to be able to read, understand, and use these materials (Kirsch et al., 1993). Table 3.1 Average Basic Skills of U.S. Adults Average score Distribution of skills Very low skills Low skills Moderate skills High skills Very high skills Quantitative Skills 271 (0.7) Document Skills 267 (0.7) Prose Skills 272 (0.6) 22 23 21 25 28 27 31 31 32 17 15 17 4 33 SOURCE: NALS; full sample. NOTE: Very low skills correspond to scores less than 225, low skills correspond to scores between 226 and 275, moderate skills correspond to scores between 276 and 325, high skills correspond to scores between 326 and 375, and very high skills correspond to scores above 375. Standard errors in parentheses. The average basic skills scores for Californians are slightly lower than for adults in the rest of the country. Indeed, of the 12 states that participated with NCES to increase sample sizes, California ranked ninth in terms of literacy scores (see Table 3.2). Also, the distribution of basic skills is more extreme in California than in the rest of the nation. As shown in Table 3.3, the proportions of Californians at the very lowest skill level and at the very highest skill level are slightly higher than in the nation as a whole. None of the other states with expanded samples, with the possible exception of Illinois, show a similar pattern. For example, although Iowa and Washington have relatively high proportions of adults 18 Table 3.2 Average Basic Skills Scores, by State and for the Nation (Ranked by Quantitative Mean) State United States Quantitative Mean 271 (0.7) Document Mean 267 (0.7) Washington Iowa Indiana Ohio Illinois Pennsylvania New Jersey Florida California Louisiana New York Texas 293 (4.1) 287 (3.4) 282 (2.3) 280 (2.7) 274 (1.8) 274 (2.5) 273 (2.3) 271 (4.1) 269 (1.7) 261 (4.3) 258 (2.1) 258 (1.9) 288 (3.4) 280 (2.8) 276 (1.7) 276 (2.4) 269 (1.6) 270 (1.9) 268 (1.9) 264 (4.2) 263 (1.8) 257 (3.0) 257 (2.1) 255 (2.0) SOURCE: NALS; full sample. NOTE: Standard errors in parentheses. Prose Mean 272 (0.6) 291 (4.3) 285 (3.0) 281 (1.5) 280 (2.3) 274 (1.5) 275 (1.5) 273 (1.6) 269 (3.2) 270 (1.7) 263 (3.7) 262 (1.9) 259 (2.0) at the highest skill levels, they have relatively low proportions at the lowest skill levels. The distributions in Louisiana, New York, and Texas, on the other hand, are skewed toward the low end of the scale. California’s relatively bipolar distribution mirrors the greater income inequality in the state than in the rest of the nation, suggesting that at least part of the reason for the relatively high income inequality in the state is related to the large variation in basic skills of California residents. The Skills Gap: Basic Skills of Welfare Recipients and Workers Although the basic skills of the adult population in California and the nation are fairly low, the basic skills of welfare recipients are even 19 Table 3.3 Distribution of Basic Skills Scores, by State and for the Nation State United States Percentage of Adults with: High or Very Low Low Moderate Very High Skills Skills Skills Skills 22 25 31 21 California 24 22 30 Florida 21 27 31 Illinois 22 23 31 Indiana 16 27 35 Iowa 15 22 36 Louisiana 26 28 29 New Jersey 24 25 31 New York 28 26 28 Ohio 17 27 33 Pennsylvania 21 25 33 Texas 28 25 29 Washington 10 22 40 24 21 23 23 27 16 20 18 23 21 18 29 SOURCE: NALS; full sample. NOTE: Results are for quantitative skills. Similar results were found for document and prose skills. lower. Not only do welfare recipients tend to be less skilled than the general adult population, they tend to be much less skilled than employed people not receiving aid. In addition, people heavily dependent on welfare, defined as welfare recipients who did not work in the prior year, tend to have even lower skill levels than other welfare recipients. As shown in Table 3.4, welfare recipients scored 55 points lower on average than employed persons on the test of quantitative skills, and persons heavily dependent on welfare scored 72 points lower on average than employed persons. These very low scores mean that the 20 Table 3.4 Average Basic Skills, by Welfare Status: United States Received welfare Heavily welfare dependent Did not receive welfare Employed full time Average Quantitative Score 239 (2.0) 222 (2.6) 287 (0.7) 294 (0.8) SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. Persons are defined as heavily welfare dependent if they did not receive any wage income in the 12 months before the survey. NOTE: Standard errors in parentheses. average person heavily dependent on welfare has difficulty performing simple arithmetic operations, such as addition, and generally cannot perform tasks requiring a single mathematical operation that is not specified in the question (see Appendix A for sample questions and their difficulty level). Half of all welfare recipients in the nation were heavily dependent on welfare. Similar results were obtained for the other types of skills measured by the NALS. Another way to compare the skills of welfare recipients to other persons is to examine the distribution of scores by welfare status. As shown in Table 3.5, 60 percent of welfare recipients and 81 percent of persons heavily dependent on welfare have either low basic skills or very low basic skills. Welfare recipients in California tend to have substantially lower basic skills than welfare recipients in the rest of the nation (see Table 3.6), whereas people heavily dependent on welfare and employed people have 21 Table 3.5 Distribution of Basic Skills, by Welfare Status: United States Very Low Skills Adults 16–55, not in high school (not on welfare) 17.7 Persons employed full time (not on welfare) 10.6 All welfare recipients 26.2 Persons heavily dependent on welfare 49.3 Percentage with: Low Moderate High Skills Skills Skills 25.8 35.5 18.9 20.0 38.2 33.3 30.7 26.4 9.7 31.9 16.1 2.6 Very High Skills 2.1 4.8 0.1 0.0 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. Table 3.6 Average Basic Skills, by Welfare Status: California Compared to the Rest of the Nation Received welfare Heavily welfare dependent Did not receive welfare Employed full time Average Quantitative Score California Rest of Nation 221 (6.2) 242 (2.1) 221 (8.0) 222 (2.7) 279 (2.6) 288 (0.7) 287 (3.1) 295 (0.8) SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Persons are defined as heavily welfare dependent if they did not receive any wage income in the 12 months before the survey. Standard errors in parentheses. only slightly lower basic skills. Thus, the basic skills gap between welfare recipients and employed people is greater in California than in the rest of the nation. This finding is somewhat surprising. Because California provides more generous welfare payments than most states and has a 22 higher proportion of its population receiving welfare,2 we would expect that the California welfare population would include a greater share of moderately skilled persons than the rest of the country. The selection effect into welfare should be less in California than in most other states. However, as shown in Table 3.7, the proportion of welfare recipients with very low skills is substantially higher in California than in the rest of the nation (41 percent compared to 24 percent). In California, almost four of every five welfare recipients have either low or very low basic skills. Table 3.7 Distribution of Basic Skills, by Welfare Status: California and the Rest of the Nation Percentage with: Very Low Low Moderate High Very High Skills Skills Skills Skills Skills California Adults 16–55, not in high school (not on welfare) 22.7 20.4 32.8 21.6 2.6 Persons employed full time (not on welfare) 15.8 18.2 34.4 25.6 6.0 All welfare recipients 41.3 35.6 20.3 2.9 0.0 Persons heavily dependent on welfare 46.5 26.9 23.4 3.2 0.0 Rest of the Nation Adults 16–55, not in high school (not on welfare) 16.9 26.7 35.9 18.5 2.0 Persons employed full time (not on welfare) 9.9 20.3 38.8 26.4 4.6 All welfare recipients 24.1 33.0 32.1 10.7 0.1 Persons heavily dependent on welfare 49.8 32.8 14.8 2.5 0.0 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. ____________ 2In the restricted NALS sample we used, 8.2 percent of adults in California reported receiving welfare compared to 7.7 percent in the rest of the nation. 23 The fact that the skills gap between welfare recipients and employed persons is greater in California coupled with the very low basic skills levels of most California welfare recipients suggests that California will have a more difficult task than most states in moving persons from welfare to full-time work. Indeed, although welfare rolls have declined in California since 1993, the drop has been much less precipitous than in most other states (see Figure 3.1). Only three states have experienced declines that were smaller than California’s between January 1993 and September 1997. Although the relatively small decline in the welfare rolls in California is probably primarily a function of the state’s economy (i.e., the availability of jobs), the state’s more generous welfare benefits, and slower implementation of welfare reform than in some other states, the relatively weak decrease is probably also a reflection of the very low skills of welfare recipients in the state. On the other hand, persons who are heavily dependent on welfare in California are not much less skilled than other welfare recipients in the state. This is in sharp contrast to the rest of the country, where persons heavily dependent on welfare have substantially lower basic skills than other welfare recipients. The small basic skills gap between persons heavily dependent on welfare and other welfare recipients in California is not due to relatively high skills of heavily dependent welfare users in California; rather, it is due to the very low average skill level of all welfare recipients in the state. In addition, California has a higher proportion of heavily dependent welfare users among its welfare population than does the nation (55 percent compared to 48 percent). 24 Wyoming Idaho Wisconsin Oregon Mississippi Alabama South Carolina Oklahoma Tennessee Florida Indiana Louisiana North Dakota Kansas New Mexico Colorado South Dakota Texas Utah Arkansas Michigan New Hampshire Ohio Georgia Massachusetts Virginia Maryland Montana West Virginia North Carolina Maine Kentucky Pennsylvania Arizona Missouri New Jersey Iowa Minnesota Vermont Delaware Illinois Nebraska New York Washington Nevada Rhode Island California Connecticut Alaska Hawaii –100 –80 –60 –40 –20 Percentage change SOURCE: U.S. DHHS (1998). 0 20 40 Figure 3.1—Percentage Change in AFDC Families, by State, 1993–1997 25 Educational Attainment and the Basic Skills Gap Because of a lack of data on the basic skills of welfare recipients, researchers and policymakers have used educational attainment as a proxy for skills. However, it is not clear to what extent educational attainment is an adequate indicator of a welfare recipient’s basic skills. It seems reasonable to expect that welfare recipients have lower basic skills than similarly educated adults who are not on welfare, and the NALS provides us with the opportunity to evaluate the extent to which educational attainment overstates basic skills of welfare recipients compared to other adults. In general, we want to determine whether the skills gap between welfare recipients and the rest of the population can be understood through differences in education and demographic characteristics. Do welfare recipients tend to have low literacy scores solely because they are poorly educated, are more likely to have a disability, and are younger than the general population? As noted previously, the NALS provides us with the unique opportunity to examine this question, since it is the only nationally representative sample of welfare recipients and workers that measures basic skills contemporaneously with labor force and welfare status. In this section, we first examine the basic skills gap between welfare recipients and other adults within educational attainment levels. We then develop regression models to examine the relationship between proficiency in basic skills, education, and welfare, controlling for a host of sociodemographic factors such as gender, age, marital status, California residence, language spoken at home, and mental or physical disabilities. 26 Table 3.8 shows the skills gap by educational attainment level and the distribution of welfare recipients and other adults by educational attainment. As shown in the first two columns of the table, welfare recipients are less educated than other adults. Because people with lower levels of education tend to have lower basic skills, some of the skills gap can be explained by the lower levels of education of welfare recipients. However, as shown in the last three columns, welfare recipients with the same levels of education as other adults tend to have substantially lower basic skills.3 For example, we find that welfare recipients with a high school diploma or GED have quantitative basic skills scores that are 24 points lower on average than those of other adults with a high school diploma or GED. A simple decomposition reveals that if welfare recipients had the same educational attainment distribution as other Table 3.8 The Skills Gap and Educational Attainment Levels in the Nation Percent by Educational Attainment Mean Quantitative Score Difference in Scores Educational Attainment Level 0–8 years 9–12 years High school graduate or GED Some college College graduate Welfare Recipients 11 29 All Other Adults 5 11 Welfare Recipients 158 (7.0) 209 (2.9) All Other Adults 156 (3.4) 227 (2.0) 45 39 251 (2.4) 275 (0.9) 13 24 275 (3.5) 303 (0.9) 2 22 286 (11.6) 332 (0.9) (the Skills Gap) –2 18* 24* 28* 46* SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: * indicates significance at the .01 level. Standard errors in parentheses. ____________ 3The one exception is for adults with 0–8 years of education. For adults with this lowest level of educational attainment, average basic skills scores are extremely low (less than 160) for both welfare recipients and other adults. 27 adults, the basic skills gap would have been reduced by just over 40 percent. In other words, the basic skills gap is partially, but not primarily, explained by differences in education between welfare recipients and other adults. In particular, research that uses education as a proxy for the basic skills of welfare recipients substantially underestimates the skills gap between welfare recipients and other adults. A similar decomposition for California suggests that the basic skills gap between welfare recipients and other adults would be reduced by about 30 points (40 percent of the total difference) if California welfare recipients had the same levels of educational attainment as other adults in the state. Thus, the majority of the skills gap remains unexplained if one considers education alone. Using a regression framework, we also explore whether differences in demographic characteristics, in addition to educational attainment, might explain the differences in basic skills between welfare recipients and other adults (see Appendix D for a discussion of the model). We evaluate differences in basic skills that might be due to differences in age, gender, language spoken at home, and physical and mental disabilities. We find that the basic skills gap between welfare recipients and others persists even when we control for all of these characteristics in addition to educational attainment. In other words, the skills gap between welfare recipients and other adults cannot be fully explained by a host of sociodemographic factors. Even after controlling for mental and physical disabilities, age, gender, language, and marital status, we still find significant differences in basic skills between welfare recipients and similarly educated persons not receiving welfare (see Table 3.9). This persistence in the skills gap indicates that the gap is not merely a population composition effect: Welfare recipients with characteristics 28 Table 3.9 The Skills Gap and Population Composition Effects in the Nation Cumulative Controls No controls within education group Physical/mental disabilities Language Gender and age Skills Gap Between Welfare Recipients and Others Within Specified Educational Attainment Level High School 0–8 9–12 Graduate Some College Years Years or GED College Graduate –2 18 24 28 46 –2 19 23 28 49 9 20 23 28 43 11 24 22 21 32 SOURCE: Authors’ regression models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: The first row corresponds with the last column in the previous table. similar to other adults have lower skills than those other adults. These findings have important implications for designing programs to improve the skill levels of welfare recipients. They suggest that the basic skills deficiencies of most welfare recipients are not due to easily identifiable problems such as English proficiency (or mental disabilities). Similar analyses for California are shown in Tables 3.10 and 3.11. The primary finding that the skills gap persists even controlling for educational attainment and other factors is also true for California. However, we do see some unique California patterns. First, although the overall skills gap between welfare recipients and other adults is larger in California than in the nation, the skills gaps within educational attainment groupings are similar to those in the rest of the nation. Second, less-educated Californians (those who have not attended or graduated from college) have relatively lower basic skills than less- 29 Table 3.10 The Skills Gap and Educational Attainment Levels in California Educational Attainment Level 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent by Educational Attainment Welfare All Other Recipients Adults 16 8 23 10 44 30 15 29 1 24 Mean Quantitative Score Welfare All Other Recipients Adults 129 (11.5) 125 (5.3) 204 (8.5) 212 (6.8) 237 (7.8) 257 (3.9) 281 (6.8) 301 (2.6) 242 (71.9) 332 (2.8) Difference in Scores (the Skills Gap) –4 8 20* 20* 91 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: * indicates significance at the .01 level. Standard errors in parentheses. Table 3.11 The Skills Gap and Population Composition Effects in California Cumulative Controls No controls within education group Physical/mental disabilities Language Gender and age Skills Gap Between Welfare Recipients and Others Within Specified Educational Attainment Level High School 0–8 9–12 Graduate Some College Years Years or GED College Graduate –4 8 20 20 91 –4 7 17 20 91 36 30 28 26 84 36 27 17 11 67 SOURCE: Authors’ regression models from the NALS: sample restricted to adults aged 16 to 55 not in high school. NOTE: The first row corresponds with the last column in the previous table. educated adults in the rest of the nation (compare mean quantitative scores in Tables 3.8 and 3.10). The lower basic skills levels of adults in California compared to the nation can largely be attributed to language differences. California has a greater share of people for whom English is 30 a second language; such people tend to have lower basic skills (as measured in English) than do native English speakers. The difference in scores between less-educated adults in California and the rest of the nation is greatly diminished or eliminated once we control for language. Finally, after controlling for language, the skills gap between welfare recipients and other adults in California is especially large for poorly educated adults. In other words, when we compare welfare recipients with other adults who speak the same language, we observe that the skills gap is quite large among those with little education. Thus, the low basic skills of poorly educated welfare recipients in California is not due to an inability to speak English. Estimating the Employment Prospects of Welfare Recipients CalWORKs requires welfare recipients to work after receiving aid for no more than 24 months and no more than 18 months in the case of new applicants. Given these work requirements, it is important to consider what kinds of work welfare recipients might be able to find, given their skill levels. Determining the likely experience of welfare recipients as they move off assistance is an uncertain undertaking. The success of welfare recipients in the labor force is a function not only of their individual characteristics and circumstances but also of local and nationwide economic conditions (particularly the availability of jobs). Projecting the demand for labor is beyond the scope of this research, but the NALS data do allow us to examine characteristics of welfare recipients and hence enable us to assess welfare recipients’ potential for success in the labor force. 31 To assess the potential labor force outcomes of welfare recipients, we look at two other groups: • Welfare workers —persons who received welfare and who worked at some point in the 12 months before the survey.4 • Welfare counterparts —persons who did not receive welfare but who had similar basic skills and sociodemographic characteristics as welfare recipients. We contrast the labor force characteristics of those two groups with the labor force characteristics of other adults in the nation and in California.5 It is well known that many welfare recipients work while receiving welfare or cycle between work and welfare. In our sample, 48 percent of those who received welfare some time in the year before the survey also reported some earned income in that same year. The labor force ____________ 4Note that we do not know the timing of work and welfare receipt within the prior year. Some welfare recipients worked and received welfare simultaneously; others received welfare during part of the year and worked during other parts of the year. 5We considered whether all low-skill workers might also serve as a proxy for the labor force prospects of welfare recipients. Most welfare recipients, welfare workers, and welfare counterparts are low-skilled. However, most low-skill adults are not welfare recipients and are not in our welfare counterparts group. We rejected all low-skill workers as a proxy for the potential labor force outcomes of welfare recipients. Persons with low skills who do not receive welfare constitute a very different population from welfare recipients, especially in California. Low-skill workers are more likely to be high school graduates, immigrants, male, married, and older than welfare recipients. Low-skill workers are less likely to have children than welfare recipients. Some of these differences are programmatic (for example, it is necessary to have children to receive welfare), but some of these differences, and, in fact, some of the programmatic differences, indicate that there might be very different labor markets for low-skill welfare recipients than for low-skill workers, and also different obstacles in finding work. Together, these differences are substantial and suggest that the type of work that many low-skill workers engage in might not be available to welfare recipients. Our welfare counterparts and welfare workers groups are better proxies for the potential labor force outcomes of welfare recipients because in addition to being primarily low-skill, they also have other characteristics similar to welfare recipients. 32 experience of these welfare workers could be a proxy for the labor force experience of welfare recipients who did not work in the prior year. The labor force experience of welfare workers probably represents an optimistic scenario for the potential labor force experience of recipients who did not work. For example, welfare workers in the nation have substantially higher basic skills, on average, and higher educational attainment levels than the welfare recipients who did not work in the year before the survey (see Table 3.12). In addition, welfare workers Table 3.12 Characteristics of Welfare Recipients, by Work Status Mean basic skills score Educational attainment (%) 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent immigrants Percent female Percent married Percent aged 16–24 25–39 40–54 Percent with mental/physical disability Percent with children aged < 6a California Other Welfare Welfare Workers Recipients 226 (12.0) 219 (7.3) Rest of the Nation Other Welfare Welfare Workers Recipients 258 (3.4) 227 (2.4) 16 17 8 14 21 24 23 34 44 45 49 41 16 14 16 10 3 0 41 34 30 10 13 47 89 58 87 45 24 38 25 30 32 31 28 53 55 51 55 16 13 18 17 0 4 12 58 71 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Welfare recipients who reported income from wages in the 12 months before the survey are classified as welfare workers. Standard errors in parentheses. aCalifornia sample only. 33 were more likely to be married, less likely to have children, and much more likely to be male. We define welfare counterparts as persons in the NALS who were very similar to welfare recipients in terms of basic skills, education, and demographic characteristics but who were not receiving welfare. Welfare counterparts were identified using a statistical model that controlled for quantitative skills score, education, age, disabilities, gender, marital status, and immigrant status.6 As shown in Table 3.13, welfare counterparts are very similar to welfare recipients, with one important exception: Welfare counterparts do not receive welfare. Because welfare counterparts are similar to welfare recipients in terms of basic skills, education, and demographic characteristics, their labor force experience can serve as a proxy for the likely labor force experience of welfare recipients. Of course, there are differences between welfare recipients and workers that are either not measurable or that are not measured in the survey. For example, welfare counterparts might have better access to transportation, live in areas with numerous employment opportunities, have family members who can provide child care, have alternative sources of income, or have healthier or fewer dependents than welfare recipients. Such differences might allow our welfare counterparts to work rather than receive welfare (or to not work and not rely on welfare), but these factors were not measured by the NALS. Because these might be important determinants of welfare dependence, our findings probably represent a best case scenario. That is, the characteristics of jobs held by ____________ 6For California, we also controlled for the presence of children younger than 6 years of age. See Appendix E for a complete discussion of the model. 34 Table 3.13 Characteristics of Welfare Recipients and Welfare Counterparts California Welfare Welfare Counterparts Recipients Entire Nation Welfare Welfare Counterparts Recipients Mean basic skills score Educational attainment (%) 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent immigrants Percent female Percent married Percent aged 16–24 25–39 40–54 Percent with mental/physical disability Percent with children aged < 6a 202 (7.5) 20 26 43 11 0 37 79 33 32 60 8 2 73 221 (6.2) 16 23 44 15 1 32 70 34 31 54 15 2 65 217 (2.0) 13 38 45 4 0 11 86 17 35 55 9 2 239 (2.0) 11 29 45 13 2 11 72 32 30 53 18 1 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Welfare recipients who reported income from wages in the 12 months before the survey are classified as welfare workers. aCalifornia sample only. welfare counterparts and welfare workers are optimistic measures of the employment potential of welfare recipients in general.7 ____________ 7Some of this bias is offset by the slightly lower skill levels of welfare counterparts. As noted in Table 3.13, mean basic skills scores are slightly lower for welfare counterparts than for welfare recipients. 35 Labor Force Status8 Many welfare recipients will have difficulty finding work. Table 3.14 shows that unemployment rates are substantially higher for welfare counterparts than for other adults, and labor force participation rates are substantially lower. In California, almost 40 percent of welfare counterparts were either unemployed or out of the labor force (i.e., not employed and not looking for work) at the time of the survey. An Table 3.14 Labor Force Status of Welfare Counterparts and Other Adults All Welfare Counterparts California Not in the labor force 27.5% In the labor force, unemployed 14.4% In the labor force, employed part time 16.5% In the labor force, employed semi- permanently full time 8.6% In the labor force, permanently employed full time 33.0% Rest of Nation Not in the labor force 21.4% In the labor force, unemployed 12.3% In the labor force, employed part time 15.9% In the labor force, employed semi- permanently full time 10.4% In the labor force, permanently employed full time 40.1% Welfare Counterparts with Very Low Basic Skills 33.1% 11.2% 17.0% 10.8% 28.0% 29.1% 14.3% 13.8% 11.1% 31.6% Other NonWelfare Adults 11.8% 8.7% 13.8% 7.5% 58.3% 13.1% 6.7% 12.2% 7.6% 60.4% SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Full-time employment is 35 hours per week or more; permanent employment is 40 weeks per year or more. Very low basic skills correspond to basic quantitative skills scores of less than 225. ____________ 8Individuals are either in the labor force employed, in the labor force unemployed, or not in the labor force. 36 additional 23 percent were employed either part time or semipermanently. Welfare counterparts with very low basic skills have especially weak attachments to the labor force. Only 28 percent of very low-skill welfare counterparts in California were permanently employed full time in the year before the survey (compared to 33 percent for all welfare counterparts and 58 percent for the rest of adults in the state). Unemployment rates for welfare counterparts were more than twice those of other adults (20 percent compared to 9 percent). The low labor force participation rates and high unemployment of welfare counterparts might overstate the difficulty of welfare recipients in finding work, since social support systems available to welfare counterparts might not be available to welfare recipients. These social support systems might provide financial support and could lessen the urgency of finding employment for welfare counterparts. However, such support systems might also be important sources of job information and referrals. Still, the very low labor force participation rates of welfare counterparts suggest that many welfare recipients might not transition from welfare to work, but might instead transition from welfare to dependence on friends or family (or, perhaps, homelessness if they lack such support networks). Early reviews of the decline in welfare caseloads indicate that many former welfare recipients do not seem to be employed.9 ____________ 9For example, an analysis of New York state welfare and employment data revealed that a substantial share of former welfare recipients did not appear to be employed in the state of New York (“Most Dropped from Welfare Don’t Get Jobs,” New York Times, March 23, 1998). 37 Earnings Even when persons with the basic skills and sociodemographic characteristics of welfare recipients do find work, their earnings are often not enough to lift them out of poverty. Over the course of an entire year, welfare counterparts in California who worked earned an average income of $12,400, and over half did not have sufficient earnings to lift a family of three out of poverty.10 Table 3.15 shows the distribution of annual income for welfare counterparts, welfare workers, and other workers in California and the rest of the nation. Even if we restrict our analysis to full-time workers, we observe very low average weekly wages for welfare counterparts working full time and for full-time workers who received welfare some time in the past year (see Table 3.16). As with labor force status, the findings are particularly bleak for persons with very low basic skills. Welfare counterparts in California with very low basic skills earned less than $10,000 per year on average, and fully 70 percent did not earn enough to lift a family of three out of poverty (see Table 3.17).11 The low annual earnings of this group reflect, in part, their lack of year-round full-time employment. Intermittent employment is a problem common to many low-skill workers. However, even when we consider weekly earnings of welfare counterparts with very low basic skills who work full time, we still observe very low wage levels. ____________ 10Converting these earnings to 1998 dollars and using the 1998 Earned Income Tax Credit lowers this figure to 44 percent. 11Converting these earnings to 1998 dollars and using the 1998 Earned Income Tax Credit lowers this figure to 52 percent. 38 Table 3.15 Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Workers California Welfare Counterparts 7,917 48.5 17.1 16.1 9.8 6.0 0.9 1.5 Rest of Nation 12,383 33.6 17.5 16.5 11.5 6.4 4.9 9.7 8,937 45.6 23.1 13.2 8.5 4.6 0.6 4.1 10,360 28.8 27.2 22.5 11.2 5.5 2.3 2.6 Other NonWelfare Adults 26,830 12.2 9.4 14.3 12.9 9.3 8.2 33.7 22,445 14.5 12.7 14.9 13.1 10.8 8.7 25.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. Occupations Welfare workers and welfare counterparts are concentrated in occupations that typically consist of low-skill, low-wage, high-turnover jobs. Relative to other adult workers, welfare counterparts and welfare workers are vastly underrepresented in managerial and professional occupations and are especially overrepresented in service sector jobs (see Table 3.18). Although a detailed delineation of the occupations within the broad categories shown in Table 3.18 is not possible given our 39 Table 3.16 Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Currently Working Full Time Average annual earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Full-Time Counterparts Welfare Working Full Workers Time California 13,347 17,468 16.9 11.6 22.4 13.4 26.7 21.7 12.5 15.5 14.2 10.1 2.8 9.3 4.5 15.5 Rest of Nation 14,161 13,187 20.2 11.0 25.4 27.3 20.4 30.6 14.8 15.9 9.5 8.3 1.0 3.6 8.7 3.2 Other Full-Time Workers 32,381 3.5 5.2 13.4 14.2 11.4 9.6 42.7 26,732 4.6 8.9 15.4 15.2 13.1 10.7 32.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Full-time employment is 35 hours per week or more. sample size, we do note that the average wage of welfare counterparts and welfare workers is much lower than that of other workers within the same occupational category (see Table 3.18). For example, welfare counterparts earned about one-third less per week than other adults in service occupations. In no occupational category did welfare counterparts earn more than 70 percent of the earnings of other adults. Weekly earnings of welfare workers are even lower than the earnings of welfare counterparts. 40 Table 3.17 Earnings of Welfare Workers with Very Low Basic Skills, Welfare Counterparts with Very Low Basic Skills, and Other Non-Welfare Adults Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Workers with Very Low Basic Skills California Welfare Counterparts with Very Low Basic Skills 6,615 50.2 19.8 17.4 10.5 2.1 0 0 Rest of Nation 9,926 38.4 23.0 18.9 6.2 5.5 1.8 6.2 7,155 48.4 26.7 11.3 6.3 5.9 0.2 1.3 9,458 32.3 33.3 20.0 6.9 2.7 1.9 3.0 Other NonWelfare Adults (Any Skill Level) 28,830 12.2 9.4 14.3 12.9 9.3 8.2 33.7 26,732 14.1 12.7 14.9 13.1 10.8 8.7 25.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Very low basic skills correspond to basic quantitative skills scores of less than 225. Industry Welfare workers and welfare counterparts are also concentrated in industrial sectors of the economy that are typified by low-skill, low-wage, high-turnover jobs. As shown in Table 3.19, welfare workers and welfare counterparts are substantially overrepresented in personal services and 41 Table 3.18 Occupational Profile of Welfare Workers, Welfare Counterparts, and Other Adults Welfare Welfare Occupation Workers Counterparts Percentage Among Those with Work Service 32.3 34.6 Farming, forestry, and fishing 4.8 3.1 Technical, sales, and admin. support 21.5 35.5 Precision production, craft, repair 32.2 24.7 Managerial and professional 9.2 2.1 Average Weekly Earnings ($) Service 167 146 Farming, forestry, and fishing 212 148 Technical, sales, and admin. support 211 190 Precision production, craft, repair 312 188 Managerial and professional 421 290 Other Adults 16.6 2.6 32.4 26.3 22.1 210 241 321 337 681 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. agriculture and underrepresented in professional services. Welfare workers and welfare counterparts earn substantially lower wages than other adults employed in the same industry. Thus, not only are welfare workers and welfare recipients concentrated in low-paying industries, they tend to occupy the lower-level jobs within an industry. Summary of Findings on Employment Prospects On the basis of our analysis of the labor force characteristics of welfare workers and welfare counterparts, we find that the labor force prospects of welfare recipients are not especially promising. Welfare recipients are not likely to find jobs that would pay sufficient wages to lift them out of poverty.12 In addition, welfare recipients face a segment of ____________ 12The picture is not quite so bleak if we consider the Earned Income Tax Credit. 42 Table 3.19 Industrial Profile of Welfare Workers, Welfare Counterparts, and Other Adults Welfare Welfare Occupation Workers Counterparts Percentage Among Those with Work Agriculture, forestry, fishing, and mining 3.9 2.8 Construction 5.5 3.9 Manufacturing 19.2 17.5 Trade 26.1 33.9 Personal services 7.2 8.3 Professional service 22.3 17.4 Other 15.9 16.2 Average Weekly Earnings ($) Agriculture, forestry, fishing, and mining 220 136 Construction 243 165 Manufacturing 311 193 Trade 159 133 Personal services 163 159 Professional service 289 174 Other 310 266 Other Adults 3.0 6.2 17.4 20.3 3.6 23.4 26.1 291 374 427 261 199 427 462 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. the labor market that has relatively high unemployment and low labor force participation, suggesting that many recipients will encounter difficulty in finding employment. Still, we do find that most adults with skills and measurable characteristics similar to welfare recipients are working. Indeed, we find that over half of welfare counterparts are employed. For reasons noted above, these employment rates represent an optimistic scenario for welfare recipients. Because welfare recipients represent a substantial share of the low-skill population in California, their movement off welfare could increase already high unemployment rates and decrease already low labor force participation rates among lowskill residents of the state. Finally, we find that welfare recipients with 43 very low basic skills levels will have the greatest difficulty in transitioning from work to welfare. 44 4. Policy Implications If the goal of welfare reform is to move people off welfare, then it can and will work, even if only in a deterministic manner.1 Huge reductions in caseloads nationwide and in many states suggest that welfare reform has played a part in reducing caseloads even before most individuals are subject to work requirements and elimination from the rolls.2 However, if the goal of welfare reform is to move people from welfare to work, our findings suggest that welfare reform will have mixed results. Some, perhaps the majority, of welfare recipients will find work, but a substantial share will not. It is not clear how those who do not find work will respond to reductions in welfare benefits or outright elimination in eligibility for welfare. If the goal of welfare reform is to improve basic skills via the workplace, our findings suggest that welfare reform will probably not work. The types of jobs welfare recipients are likely to ____________ 1By deterministic we mean the elimination of aid via eligibility requirements. That is, after a welfare recipient has received aid for a lifetime total of 60 months, states can and some will deny further benefits regardless of the recipient’s employment status. 2Of course, the strong economy might account for most of the decline. 45 qualify for do not generally provide the kind of training that could lead to improvements in basic skills and better employment prospects in the future. Finally, if the goal is to lift people out of poverty, our findings indicate limited success. Welfare reform has changed the standard for success in transitioning people from welfare to work. Under TANF and CalWORKs, every ablebodied welfare recipient is expected to work. Failure to work will result in either a reduction in welfare payments or elimination from welfare altogether. Although the ultimate success of welfare reform will be determined after time limits are encountered, the social and individual costs of failure require that we anticipate and respond to potential impediments to success before that time. Our findings suggest that without improvement in basic skills, many welfare recipients will not be successfully integrated into the labor force. California faces an even greater challenge. The basic skills of welfare recipients in the state are lower than those of welfare recipients in the rest of the country, and the skills gap between workers and welfare recipients is larger. The low skills of welfare recipients are not easily amenable to change. Many recipients have graduated from high school, yet even after a dozen years of schooling they are unable to perform simple tasks commonly encountered in the workplace. The track record of training programs is not especially promising.3 We are also skeptical that on the job training ____________ 3Even among programs cited as successful, it is not clear how appropriate they are as a basis for comparison. For example, the Center for Education and Training (in San Jose, California) is commonly cited as a successful program. However, it is a voluntary program for minority female single parents; approximately one-third of the past recipients have never been on welfare. Five years after enrolling in the program, increases in earnings relative to a control group were substantial only for women who entered the program with 12 or more years of education (Zambroski and Gordon, 1993). 46 will provide these skills—especially considering the types of jobs that welfare recipients might hold. The difficulty in improving the basic skills of welfare recipients does not mean that we should not try. It does mean that we need to be realistic about the costs of improving basic skills and of providing meaningful training. Basic skills and training programs need to be critically assessed, with their costs weighed against their benefits. Programs that seem most promising are those that focus on employment and integrate real job situations into the vocational and basic skills training (U.S. Department of Labor, 1995). Those programs should be pursued on a wider basis. Ultimately, we might need to accept that a substantial portion of welfare recipients will continue to need some form of income support, either because their very low skills make them virtually unemployable or because the work they find is of such low quality (and quantity) that they are still living in poverty. 47 Appendix A The National Adult Literacy Survey: Examples of Tasks and Difficulty Levels1 The NALS defines literacy as the “ability to understand and employ printed information in daily activities at home, at work, and in the community to achieve one’s goals and develop one’s knowledge and potential.” As noted in the text of this report, we prefer the term “basic skills.” In the NALS, basic skills are measured on three scales: • Prose skills: the knowledge and skills needed to understand and use information from texts that include editorials, news stories, poems, and fiction—for example, finding a piece of information in a newspaper article, interpreting instructions from a warranty, inferring a theme from a poem, or contrasting views expressed in an editorial. ____________ 1The following examples and discussions are taken from Kirsch et al. (1993). 49 • Document skills: the knowledge and skills required to locate and use information contained in materials that include job applications, payroll forms, transportation schedules, maps, tables and graphs—for example, locating a particular intersection on a street map, using a schedule to choose the appropriate bus, or entering information on an application form. • Quantitative skills: the knowledge and skills required to apply arithmetic operations, either alone or sequentially, using numbers embedded in printed materials—for example, balancing a checkbook, figuring out a tip, completing an order form, or determining the amount of interest from a loan advertisement. Skills levels are grouped by NALS into five categories. The outline below describes those categories for quantitative skills, and provides examples of tasks. Quantitative Level 1 Scale Range: 0–225 Tasks in this level require participants to perform single, relatively simple arithmetic operations such as addition. The numbers to be used are provided and the arithmetic operation to be performed is specified. Example: The respondent is shown a bank deposit slip and asked to figure the total amount of two checks being deposited. They are asked to enter the amount on the form in the space labeled “TOTAL.” Quantitative Level 2 Scale Range: 226–275 Tasks in this level typically require readers to perform a single operation using numbers that are either stated in the task or easily 50 located in the material. The operation to be performed may be stated in the question or easily determined from the format of the material. Example: The respondent is directed to complete an order form for office supplies using a page from a catalogue. No other specific instructions as to what parts of the form should be completed are given in the directive. Quantitative Level 3 Scale Range: 276–325 In tasks in this level, two or more numbers are typically needed to solve the problem, and these must be found in the material. The operation(s) needed can be determined from the arithmetic terms used in the question or directive. Example: The respondent is given a bus schedule and asked the following question. “Suppose that you took the 12:45 p.m. bus from U.A.L.R. Student Union to 17th and Main on a Saturday. According to the schedule, how many minutes is the bus ride?” Quantitative Level 4 Scale Range: 326–375 These tasks tend to require that readers perform two or more sequential operations or a single operation in which the quantities are found in different types of displays, or the operations must be inferred from the semantic information given or drawn from prior knowledge. Example: The respondent is asked to select the information necessary from two price labels to estimate the cost per ounce of creamy peanut butter. The price required for the calculation is given in 51 dollars/lb: The price on the labels is given in dollars and the quantity is given in ounces. Quantitative Level 5 Scale Range: 376–500 These tasks require readers to perform multiple operations sequentially. They must disembed the features of the problem from text or rely on background knowledge to determine the quantities or operations needed. Example: The respondent is asked to look at an advertisement for a home equity loan and then, using the information given, explain how they would calculate the total amount of interest charges associated with the loan. 52 Appendix B The National Adult Literacy Survey: Sampling Design and Scoring The national and state samples were drawn using a four-stage, stratified sampling procedure. The four stages were the primary sampling unit level, followed by the census block level, the household level, and finally the selection of age-eligible individuals. The primary sampling units consisted of counties or groups of counties and were stratified according to census region, metropolitan status, percentage of Black residents, percentage of Hispanic residents, and, whenever possible, per capita income. In the national sample, Black and Hispanic individuals were sampled at a higher rate to increase their representation in the survey. Table B.1 shows response rates for the national and California samples. Although the NALS exam was given only in English, the screener survey was given in both English and Spanish. Response rates in California were similar to those for the nation. The ETS did attempt to correct for non-participation and for non-completion of the 53 Table B.1 NALS Response Rates Instrument Screener Background questionnaire Exercise booklet Overall Percent Completing in Nation California 88.8 87.6 81.9 79.0 95.3 95.3 69.3 66.0 SOURCE: Kolstad et al. (forthcoming). NOTE: Weighted to reflect national adult population. exam. Because low basic skills due to poor proficiency in English should be understood and addressed differently than low basic skills for native English speakers, we control for language in our analyses. Because the goal of the NALS was to produce accurate population estimates of basic skills, a broad range of simulation tasks (165 in total) were administered. However, time did not permit each respondent to answer every question. Thus, each participant responded to a subset of questions (approximately 39 tasks per test booklet), selected such that the 165 tasks were administered to a nationally representative sample. Since some subsets of tasks may have been more difficult than others, basic skills proficiencies could not be reported as a percentage of correct answers. Moreover task-by-task reporting ignores the similarities of subgroups’ response patterns across tasks. These limitations were addressed by using item response theory scaling. The idea behind this scaling is that when several tasks require similar skills, the response patterns should have some regularity. This regularity can be used to characterize both respondents and tasks in terms of a common standard scale. 54 Although each individual completed only a subset of the total number of basic skills tasks, the NALS design allowed for a wide range of content representation when responses are summed for all respondents. The advantage of this design is that it yields more precise population estimates; however, this advantage is offset by the fact that it yields less precise individual estimates. Thus, NALS individual scores are not test scores in the usual sense; rather, they consist of five plausible scores for each of the three basic skills scales. We report the average of these five scores for each individual. Plausible scores were drawn from a posteriori distributions that were a function of the task difficulty of items answered correctly and background variables (gender, ethnicity, languages spoken, region of country, education, parents’ education, occupation, and reading practices). Because these background variables do not include the receipt of public aid, the scoring approach used by ETS reduces our ability to discern differences in the basic skills between welfare recipients and other adults. Because of the complexity of the NALS scoring procedures, even the calculation of descriptive statistics is not entirely straightforward. Individual scores are estimated as the mean of the five plausible values for the given type of skill. Population means are calculated as the weighted mean of individual scores. We report standard errors that are corrected using a design effect of 2.0. The design effect is derived via bootstrap procedures that take into account both the sampling design and the within-individual variation in plausible scores. 55 Appendix C NALS and Labor Force Outcomes We performed a series of regressions to identify the association between NALS scores and earnings. In our regression models using NALS scores as predictors of the log of earnings, we find that NALS scores are at least as strong predictors of earnings as educational attainment. For example, using a restricted sample of males currently working full time, we performed two separate regressions on the log of earnings. In the first regression, using only age and the quantitative skills score as the independent variables, we obtained an R2 value of .24; the second regression, using only age and educational attainment levels as the independent variables, resulted in an R2 value of .21 (see Table C.1). We also performed separate regressions by educational attainment group on the log of earnings, using only age and the quantitative skills score as the independent variables. We find that quantitative skills are a significant predictor of wages within educational attainment groups, with 57 Table C.1 Wage Equations Using Quantitative Basic Skills and Education as Dependent Variables Model 1: Log of earnings as dependent variable and age and quantitative basic skills score as independent variables; full-time male workers. Parameter Estimates Variable INTERCEP DAGE Q5MEAN R2 DF 1 1 1 0.2381 Parameter Estimate 7.550736 0.029478 0.004852 Standard T for HO: Error Parameter=0 0.06075985 124.272 0.00111297 26.486 0.00016285 29.791 Prob > |T| 0.0001 0.0001 0.0001 Model 2: Log of earnings as dependent variable and age and educational attainment (four dichotomous variables with high school graduates as the reference group) as independent variables; full-time male workers. Parameter Estimates Variable INTERCEP DAGE A08 A912 ASOCOLL ACOLPOST R2 DF 1 1 1 1 1 1 0.2234 Parameter Estimate 8.845423 0.029048 –0.523481 –0.268713 0.188651 0.530805 Standard Error 0.04353709 0.00113192 0.05399125 0.04026412 0.02710818 0.02615461 T for HO: Parameter=0 203.170 25.663 –9.696 –6.674 6.959 20.295 Prob > |T| 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 R2 values ranging from .14 to .19 depending on the group.1 In each regression, the quantitative skills coefficient was significant at the .01 level and ranged in value from .0026 to .0046. ____________ 1The lone exception was for individuals with less than an eighth-grade education. In that group, the R2 was only .09. 58 Although the prose, document, and quantitative scores are highly correlated (.93–.95), we selected the quantitative scores in our analyses because Reder and Wikelund (1994) show that higher math gains are associated with lower subsequent welfare utilization. 59 Appendix D Regressions on Quantitative Basic Skills Score To determine whether the basic skills gap between welfare recipients and other adults can be explained as a population composition effect, we performed a series of regressions. In all the models, the dependent variable is quantitative literacy score. Independent variables represent demographic and other individual characteristics. Independent variables were added consecutively to the models, thereby introducing a series of cumulative controls. Separate models were developed for each educational attainment level and by welfare status. To evaluate the population composition effect, we predicted the mean quantitative literacy score for welfare recipients, assuming they had the same population composition characteristics as persons who did not receive welfare. This was done by applying the coefficients from the regressions for welfare recipients to the non-welfare means of the values of the population composition variables. The difference between the mean 61 literacy score for non-welfare adults and the predicted mean literacy score for welfare recipients is taken as the difference in basic skills after adjusting for population composition and is reported in Tables 3.9 and 3.11. Table D.1 describes the variables and in Table D.2 we report variable means and parameter estimates from the models . Table D.1 Variables Used in Regressions to Evaluate Basic Skills Gap Description Dependent variable Q5MEAN Mean of five plausible values for quantitative basic skills score from the NALS Groups of models run separately for educational attainment levels _08yr 1= 0–8 years of education _912yr 1 = 9–12 years of education hsGEDtr 1 = high school graduate or GED completion somecol 1 = attended some college colpost 1 = completed a bachelor’s degree or more Independent variables CA 1 = California resident Disability MENTAL 1 = mental disability PHYSICAL 1 = physical disability Language HBIENG 1= speak English and another language at home HSP 1 = speak Spanish at home HOTHR 1= speak a language other than English or Spanish at home Demographic MALE 1 = male MARITAL 1 = married Age _1618_ 1 = aged 16–18 _1924_ 1 = aged 19–24 _4054_ 1 = aged 40–54 62 Table D.2 Descriptive Statistics and Regression Results Variables INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Sum Mean Model 1 Model 2 Model 3 Educational attainment: _08yr; Welfare: No 726 1.000 166.45*** 169.20*** 199.38*** 174 0.240 –41.41*** –44.16*** –9.21 14 0.019 –8.18 –28.42 19 0.026 –73.79*** –97.86*** 26 0.036 5.26 382 0.526 –69.36*** 41 0.056 –21.83 373 0.514 422 0.581 23 0.032 61 0.084 331 0.456 156.529 Educational attainment: _08yr; Welfare: Yes 173 1.000 166.78*** 167.80*** 195.02*** 40 0.231 –37.59*** –38.61*** 8.66 5 0.029 –27.04*** –54.26*** 0. .. 19 0.110 –4.88 70 0.405 –86.51*** 6 0.035 –52.77*** 45 0.260 58 0.335 12 0.069 26 0.150 48 0.277 158.089 Educational attainment: _912yr; Welfare: No 1576 1.000 229.35*** 230.67*** 237.83*** 179 0.114 –17.59*** –18.91*** 4.80 15 0.010 –24.92 –24.43 13 0.008 –113.12 –120.50 145 0.092 –4.57 214 0.136 –65.78 25 0.016 –27.83 747 0.474 761 0.483 117 0.074 291 0.185 494 0.313 227.356 Model 4 192.80*** –11.87 –26.16 –90.73*** 3.47 –71.04*** –24.99 –0.19 16.22*** 28.55 7.51 –6.03 196.04*** –1.03 –40.08*** . 0.75 –83.03*** –63.28** –1.63 22.37 5.98 4.04 –32.25** 230.99*** 4.24 –20.40 –120.05*** –5.88 –68.93*** –28.49* 1.91 14.64*** 20.82*** 5.66 –10.03** 63 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: _912yr; Welfare: Yes 523 1.000 209.64*** 209.94*** 213.03*** 211.25*** 59 0.113 –5.54 –4.29 6.04 5.49 7 0.013 –9.86 –12.95 –0.93 2 0.004 –81.54 –86.71 –98.18* 37 0.071 4.86 1.82 52 0.099 –46.19*** –45.08*** 3 0.006 8.84 21.60 79 0.151 –5.07 121 0.231 8.79 37 0.071 8.10 150 0.287 8.06 64 0.122 –19.53** 209.012 Educational attainment: hsGEDtr; Welfare: No 5877 1.000 276.34*** 276.95*** 280.82*** 270.78*** 462 0.079 –19.24*** –19.22*** –6.95** –5.54* 24 0.004 –49.86*** –47.04*** –43.46*** 17 0.003 –138.57*** –143.66*** –137.48*** 405 0.069 –6.40* –7.66** 298 0.051 –65.81*** –67.51*** 128 0.022 –48.57*** –49.86*** 2619 0.446 3.62** 3260 0.555 15.95*** 101 0.017 15.53* 851 0.145 2.00 2025 0.345 –2.65 274.832 Educational attainment: hsGEDtr; Welfare: Yes 761 1.000 252.33*** 252.86*** 255.48*** 250.27*** 87 0.114 –15.48* –11.18 –6.58 –5.40 4 0.005 –33.53 –34.88 –35.25 5 0.007 –128.82*** –133.66*** –130.90*** 39 0.051 –8.38 –7.87 56 0.074 –34.63*** –34.81*** 4 0.005 –24.64 –28.34 157 0.206 6.15 192 0.252 6.34 7 0.009 7.15 180 0.237 10.67* 99 0.130 –3.02 250.560 64 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: somecol; Welfare: No 4586 1.000 303.28*** 303.50*** 306.42*** 297.33*** 563 0.123 –2.38 –2.30 1.53 1.82 11 0.002 –54.20*** –50.78*** –48.44*** 4 0.001 –114.73*** –120.31*** –119.80*** 374 0.082 –10.25*** –10.21*** 147 0.032 –52.08*** –52.92*** 100 0.022 –40.51*** –42.04*** 2032 0.443 7.52*** 2250 0.491 9.76*** 24 0.005 6.04 1084 0.236 6.14*** 1305 0.285 –1.73 302.993 Educational attainment: somecol; Welfare: Yes 296 1.000 274.08*** 274.28*** 277.02*** 268.04*** 40 0.135 6.87 6.67 7.81 7.01 2 0.007 –26.23 –28.97 –36.33 0– 0.00 0.00 0.00 24 0.081 –15.50 –13.09 11 0.037 –41.60** –44.59*** 1 0.003 –20.42 –43.32 60 0.203 7.97 76 0.257 24.71*** 1 0.003 6.51 70 0.236 4.84 49 0.166 –0.01 275.005 Educational attainment: colpost; Welfare: No 3940 1.000 331.51*** 331.62*** 335.05*** 324.54*** 442 0.112 0.89 0.78 3.15 3.63 6 0.002 –35.05 –38.48* –41.33* 1 0.000 –166.40*** –157.11*** –145.29*** 314 0.080 –12.72*** –12.97*** 68 0.017 –51.99*** –52.21*** 176 0.045 –39.94*** –41.89*** 1928 0.489 11.49*** 2321 0.589 9.62*** 0. . 242 0.061 –4.85 1562 0.396 –1.06 331.608 65 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: colpost; Welfare: Yes 43 1.000 287.81*** 287.81*** 298.94*** 2 0.047 –46.08 –46.08 –38.74 0. .. 0. .. 4 0.093 –36.93 3 0.070 –79.07* 4 0.093 –27.07 16 0.372 16 0.372 0. 3 0.070 25 0.581 285.667 284.55*** –43.59 . . –34.87 –85.43** –53.73 –12.92 52.56*** . –3.18 4.84 66 Appendix E Determination of Welfare Counterparts We use logistic regression models to predict the probability of receiving welfare. We defined welfare counterparts as persons who did not receive welfare but who were predicted to be welfare recipients by the model. In the logistic regression framework, welfare counterparts are false positives. The goal of the regressions is to identify persons who are not welfare recipients but who have characteristics associated with the receipt of welfare. That is, we want to identify a population that is very like welfare recipients to determine what kinds of labor force outcomes welfare recipients might achieve as they move off welfare. We developed two models: one for California, and one for the rest of the United States. The models were developed separately because the California sample includes a question on the presence of children younger than six years of age in the household, whereas the sample in the rest of the nation does not include such information. The presence of children younger than six years old is an important predictor of welfare 67 receipt. Variables used in the models are described in Table E.1 and the results are shown in Table E.2. Some variables that might be highly predictive of welfare receipt were intentionally left out of the model. For example, although income is a strong predictor of welfare receipt, to place it in the model would inappropriately prescribe our findings (apart from problems of endogeneity). Table E.1 Variables Used in Logit Regressions to Identify Welfare Counterparts Description Dependent variable AFDCPAPW 0 = did not receive welfare in the prior year 1 = received welfare in the prior year Independent variables Education A08 1 = 0–8 years of education A912 1 = 9–12 years of education ASOCOLL 1 = some college ACOLPOST 1 = completed a bachelor’s degree or more Age _1618_ 1 = aged 16–18 _1924_ 1 = aged 19–24 _4054_ 1 = aged 40–54 Q5MEAN MENTAL PHYSICAL MALE MARITAL USA KIDS6 Mean of five plausible values for quantitative basic skills score from NALS 1 = mental disability 1 = physical disability 1 = male 1 = married 1 = U.S. born 1 = has children younger than six years of age (California regression only) NOTE: Omitted or reference categories are high school graduates and persons aged 25–39. 68 Table E.2 Logistic Regressions Used to Identify Welfare Counterparts Model 1: Logistic Regression for the Rest of the United States Number of Observations: 16485 Response Profile Ordered Value 1 2 AFDCPAPW 1 0 Count 1572 14913 Model Fitting Information and Testing Global Null Hypothesis BETA = 0 Intercept Criterion Only –2 LOG L 10377.810 R2 = 0.1284 Intercept and Covariates Chi-Square for Covariates 8112.043 2265.768 with 13 DF (p = 0.0001) Max-rescaled R2 = 0.2749 Variable INTERCPT A08 A912 ASOCOLL ACOLPOST _1618_ _1924_ _4054_ Q5MEAN MENTAL PHYSICAL MALE MARITAL USA DF 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Analysis of Maximum Likelihood Estimates Parameter Estimate 0.7698 0.2237 0.5343 –0.5584 –1.7277 –0.8006 –0.00961 –1.0755 –0.00975 –2.8920 0.5078 –1.3033 –1.0112 0.8666 Wald Standard Chi- Error Square 0.1752 19.3072 0.1321 2.8688 0.0782 46.6644 0.0814 47.0306 0.1679 105.8286 0.1845 18.8214 0.0735 0.0171 0.0804 178.9236 0.00062 247.3045 0.6302 21.0576 0.3367 2.2741 0.0701 345.7192 0.0660 234.7345 0.1240 48.8696 Pr > ChiSquare 0.0001 0.0903 0.0001 0.0001 0.0001 0.0001 0.8959 0.0001 0.0001 0.0001 0.1316 0.0001 0.0001 0.0001 Standardized Estimate . 0.024616 0.093225 –0.134960 –0.391130 –0.057927 –0.001936 –0.278323 –0.331453 –0.091945 0.019697 –0.355978 –0.278607 0.131319 Odds Ratio . 1.251 1.706 0.572 0.178 0.449 0.990 0.341 0.990 0.055 1.662 0.272 0.364 2.379 69 Table E.2 (continued) Model 2: Logistic Regression for California Number of Observations: 2071 Ordered Value 1 2 Response Profile AFDCPAPW 1 0 Count 233 1838 Model Fitting Information and Testing Global Null Hypothesis BETA = 0 Intercept Criterion Only –2 LOG L 1456.836 R2 = 0.1871 Intercept and Covariates Chi-Square for Covariates 1027.799 429.038 with 14 DF (p = 0.0001) Max-rescaled R2 = 0.3704 Analysis of Maximum Likelihood Estimates Variable INTERCPT A08 A912 ASOCOLL ACOLPOST _1618_ _1924_ _4054_ Q5MEANCA MENTAL PHYSICAL MALE MARITAL USA KIDS6 Wald Parameter Standard ChiDF Estimate Error Square 1 –0.5536 0.4196 1.7404 1 –0.1170 0.3032 0.1489 1 0.0644 0.2300 0.0785 1 –0.6243 0.2293 7.4099 1 –2.8346 0.7362 14.8260 1 1.1970 0.4223 8.0356 1 –0.1228 0.2128 0.3331 1 –0.2389 0.2370 1.0161 1 –0.00726 0.00175 17.1394 1 1.0530 1.0546 0.9971 1 0.0675 1.2346 0.0030 1 –0.7870 0.1821 18.6720 1 –1.6156 0.1936 69.6693 1 0.9102 0.2477 13.4985 1 1.8693 0.1911 95.7177 Pr > ChiSquare 0.1871 0.6995 0.7794 0.0065 0.0001 0.0046 0.5638 0.3135 0.0001 0.3180 0.9564 0.0001 0.0001 0.0002 0.0001 Standardized Estimate . –0.019640 0.011333 –0.156393 –0.641525 0.085084 –0.025522 –0.060130 –0.320705 0.031211 0.002002 –0.216368 –0.445121 0.226330 0.485542 Odds Ratio . 0.890 1.067 0.536 0.059 3.310 0.884 0.788 0.993 2.866 1.070 0.455 0.199 2.485 6.484 70 References Barton, Paul E., and Lynn Jenkins, Literacy and Dependency: The Literacy Skills of Welfare Recipients in the United States, Educational Testing Service, Princeton, New Jersey, 1995. Brady, Peter, and Michael Wiseman, Welfare Reform and the Labor Market: Earnings Potential and Welfare Benefits in California, 1972–1994, Institute for Research on Poverty, University of Wisconsin-Madison, Wisconsin, 1997. Brock, Thomas, David Butler, and David Long, Unpaid Work Experience for Welfare Recipients: Findings and Lessons from MDRC Research, Manpower Demonstration Research Corporation, New York, 1993. Burtless, Gary, “Employment Prospects of Welfare Recipients,” in Demetra Smith Nightingales and Robert H. Haveman (eds.), The Work Alternative, The Urban Institute, Washington, D.C., 1995. California Department of Social Services, “Fact Sheets,” 1998. http://www.dss.cahwnet.gov/calworks/default.htm California Department of Social Services, “Fact Sheet: California Welfare Population, September 1997,” 1997. www.dss.cahwnet. gov. 71 California State Senate, Senate Health and Human Services Committee, Senate Floor Committee Analysis, Bill No. AB 1542 [Welfare Reform], August 1997. California State Legislature, Welfare Reform Conference Committee, Job Creation Package, June 1997. Edin, Kathryn, and Laura Lein, Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work, The Russell Sage Foundation, New York, 1997. Freedman, Stephen, Daniel Friedlander, Winston Lin, and Amanda Schweder,The GAIN Evaluation Working Paper 96.1: Five-Year Impacts on Employment, Earnings and AFDC Receipt, Manpower Demonstration Research Corporation, New York, 1996. Friedlander, Daniel, D. H. Greenberg, and P. K. Robins, “Evaluating Government Training Programs for the Economically Disadvantaged,” Journal of Economic Literature, Vol. 35, December 1997, pp. 1809–1855. Friedlander, Daniel, and Gary Burtless, Five Years After: The Long Term Effects of Welfare-to-Work Programs, Russell Sage Foundation, New York, 1995. Gueron, Judith M., and Edward Pauly, From Welfare to Work, Russell Sage Foundation, New York, 1991. Jansson, Bruce S., The Reluctant Welfare State: American Social Welfare Policies—Past, Present, and Future, Brooks/Cole Publishing, Pacific Grove, California, 1997. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in California: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Florida: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. 72 Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Illinois: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Indiana: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Iowa: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Louisiana: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in New Jersey: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in New York: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Ohio: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Pennsylvania: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Texas: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Washington: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Kirsch, Irwin S., Ann Jungeblut, Lynn Jenkins, and Andrew Kolstad, Adult Literacy in America: A First Look at the Results of the National 73 Adult Literacy Survey, National Center for Education Statistics, September 1993. Kolstad, Andrew, et al., Technical Report and Data File Users Manual for the National Adult Literacy Survey, National Center for Educational Statistics, Washington, D.C., forthcoming. Legislative Analyst’s Office, Welfare Reform in California: A Welfare-toWork Approach, Sacramento, California, 1997. Lord, Frederic M., Applications of Item Response Theory to Practical Testing Problems, Educational Testing Service, Hillsdale, New Jersey, 1980. MaCurdy, Thomas, and Margaret O’Brien-Strain, Who Will Be Affected by Welfare Reform in California? Public Policy Institute of California, San Francisco, California, 1997. Majority Staff of the Committee on Ways and Means, 1996 Green Book, November 4, 1996. Martinson, Karin, and Daniel Friedlander, GAIN Basic Education in a Welfare to Work Program, Manpower Demonstration Research Corporation, New York, 1994. Mislevy, R. J., A. E. Beaton, B. Kaplan, and K. M. Sheehan, “Estimating Population Characteristics from Sparse Matrix Samples of Item Responses,” Journal of Educational Measurement, Vol. 29, No. 2, Summer 1992, pp. 133–161. “Most Dropped from Welfare Don’t Get Jobs,” New York Times, March 23, 1998. Olson, K., and La Donna Pavetti, Personal and Family Challenges to the Successful Transition from Welfare to Work, The Urban Institute, Washington, D.C., 1996. O’Neill, Dave M., and June Ellenoff O’Neill, Lessons for Welfare Reform: An Analysis of the AFDC Caseload and Past Welfare-to-Work Programs, W. E. Upjohn Institute for Employment Research, Kalamazoo, Michigan, 1997. 74 Pavetti, LaDonna, How Much More Can They Work? Setting Realistic Expectations for Welfare Mothers, The Urban Institute, Washington, D.C., 1997. Reder, Stephen, and K. R. Wikelund, Steps to Success: Literacy Development in a Welfare-to-Work Program, Northwest Regional Educational Laboratory, Portland, Oregon, 1994. Riccio, James, Daniel Friedlander, and Stephen Freedman, GAIN: Benefits, Costs and Three-Year Impacts of a Welfare-to-Work Program, Manpower Demonstration Research Corporation, September 1994. Strawn, Julie, Beyond Job Search or Basic Education: Rethinking the Role of Skills in Welfare Reform, Center for Law and Social Policy, Washington, D.C., 1998. U.S. Department of Health and Human Services (DHHS), Administration for Children and Families, “Change in Welfare Caseloads as of September 1997,” Table at http://www.acf.dhhs. gov/news/case-fam.htm. U.S. Department of Labor, What’s Working and What’s Not: A Summary of the Research on the Economic Impacts of Employment and Training Programs, U.S. Department of Labor, Washington, D.C., 1995. U.S. General Accounting Office, Welfare to Work: Most AFDC Training Programs Not Emphasizing Job Placement, Report to the Ranking Minority Member, Committee on Finance, U.S. Senate, GAO/HEHS-95-113, 1995. Zambrowski, Amy, and Anne Gordon, Evaluation of the Minority Female Single Parent Demonstration: Fifth-Year Impacts at CET, Mathematica Policy Research, Princeton, New Jersey, 1993. 75 About the Authors HANS P. JOHNSON Hans P. Johnson is a research fellow at the Public Policy Institute of California. In addition to adult literacy, his research interests include international and domestic migration, population estimates and projections, and state and local demography. He was previously the senior demographer at the California Research Bureau, where he conducted research for the State Legislature and Governor’s Office on population issues, authoring several publications on migration. 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. SONYA M. TAFOYA Sonya M. Tafoya is a research associate at the Public Policy Institute of California. Her research interests include immigration and California demography. Before joining PPIC, she worked as a biology lecturer and postgraduate researcher at the University of California, Davis. While working as a researcher at Davis, she evaluated the effectiveness of an undergraduate science enrichment program. She holds a B.S. in biology and an M.S. in plant biology from the University of California, Davis. 77" } ["___content":protected]=> string(102) "

R 499HJR

" ["_permalink":protected]=> string(113) "https://www.ppic.org/publication/the-basic-skills-of-welfare-recipients-implications-for-welfare-reform/r_499hjr/" ["_next":protected]=> array(0) { } ["_prev":protected]=> array(0) { } ["_css_class":protected]=> NULL ["id"]=> int(8102) ["ID"]=> int(8102) ["post_author"]=> string(1) "1" ["post_content"]=> string(0) "" ["post_date"]=> string(19) "2017-05-20 02:34:54" ["post_excerpt"]=> string(0) "" ["post_parent"]=> int(3202) ["post_status"]=> string(7) "inherit" ["post_title"]=> string(8) "R 499HJR" ["post_type"]=> string(10) "attachment" ["slug"]=> string(8) "r_499hjr" ["__type":protected]=> NULL ["_wp_attached_file"]=> string(12) "R_499HJR.pdf" ["wpmf_size"]=> string(6) "161736" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(117980) "The Basic Skills of Welfare Recipients: Implications for Welfare Reform Hans P. Johnson Sonya M. Tafoya 1999 Copyright © 1999 Public Policy Institute of California, San Francisco, CA. All rights reserved. PPIC permits short sections of text, not to exceed three paragraphs, to be quoted without written permission, provided that full attribution is given to the source and the above copyright notice is included. Foreword This report is the fourth in a series of studies undertaken by PPIC to understand the consequences of the Personal Responsibility and Work Opportunity and Reconciliation Act of 1996. The authors, Hans P. Johnson and Sonya M. Tafoya, analyze the National Adult Literacy Survey to assess the basic skills of adults on welfare and the likelihood that welfare recipients will be able to find and hold full-time jobs, given their educational background and skill level. In spite of the remarkable reduction in welfare rolls since the reform legislation of 1996, and the sustained growth of the California economy, the findings do not augur well for the poor still on the rolls. Welfare recipients in California are found to have substantially lower basic skills than other adults in the state and the nation, even when compared to other adults with the same level of education. Why, then, are the rolls shrinking and applications for assistance continuing to decline? The authors do not have a direct answer, but they do find that over 50 percent of the adults in California who have basic skills and iii demographic characteristics similar to welfare recipients, but who are not receiving welfare, work at least part time. Most hold jobs intermittently, and the jobs are low-paying. These findings suggest that some welfare recipients could be similarly employed. The authors wave a flag of caution, however, and note that any softening of the economy for a sustained period could hit these workers—the ones with the lowest levels of basic skills—the hardest. There is no major reform of public policy that has come under closer scrutiny than welfare reform. For those who cheer the strong economy and the declining caseload, there are others who see a grim tale of poorly educated and undernourished children whose parents will return to the rolls with the first downturn in the California job market. The authors suggest that improving the basic skills of welfare recipients, although difficult, merits public policy attention; some contact with the job market, however unsteady, is a realistic option for some, if not all, of those currently receiving assistance. Future publications by PPIC will explore this welfare/work relationship in further detail. David W. Lyon President and CEO Public Policy Institute of California iv Summary Large reductions in welfare caseloads have led many to conclude that welfare reform initiated by the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 has been a success. In California, for example, the number of families receiving welfare declined 22 percent from January 1997 to September 1998. Although impressive and certainly one indicator of success, this decline has occurred during a period of strong economic growth. The ability of welfare recipients to transition from welfare to work during a recession is less certain. Even the large decline in welfare caseloads during this current period of economic growth is not necessarily due to welfare recipients’ success in finding work. Some of the decline is due to a drop in the number of applications for welfare. Additionally, some may have left the welfare rolls but not to work, relying instead on friends and family for financial support. Many questions remain about the ability of welfare recipients to find work and the quality of jobs that they do find. v This report addresses the prospects of the nation’s and particularly California’s welfare population as it faces a new welfare system of work requirements, sanctions, and time limits. The report describes the basic skills of welfare recipients and evaluates the employment implications of such skills. In this report, we seek to answer three questions: • How do the basic skills of welfare recipients differ from those of other adults in general and workers in particular? • How much of the gap in skills between workers and welfare recipients can be explained by educational attainment? • What are the labor force status and characteristics of jobs held by persons with skills and characteristics similar to the skills and characteristics of welfare recipients? To answer these questions, we use data from the National Adult Literacy Survey. This nationally representative survey, conducted in 1992, includes a test of basic skills. It assessed the ability of respondents to perform tasks commonly encountered in daily living (e.g., understanding the argument in a newspaper editorial) and tasks that could be encountered in the workplace (e.g., completing a job application). We use several methods to answer the questions posited above, from simple descriptive statistics to logistic regression. These are our major findings: • Welfare recipients have substantially lower basic skills than other adults. In California, for example, almost 80 percent of welfare recipients have either low or very low basic skills, compared to 34 percent of full-time workers in the state. With such poor basic skills, most welfare recipients have difficulty successfully completing tasks commonly encountered in daily living. For example, the average welfare recipient in California has difficulty vi following simple written directions to perform a single mathematical operation (such as addition) using numbers easily located in the text. • Differences in educational attainment between welfare recipients and other adults explain some of the skills gap but not the majority of the gap. About 40 percent of the difference in basic skills scores between welfare recipients and other adults can be attributed to lower educational attainment levels of welfare recipients. However, welfare recipients have substantially lower basic skills than other adults with the same level of education. • We have some cause for optimism: In California, a substantial proportion (58 percent) of adults with basic skills and demographic characteristics similar to welfare recipients are working at least part time. • We also have cause for concern: The jobs held by people whose basic skills are similar to those of welfare recipients are characterized by low wages, intermittent employment, and less than full-time hours. In California, only one-third of adults with basic skills similar to welfare recipients were employed full time year-round. Although the ultimate success of welfare reform will be determined as recipients encounter time limits, the social and individual costs of failure require that we anticipate and respond to potential impediments to success before that time. Our findings suggest that although many welfare recipients can and will find work, a substantial proportion lack the skills for successful integration into the labor force. California faces a greater challenge than most other states: The basic skills of welfare recipients in California are lower than those of welfare recipients in the rest of the nation, and the skills gap between workers and welfare recipients is greater in California than in the rest of the nation. vii The low skills of welfare recipients are not easily amenable to change. Many welfare recipients have graduated from high school, yet even after a dozen years of schooling they are unable to perform simple tasks commonly encountered in the workplace. The track record of training programs is not especially promising. We are also skeptical that on-thejob training will provide these skills—especially considering the types of jobs that welfare recipients might hold. The difficulty in improving the basic skills of welfare recipients does not mean that we should not try. It does mean that we need to be realistic about the costs of providing meaningful training and of improving basic skills. Training programs for improving basic skills need to be critically assessed, with their costs weighed against their benefits. The most promising programs seem to be those that focus on employment and that integrate real job situations into the vocational and basic skills training. Ultimately, we might need to accept that a substantial portion of welfare recipients will continue to need some form of income support, either because their very low skills make them virtually unemployable or because the work they find is of such low quality (and quantity) that they are still living in poverty. viii Contents Foreword ..................................... Summary..................................... Figure and Tables ............................... Acknowledgments ............................... Acronyms .................................... iii v xi xiii xv 1. INTRODUCTION ........................... Historical Welfare Context ....................... California Welfare Context....................... Scope of This Research ......................... 2. DATA AND METHODOLOGY .................. The National Adult Literacy Survey ................. Study Methods .............................. 3. FINDINGS ................................ General Findings ............................. The Skills Gap: Basic Skills of Welfare Recipients and Workers ............................... Educational Attainment and the Basic Skills Gap......... Estimating the Employment Prospects of Welfare Recipients .............................. Labor Force Status .......................... Earnings ................................. 1 3 5 8 11 11 13 17 17 19 26 31 36 38 ix Occupations .............................. Industry ................................. Summary of Findings on Employment Prospects ....... 4. POLICY IMPLICATIONS ...................... Appendix A. The National Adult Literacy Survey: Examples of Tasks and Difficulty Levels ............................. B. The National Adult Literacy Survey: Sampling Design and Scoring ................................... C. NALS and Labor Force Outcomes .................. D. Regressions on Quantitative Basic Skills Score .......... E. Determination of Welfare Counterparts .............. References .................................... About the Authors ............................... 39 41 42 45 49 53 57 61 67 71 77 x Figure 3.1. Percentage Change in AFDC Families, by State, 1993– 1997 .................................. 25 Tables 3.1. Average Basic Skills of U.S. Adults ............... 3.2. Average Basic Skills Scores, by State and for the Nation .. 3.3. Distribution of Basic Skills Scores, by State and for the Nation ................................. 3.4. Average Basic Skills, by Welfare Status: United States ... 3.5. Distribution of Basic Skills, by Welfare Status: United States .................................. 3.6. Average Basic Skills, by Welfare Status: California Compared to the Rest of the Nation .............. 3.7. Distribution of Basic Skills, by Welfare Status: California and the Rest of the Nation ............. 3.8. The Skills Gap and Educational Attainment Levels in the Nation ................................ 18 19 20 21 22 22 23 27 xi 3.9. The Skills Gap and Population Composition Effects in the Nation .............................. 3.10. The Skills Gap and Educational Attainment Levels in California ............................... 3.11. The Skills Gap and Population Composition Effects in California ............................... 3.12. Characteristics of Welfare Recipients, by Work Status ... 3.13. Characteristics of Welfare Recipients and Welfare Counterparts ............................. 3.14. Labor Force Status of Welfare Counterparts and Other Adults ................................. 3.15. Earnings of Welfare Workers, Welfare Counterparts, and Other Adults ............................. 3.16. Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Currently Working Full Time ......... 3.17. Earnings of Welfare Workers with Very Low Basic Skills, Welfare Counterparts with Very Low Basic Skills, and Other Non-Welfare Adults .................... 3.18. Occupational Profile of Welfare Workers, Welfare Counterparts, and Other Adults ................. 3.19. Industrial Profile of Welfare Workers, Welfare Counterparts, and Other Adults ................. B.1. NALS Response Rates ....................... C.1. Wage Equations Using Quantitative Basic Skills and Education as Dependent Variables ............... D.1. Variables Used in Regressions to Evaluate Basic Skills Gap................................... D.2. Descriptive Statistics and Regression Results ......... E.1. Variables Used in Logit Regressions to Identify Welfare Counterparts ............................. E.2. Logistic Regressions Used to Identify Welfare Counterparts ............................. 29 30 30 33 35 36 39 40 41 42 43 54 58 62 63 68 69 xii Acknowledgments We are grateful to Hans Bos, Steve Reder, Margaret O’Brien-Strain, Michael Teitz, and Kim Rueben for their thoughtful reviews of an earlier version of this report. Andrew Kolstad was generous in sharing his time and expertise on the National Adult Literacy Survey. The California Research Bureau of the California State Library generously provided us with the California State Adult Literacy Survey. David Illig and others at the California Research Bureau provided helpful comments when this research was at various stages of development. Once again, Gary Bjork and Joyce Peterson proved to be the best of stylistic reviewers and Patricia Bedrosian an excellent editor. Although this report reflects the contributions of many people, the authors are solely responsible for its content. xiii Acronyms ADC AFDC AFDC-UP CalWORKs DHHS ETS FSA GAIN GED JOBS NALS NCES PRWORA SALS Aid to Dependent Children Aid to Families with Dependent Children Aid to Families with Dependent Children— Unemployed Parent California Work Opportunity and Responsibility to Kids Department of Health and Human Services Educational Testing Service Family Support Act Greater Avenues for Independence Program General Equivalency Diploma Job Opportunities and Basic Skills Training Program National Adult Literacy Survey National Center for Education Statistics Personal Responsibility and Work Opportunity Reconciliation Act of 1996 State Adult Literacy Survey xv SSI TANF WIN Supplemental Security Income Temporary Assistance to Needy Families Work Incentive Program xvi 1. Introduction Since the inception of welfare programs in the United States, one primary goal of policymakers has been to reduce the number of welfare recipients. Particularly over the past few decades, numerous programs have been devised to improve the employment prospects of welfare recipients and lead them forward to self-sufficiency. These welfare-towork programs have focused variously on job searches, unpaid work experience, monetary incentives (e.g., earnings disregards),1 classroom training, and remedial education. The latest and most dramatic incarnation of welfare reform, operating partly under the assumption that welfare recipients lack the proper motivation to work, requires welfare recipients to work after a certain amount of time on aid and limits the total amount of time an individual can receive assistance. The success of welfare reform largely depends on moving people from welfare ____________ 1Earnings disregards provide monetary work incentives for welfare recipients. Rather than reducing welfare benefits by the full amount of earnings, under earnings disregard programs, some welfare recipients who work are able to continue to receive full or only partially reduced benefits. 1 to work. However, it also depends on the duration and wages of that work. Income support programs might still be necessary if one goal of welfare reform is to lift welfare recipients out of poverty. Ascertaining the ability of welfare recipients to find work and determining the quality of the jobs they find are essential to assessing the effectiveness of welfare reform. However, projecting such labor force outcomes is difficult. It is well known that welfare recipients are less educated and less skilled than other adults in the labor force (see, for example, Burtless, 1995; Barton and Jenkins, 1995; MaCurdy and O’Brien-Strain, 1997; Pavetti, 1997; and Reder and Wikelund, 1994). However, it is not clear to what extent these low levels of skills are an impediment to employment. In this report, we use data on basic skills from a national survey of adults to determine the basic skills gap between welfare recipients and other adults and to estimate the employment implications for welfare recipients. We conduct our analyses for both California and the United States. California is an important state to single out for the study of basic skills as they relate to welfare recipients. As of September 1997, California was home to 23 percent of the nation’s welfare (Temporary Assistance to Needy Families—TANF) recipients (U.S. DHHS, 1998). This group, totaling 2,225,893 people, consisted of 663,396 adults and 1,559,497 children. Eighteen percent of families in this group were twoparent families. California spent $4.8 billion on its welfare program in fiscal year 1996–97 (California Department of Social Services, 1998). If the nation is to successfully reform welfare, California, with its large and diverse population, must be considered as a crucial factor in the equation. 2 Historical Welfare Context Aid to Dependent Children (ADC)2 was created in 1935 to ensure income security for mothers who had lost the income of a spouse as a result of death or disability (O’Neill and O’Neill, 1997). Work requirements and work skills, topics now central to welfare policy debates, were not among initial policy concerns, as mothers were not expected to work. However, in the 1960s, when women from every social class began to enter the labor force in large numbers, support for policies that allowed parents to receive public assistance rather than working to support their children began to decline (Jansson, 1997). In 1962, the first federally sponsored work requirement was instituted (Brock, Butler, and Long, 1993). Though small, it was followed by larger federal programs that stressed training and work requirements, thereby introducing the basic skills of welfare recipients as a factor in formulating welfare policy. In 1967, Congress created the Work Incentive Program (WIN), which introduced mandatory training programs for some welfare recipients.3 The program was intended to “reorient welfare toward work” (Gueron and Pauly, 1991). Supervised job searches and unpaid work experience were the main activities of the programs, and earnings disregards were instituted to encourage recipients to work their way off welfare. In practice, however, low enrollment and lack of adequate funding meant that the hope of “reorienting welfare to work” went unfulfilled (O’Neill and O’Neill, 1997; Friedlander, Greenberg, and ____________ 2ADC was the precursor to Aid to Families with Dependent Children (AFDC). 3Those recipients were heads of single-parent AFDC families without preschoolaged children, and heads of two-parent AFDC-UP (-unemployed parent) families (Friedlander, Greenberg, and Robins, 1997). 3 Robins, 1997). AFDC caseloads did not decline and welfare rolls swelled in the early 1970s (O’Neill and O’Neill, 1997; Majority Staff of the Committee on Ways and Means, 1996). In an effort to encourage innovative and cost-effective programs, the Omnibus Budget Reconciliation Act (1981) granted states the flexibility to design their own WIN demonstration projects (O’Neill and O’Neill, 1997; Friedlander, Greenberg, and Robins, 1997; also, see Gueron and Pauly, 1991, for a review of these projects). Based on the most promising of these demonstration projects, the Family Support Act (FSA) was passed in 1988 and the Job Opportunities and Basic Skills Training Program (JOBS) was established to replace WIN in providing federal funds for welfare-to-work program services (Gueron and Pauly, 1991). The FSA stressed the primary responsibility of parents to financially support their children, without changing the entitlement nature of AFDC. JOBS broadened the population of recipients mandated to participate in training and work, increased sanctions for nonparticipation, and committed federal funds to remedial and basic education in welfare-to-work programs (Friedlander, Greenberg, and Robins, 1997; Brock, Butler, and Long, 1993). The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) replaced both the AFDC and JOBS programs with TANF, effectively ending the entitlement nature of AFDC. PRWORA was based on the notion that welfare benefits have had the perverse effect of encouraging a cycle of dependency. The reform emphasizes time limits and work requirements (rather than education and job training), imposing a lifetime limit of 60 months of benefits and a work requirement after a maximum of two years of assistance. States that are unable to move welfare recipients to work face penalties. 4 California Welfare Context Under TANF, each state is given a block grant and some flexibility to design its own welfare program. California’s program is entitled California Work Opportunity and Responsibility to Kids (CalWORKs). New applicants to CalWORKs may receive aid for 18 continuous months, although counties may extend aid for an additional six months.4 As required by federal law, there is a five-year cumulative lifetime limit on aid, although children of adults who reach the lifetime limit will continue to receive aid (California Department of Social Services, 1997). Most welfare applicants are required to first engage in a job search.5 If the job search is unsuccessful, a county employee will conduct an assessment interview with the applicant, during which the applicant and the county will enter an agreement written up as a welfare-to-work plan. Applicants will then participate in welfare-to-work activities for the period specified in the plan. If the time period expires and the applicant has not found unsubsidized work, the county may extend the plan by six months. For new adult applicants, Adult Basic Education, vocational education, and education directly related to employment will qualify as work activities but only in cases where the education is needed to become employed (California Welfare and Institutions Code). Thus, CalWORKs supports the development of basic skills only to the extent that it is necessary to qualify an applicant to enter the workforce. The structure of CalWORKs’ welfare-to-work component has its roots in California’s Greater Avenues for Independence Program (GAIN), which was instituted in 1985, and began operating under the ____________ 4Current recipients may receive aid for 24 consecutive months. 5See the California Welfare and Institutions Code for a list of individuals exempt from this sequence of activities. 5 federal JOBS program in 1989. GAIN has been evaluated closely in six California counties—Alameda, Butte, Los Angeles, Riverside, San Diego, and Tulare. The successes of GAIN, particularly in Riverside County, were cited as examples for the state to follow in establishing welfare-towork programs under PRWORA (Legislative Analyst’s Office, 1997). Initially, the GAIN program differed from previous programs in that it made basic education mandatory for the subset of welfare recipients deemed “in need of basic education.”6 This group of registrants could either attend basic education classes or elect a job search activity. If they chose a job search and failed to obtain employment, they were required to attend basic education classes. The six counties evaluated varied in their emphasis on basic skills’ development versus quick employment (Riccio, Friedlander, and Freedman, 1994). The Riverside program, with its emphasis on “quick employment,” is one of the most successful welfare-to-work programs to date. The Riverside program increased the five-year average of those ever employed by 16 percent, increased average total earnings over five years by 42 percent, and reduced five-year average total AFDC payments by 15 percent (Freedman et al., 1996). The Riverside GAIN “quick employment” strategy has been adopted by CalWORKs. Yet the GAIN results also raise concerns about time limits set by PRWORA, especially for recipients with relatively low basic skills and numerous obstacles to employment. Although federal law allows a state to exempt up to 20 percent of its caseload from the five-year time limit, ____________ 6AB 1371 in 1996 repealed the mandate for basic education and required job search activities as the first activity, except for individuals who lack the education to succeed in even the most unskilled employment (California Assembly Bill 1371 at www.leginfo.ca. gov). 6 results from the GAIN program indicate that a larger exemption may be necessary. For example, by the last quarter of the fifth year, nearly onethird of those in the experimental group in Riverside were collecting AFDC payments—about the same number as in the control group. Similarly, in Los Angeles County,7 approximately half of both the control and experimental groups were collecting AFDC payments in the last quarter of the fifth year (Freedman et al., 1996). This is a discouraging finding, given that about one-third of the state welfare caseload is located in Los Angeles County (Riccio, Friedlander, and Freedman, 1994). When all six counties in the GAIN evaluation were taken into account, and the experimental group was compared with the control group, the five-year average of those ever employed increased by only 7 percent. The average total earnings over five years increased by 23 percent, and the five-year average total AFDC payments fell by 7 percent. The proportion of the experimental group collecting welfare at the end of the fifth year was 39 percent (Riccio, Friedlander, and Freedman, 1994). These results demonstrate that the success of GAIN in California was not universal. Strawn (1998) asserts that low earnings and lack of steady employment account for high levels of AFDC receipt in the fifth year of GAIN. She suggests that these outcomes are the result of quick employment programs, which increase average earnings mostly by helping recipients to work more, rather than helping them to find better jobs. Prior evaluations of GAIN and like programs have yielded similar conclusions, adding that even though these programs have shown ____________ 7In Los Angeles County, GAIN focused exclusively on long-term welfare recipients. 7 success, they have not lifted large numbers of their participants above poverty (U.S. Department of Labor, 1995). It is well known that the earnings capacity of both men and women at risk of need for public assistance has been declining since the 1970s and that the decline has been especially steep since the late 1980s (Brady and Wiseman, 1997). The earlier results of the GAIN program, combined with the realities of the low-skill labor market, highlight the importance of earnings disregards and the Earned Income Tax Credit in alleviating poverty. They also highlight the importance of job skills in successfully making the transition from welfare to work. Scope of This Research Although a primary tenet of the TANF legislation is that able-bodied welfare recipients should work, we do not know much about the labor force skills of welfare recipients. Educational attainment levels of welfare recipients are well known but might not be an adequate measure of a welfare recipient’s employability. In particular, educational attainment levels probably overstate the skills of welfare recipients. For example, high school graduates who are welfare recipients can be expected to be less skilled than high school graduates who are in the labor force. Some past research has compared the skills of welfare recipients to those in the rest of the population (see O’Neill and O’Neill, 1997, for a summary). However, such research has been limited because of relatively small sample sizes, and the skills test, the Armed Forces Qualifying Test, was administered years before entry into the labor force or receipt of welfare. To assess the employment prospects of welfare recipients, a measure of basic skills is necessary, preferably one that is contemporaneous with 8 labor force experience and that also captures the types of skills employers might value. In 1992, the National Center for Education Statistics (NCES) and the Educational Testing Service (ETS) conducted the National Adult Literacy Survey (NALS), administering the survey to a nationally representative group of adults, including welfare recipients. It was the first national scale survey to measure the basic skills of working-age persons contemporaneously with their labor force experience. Twelve states, including California, sponsored increased sample sizes for their states to obtain reliable information at the state level. The goal of NCES and ETS was to assess people’s ability to succeed in dealing with practical analytical problems involving reading, writing, and calculating— problems that they could be expected to encounter in their work, home, and civic lives. For example, the exam included such tasks as completing a job application, calculating the total cost of a purchase from an order form, totaling a bank deposit entry, using a bus schedule, and writing a brief letter explaining an error on a credit card bill (see Appendix A for examples of tasks and levels of difficulty). Because the NALS included a questionnaire rich in demographic and socioeconomic information, we have a source of information that is well suited to the study of the basic skills of the employed, the working poor, and welfare recipients. In this report, we address the prospects of the nation’s, and particularly California’s, welfare population as it faces a new welfare system of work requirements, sanctions, and time limits. Using the NALS database, we examine the characteristics of several groups of respondents, including welfare recipients, heavily dependent welfare recipients, workers not receiving public aid, and other adults. We analyze how the basic skills of welfare recipients differ from those of 9 other adults in general and from workers in particular. Although it is known that welfare recipients are, in general, less educated than workers, we determine how much of the gap in skills between workers and welfare recipients can be explained by educational attainment. Additionally, we identify the types of jobs held by persons with skills and characteristics similar to the skills and characteristics of welfare recipients. Finally, we discuss some of the policy implications of this research. 10 2. Data and Methodology The National Adult Literacy Survey The NALS was conducted in 1992 and included both a national household sample and supplemental household samples for 12 states, including California. The NALS gathered descriptive information and examined proficiency in basic skills for 26,091 respondents aged 16 and older.1 In California, the total sample size was 2,665. All respondents completed a background questionnaire, which provides demographic, linguistic, educational, and socioeconomic information including data on income, work, and public aid. This information was used to characterize the adult population of the United States, to understand factors related to the distribution of basic skills ____________ 1The total number includes the additional samples of approximately 1,000 people per state for each of 12 states that chose to fund additional sampling in their respective states (state samples are referred to as State Adult Literacy Surveys or SALS). These supplemental state samples allow for state-level analyses. See Appendix B for a more complete discussion of the samples and the data. 11 scores, and to compare the NALS results with previous studies. It was also used to summarize the data by various demographic groups and to increase the accuracy of the basic skills estimates for various subpopulations (see Appendix B). Respondents spent approximately 20 minutes completing the background questionnaire and 45 minutes completing a booklet of tasks measuring their prose, document, and quantitative skills. These groups of tasks were scored separately, so that each individual received scores along a prose scale, a document scale, and a quantitative scale. The tasks were designed to measure an individual’s ability to succeed in common, practical, analytical problems. Examples of the tasks are presented in Appendix A. In previous literacy surveys, adult skills were measured by grade-level criteria, such as understanding a sixth-grade vocabulary list, or correctly completing an eighth-grade mathematical exercise. Because such tasks do not reflect the kinds of tasks that adults must routinely perform, they are neither appropriate nor adequate for assessing adult basic skills in the context of assessing employment prospects (Kirsch et al., 1993). Thus, our analysis and discussion of basic skills in this report are based on the scores derived from the NALS.2 Our analyses indicate that basic skills scores are a good predictor of labor force outcomes (see Appendix C). Indeed, basic skills scores are at least as good a predictor of labor force ____________ 2NCES and ETS use the term literacy rather than basic skills. Our experience has been that many people understand literacy as a dichotomous skill (the ability to read and write). NCES and ETS, however, consider literacy to be much less discrete, noting that the NALS shows a wide range of literacy proficiencies. We chose the term basic skills rather than literacy because it is more readily understood as consisting of a range of abilities, and to make explicit that, more than simply testing for the ability to read and write, the exam contained practical reading, writing, calculating, analyzing, and reasoning tasks that adults face in their everyday lives. 12 outcomes as education. This suggests that the NALS exam measures skills that employers value. One potential problem of the survey is its age; although we would not expect a substantial change in the literacy proficiencies of adults since 1992, we might expect changes in the population receiving welfare. In 1992, the nation, especially California, was experiencing a recession. Welfare caseloads were substantially higher in 1992 than they are today. It is reasonable to expect that those most likely to leave welfare in the intervening years were those most skilled. Thus, we would expect our findings to understate the current difference in skills between welfare recipients and other adults. On the other hand, the labor force characteristics of persons not on welfare might have been depressed, in terms of both wages and employment rates during the recession. Because we analyze the employment prospects of welfare recipients by looking at the labor force status and characteristics of jobs of certain adults not on welfare, the survey’s timing might lead to an overstatement of the problems welfare recipients might face. The net effect is uncertain. Study Methods We use several methods to accomplish the various goals of our analyses. In almost all of our analyses, we use the quantitative literacy score as our measure of basic skills and restrict the sample to adults between the ages of 16 and 55 who are not enrolled in high school. We identify welfare recipients as persons who report living in a household that received AFDC, public assistance, or public welfare in the past 12 months.3 Because one of our primary goals is to evaluate the ability of ____________ 3The survey asked separate questions for Supplemental Security Income (SSI) and for Food Stamps. 13 welfare recipients to move off aid and into employment, we chose not to consider persons over the age of 55. Welfare recipients beyond 55 years of age will soon be, if they are not already, eligible for other forms of public assistance. We chose to exclude students still in high school, because those students are not in the labor force, generally not on welfare, and their basic skills are subject to substantial change as they complete more schooling. We might have chosen to exclude college students for the same reasons; however, we did not want to exclude a group of such substantial size and in the same age groups as many welfare recipients. Generally, we focus on quantitative basic skills because they are slightly better predictors of labor force outcomes than either document or prose skills (see Appendix C). In any event, the three types of basic skills are highly correlated and our results did not change in any substantial way when we considered one of the other types of skills. In this report, we first present general findings of the NALS. In describing and comparing the basic skills scores of adults and certain subgroups, we provide simple statistics such as means and distributions. These statistics are weighted to reflect state and national adult populations. Population means are calculated as the weighted mean of individual scores, and standard errors were adjusted to take into account the sampling design and the NALS scoring procedure (see Appendix B). We then develop one set of regression models to evaluate the difference in basic skills between welfare recipients and other adults. Our goal in these regression models is to determine if differences in basic skills can be ascribed to population composition differences between welfare recipients and other adults. Thus, the models we consider attempt to predict an individual’s basic skills score using a prescribed and limited set of variables that identify certain demographic and social characteristics. 14 We conduct separate regressions by welfare status and educational attainment level. The substantive results of this set of regressions are discussed in Chapter 3; Appendix D contains the regression results themselves. Finally, we develop a logistic regression model to predict the receipt of welfare. Our goal in this model is to identify persons similar to welfare recipients in terms of basic skills (and other characteristics) but who did not receive welfare. We seek to characterize the employment status and types of jobs of persons who are similar to welfare recipients in terms of education, basic skills, and some demographic characteristics. The model is described in Appendix E and the findings from the model are discussed in Chapter 3. 15 3. Findings General Findings The results of the NALS suggest that a substantial number of Americans lack fundamental basic skills.1 As shown in Table 3.1, almost one in four American adults has very low basic skills. People at this lowest level can be expected to fail at tasks that are often encountered in an increasingly technical workplace that demands mental rather than physical skills. For example, people in the lowest basic skills level are generally unable to follow simple directions for performing a single mathematical operation using numbers that are easily located in a text (see Appendix A for examples of tasks in the survey). About half of Americans have either very low basic skills or low basic skills. The large number of Americans with low basic skills is disturbing. In their review of the results of the NALS, the authors of the study state: ____________ 1Detailed tables of descriptive statistics for the general adult population are available in Kirsch et al. (1993). 17 If large percentages of adults had to do little more than be able to sign their name on a form or locate a single fact in a newspaper or table, then the levels of literacy seen in this survey might not warrant concern. We live in a nation, however, where both the volume and variety of written information are growing and where increasing numbers of citizens are expected to be able to read, understand, and use these materials (Kirsch et al., 1993). Table 3.1 Average Basic Skills of U.S. Adults Average score Distribution of skills Very low skills Low skills Moderate skills High skills Very high skills Quantitative Skills 271 (0.7) Document Skills 267 (0.7) Prose Skills 272 (0.6) 22 23 21 25 28 27 31 31 32 17 15 17 4 33 SOURCE: NALS; full sample. NOTE: Very low skills correspond to scores less than 225, low skills correspond to scores between 226 and 275, moderate skills correspond to scores between 276 and 325, high skills correspond to scores between 326 and 375, and very high skills correspond to scores above 375. Standard errors in parentheses. The average basic skills scores for Californians are slightly lower than for adults in the rest of the country. Indeed, of the 12 states that participated with NCES to increase sample sizes, California ranked ninth in terms of literacy scores (see Table 3.2). Also, the distribution of basic skills is more extreme in California than in the rest of the nation. As shown in Table 3.3, the proportions of Californians at the very lowest skill level and at the very highest skill level are slightly higher than in the nation as a whole. None of the other states with expanded samples, with the possible exception of Illinois, show a similar pattern. For example, although Iowa and Washington have relatively high proportions of adults 18 Table 3.2 Average Basic Skills Scores, by State and for the Nation (Ranked by Quantitative Mean) State United States Quantitative Mean 271 (0.7) Document Mean 267 (0.7) Washington Iowa Indiana Ohio Illinois Pennsylvania New Jersey Florida California Louisiana New York Texas 293 (4.1) 287 (3.4) 282 (2.3) 280 (2.7) 274 (1.8) 274 (2.5) 273 (2.3) 271 (4.1) 269 (1.7) 261 (4.3) 258 (2.1) 258 (1.9) 288 (3.4) 280 (2.8) 276 (1.7) 276 (2.4) 269 (1.6) 270 (1.9) 268 (1.9) 264 (4.2) 263 (1.8) 257 (3.0) 257 (2.1) 255 (2.0) SOURCE: NALS; full sample. NOTE: Standard errors in parentheses. Prose Mean 272 (0.6) 291 (4.3) 285 (3.0) 281 (1.5) 280 (2.3) 274 (1.5) 275 (1.5) 273 (1.6) 269 (3.2) 270 (1.7) 263 (3.7) 262 (1.9) 259 (2.0) at the highest skill levels, they have relatively low proportions at the lowest skill levels. The distributions in Louisiana, New York, and Texas, on the other hand, are skewed toward the low end of the scale. California’s relatively bipolar distribution mirrors the greater income inequality in the state than in the rest of the nation, suggesting that at least part of the reason for the relatively high income inequality in the state is related to the large variation in basic skills of California residents. The Skills Gap: Basic Skills of Welfare Recipients and Workers Although the basic skills of the adult population in California and the nation are fairly low, the basic skills of welfare recipients are even 19 Table 3.3 Distribution of Basic Skills Scores, by State and for the Nation State United States Percentage of Adults with: High or Very Low Low Moderate Very High Skills Skills Skills Skills 22 25 31 21 California 24 22 30 Florida 21 27 31 Illinois 22 23 31 Indiana 16 27 35 Iowa 15 22 36 Louisiana 26 28 29 New Jersey 24 25 31 New York 28 26 28 Ohio 17 27 33 Pennsylvania 21 25 33 Texas 28 25 29 Washington 10 22 40 24 21 23 23 27 16 20 18 23 21 18 29 SOURCE: NALS; full sample. NOTE: Results are for quantitative skills. Similar results were found for document and prose skills. lower. Not only do welfare recipients tend to be less skilled than the general adult population, they tend to be much less skilled than employed people not receiving aid. In addition, people heavily dependent on welfare, defined as welfare recipients who did not work in the prior year, tend to have even lower skill levels than other welfare recipients. As shown in Table 3.4, welfare recipients scored 55 points lower on average than employed persons on the test of quantitative skills, and persons heavily dependent on welfare scored 72 points lower on average than employed persons. These very low scores mean that the 20 Table 3.4 Average Basic Skills, by Welfare Status: United States Received welfare Heavily welfare dependent Did not receive welfare Employed full time Average Quantitative Score 239 (2.0) 222 (2.6) 287 (0.7) 294 (0.8) SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. Persons are defined as heavily welfare dependent if they did not receive any wage income in the 12 months before the survey. NOTE: Standard errors in parentheses. average person heavily dependent on welfare has difficulty performing simple arithmetic operations, such as addition, and generally cannot perform tasks requiring a single mathematical operation that is not specified in the question (see Appendix A for sample questions and their difficulty level). Half of all welfare recipients in the nation were heavily dependent on welfare. Similar results were obtained for the other types of skills measured by the NALS. Another way to compare the skills of welfare recipients to other persons is to examine the distribution of scores by welfare status. As shown in Table 3.5, 60 percent of welfare recipients and 81 percent of persons heavily dependent on welfare have either low basic skills or very low basic skills. Welfare recipients in California tend to have substantially lower basic skills than welfare recipients in the rest of the nation (see Table 3.6), whereas people heavily dependent on welfare and employed people have 21 Table 3.5 Distribution of Basic Skills, by Welfare Status: United States Very Low Skills Adults 16–55, not in high school (not on welfare) 17.7 Persons employed full time (not on welfare) 10.6 All welfare recipients 26.2 Persons heavily dependent on welfare 49.3 Percentage with: Low Moderate High Skills Skills Skills 25.8 35.5 18.9 20.0 38.2 33.3 30.7 26.4 9.7 31.9 16.1 2.6 Very High Skills 2.1 4.8 0.1 0.0 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. Table 3.6 Average Basic Skills, by Welfare Status: California Compared to the Rest of the Nation Received welfare Heavily welfare dependent Did not receive welfare Employed full time Average Quantitative Score California Rest of Nation 221 (6.2) 242 (2.1) 221 (8.0) 222 (2.7) 279 (2.6) 288 (0.7) 287 (3.1) 295 (0.8) SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Persons are defined as heavily welfare dependent if they did not receive any wage income in the 12 months before the survey. Standard errors in parentheses. only slightly lower basic skills. Thus, the basic skills gap between welfare recipients and employed people is greater in California than in the rest of the nation. This finding is somewhat surprising. Because California provides more generous welfare payments than most states and has a 22 higher proportion of its population receiving welfare,2 we would expect that the California welfare population would include a greater share of moderately skilled persons than the rest of the country. The selection effect into welfare should be less in California than in most other states. However, as shown in Table 3.7, the proportion of welfare recipients with very low skills is substantially higher in California than in the rest of the nation (41 percent compared to 24 percent). In California, almost four of every five welfare recipients have either low or very low basic skills. Table 3.7 Distribution of Basic Skills, by Welfare Status: California and the Rest of the Nation Percentage with: Very Low Low Moderate High Very High Skills Skills Skills Skills Skills California Adults 16–55, not in high school (not on welfare) 22.7 20.4 32.8 21.6 2.6 Persons employed full time (not on welfare) 15.8 18.2 34.4 25.6 6.0 All welfare recipients 41.3 35.6 20.3 2.9 0.0 Persons heavily dependent on welfare 46.5 26.9 23.4 3.2 0.0 Rest of the Nation Adults 16–55, not in high school (not on welfare) 16.9 26.7 35.9 18.5 2.0 Persons employed full time (not on welfare) 9.9 20.3 38.8 26.4 4.6 All welfare recipients 24.1 33.0 32.1 10.7 0.1 Persons heavily dependent on welfare 49.8 32.8 14.8 2.5 0.0 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. ____________ 2In the restricted NALS sample we used, 8.2 percent of adults in California reported receiving welfare compared to 7.7 percent in the rest of the nation. 23 The fact that the skills gap between welfare recipients and employed persons is greater in California coupled with the very low basic skills levels of most California welfare recipients suggests that California will have a more difficult task than most states in moving persons from welfare to full-time work. Indeed, although welfare rolls have declined in California since 1993, the drop has been much less precipitous than in most other states (see Figure 3.1). Only three states have experienced declines that were smaller than California’s between January 1993 and September 1997. Although the relatively small decline in the welfare rolls in California is probably primarily a function of the state’s economy (i.e., the availability of jobs), the state’s more generous welfare benefits, and slower implementation of welfare reform than in some other states, the relatively weak decrease is probably also a reflection of the very low skills of welfare recipients in the state. On the other hand, persons who are heavily dependent on welfare in California are not much less skilled than other welfare recipients in the state. This is in sharp contrast to the rest of the country, where persons heavily dependent on welfare have substantially lower basic skills than other welfare recipients. The small basic skills gap between persons heavily dependent on welfare and other welfare recipients in California is not due to relatively high skills of heavily dependent welfare users in California; rather, it is due to the very low average skill level of all welfare recipients in the state. In addition, California has a higher proportion of heavily dependent welfare users among its welfare population than does the nation (55 percent compared to 48 percent). 24 Wyoming Idaho Wisconsin Oregon Mississippi Alabama South Carolina Oklahoma Tennessee Florida Indiana Louisiana North Dakota Kansas New Mexico Colorado South Dakota Texas Utah Arkansas Michigan New Hampshire Ohio Georgia Massachusetts Virginia Maryland Montana West Virginia North Carolina Maine Kentucky Pennsylvania Arizona Missouri New Jersey Iowa Minnesota Vermont Delaware Illinois Nebraska New York Washington Nevada Rhode Island California Connecticut Alaska Hawaii –100 –80 –60 –40 –20 Percentage change SOURCE: U.S. DHHS (1998). 0 20 40 Figure 3.1—Percentage Change in AFDC Families, by State, 1993–1997 25 Educational Attainment and the Basic Skills Gap Because of a lack of data on the basic skills of welfare recipients, researchers and policymakers have used educational attainment as a proxy for skills. However, it is not clear to what extent educational attainment is an adequate indicator of a welfare recipient’s basic skills. It seems reasonable to expect that welfare recipients have lower basic skills than similarly educated adults who are not on welfare, and the NALS provides us with the opportunity to evaluate the extent to which educational attainment overstates basic skills of welfare recipients compared to other adults. In general, we want to determine whether the skills gap between welfare recipients and the rest of the population can be understood through differences in education and demographic characteristics. Do welfare recipients tend to have low literacy scores solely because they are poorly educated, are more likely to have a disability, and are younger than the general population? As noted previously, the NALS provides us with the unique opportunity to examine this question, since it is the only nationally representative sample of welfare recipients and workers that measures basic skills contemporaneously with labor force and welfare status. In this section, we first examine the basic skills gap between welfare recipients and other adults within educational attainment levels. We then develop regression models to examine the relationship between proficiency in basic skills, education, and welfare, controlling for a host of sociodemographic factors such as gender, age, marital status, California residence, language spoken at home, and mental or physical disabilities. 26 Table 3.8 shows the skills gap by educational attainment level and the distribution of welfare recipients and other adults by educational attainment. As shown in the first two columns of the table, welfare recipients are less educated than other adults. Because people with lower levels of education tend to have lower basic skills, some of the skills gap can be explained by the lower levels of education of welfare recipients. However, as shown in the last three columns, welfare recipients with the same levels of education as other adults tend to have substantially lower basic skills.3 For example, we find that welfare recipients with a high school diploma or GED have quantitative basic skills scores that are 24 points lower on average than those of other adults with a high school diploma or GED. A simple decomposition reveals that if welfare recipients had the same educational attainment distribution as other Table 3.8 The Skills Gap and Educational Attainment Levels in the Nation Percent by Educational Attainment Mean Quantitative Score Difference in Scores Educational Attainment Level 0–8 years 9–12 years High school graduate or GED Some college College graduate Welfare Recipients 11 29 All Other Adults 5 11 Welfare Recipients 158 (7.0) 209 (2.9) All Other Adults 156 (3.4) 227 (2.0) 45 39 251 (2.4) 275 (0.9) 13 24 275 (3.5) 303 (0.9) 2 22 286 (11.6) 332 (0.9) (the Skills Gap) –2 18* 24* 28* 46* SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: * indicates significance at the .01 level. Standard errors in parentheses. ____________ 3The one exception is for adults with 0–8 years of education. For adults with this lowest level of educational attainment, average basic skills scores are extremely low (less than 160) for both welfare recipients and other adults. 27 adults, the basic skills gap would have been reduced by just over 40 percent. In other words, the basic skills gap is partially, but not primarily, explained by differences in education between welfare recipients and other adults. In particular, research that uses education as a proxy for the basic skills of welfare recipients substantially underestimates the skills gap between welfare recipients and other adults. A similar decomposition for California suggests that the basic skills gap between welfare recipients and other adults would be reduced by about 30 points (40 percent of the total difference) if California welfare recipients had the same levels of educational attainment as other adults in the state. Thus, the majority of the skills gap remains unexplained if one considers education alone. Using a regression framework, we also explore whether differences in demographic characteristics, in addition to educational attainment, might explain the differences in basic skills between welfare recipients and other adults (see Appendix D for a discussion of the model). We evaluate differences in basic skills that might be due to differences in age, gender, language spoken at home, and physical and mental disabilities. We find that the basic skills gap between welfare recipients and others persists even when we control for all of these characteristics in addition to educational attainment. In other words, the skills gap between welfare recipients and other adults cannot be fully explained by a host of sociodemographic factors. Even after controlling for mental and physical disabilities, age, gender, language, and marital status, we still find significant differences in basic skills between welfare recipients and similarly educated persons not receiving welfare (see Table 3.9). This persistence in the skills gap indicates that the gap is not merely a population composition effect: Welfare recipients with characteristics 28 Table 3.9 The Skills Gap and Population Composition Effects in the Nation Cumulative Controls No controls within education group Physical/mental disabilities Language Gender and age Skills Gap Between Welfare Recipients and Others Within Specified Educational Attainment Level High School 0–8 9–12 Graduate Some College Years Years or GED College Graduate –2 18 24 28 46 –2 19 23 28 49 9 20 23 28 43 11 24 22 21 32 SOURCE: Authors’ regression models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: The first row corresponds with the last column in the previous table. similar to other adults have lower skills than those other adults. These findings have important implications for designing programs to improve the skill levels of welfare recipients. They suggest that the basic skills deficiencies of most welfare recipients are not due to easily identifiable problems such as English proficiency (or mental disabilities). Similar analyses for California are shown in Tables 3.10 and 3.11. The primary finding that the skills gap persists even controlling for educational attainment and other factors is also true for California. However, we do see some unique California patterns. First, although the overall skills gap between welfare recipients and other adults is larger in California than in the nation, the skills gaps within educational attainment groupings are similar to those in the rest of the nation. Second, less-educated Californians (those who have not attended or graduated from college) have relatively lower basic skills than less- 29 Table 3.10 The Skills Gap and Educational Attainment Levels in California Educational Attainment Level 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent by Educational Attainment Welfare All Other Recipients Adults 16 8 23 10 44 30 15 29 1 24 Mean Quantitative Score Welfare All Other Recipients Adults 129 (11.5) 125 (5.3) 204 (8.5) 212 (6.8) 237 (7.8) 257 (3.9) 281 (6.8) 301 (2.6) 242 (71.9) 332 (2.8) Difference in Scores (the Skills Gap) –4 8 20* 20* 91 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: * indicates significance at the .01 level. Standard errors in parentheses. Table 3.11 The Skills Gap and Population Composition Effects in California Cumulative Controls No controls within education group Physical/mental disabilities Language Gender and age Skills Gap Between Welfare Recipients and Others Within Specified Educational Attainment Level High School 0–8 9–12 Graduate Some College Years Years or GED College Graduate –4 8 20 20 91 –4 7 17 20 91 36 30 28 26 84 36 27 17 11 67 SOURCE: Authors’ regression models from the NALS: sample restricted to adults aged 16 to 55 not in high school. NOTE: The first row corresponds with the last column in the previous table. educated adults in the rest of the nation (compare mean quantitative scores in Tables 3.8 and 3.10). The lower basic skills levels of adults in California compared to the nation can largely be attributed to language differences. California has a greater share of people for whom English is 30 a second language; such people tend to have lower basic skills (as measured in English) than do native English speakers. The difference in scores between less-educated adults in California and the rest of the nation is greatly diminished or eliminated once we control for language. Finally, after controlling for language, the skills gap between welfare recipients and other adults in California is especially large for poorly educated adults. In other words, when we compare welfare recipients with other adults who speak the same language, we observe that the skills gap is quite large among those with little education. Thus, the low basic skills of poorly educated welfare recipients in California is not due to an inability to speak English. Estimating the Employment Prospects of Welfare Recipients CalWORKs requires welfare recipients to work after receiving aid for no more than 24 months and no more than 18 months in the case of new applicants. Given these work requirements, it is important to consider what kinds of work welfare recipients might be able to find, given their skill levels. Determining the likely experience of welfare recipients as they move off assistance is an uncertain undertaking. The success of welfare recipients in the labor force is a function not only of their individual characteristics and circumstances but also of local and nationwide economic conditions (particularly the availability of jobs). Projecting the demand for labor is beyond the scope of this research, but the NALS data do allow us to examine characteristics of welfare recipients and hence enable us to assess welfare recipients’ potential for success in the labor force. 31 To assess the potential labor force outcomes of welfare recipients, we look at two other groups: • Welfare workers —persons who received welfare and who worked at some point in the 12 months before the survey.4 • Welfare counterparts —persons who did not receive welfare but who had similar basic skills and sociodemographic characteristics as welfare recipients. We contrast the labor force characteristics of those two groups with the labor force characteristics of other adults in the nation and in California.5 It is well known that many welfare recipients work while receiving welfare or cycle between work and welfare. In our sample, 48 percent of those who received welfare some time in the year before the survey also reported some earned income in that same year. The labor force ____________ 4Note that we do not know the timing of work and welfare receipt within the prior year. Some welfare recipients worked and received welfare simultaneously; others received welfare during part of the year and worked during other parts of the year. 5We considered whether all low-skill workers might also serve as a proxy for the labor force prospects of welfare recipients. Most welfare recipients, welfare workers, and welfare counterparts are low-skilled. However, most low-skill adults are not welfare recipients and are not in our welfare counterparts group. We rejected all low-skill workers as a proxy for the potential labor force outcomes of welfare recipients. Persons with low skills who do not receive welfare constitute a very different population from welfare recipients, especially in California. Low-skill workers are more likely to be high school graduates, immigrants, male, married, and older than welfare recipients. Low-skill workers are less likely to have children than welfare recipients. Some of these differences are programmatic (for example, it is necessary to have children to receive welfare), but some of these differences, and, in fact, some of the programmatic differences, indicate that there might be very different labor markets for low-skill welfare recipients than for low-skill workers, and also different obstacles in finding work. Together, these differences are substantial and suggest that the type of work that many low-skill workers engage in might not be available to welfare recipients. Our welfare counterparts and welfare workers groups are better proxies for the potential labor force outcomes of welfare recipients because in addition to being primarily low-skill, they also have other characteristics similar to welfare recipients. 32 experience of these welfare workers could be a proxy for the labor force experience of welfare recipients who did not work in the prior year. The labor force experience of welfare workers probably represents an optimistic scenario for the potential labor force experience of recipients who did not work. For example, welfare workers in the nation have substantially higher basic skills, on average, and higher educational attainment levels than the welfare recipients who did not work in the year before the survey (see Table 3.12). In addition, welfare workers Table 3.12 Characteristics of Welfare Recipients, by Work Status Mean basic skills score Educational attainment (%) 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent immigrants Percent female Percent married Percent aged 16–24 25–39 40–54 Percent with mental/physical disability Percent with children aged < 6a California Other Welfare Welfare Workers Recipients 226 (12.0) 219 (7.3) Rest of the Nation Other Welfare Welfare Workers Recipients 258 (3.4) 227 (2.4) 16 17 8 14 21 24 23 34 44 45 49 41 16 14 16 10 3 0 41 34 30 10 13 47 89 58 87 45 24 38 25 30 32 31 28 53 55 51 55 16 13 18 17 0 4 12 58 71 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Welfare recipients who reported income from wages in the 12 months before the survey are classified as welfare workers. Standard errors in parentheses. aCalifornia sample only. 33 were more likely to be married, less likely to have children, and much more likely to be male. We define welfare counterparts as persons in the NALS who were very similar to welfare recipients in terms of basic skills, education, and demographic characteristics but who were not receiving welfare. Welfare counterparts were identified using a statistical model that controlled for quantitative skills score, education, age, disabilities, gender, marital status, and immigrant status.6 As shown in Table 3.13, welfare counterparts are very similar to welfare recipients, with one important exception: Welfare counterparts do not receive welfare. Because welfare counterparts are similar to welfare recipients in terms of basic skills, education, and demographic characteristics, their labor force experience can serve as a proxy for the likely labor force experience of welfare recipients. Of course, there are differences between welfare recipients and workers that are either not measurable or that are not measured in the survey. For example, welfare counterparts might have better access to transportation, live in areas with numerous employment opportunities, have family members who can provide child care, have alternative sources of income, or have healthier or fewer dependents than welfare recipients. Such differences might allow our welfare counterparts to work rather than receive welfare (or to not work and not rely on welfare), but these factors were not measured by the NALS. Because these might be important determinants of welfare dependence, our findings probably represent a best case scenario. That is, the characteristics of jobs held by ____________ 6For California, we also controlled for the presence of children younger than 6 years of age. See Appendix E for a complete discussion of the model. 34 Table 3.13 Characteristics of Welfare Recipients and Welfare Counterparts California Welfare Welfare Counterparts Recipients Entire Nation Welfare Welfare Counterparts Recipients Mean basic skills score Educational attainment (%) 0–8 years 9–12 years High school graduate or GED Some college College graduate Percent immigrants Percent female Percent married Percent aged 16–24 25–39 40–54 Percent with mental/physical disability Percent with children aged < 6a 202 (7.5) 20 26 43 11 0 37 79 33 32 60 8 2 73 221 (6.2) 16 23 44 15 1 32 70 34 31 54 15 2 65 217 (2.0) 13 38 45 4 0 11 86 17 35 55 9 2 239 (2.0) 11 29 45 13 2 11 72 32 30 53 18 1 SOURCE: Authors’ tabulations from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Welfare recipients who reported income from wages in the 12 months before the survey are classified as welfare workers. aCalifornia sample only. welfare counterparts and welfare workers are optimistic measures of the employment potential of welfare recipients in general.7 ____________ 7Some of this bias is offset by the slightly lower skill levels of welfare counterparts. As noted in Table 3.13, mean basic skills scores are slightly lower for welfare counterparts than for welfare recipients. 35 Labor Force Status8 Many welfare recipients will have difficulty finding work. Table 3.14 shows that unemployment rates are substantially higher for welfare counterparts than for other adults, and labor force participation rates are substantially lower. In California, almost 40 percent of welfare counterparts were either unemployed or out of the labor force (i.e., not employed and not looking for work) at the time of the survey. An Table 3.14 Labor Force Status of Welfare Counterparts and Other Adults All Welfare Counterparts California Not in the labor force 27.5% In the labor force, unemployed 14.4% In the labor force, employed part time 16.5% In the labor force, employed semi- permanently full time 8.6% In the labor force, permanently employed full time 33.0% Rest of Nation Not in the labor force 21.4% In the labor force, unemployed 12.3% In the labor force, employed part time 15.9% In the labor force, employed semi- permanently full time 10.4% In the labor force, permanently employed full time 40.1% Welfare Counterparts with Very Low Basic Skills 33.1% 11.2% 17.0% 10.8% 28.0% 29.1% 14.3% 13.8% 11.1% 31.6% Other NonWelfare Adults 11.8% 8.7% 13.8% 7.5% 58.3% 13.1% 6.7% 12.2% 7.6% 60.4% SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Full-time employment is 35 hours per week or more; permanent employment is 40 weeks per year or more. Very low basic skills correspond to basic quantitative skills scores of less than 225. ____________ 8Individuals are either in the labor force employed, in the labor force unemployed, or not in the labor force. 36 additional 23 percent were employed either part time or semipermanently. Welfare counterparts with very low basic skills have especially weak attachments to the labor force. Only 28 percent of very low-skill welfare counterparts in California were permanently employed full time in the year before the survey (compared to 33 percent for all welfare counterparts and 58 percent for the rest of adults in the state). Unemployment rates for welfare counterparts were more than twice those of other adults (20 percent compared to 9 percent). The low labor force participation rates and high unemployment of welfare counterparts might overstate the difficulty of welfare recipients in finding work, since social support systems available to welfare counterparts might not be available to welfare recipients. These social support systems might provide financial support and could lessen the urgency of finding employment for welfare counterparts. However, such support systems might also be important sources of job information and referrals. Still, the very low labor force participation rates of welfare counterparts suggest that many welfare recipients might not transition from welfare to work, but might instead transition from welfare to dependence on friends or family (or, perhaps, homelessness if they lack such support networks). Early reviews of the decline in welfare caseloads indicate that many former welfare recipients do not seem to be employed.9 ____________ 9For example, an analysis of New York state welfare and employment data revealed that a substantial share of former welfare recipients did not appear to be employed in the state of New York (“Most Dropped from Welfare Don’t Get Jobs,” New York Times, March 23, 1998). 37 Earnings Even when persons with the basic skills and sociodemographic characteristics of welfare recipients do find work, their earnings are often not enough to lift them out of poverty. Over the course of an entire year, welfare counterparts in California who worked earned an average income of $12,400, and over half did not have sufficient earnings to lift a family of three out of poverty.10 Table 3.15 shows the distribution of annual income for welfare counterparts, welfare workers, and other workers in California and the rest of the nation. Even if we restrict our analysis to full-time workers, we observe very low average weekly wages for welfare counterparts working full time and for full-time workers who received welfare some time in the past year (see Table 3.16). As with labor force status, the findings are particularly bleak for persons with very low basic skills. Welfare counterparts in California with very low basic skills earned less than $10,000 per year on average, and fully 70 percent did not earn enough to lift a family of three out of poverty (see Table 3.17).11 The low annual earnings of this group reflect, in part, their lack of year-round full-time employment. Intermittent employment is a problem common to many low-skill workers. However, even when we consider weekly earnings of welfare counterparts with very low basic skills who work full time, we still observe very low wage levels. ____________ 10Converting these earnings to 1998 dollars and using the 1998 Earned Income Tax Credit lowers this figure to 44 percent. 11Converting these earnings to 1998 dollars and using the 1998 Earned Income Tax Credit lowers this figure to 52 percent. 38 Table 3.15 Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Workers California Welfare Counterparts 7,917 48.5 17.1 16.1 9.8 6.0 0.9 1.5 Rest of Nation 12,383 33.6 17.5 16.5 11.5 6.4 4.9 9.7 8,937 45.6 23.1 13.2 8.5 4.6 0.6 4.1 10,360 28.8 27.2 22.5 11.2 5.5 2.3 2.6 Other NonWelfare Adults 26,830 12.2 9.4 14.3 12.9 9.3 8.2 33.7 22,445 14.5 12.7 14.9 13.1 10.8 8.7 25.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. Occupations Welfare workers and welfare counterparts are concentrated in occupations that typically consist of low-skill, low-wage, high-turnover jobs. Relative to other adult workers, welfare counterparts and welfare workers are vastly underrepresented in managerial and professional occupations and are especially overrepresented in service sector jobs (see Table 3.18). Although a detailed delineation of the occupations within the broad categories shown in Table 3.18 is not possible given our 39 Table 3.16 Earnings of Welfare Workers, Welfare Counterparts, and Other Adults Currently Working Full Time Average annual earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Full-Time Counterparts Welfare Working Full Workers Time California 13,347 17,468 16.9 11.6 22.4 13.4 26.7 21.7 12.5 15.5 14.2 10.1 2.8 9.3 4.5 15.5 Rest of Nation 14,161 13,187 20.2 11.0 25.4 27.3 20.4 30.6 14.8 15.9 9.5 8.3 1.0 3.6 8.7 3.2 Other Full-Time Workers 32,381 3.5 5.2 13.4 14.2 11.4 9.6 42.7 26,732 4.6 8.9 15.4 15.2 13.1 10.7 32.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Full-time employment is 35 hours per week or more. sample size, we do note that the average wage of welfare counterparts and welfare workers is much lower than that of other workers within the same occupational category (see Table 3.18). For example, welfare counterparts earned about one-third less per week than other adults in service occupations. In no occupational category did welfare counterparts earn more than 70 percent of the earnings of other adults. Weekly earnings of welfare workers are even lower than the earnings of welfare counterparts. 40 Table 3.17 Earnings of Welfare Workers with Very Low Basic Skills, Welfare Counterparts with Very Low Basic Skills, and Other Non-Welfare Adults Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Average annual earnings among those with earnings ($) < 4,999 5,000–9,999 10,000–14,999 15,000–19,999 20,000–24,999 25,000–29,999 30,000 + Welfare Workers with Very Low Basic Skills California Welfare Counterparts with Very Low Basic Skills 6,615 50.2 19.8 17.4 10.5 2.1 0 0 Rest of Nation 9,926 38.4 23.0 18.9 6.2 5.5 1.8 6.2 7,155 48.4 26.7 11.3 6.3 5.9 0.2 1.3 9,458 32.3 33.3 20.0 6.9 2.7 1.9 3.0 Other NonWelfare Adults (Any Skill Level) 28,830 12.2 9.4 14.3 12.9 9.3 8.2 33.7 26,732 14.1 12.7 14.9 13.1 10.8 8.7 25.2 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. NOTE: Very low basic skills correspond to basic quantitative skills scores of less than 225. Industry Welfare workers and welfare counterparts are also concentrated in industrial sectors of the economy that are typified by low-skill, low-wage, high-turnover jobs. As shown in Table 3.19, welfare workers and welfare counterparts are substantially overrepresented in personal services and 41 Table 3.18 Occupational Profile of Welfare Workers, Welfare Counterparts, and Other Adults Welfare Welfare Occupation Workers Counterparts Percentage Among Those with Work Service 32.3 34.6 Farming, forestry, and fishing 4.8 3.1 Technical, sales, and admin. support 21.5 35.5 Precision production, craft, repair 32.2 24.7 Managerial and professional 9.2 2.1 Average Weekly Earnings ($) Service 167 146 Farming, forestry, and fishing 212 148 Technical, sales, and admin. support 211 190 Precision production, craft, repair 312 188 Managerial and professional 421 290 Other Adults 16.6 2.6 32.4 26.3 22.1 210 241 321 337 681 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. agriculture and underrepresented in professional services. Welfare workers and welfare counterparts earn substantially lower wages than other adults employed in the same industry. Thus, not only are welfare workers and welfare recipients concentrated in low-paying industries, they tend to occupy the lower-level jobs within an industry. Summary of Findings on Employment Prospects On the basis of our analysis of the labor force characteristics of welfare workers and welfare counterparts, we find that the labor force prospects of welfare recipients are not especially promising. Welfare recipients are not likely to find jobs that would pay sufficient wages to lift them out of poverty.12 In addition, welfare recipients face a segment of ____________ 12The picture is not quite so bleak if we consider the Earned Income Tax Credit. 42 Table 3.19 Industrial Profile of Welfare Workers, Welfare Counterparts, and Other Adults Welfare Welfare Occupation Workers Counterparts Percentage Among Those with Work Agriculture, forestry, fishing, and mining 3.9 2.8 Construction 5.5 3.9 Manufacturing 19.2 17.5 Trade 26.1 33.9 Personal services 7.2 8.3 Professional service 22.3 17.4 Other 15.9 16.2 Average Weekly Earnings ($) Agriculture, forestry, fishing, and mining 220 136 Construction 243 165 Manufacturing 311 193 Trade 159 133 Personal services 163 159 Professional service 289 174 Other 310 266 Other Adults 3.0 6.2 17.4 20.3 3.6 23.4 26.1 291 374 427 261 199 427 462 SOURCE: Authors’ tabulations and models from the NALS; sample restricted to adults aged 16 to 55 not in high school. the labor market that has relatively high unemployment and low labor force participation, suggesting that many recipients will encounter difficulty in finding employment. Still, we do find that most adults with skills and measurable characteristics similar to welfare recipients are working. Indeed, we find that over half of welfare counterparts are employed. For reasons noted above, these employment rates represent an optimistic scenario for welfare recipients. Because welfare recipients represent a substantial share of the low-skill population in California, their movement off welfare could increase already high unemployment rates and decrease already low labor force participation rates among lowskill residents of the state. Finally, we find that welfare recipients with 43 very low basic skills levels will have the greatest difficulty in transitioning from work to welfare. 44 4. Policy Implications If the goal of welfare reform is to move people off welfare, then it can and will work, even if only in a deterministic manner.1 Huge reductions in caseloads nationwide and in many states suggest that welfare reform has played a part in reducing caseloads even before most individuals are subject to work requirements and elimination from the rolls.2 However, if the goal of welfare reform is to move people from welfare to work, our findings suggest that welfare reform will have mixed results. Some, perhaps the majority, of welfare recipients will find work, but a substantial share will not. It is not clear how those who do not find work will respond to reductions in welfare benefits or outright elimination in eligibility for welfare. If the goal of welfare reform is to improve basic skills via the workplace, our findings suggest that welfare reform will probably not work. The types of jobs welfare recipients are likely to ____________ 1By deterministic we mean the elimination of aid via eligibility requirements. That is, after a welfare recipient has received aid for a lifetime total of 60 months, states can and some will deny further benefits regardless of the recipient’s employment status. 2Of course, the strong economy might account for most of the decline. 45 qualify for do not generally provide the kind of training that could lead to improvements in basic skills and better employment prospects in the future. Finally, if the goal is to lift people out of poverty, our findings indicate limited success. Welfare reform has changed the standard for success in transitioning people from welfare to work. Under TANF and CalWORKs, every ablebodied welfare recipient is expected to work. Failure to work will result in either a reduction in welfare payments or elimination from welfare altogether. Although the ultimate success of welfare reform will be determined after time limits are encountered, the social and individual costs of failure require that we anticipate and respond to potential impediments to success before that time. Our findings suggest that without improvement in basic skills, many welfare recipients will not be successfully integrated into the labor force. California faces an even greater challenge. The basic skills of welfare recipients in the state are lower than those of welfare recipients in the rest of the country, and the skills gap between workers and welfare recipients is larger. The low skills of welfare recipients are not easily amenable to change. Many recipients have graduated from high school, yet even after a dozen years of schooling they are unable to perform simple tasks commonly encountered in the workplace. The track record of training programs is not especially promising.3 We are also skeptical that on the job training ____________ 3Even among programs cited as successful, it is not clear how appropriate they are as a basis for comparison. For example, the Center for Education and Training (in San Jose, California) is commonly cited as a successful program. However, it is a voluntary program for minority female single parents; approximately one-third of the past recipients have never been on welfare. Five years after enrolling in the program, increases in earnings relative to a control group were substantial only for women who entered the program with 12 or more years of education (Zambroski and Gordon, 1993). 46 will provide these skills—especially considering the types of jobs that welfare recipients might hold. The difficulty in improving the basic skills of welfare recipients does not mean that we should not try. It does mean that we need to be realistic about the costs of improving basic skills and of providing meaningful training. Basic skills and training programs need to be critically assessed, with their costs weighed against their benefits. Programs that seem most promising are those that focus on employment and integrate real job situations into the vocational and basic skills training (U.S. Department of Labor, 1995). Those programs should be pursued on a wider basis. Ultimately, we might need to accept that a substantial portion of welfare recipients will continue to need some form of income support, either because their very low skills make them virtually unemployable or because the work they find is of such low quality (and quantity) that they are still living in poverty. 47 Appendix A The National Adult Literacy Survey: Examples of Tasks and Difficulty Levels1 The NALS defines literacy as the “ability to understand and employ printed information in daily activities at home, at work, and in the community to achieve one’s goals and develop one’s knowledge and potential.” As noted in the text of this report, we prefer the term “basic skills.” In the NALS, basic skills are measured on three scales: • Prose skills: the knowledge and skills needed to understand and use information from texts that include editorials, news stories, poems, and fiction—for example, finding a piece of information in a newspaper article, interpreting instructions from a warranty, inferring a theme from a poem, or contrasting views expressed in an editorial. ____________ 1The following examples and discussions are taken from Kirsch et al. (1993). 49 • Document skills: the knowledge and skills required to locate and use information contained in materials that include job applications, payroll forms, transportation schedules, maps, tables and graphs—for example, locating a particular intersection on a street map, using a schedule to choose the appropriate bus, or entering information on an application form. • Quantitative skills: the knowledge and skills required to apply arithmetic operations, either alone or sequentially, using numbers embedded in printed materials—for example, balancing a checkbook, figuring out a tip, completing an order form, or determining the amount of interest from a loan advertisement. Skills levels are grouped by NALS into five categories. The outline below describes those categories for quantitative skills, and provides examples of tasks. Quantitative Level 1 Scale Range: 0–225 Tasks in this level require participants to perform single, relatively simple arithmetic operations such as addition. The numbers to be used are provided and the arithmetic operation to be performed is specified. Example: The respondent is shown a bank deposit slip and asked to figure the total amount of two checks being deposited. They are asked to enter the amount on the form in the space labeled “TOTAL.” Quantitative Level 2 Scale Range: 226–275 Tasks in this level typically require readers to perform a single operation using numbers that are either stated in the task or easily 50 located in the material. The operation to be performed may be stated in the question or easily determined from the format of the material. Example: The respondent is directed to complete an order form for office supplies using a page from a catalogue. No other specific instructions as to what parts of the form should be completed are given in the directive. Quantitative Level 3 Scale Range: 276–325 In tasks in this level, two or more numbers are typically needed to solve the problem, and these must be found in the material. The operation(s) needed can be determined from the arithmetic terms used in the question or directive. Example: The respondent is given a bus schedule and asked the following question. “Suppose that you took the 12:45 p.m. bus from U.A.L.R. Student Union to 17th and Main on a Saturday. According to the schedule, how many minutes is the bus ride?” Quantitative Level 4 Scale Range: 326–375 These tasks tend to require that readers perform two or more sequential operations or a single operation in which the quantities are found in different types of displays, or the operations must be inferred from the semantic information given or drawn from prior knowledge. Example: The respondent is asked to select the information necessary from two price labels to estimate the cost per ounce of creamy peanut butter. The price required for the calculation is given in 51 dollars/lb: The price on the labels is given in dollars and the quantity is given in ounces. Quantitative Level 5 Scale Range: 376–500 These tasks require readers to perform multiple operations sequentially. They must disembed the features of the problem from text or rely on background knowledge to determine the quantities or operations needed. Example: The respondent is asked to look at an advertisement for a home equity loan and then, using the information given, explain how they would calculate the total amount of interest charges associated with the loan. 52 Appendix B The National Adult Literacy Survey: Sampling Design and Scoring The national and state samples were drawn using a four-stage, stratified sampling procedure. The four stages were the primary sampling unit level, followed by the census block level, the household level, and finally the selection of age-eligible individuals. The primary sampling units consisted of counties or groups of counties and were stratified according to census region, metropolitan status, percentage of Black residents, percentage of Hispanic residents, and, whenever possible, per capita income. In the national sample, Black and Hispanic individuals were sampled at a higher rate to increase their representation in the survey. Table B.1 shows response rates for the national and California samples. Although the NALS exam was given only in English, the screener survey was given in both English and Spanish. Response rates in California were similar to those for the nation. The ETS did attempt to correct for non-participation and for non-completion of the 53 Table B.1 NALS Response Rates Instrument Screener Background questionnaire Exercise booklet Overall Percent Completing in Nation California 88.8 87.6 81.9 79.0 95.3 95.3 69.3 66.0 SOURCE: Kolstad et al. (forthcoming). NOTE: Weighted to reflect national adult population. exam. Because low basic skills due to poor proficiency in English should be understood and addressed differently than low basic skills for native English speakers, we control for language in our analyses. Because the goal of the NALS was to produce accurate population estimates of basic skills, a broad range of simulation tasks (165 in total) were administered. However, time did not permit each respondent to answer every question. Thus, each participant responded to a subset of questions (approximately 39 tasks per test booklet), selected such that the 165 tasks were administered to a nationally representative sample. Since some subsets of tasks may have been more difficult than others, basic skills proficiencies could not be reported as a percentage of correct answers. Moreover task-by-task reporting ignores the similarities of subgroups’ response patterns across tasks. These limitations were addressed by using item response theory scaling. The idea behind this scaling is that when several tasks require similar skills, the response patterns should have some regularity. This regularity can be used to characterize both respondents and tasks in terms of a common standard scale. 54 Although each individual completed only a subset of the total number of basic skills tasks, the NALS design allowed for a wide range of content representation when responses are summed for all respondents. The advantage of this design is that it yields more precise population estimates; however, this advantage is offset by the fact that it yields less precise individual estimates. Thus, NALS individual scores are not test scores in the usual sense; rather, they consist of five plausible scores for each of the three basic skills scales. We report the average of these five scores for each individual. Plausible scores were drawn from a posteriori distributions that were a function of the task difficulty of items answered correctly and background variables (gender, ethnicity, languages spoken, region of country, education, parents’ education, occupation, and reading practices). Because these background variables do not include the receipt of public aid, the scoring approach used by ETS reduces our ability to discern differences in the basic skills between welfare recipients and other adults. Because of the complexity of the NALS scoring procedures, even the calculation of descriptive statistics is not entirely straightforward. Individual scores are estimated as the mean of the five plausible values for the given type of skill. Population means are calculated as the weighted mean of individual scores. We report standard errors that are corrected using a design effect of 2.0. The design effect is derived via bootstrap procedures that take into account both the sampling design and the within-individual variation in plausible scores. 55 Appendix C NALS and Labor Force Outcomes We performed a series of regressions to identify the association between NALS scores and earnings. In our regression models using NALS scores as predictors of the log of earnings, we find that NALS scores are at least as strong predictors of earnings as educational attainment. For example, using a restricted sample of males currently working full time, we performed two separate regressions on the log of earnings. In the first regression, using only age and the quantitative skills score as the independent variables, we obtained an R2 value of .24; the second regression, using only age and educational attainment levels as the independent variables, resulted in an R2 value of .21 (see Table C.1). We also performed separate regressions by educational attainment group on the log of earnings, using only age and the quantitative skills score as the independent variables. We find that quantitative skills are a significant predictor of wages within educational attainment groups, with 57 Table C.1 Wage Equations Using Quantitative Basic Skills and Education as Dependent Variables Model 1: Log of earnings as dependent variable and age and quantitative basic skills score as independent variables; full-time male workers. Parameter Estimates Variable INTERCEP DAGE Q5MEAN R2 DF 1 1 1 0.2381 Parameter Estimate 7.550736 0.029478 0.004852 Standard T for HO: Error Parameter=0 0.06075985 124.272 0.00111297 26.486 0.00016285 29.791 Prob > |T| 0.0001 0.0001 0.0001 Model 2: Log of earnings as dependent variable and age and educational attainment (four dichotomous variables with high school graduates as the reference group) as independent variables; full-time male workers. Parameter Estimates Variable INTERCEP DAGE A08 A912 ASOCOLL ACOLPOST R2 DF 1 1 1 1 1 1 0.2234 Parameter Estimate 8.845423 0.029048 –0.523481 –0.268713 0.188651 0.530805 Standard Error 0.04353709 0.00113192 0.05399125 0.04026412 0.02710818 0.02615461 T for HO: Parameter=0 203.170 25.663 –9.696 –6.674 6.959 20.295 Prob > |T| 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 R2 values ranging from .14 to .19 depending on the group.1 In each regression, the quantitative skills coefficient was significant at the .01 level and ranged in value from .0026 to .0046. ____________ 1The lone exception was for individuals with less than an eighth-grade education. In that group, the R2 was only .09. 58 Although the prose, document, and quantitative scores are highly correlated (.93–.95), we selected the quantitative scores in our analyses because Reder and Wikelund (1994) show that higher math gains are associated with lower subsequent welfare utilization. 59 Appendix D Regressions on Quantitative Basic Skills Score To determine whether the basic skills gap between welfare recipients and other adults can be explained as a population composition effect, we performed a series of regressions. In all the models, the dependent variable is quantitative literacy score. Independent variables represent demographic and other individual characteristics. Independent variables were added consecutively to the models, thereby introducing a series of cumulative controls. Separate models were developed for each educational attainment level and by welfare status. To evaluate the population composition effect, we predicted the mean quantitative literacy score for welfare recipients, assuming they had the same population composition characteristics as persons who did not receive welfare. This was done by applying the coefficients from the regressions for welfare recipients to the non-welfare means of the values of the population composition variables. The difference between the mean 61 literacy score for non-welfare adults and the predicted mean literacy score for welfare recipients is taken as the difference in basic skills after adjusting for population composition and is reported in Tables 3.9 and 3.11. Table D.1 describes the variables and in Table D.2 we report variable means and parameter estimates from the models . Table D.1 Variables Used in Regressions to Evaluate Basic Skills Gap Description Dependent variable Q5MEAN Mean of five plausible values for quantitative basic skills score from the NALS Groups of models run separately for educational attainment levels _08yr 1= 0–8 years of education _912yr 1 = 9–12 years of education hsGEDtr 1 = high school graduate or GED completion somecol 1 = attended some college colpost 1 = completed a bachelor’s degree or more Independent variables CA 1 = California resident Disability MENTAL 1 = mental disability PHYSICAL 1 = physical disability Language HBIENG 1= speak English and another language at home HSP 1 = speak Spanish at home HOTHR 1= speak a language other than English or Spanish at home Demographic MALE 1 = male MARITAL 1 = married Age _1618_ 1 = aged 16–18 _1924_ 1 = aged 19–24 _4054_ 1 = aged 40–54 62 Table D.2 Descriptive Statistics and Regression Results Variables INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Sum Mean Model 1 Model 2 Model 3 Educational attainment: _08yr; Welfare: No 726 1.000 166.45*** 169.20*** 199.38*** 174 0.240 –41.41*** –44.16*** –9.21 14 0.019 –8.18 –28.42 19 0.026 –73.79*** –97.86*** 26 0.036 5.26 382 0.526 –69.36*** 41 0.056 –21.83 373 0.514 422 0.581 23 0.032 61 0.084 331 0.456 156.529 Educational attainment: _08yr; Welfare: Yes 173 1.000 166.78*** 167.80*** 195.02*** 40 0.231 –37.59*** –38.61*** 8.66 5 0.029 –27.04*** –54.26*** 0. .. 19 0.110 –4.88 70 0.405 –86.51*** 6 0.035 –52.77*** 45 0.260 58 0.335 12 0.069 26 0.150 48 0.277 158.089 Educational attainment: _912yr; Welfare: No 1576 1.000 229.35*** 230.67*** 237.83*** 179 0.114 –17.59*** –18.91*** 4.80 15 0.010 –24.92 –24.43 13 0.008 –113.12 –120.50 145 0.092 –4.57 214 0.136 –65.78 25 0.016 –27.83 747 0.474 761 0.483 117 0.074 291 0.185 494 0.313 227.356 Model 4 192.80*** –11.87 –26.16 –90.73*** 3.47 –71.04*** –24.99 –0.19 16.22*** 28.55 7.51 –6.03 196.04*** –1.03 –40.08*** . 0.75 –83.03*** –63.28** –1.63 22.37 5.98 4.04 –32.25** 230.99*** 4.24 –20.40 –120.05*** –5.88 –68.93*** –28.49* 1.91 14.64*** 20.82*** 5.66 –10.03** 63 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: _912yr; Welfare: Yes 523 1.000 209.64*** 209.94*** 213.03*** 211.25*** 59 0.113 –5.54 –4.29 6.04 5.49 7 0.013 –9.86 –12.95 –0.93 2 0.004 –81.54 –86.71 –98.18* 37 0.071 4.86 1.82 52 0.099 –46.19*** –45.08*** 3 0.006 8.84 21.60 79 0.151 –5.07 121 0.231 8.79 37 0.071 8.10 150 0.287 8.06 64 0.122 –19.53** 209.012 Educational attainment: hsGEDtr; Welfare: No 5877 1.000 276.34*** 276.95*** 280.82*** 270.78*** 462 0.079 –19.24*** –19.22*** –6.95** –5.54* 24 0.004 –49.86*** –47.04*** –43.46*** 17 0.003 –138.57*** –143.66*** –137.48*** 405 0.069 –6.40* –7.66** 298 0.051 –65.81*** –67.51*** 128 0.022 –48.57*** –49.86*** 2619 0.446 3.62** 3260 0.555 15.95*** 101 0.017 15.53* 851 0.145 2.00 2025 0.345 –2.65 274.832 Educational attainment: hsGEDtr; Welfare: Yes 761 1.000 252.33*** 252.86*** 255.48*** 250.27*** 87 0.114 –15.48* –11.18 –6.58 –5.40 4 0.005 –33.53 –34.88 –35.25 5 0.007 –128.82*** –133.66*** –130.90*** 39 0.051 –8.38 –7.87 56 0.074 –34.63*** –34.81*** 4 0.005 –24.64 –28.34 157 0.206 6.15 192 0.252 6.34 7 0.009 7.15 180 0.237 10.67* 99 0.130 –3.02 250.560 64 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: somecol; Welfare: No 4586 1.000 303.28*** 303.50*** 306.42*** 297.33*** 563 0.123 –2.38 –2.30 1.53 1.82 11 0.002 –54.20*** –50.78*** –48.44*** 4 0.001 –114.73*** –120.31*** –119.80*** 374 0.082 –10.25*** –10.21*** 147 0.032 –52.08*** –52.92*** 100 0.022 –40.51*** –42.04*** 2032 0.443 7.52*** 2250 0.491 9.76*** 24 0.005 6.04 1084 0.236 6.14*** 1305 0.285 –1.73 302.993 Educational attainment: somecol; Welfare: Yes 296 1.000 274.08*** 274.28*** 277.02*** 268.04*** 40 0.135 6.87 6.67 7.81 7.01 2 0.007 –26.23 –28.97 –36.33 0– 0.00 0.00 0.00 24 0.081 –15.50 –13.09 11 0.037 –41.60** –44.59*** 1 0.003 –20.42 –43.32 60 0.203 7.97 76 0.257 24.71*** 1 0.003 6.51 70 0.236 4.84 49 0.166 –0.01 275.005 Educational attainment: colpost; Welfare: No 3940 1.000 331.51*** 331.62*** 335.05*** 324.54*** 442 0.112 0.89 0.78 3.15 3.63 6 0.002 –35.05 –38.48* –41.33* 1 0.000 –166.40*** –157.11*** –145.29*** 314 0.080 –12.72*** –12.97*** 68 0.017 –51.99*** –52.21*** 176 0.045 –39.94*** –41.89*** 1928 0.489 11.49*** 2321 0.589 9.62*** 0. . 242 0.061 –4.85 1562 0.396 –1.06 331.608 65 INTERCEP CA PHYSICAL MENTAL HBIENG HSP HOTHR MALE MARITAL _1618_ _1924_ _4054_ Q5MEAN Table D.2 (continued) Educational attainment: colpost; Welfare: Yes 43 1.000 287.81*** 287.81*** 298.94*** 2 0.047 –46.08 –46.08 –38.74 0. .. 0. .. 4 0.093 –36.93 3 0.070 –79.07* 4 0.093 –27.07 16 0.372 16 0.372 0. 3 0.070 25 0.581 285.667 284.55*** –43.59 . . –34.87 –85.43** –53.73 –12.92 52.56*** . –3.18 4.84 66 Appendix E Determination of Welfare Counterparts We use logistic regression models to predict the probability of receiving welfare. We defined welfare counterparts as persons who did not receive welfare but who were predicted to be welfare recipients by the model. In the logistic regression framework, welfare counterparts are false positives. The goal of the regressions is to identify persons who are not welfare recipients but who have characteristics associated with the receipt of welfare. That is, we want to identify a population that is very like welfare recipients to determine what kinds of labor force outcomes welfare recipients might achieve as they move off welfare. We developed two models: one for California, and one for the rest of the United States. The models were developed separately because the California sample includes a question on the presence of children younger than six years of age in the household, whereas the sample in the rest of the nation does not include such information. The presence of children younger than six years old is an important predictor of welfare 67 receipt. Variables used in the models are described in Table E.1 and the results are shown in Table E.2. Some variables that might be highly predictive of welfare receipt were intentionally left out of the model. For example, although income is a strong predictor of welfare receipt, to place it in the model would inappropriately prescribe our findings (apart from problems of endogeneity). Table E.1 Variables Used in Logit Regressions to Identify Welfare Counterparts Description Dependent variable AFDCPAPW 0 = did not receive welfare in the prior year 1 = received welfare in the prior year Independent variables Education A08 1 = 0–8 years of education A912 1 = 9–12 years of education ASOCOLL 1 = some college ACOLPOST 1 = completed a bachelor’s degree or more Age _1618_ 1 = aged 16–18 _1924_ 1 = aged 19–24 _4054_ 1 = aged 40–54 Q5MEAN MENTAL PHYSICAL MALE MARITAL USA KIDS6 Mean of five plausible values for quantitative basic skills score from NALS 1 = mental disability 1 = physical disability 1 = male 1 = married 1 = U.S. born 1 = has children younger than six years of age (California regression only) NOTE: Omitted or reference categories are high school graduates and persons aged 25–39. 68 Table E.2 Logistic Regressions Used to Identify Welfare Counterparts Model 1: Logistic Regression for the Rest of the United States Number of Observations: 16485 Response Profile Ordered Value 1 2 AFDCPAPW 1 0 Count 1572 14913 Model Fitting Information and Testing Global Null Hypothesis BETA = 0 Intercept Criterion Only –2 LOG L 10377.810 R2 = 0.1284 Intercept and Covariates Chi-Square for Covariates 8112.043 2265.768 with 13 DF (p = 0.0001) Max-rescaled R2 = 0.2749 Variable INTERCPT A08 A912 ASOCOLL ACOLPOST _1618_ _1924_ _4054_ Q5MEAN MENTAL PHYSICAL MALE MARITAL USA DF 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Analysis of Maximum Likelihood Estimates Parameter Estimate 0.7698 0.2237 0.5343 –0.5584 –1.7277 –0.8006 –0.00961 –1.0755 –0.00975 –2.8920 0.5078 –1.3033 –1.0112 0.8666 Wald Standard Chi- Error Square 0.1752 19.3072 0.1321 2.8688 0.0782 46.6644 0.0814 47.0306 0.1679 105.8286 0.1845 18.8214 0.0735 0.0171 0.0804 178.9236 0.00062 247.3045 0.6302 21.0576 0.3367 2.2741 0.0701 345.7192 0.0660 234.7345 0.1240 48.8696 Pr > ChiSquare 0.0001 0.0903 0.0001 0.0001 0.0001 0.0001 0.8959 0.0001 0.0001 0.0001 0.1316 0.0001 0.0001 0.0001 Standardized Estimate . 0.024616 0.093225 –0.134960 –0.391130 –0.057927 –0.001936 –0.278323 –0.331453 –0.091945 0.019697 –0.355978 –0.278607 0.131319 Odds Ratio . 1.251 1.706 0.572 0.178 0.449 0.990 0.341 0.990 0.055 1.662 0.272 0.364 2.379 69 Table E.2 (continued) Model 2: Logistic Regression for California Number of Observations: 2071 Ordered Value 1 2 Response Profile AFDCPAPW 1 0 Count 233 1838 Model Fitting Information and Testing Global Null Hypothesis BETA = 0 Intercept Criterion Only –2 LOG L 1456.836 R2 = 0.1871 Intercept and Covariates Chi-Square for Covariates 1027.799 429.038 with 14 DF (p = 0.0001) Max-rescaled R2 = 0.3704 Analysis of Maximum Likelihood Estimates Variable INTERCPT A08 A912 ASOCOLL ACOLPOST _1618_ _1924_ _4054_ Q5MEANCA MENTAL PHYSICAL MALE MARITAL USA KIDS6 Wald Parameter Standard ChiDF Estimate Error Square 1 –0.5536 0.4196 1.7404 1 –0.1170 0.3032 0.1489 1 0.0644 0.2300 0.0785 1 –0.6243 0.2293 7.4099 1 –2.8346 0.7362 14.8260 1 1.1970 0.4223 8.0356 1 –0.1228 0.2128 0.3331 1 –0.2389 0.2370 1.0161 1 –0.00726 0.00175 17.1394 1 1.0530 1.0546 0.9971 1 0.0675 1.2346 0.0030 1 –0.7870 0.1821 18.6720 1 –1.6156 0.1936 69.6693 1 0.9102 0.2477 13.4985 1 1.8693 0.1911 95.7177 Pr > ChiSquare 0.1871 0.6995 0.7794 0.0065 0.0001 0.0046 0.5638 0.3135 0.0001 0.3180 0.9564 0.0001 0.0001 0.0002 0.0001 Standardized Estimate . –0.019640 0.011333 –0.156393 –0.641525 0.085084 –0.025522 –0.060130 –0.320705 0.031211 0.002002 –0.216368 –0.445121 0.226330 0.485542 Odds Ratio . 0.890 1.067 0.536 0.059 3.310 0.884 0.788 0.993 2.866 1.070 0.455 0.199 2.485 6.484 70 References Barton, Paul E., and Lynn Jenkins, Literacy and Dependency: The Literacy Skills of Welfare Recipients in the United States, Educational Testing Service, Princeton, New Jersey, 1995. Brady, Peter, and Michael Wiseman, Welfare Reform and the Labor Market: Earnings Potential and Welfare Benefits in California, 1972–1994, Institute for Research on Poverty, University of Wisconsin-Madison, Wisconsin, 1997. Brock, Thomas, David Butler, and David Long, Unpaid Work Experience for Welfare Recipients: Findings and Lessons from MDRC Research, Manpower Demonstration Research Corporation, New York, 1993. Burtless, Gary, “Employment Prospects of Welfare Recipients,” in Demetra Smith Nightingales and Robert H. Haveman (eds.), The Work Alternative, The Urban Institute, Washington, D.C., 1995. California Department of Social Services, “Fact Sheets,” 1998. http://www.dss.cahwnet.gov/calworks/default.htm California Department of Social Services, “Fact Sheet: California Welfare Population, September 1997,” 1997. www.dss.cahwnet. gov. 71 California State Senate, Senate Health and Human Services Committee, Senate Floor Committee Analysis, Bill No. AB 1542 [Welfare Reform], August 1997. California State Legislature, Welfare Reform Conference Committee, Job Creation Package, June 1997. Edin, Kathryn, and Laura Lein, Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work, The Russell Sage Foundation, New York, 1997. Freedman, Stephen, Daniel Friedlander, Winston Lin, and Amanda Schweder,The GAIN Evaluation Working Paper 96.1: Five-Year Impacts on Employment, Earnings and AFDC Receipt, Manpower Demonstration Research Corporation, New York, 1996. Friedlander, Daniel, D. H. Greenberg, and P. K. Robins, “Evaluating Government Training Programs for the Economically Disadvantaged,” Journal of Economic Literature, Vol. 35, December 1997, pp. 1809–1855. Friedlander, Daniel, and Gary Burtless, Five Years After: The Long Term Effects of Welfare-to-Work Programs, Russell Sage Foundation, New York, 1995. Gueron, Judith M., and Edward Pauly, From Welfare to Work, Russell Sage Foundation, New York, 1991. Jansson, Bruce S., The Reluctant Welfare State: American Social Welfare Policies—Past, Present, and Future, Brooks/Cole Publishing, Pacific Grove, California, 1997. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in California: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Florida: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. 72 Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Illinois: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Indiana: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Iowa: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Louisiana: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in New Jersey: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in New York: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Ohio: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Pennsylvania: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Texas: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Jenkins, Lynn B., and Irwin S. Kirsch, Adult Literacy in Washington: Results of the State Adult Literacy Survey, Educational Testing Service, Princeton, New Jersey, May 1994. Kirsch, Irwin S., Ann Jungeblut, Lynn Jenkins, and Andrew Kolstad, Adult Literacy in America: A First Look at the Results of the National 73 Adult Literacy Survey, National Center for Education Statistics, September 1993. Kolstad, Andrew, et al., Technical Report and Data File Users Manual for the National Adult Literacy Survey, National Center for Educational Statistics, Washington, D.C., forthcoming. Legislative Analyst’s Office, Welfare Reform in California: A Welfare-toWork Approach, Sacramento, California, 1997. Lord, Frederic M., Applications of Item Response Theory to Practical Testing Problems, Educational Testing Service, Hillsdale, New Jersey, 1980. MaCurdy, Thomas, and Margaret O’Brien-Strain, Who Will Be Affected by Welfare Reform in California? Public Policy Institute of California, San Francisco, California, 1997. Majority Staff of the Committee on Ways and Means, 1996 Green Book, November 4, 1996. Martinson, Karin, and Daniel Friedlander, GAIN Basic Education in a Welfare to Work Program, Manpower Demonstration Research Corporation, New York, 1994. Mislevy, R. J., A. E. Beaton, B. Kaplan, and K. M. Sheehan, “Estimating Population Characteristics from Sparse Matrix Samples of Item Responses,” Journal of Educational Measurement, Vol. 29, No. 2, Summer 1992, pp. 133–161. “Most Dropped from Welfare Don’t Get Jobs,” New York Times, March 23, 1998. Olson, K., and La Donna Pavetti, Personal and Family Challenges to the Successful Transition from Welfare to Work, The Urban Institute, Washington, D.C., 1996. O’Neill, Dave M., and June Ellenoff O’Neill, Lessons for Welfare Reform: An Analysis of the AFDC Caseload and Past Welfare-to-Work Programs, W. E. Upjohn Institute for Employment Research, Kalamazoo, Michigan, 1997. 74 Pavetti, LaDonna, How Much More Can They Work? Setting Realistic Expectations for Welfare Mothers, The Urban Institute, Washington, D.C., 1997. Reder, Stephen, and K. R. Wikelund, Steps to Success: Literacy Development in a Welfare-to-Work Program, Northwest Regional Educational Laboratory, Portland, Oregon, 1994. Riccio, James, Daniel Friedlander, and Stephen Freedman, GAIN: Benefits, Costs and Three-Year Impacts of a Welfare-to-Work Program, Manpower Demonstration Research Corporation, September 1994. Strawn, Julie, Beyond Job Search or Basic Education: Rethinking the Role of Skills in Welfare Reform, Center for Law and Social Policy, Washington, D.C., 1998. U.S. Department of Health and Human Services (DHHS), Administration for Children and Families, “Change in Welfare Caseloads as of September 1997,” Table at http://www.acf.dhhs. gov/news/case-fam.htm. U.S. Department of Labor, What’s Working and What’s Not: A Summary of the Research on the Economic Impacts of Employment and Training Programs, U.S. Department of Labor, Washington, D.C., 1995. U.S. General Accounting Office, Welfare to Work: Most AFDC Training Programs Not Emphasizing Job Placement, Report to the Ranking Minority Member, Committee on Finance, U.S. Senate, GAO/HEHS-95-113, 1995. Zambrowski, Amy, and Anne Gordon, Evaluation of the Minority Female Single Parent Demonstration: Fifth-Year Impacts at CET, Mathematica Policy Research, Princeton, New Jersey, 1993. 75 About the Authors HANS P. JOHNSON Hans P. Johnson is a research fellow at the Public Policy Institute of California. In addition to adult literacy, his research interests include international and domestic migration, population estimates and projections, and state and local demography. He was previously the senior demographer at the California Research Bureau, where he conducted research for the State Legislature and Governor’s Office on population issues, authoring several publications on migration. 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. SONYA M. TAFOYA Sonya M. Tafoya is a research associate at the Public Policy Institute of California. Her research interests include immigration and California demography. Before joining PPIC, she worked as a biology lecturer and postgraduate researcher at the University of California, Davis. While working as a researcher at Davis, she evaluated the effectiveness of an undergraduate science enrichment program. She holds a B.S. in biology and an M.S. in plant biology from the University of California, Davis. 77" ["post_date_gmt"]=> string(19) "2017-05-20 09:34:54" ["comment_status"]=> string(4) "open" ["ping_status"]=> string(6) "closed" ["post_password"]=> string(0) "" ["post_name"]=> string(8) "r_499hjr" ["to_ping"]=> string(0) "" ["pinged"]=> string(0) "" ["post_modified"]=> string(19) "2017-05-20 02:34:54" ["post_modified_gmt"]=> string(19) "2017-05-20 09:34:54" ["post_content_filtered"]=> string(0) "" ["guid"]=> string(50) "http://148.62.4.17/wp-content/uploads/R_499HJR.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) }