Donate
Independent, objective, nonpartisan research

Stackable Credentials in Career Education at California Community Colleges, Technical Appendix

Database

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

View live version

object(Timber\Post)#3742 (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(20) "1018sbr-appendix.pdf" ["wpmf_size"]=> string(6) "429818" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(43266) "Stackable Credentials in Career Education at California Community Colleges Technical Appendices CONTENTS Appendix A. Data and Sample Construction Appendix B. Analysis of Student Outcomes Sarah Bohn and Shannon McConville With research support from Bonnie Brooks and Sophie McGuinness Supported with funding from the ECMC Foundation, the James Irvine Foundation, and the Sutton Family Fund Appendix A. Data and Sample Construction Data sources The data used in this report come from the California Community College Chancellor’s Office (CCCCO). The Chancellor’s Office Management Information System (COMIS) is a longitudinal dataset that includes information on all students enrolled in colleges across the California Community College system. This dataset includes detailed information on student characteristics, including demographics, measures of economic and academic disadvantage, and disability, along with course enrollment, financial aid receipt, and award completion. We have these records for all students between the fall term of 1993 and the spring term of 2017. The data system also contains information on the awards or credentials that a student completes.1We classify all awards that students earn in the system into three categories—short-term certificates, long-term certificates, and associate degrees. Short-term and long-term certificates are defined based on the length of time, measured in terms of units and assuming full-time course loads, it takes to complete the degrees. Short-term certificates take less than 1 year to complete and include certificates requiring less than 30 units. Long-term certificates are defined as those requiring between 1 and 2 years to complete and include certificates requiring 60 or more units and those requiring 30-59 units. Associate degrees can be either associate of art or science and typically take 2 or more years to complete. All courses and awards include information that designates a specific field of study called a Taxonomy of Program (TOP) code. The TOP system of numerical codes is used to collect and report information on programs and courses in different colleges throughout the state that have similar outcomes. We use TOP codes to identify career education awards. The CCCCO designates all career education or vocational programs based on the 6-digit TOP code. The TOP codes were designed to aggregate information about programs and all courses and awards are coded with a 6-digit TOP code. The first two digits of the six-digit TOP denote the discipline and is used to define our broad career education areas which include Business and Management (05), Information Technology (07), Engineering (09), Family and Consumer Sciences (13) and Public and Protective Services (21). The first four digits are intended to denote a sub-discipline (e.g. 1305: Child Development/Early Childhood Education), and the entire six digits denote a specific field of study (e.g. 130580: Child Development Administration and Management). Defining Stackable Credential Features In order to systematically examine stackable credentials, we developed a strategy that would allow us to identify career education programs based on interrelated credentials they offer and to classify those credential features. Although programs are intuitive, akin to college departments like culinary arts, automotive technology, information technology, etc., no statewide data source tracks programs in this way. The TOP code contained in COMIS is both too broad and too detailed. A sub-discipline (4-digit TOP) typically contains multiple programs and sets of credentials that have no overlapping course requirements. The field of study (6-digit TOP) is too detailed, with credentials in different fields of study actually closely related in terms of course requirements and 1 One caveat in our research pertains to the potential underreporting of short-term certificates that are not approved by the Chancellor’s office. These ‘local’ certificates are college- and department-specific certificates less than 12 units. Because the CCCCO does not require colleges to report local certificates in COMIS—although many do—non-reported local certificates are not included in our analysis. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 2 career options. Furthermore, there is a lot of variation across colleges and program areas in what may or may not be offered in a sub-discipline or field of study. To appropriately identify career education programs and classify students according to their program awards, we first attempted to connect information available from COMIS on the types of awards students earn (i.e. TOP codes and credential length) to the inventory of programs and courses available from the Chancellor’s office. Unfortunately that approach was not workable due to the high levels of mismatch—there were too many credentials in COMIS that did not have a match to the inventory and vice-versa. In the end, we determined the best way to identify and describe programs according to their stackable pathways would require a detailed scan of college course catalogs that would allow us to explicitly connect credentials based on overlapping course requirements and other available descriptors of programs and credentials. To employ this more qualitative approach, we also needed to identify a subset of colleges on which to focus our scan in order to make the task more manageable. At the same time, we wanted to ensure that the colleges we scanned had career education programs that had enough different types of credentials available to make stackable credentials a possibility. We selected a group of colleges to conduct in-depth catalog scans for each of the five focal career education disciplines based on the number of awards they conferred by credential length (short-term, long-term, associate). First, we sorted all schools by the number of certificates (long-term and short-term) and also associate degrees they conferred in school year 2015-16 and selected colleges with the highest number of each type of credentials that cumulatively added up to 50% of the total awards systemwide that were conferred in that career education discipline. In cases where the number of schools needed to encompass 50% of associate degrees was close to that of certificates, we chose whichever had the greater number of schools. We find that this selection method organically provided a large and diverse sample of programs with stacking potential. Using this method, across all disciplines we identified 64 community colleges overall to be scanned and between 25 and 32 colleges for each career education discipline (business, IT, engineering, family/consumer science, and public/protective services). Second, from the selection of colleges identified for the scan, we used available information from COMIS and the CCCCO program inventory to enumerate all of the specific fields of study available in the career education discipline based on 6-digit TOP code and the program names and descriptors provided in the inventory. The TOP codes were necessary so that we could link the scan information back to the student-level COMIS data and the program names assisted us in finding the programs of interest in the course catalogs. With our scan information in college catalogs to answer questions aimed to identify the differing degrees to which a program fulfills the characteristics of three types of pathways to stacking credentials: PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 3 TABLE A1 Stackable pathway definitions and features Pathway type Progressive Lattice Definition Students can upgrade from lower-level credentials to higher-level credentials (vertical) by completing additional coursework within the same program Students complete a group of courses that serve as a “Launchpad” to multiple same-level (horizontal) and/or higher-level (vertical) credentials Features Are there smaller credentials that can be upgraded to higher-level credentials (certificate or associate) by adding additional units in the CTE area? If yes, is there a sequence that involves only certificates, not associate degrees? Does the catalog explicitly state there are levels or stages within a program (i.e. using terms like beginner, intermediate, advanced OR I, II, III in program titles) Are there 3 or more degrees and certificates that share a core group of (2+) classes? Is the shared group of courses called out (identified as a “core”) in the catalog to indicate that it can lead to multiple credentials? Even if the core is not directly labeled “core,” does the group of core courses alone earn the student either a local certificate or a certificate of achievement? After coding each field of study using the guiding questions, programs that share certain features (and subsequently share stacking potential) we assign a unique code called the Internal Control Number (ICN) to programs that share coursework. The ICN is intended to identify sequences of credentials by linking programs that share coursework across six-digit TOP codes in a given college. In this way we can connect multiple fields of study within a discipline that could be part of stackable credential pathway. The definition of the ICN is listed below. ICN = College ID + Two-Digit TOP Code + First-Degree Connection + Second-Degree Connection To determine if programs share either a first-degree or second-degree connection we ask the following questions. For the first-degree connection: at least 2 course requirements come from other programs? If so, assign the same digit to programs that overlap by at least 2 courses. For the second-degree connection: are programs themselves pieces of the same program sequence (and/or have 3-4 classes that overlap)? If so, assign the same digit to programs that overlap by 3-4 classes. In the final step, we link the information on pathway features at the ICN level (8-digit code that includes the 1st and 2nd degree of connection) to the COMIS data by college and 6-digit TOP code. Table A2 provides information on the pathway features for each of the two stackable pathway types we define at the ICN 8-digit level as it provides the most granular detail on pathway characteristics. In our main analysis, we use the ICN 7digit code to determine if students stack additional credential along a program pathway to allow for a more broad definition of potential connection between programs. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 4 TABLE A2 Characteristics of programs across career education disciplines Pathway characteristics Business Progressive Potential to upgrade credentials Certificate-only paths Explicit credential sequence Lattice Common set of course requirements for multiple credentials Explicit core/launchpad Core-alone earns credential Total programs identified 106 (82%) 72 (56%) 18 (14%) 101 (79%) 14 (11%) 0 (0%) 129 Engineering 137 (87%) 82 (52%) 10 (6%) 58 (37%) 3 (2%) 7 (4%) 157 Public and protective services 66 (71%) 20 (22%) 4 (4%) 48 (52%) 4 (4%) 1 (1%) 93 Information technology 41 (56%) 26 (36%) 7 (10%) 28 (39%) 5 (7%) 4 (6%) 72 Family and consumer sciences Total 95 (90%) 64 (61%) 27 (25%) 445 (80%) 264 (47%) 66 (12%) 63 (60%) 1 (1%) 6 (6%) 105 298 (54%) 27 (6%) 18 (3%) 556 Sample Construction We build our sample universe from the COMIS data file that records the awards students earn. For the purposes of this report, we restrict our sample to those students who earned their first observed award (at least since 1993) within the community college system between the years 2000 and 2014—so that we can observe students for at least three years following the completion of their first community college award. We further restrict our sample to students who have an SSN (scrambled and de-identified in our data extract) recorded as an identifier. Each community college assigns school-level student identification numbers, but these are not unique across the entire system. Restricting our sample to students with an SSN as an identifier allows us to track students across multiple colleges so we have a complete picture of their coursework and credentials. For our examination of the broad trends in career education credentials presented in the first section of the report, we include all students age 18 to 54 who earn their first community college credential between school years 2000–01 and 2013–14. We designate students earning career education credentials based on the designation of vocational programs provided in the Taxonomy of Programs (TOP) manual 6th Edition (California Community Colleges, Academic Affairs Division, July 2013). Table A3 provides descriptive statistics across different student groups—including all students who earned a credential regardless of if it was in career education, all students who earned any career education credential as their first community college award with breakdowns by the level of their first credential. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 5 TABLE A3 Sample means by group Total Age at first award Mean 18-24 25-29 30-34 35-39 40-54 Sex Male Female Race/Ethnicity White Latino African American Asian/Pacific Islander Other/Multi/Missing Citizenship status Citizen Non-Citizen Unknown Highest level of education No high school diploma High school diploma/GED Associate degree Bachelor’s degree or higher Other/Missing Markers of disadvantage Low income (Pell or age 25+ and BOGW) Limited English proficiency Academic disadvantage Student with disabilities Ever CalWORKs Markers of ability Ever developmental math or English Ever transfer math or English GPA Financial aid Ever BOG Waiver All students 1,131,711 Any career education, 1st award 512,377 Highest level of 1st career education award Associate degree Long-term certificate Short-term certificate 222,241 95,205 194,931 27.6 52.2% 17.7% 10.0% 7.0% 13.1% 40.5% 59.5% 30.1% 15.9% 7.1% 42.9% 4.0% 86.6% 9.6% 3.8% 3.7% 78.8% 7.8% 4.3% 5.4% 51.1% 8.2% 63.1% 7.6% 4.6% 43.7% 51.4% 2.07 60.9% 30.4 36.5% 20.1% 13.3% 10.0% 20.0% 43.2% 56.8% 30.6% 16.1% 7.4% 42.3% 3.6% 84.9% 10.9% 4.1% 4.4% 74.2% 7.1% 8.3% 6.0% 53.5% 10.0% 62.1% 8.1% 6.4% 36.0% 28.8% 2.13 60.3% 29.6 38.7% 22.1% 13.4% 9.4% 16.3% 38.5% 61.5% 42.8% 28.7% 6.6% 18.5% 3.4% 83.9% 11.3% 4.8% 2.8% 72.3% 12.2% 6.9% 5.8% 56.7% 9.3% 64.1% 7.1% 5.4% 39.6% 43.9% 2.15 63.3% 31.1 32.9% 20.2% 14.2% 10.6% 22.1% 45.8% 54.2% 41.3% 30.8% 8.9% 15.0% 4.0% 84.9% 11.2% 3.9% 4.9% 78.1% 3.3% 8.5% 5.3% 54.8% 9.8% 61.8% 9.1% 7.4% 31.4% 15.1% 2.13 61.2% 31.0 35.9% 17.9% 12.8% 10.4% 23.1% 47.3% 52.7% 42.2% 32.5% 7.5% 14.0% 3.7% 86.1% 10.4% 3.5% 5.9% 74.4% 3.1% 9.9% 6.7% 49.3% 10.8% 60.0% 8.8% 7.0% 34.2% 18.4% 2.09 56.4% PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 6 All students Any career education, 1st award Highest level of 1st career education award Associate degree Long-term certificate Short-term certificate Ever Pell 42.1% 41.7% 44.3% 42.4% 38.3% Ever Cal Grant B 12.7% 12.2% 13.9% 11.8% 10.5% Ever Cal Grant C 2.2% 3.6% 3.8% 4.8% 2.9% Received Pell after first award 24.5% 26.3% 23.8% 27.8% 28.4% Post First Award Transfer to four-year university, 3 years 41.3% 18.8% 32.8% 6.4% 8.9% Source: Authors’ calculations based on COMIS data. Includes students age 18-54 at the time of their first award earned between 2001 and 2014 school years. We also exclude a small group of students (7,043) who earned both career education and non-career education credentials as their first award in the community college system and the non-career education credential was of a higher level than the career education credential. Limitations There are limitations to our data and analytic approach. First, our pathway analysis relies on the subset of colleges included in our pathways database and is not necessarily representative of the whole California community college system. While we sought to include colleges that conferred a large number of career education credentials and those with different types of credential lengths that could support stackable pathways, the colleges not included could very well have different program features. In addition, only a small percentage of programs we do include in our sample explicitly define their pathways (12% of progressive pathways and 6% of lattice pathways). While we use college catalog information to flag features within programs that signal progressive or lattice pathway design, colleges may not have intended to design a program based on the features and definitions we impose. Furthermore, the college catalog information we used for our scan to categorize programs and their stackable pathway features is limited to the 2016-2017 academic year, which we use to then retroactively apply to the programs in previous academic years. Our main models present findings from 2010 forward, but it is possible that the stackable features we identified based on 2016-17 catalog information were not in place in the earlier years. Finally, colleges are not required to report short-term certificates that are not approved by the Chancellor’s Office to COMIS—although many do. Our catalog scan and pathways database include these local certificates, but to the extent that we do not observe them in the student-level COMIS data, we are likely under-estimating the number of students who earn short-term credentials as well as those that complete a stackable pathway that starts with local certificate—sometimes referred to as Certificates of Accomplishment. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 7 Appendix B. Analysis of Student Outcomes Empirical Strategy Our analysis is primarily descriptive, with the goal of quantifying the prevalence of stacking within the community college system and identifying correlates of stacking credentials. As the report describes, we focus on students whose first award in the system is a short-term career education credential. For these students, we track whether, through the period of study, they earn:  Any other credential  Any other credential in the same career education discipline as the first  Any other credential in the same career education program (as identified with ICN7/ICN8 in our pathway database) The last constitutes pure “stacking” in this analysis, as it indicates accumulating skills/credentials within a given career pathway. TABLE B1 Students earning multiple credentials by the level of their first career education credential Highest level of 1st career education credential Short-term certificate Long-term certificate Associate degree Any other credential Within 3 years of first 27.7% 20.3% 8.2% Within 6 years of first 31.9% 24.4% 9.6% Another credential in the same discipline (TOP2) Within 3 years of first 20.1% 15.3% 4.2% Within 6 years of first 22.5% 17.8% 4.7% Total students, 3 year 194,931 95,205 222,241 Total students, 6 year 149,413 74,668 161,329 Another credential in the same pathway program (ICN7) Within 3 years of first 23.8% 18.7% 5.1% Total students in colleges in pathway database, 3 year 73,433 26,174 45,939 SOURCES: Author’s calculations from COMIS data and pathway database. NOTES: Restricted to students 18 to 54 at the time of their first career education credential. The 3-year measures include students who earned their first award between school years 2000–01 and 2013–14 and the 6-year measures include students who earned their first award between school years 2000–01 and 2010–11. Next, to describe the characteristics of students according to their likelihood of stacking, we examine a number of socioeconomic characteristics, as defined in Table B2. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 8 TABLE B2 Variable definitions Variable Description Outcomes Stack in same CTE program (ICN7) Stack to higher credential This variable is 1 if a student earned an additional award within three years of earning their first award This variable is 1 if a student earned a higher-level award than the award they first earned Student demographic attributes (COMIS) Sex Categorical variables for female and unknown sex, with male as the reference category Race/Ethnicity Citizenship status Age Categorical variables for Latino, African American, Asian-Pacific Islander, other race (includes two or more races, Native American) and unknown race. White is the reference category Categorical variable for non-citizen (permanent resident, temporary resident, refugee/asylee, F-1 or M-1 student visa, other status) and unknown citizenship. US citizen is the reference category Continuous variable for the age of a student upon receiving their first award between 2001 and 2014. Range is between 16 and 70 years old Student academic preparedness proxies (COMIS) Highest level of education at first term Disability status Limited English proficiency Academically disadvantaged Categorical variable indicating the highest level of education completed in the term that the student completes their first career education award. Categories include less than high school (includes adult ED and GED), associate degree, or bachelor degree or higher. High school graduate is the reference category This variable is 1 if a student was ever recorded as having at least one primary disability during their academic career in the community college system. Primary disabilities include mobility, visual, hearing, or speech impairment; intellectual, learning, or mental health disability; brain injury, ADHD, autism spectrum or other. This variable is 1 if a student was either enrolled at some time in their academic career in a basic skills English as a Second Language (ESL) course or was identified as needing ESL services. This is a derived variable created by the CCCCO. This variable is 1 if a student had enrollment or been flagged as needing basic skills instruction or was reported on academic probation or dismissal at least once in their academic career. This is a derived variable created by the CCCCO. Student socioeconomic status (COMIS) Low-income Ever CalWORKs This variable is 1 if a student ever received a Pell grant or is over age 25 and received a Promise Grant (formerly called a Board of Governor’s (BOG) waiver) sometime during their academic career at the community college system. This variable is 1 if a student was ever identified as receiving CalWORKs at any time during their academic career at the community college system. Markers of ability (COMIS) Ever developmental math or English Ever college math or English GPA This variable is 1 if a student ever enrolled in a developmental math or English course within two years of earning their first award This variable is 1 if a student ever enrolled in a college-level math or English course within two years of earning their first award Continuous variable indicating a student’s grade point average in the two year period prior to their earning their first career education credential. Pathway characteristics (pathway scan database) Number of credentials offered Associate degrees offered Local short-term credentials offered Explicit pathway feature Pathway feature, but not explicit Continuous variable indicating the number of credentials that a given program offers Continuous variable indicating the percent of credentials a given program offers that are associate degrees in science or arts Continuous variable indicating the percent of credentials a given program offers that are short-term credentials not approved by the CCCCO, also referred to as local certificates This variable is 1 if a program is categorized as having either an explicit progressive pathway or a defined launchpad core for lattice. This variable is 1 if a program has smaller certificates that can be upgraded to higher level certificates (certificate-only paths) for progressive pathways, or for lattice pathways has at least three credentials that share a core group (2+) of courses. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 9 Variable Description Pathway with minimal features No pathway features This variable is 1 if a program only has certificates that can be upgraded to associate degrees, often with just the completion of general education requirements or in the case of lattice pathways, no feature. This variable is 1 if a program has no identified features of either a progressive or lattice pathway. NOTES: The CCCCO MIS data element dictionary provides a more detailed description of each variable used in our study (CCCCO undated). Table B3 provides sample means for our analytic sample of short-term certificate earners in career education programs included in our pathways database. The analytic sample is restricted to students age 18 to 54 at the time of their first award, earned in 2010-2014, and excludes students who eventually transferred or who had missing values for the control variables. It is also restricted only to students who completed their first career education award at colleges that were scanned and included in our pathways database. TABLE B3 Sample means of dependent and independent variables All years Stacking credentials within an ICN Gender Male Female Age Age at first award Age^2 Race/Ethnicity White Latino Asian-PI African American Other/Missing Citizenship status Citizen LPR Other/Missing Highest level of education HS or equivalent No HS AA BA+ Other/Missing Markers of disadvantage DSPS Mean 24.1% 47.4% 52.6% 31.1 (9.8) 1060.6 (676.8) 38.3% 36.3% 14.2% 7.8% 3.4% 84.4% 11.6% 3.9% 76.1% 6.0% 2.8% 8.8% 6.3% 9.0% 2010 forward Mean 27.2% 51.6% 48.4% 30.1 (9.7) 1001.3 (668.8) 36.4% 38.0% 13.5% 7.9% 4.2% 87.0% 9.5% 3.5% 79.6% 3.8% 2.2% 9.9% 4.5% 9.7% PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 10 All years 2010 forward Mean Mean Academic disadvantage 63.8% 66.4% Limited English proficiency 13.5% 11.4% Pell 41.0% 48.7% CalWORKs 8.5% 7.3% Markers of ability Developmental Math or English 35.8% 39.7% Avg. GPA 2.0 2.02 Course characteristics Prop. work-based courses 7.0% 8.9% Prop. evening courses 33.1% 33.5% Prop. online courses 16.1% 17.8% Wage return Wage return 9.47 (9.49) 8.75 (9.20) First award: discipline Business 16.4% 19.0% Information technology (07) 4.1% 5.5% Engineering and industrial technologies (09) 19.1% 23.8% Family and consumer sciences (13) 34.4% 29.9% Public and protective services (21) 26.0% 21.8% First award: school year 2001 6.0% 2002 6.1% 2003 6.8% 2004 6.2% 2005 6.9% 2006 7.5% 2007 7.4% 2008 6.9% 2009 7.6% 2010 7.0% 18.0% 2011 6.6% 17.2% 2012 7.6% 19.7% 2013 8.8% 22.9% Number of observations 58,111 22,445 SOURCES: Author calculations from COMIS data. NOTES: Summary statistics from sample used in model (5) of Table B4, showing all years and main analytic sample years (2010-forward). Standard deviations are included for continuous variables in parentheses. In order to assess individual characteristics one by one, and to identify the effect of program features ceteris paribus, we estimate a number of linear and fixed effect regression models. These models estimate the relationship between earning a second credential in the same program (same ICN7 code) and individual PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 11 characteristics, certain non-design program characteristics, and college, period, and discipline effects. Table B4 provides detailed regression results for the full analysis sample, with alternative specifications and fixed effects. Model (1) is the simple descriptive model used to calculate odds of stacking credentials across demographic groups and across major programs (see report Figure 9). For parsimony and clarity in creating predicted odds of stacking, not all covariates are included in this initial model. Also, the sample of students is much broader— including those in programs that are in our scan database and whose first award was a short-term certificate. Models 2 through 5 present full covariates used in our main models, and test alternative specifications with regard to fixed effects and sample criteria. Model (5) is the chosen specification we utilize for all subsequent analyses. To this specification we add covariates of interest on program design, summarized in a subsequent table. For the baseline specification, we include rich student-level characteristics that account for student background and ability, factors strongly correlated with completion, broadly speaking. We also incorporate a handful of program characteristics (aside from stackable design) that may be related to completion—course delivery methods including online, evening, and work-based. In addition we utilize estimates of the wage return with in a 6-digit TOP code level, estimated in Bohn, McConville and Gibson (2016). In theory, a student who earns an initial highvalue credential may be less likely to return to stack another credential. These wage returns are themselves results from a student-longitudinal regression model, but are not refined enough to signal wage returns for the specific credential a student earns, rather the program the credential is from. Lastly, fixed effects for discipline (2-digit TOP code), college, and year, are included to account for the fact that certain colleges or programs may be better at ensuring student completion (or their local industry, labor markets, etc., may provide more or less incentive to complete). Year effects, in particular, may also address the cyclicality of enrollment. We tested alternative fixed effects—more detailed TOP codes and additional interactions, for example—but given our sample size and degrees of freedom we chose specification (5) as our baseline. TABLE B4 Main linear and fixed effects models of earning a stacked credential in a program within 3 years Dependent variable: Stack credential within program Female Age Age^2 Latino Asian African American Other/missing race Legal Permanent Resident Other status Less than high school AA BA+ Other/missing education Disability Academic disadvantage (1) -0.0101** -0.0100*** 0.0001*** -0.0128*** 0.0312*** -0.0152** -0.0058 0.0095* 0.0513*** -0.0401*** -0.0156 -0.0192*** -0.0430*** 0.0291*** 0.0266*** (2) -0.0072** -0.0108*** 0.0001*** -0.0133*** 0.0300*** -0.0107 -0.0007 0.0136** 0.0498*** -0.0432*** -0.0190* -0.0339*** -0.0473*** 0.0230*** 0.0257*** (3) -0.0145*** -0.0091*** 0.0001*** -0.0181*** 0.0297*** -0.0148** -0.0061 0.0093 0.0510*** -0.0438*** -0.0148 -0.0198*** -0.0354*** 0.0202*** 0.0277*** (4) -0.0049 -0.0021 0.0000 -0.0127*** 0.0058 -0.0135** -0.0121 0.0153*** 0.0495*** -0.0350*** -0.0155 -0.0333*** -0.0241*** 0.0207*** 0.0203*** (5) -0.00702 -0.00472** 0.0001* 0.0021 0.0128 -0.0157 0.0116 0.0034 0.0452*** -0.0342** -0.0306 -0.0332*** -0.0013 0.0309*** 0.0176** PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 12 Dependent variable: Stack credential within program (1) (2) (3) (4) (5) Limited English 0.0382*** 0.0484*** 0.0452*** 0.0415*** 0.0567*** PELL 0.0665*** 0.0666*** 0.0638*** 0.0521*** 0.0431*** CalWORKs 0.0415*** 0.0361*** 0.0361*** 0.0337*** 0.0445*** Developmental math or English 0.1065*** 0.1060*** 0.103*** 0.0990*** 0.0956*** GPA 0.0279*** 0.0319*** 0.0280*** 0.0263*** 0.0424*** Work based courses -0.0379*** -0.0526*** -0.171*** -0.189*** Evening courses 0.0336*** -0.0030 0.0323*** 0.0008 Online courses -0.1170*** -0.0685*** -0.202*** -0.241*** Average wage return for TOP code -0.00346*** -0.0030*** -0.0020*** -0.0014*** IT -0.0062434 0.0079 0.0319*** 0.0517*** Engineering 0.0816474 0.0835*** 0.0356*** 0.0959*** Family and consumer 0.0649579 0.0670*** 0.0510*** 0.0723*** Public and protective -0.0145007 -0.0020 -0.0043 0.0042 Constant 0.2412 0.327*** 0.275*** 0.139*** 0.219*** Observations 66050 65,187 65,187 65,187 22,445 R-squared 0.055 0.055 0.061 0.098 0.108 Fixed effects Year XX TOP 2 X X XX College XX Exclusions Transfers included X XX In scan only X XX Only 2010 forward X SOURCES: Authors calculations from COMIS data and pathway database. NOTES: Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1 The dependent variable is a bivariate on whether a student whose initial credential was a short-term certificate in one of our scanned colleges earned a related credential in the same program (ICN7) within 3 years. To test whether program design is related to the odds a student stacks credentials, we estimate Model 5 including variables to indicate program features, as described in the text. Table B5 presents the estimates from those primary variables of interest. Each cell in the table is derived from a separate regression model. We do this to control which programs we are comparing in each case. For example, we test programs with strong progressive or launchpad/core pathways compared to programs with pathways that are just not explicit or to programs with no pathway, in turn. For this reason, the sample sizes vary greatly across models and are shown in Table B6. Our main models examining pathway types include a host of demographic and program characteristics. However, to understand how the relationships vary across key dimensions, we re-estimate models separately by race/ethnicity, gender, and 2-digit TOP code. These models allow all of the factors related to stacking (not just program design but also age, educational background, college, etc.) to vary in their importance across groups. This specification is more flexible than a model that incorporates interactions, though in those we find qualitatively similar results. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 13 TABLE B5 Main estimates from fixed effects models of program features on earning a stacked credential Dependent variable: Stack credential within program Overall Latino Asian African American White Male Female Business IT Engineering Family/ Public/ Consumer Protective Explicit pathways compared to… All others 0.0456*** 0.0834*** 0.0712*** -0.0468 0.0304* 0.0321** 0.0360** 0.0923*** 0.0396 0.0832*** -0.0294 0.198*** Programs with no pathways Programs with light pathway features Programs with pathway features (but not explicit) 0.164*** 0.173*** 0.0994*** 0.113*** 0.186*** 0.139*** 0.167* 0.00615 0.110*** 0.150*** 0.0713*** 0.117*** 0.135*** 0.0499** 0.0225 -0.0669 0.0558*** 0.0969*** 0.0650** -0.0458 0.0524*** 0.0441*** 0.0404** 0.113*** 0.104 0.105 0.150 0.160*** 0.135*** 0.0351 0.492* -0.00935 -0.0202 0.252*** 0.245*** 0.114** Other comparisons within pathway type Explicit launchpad/lattice programs vs programs with no strong pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificateAA only) Launchpad/Lattice programs vs programs with no pathway feature 0.0423*** 0.0421*** 0.0888*** 0.148*** 0.144*** 0.0672*** 0.0931*** 0.145*** 0.192*** 0.206*** 0.108*** 0.0610* 0.170*** 0.182*** 0.211*** -0.0812 0.0422* 0.0466** 0.0174 -0.00959 -0.0126 0.00505 0.0322* 0.0415* 0.114*** 0.136*** 0.0177 0.0611*** 0.162*** -0.0396 0.199*** 0.0949*** 0.148*** 0.204*** -0.0931 0.196*** 0.0753*** 0.124*** 0.203*** -0.141** 0.0382 0.106 0.0592 0.0767* 0.0560 0.0877*** 0.114*** 0.251*** 0.283*** 0.239*** -0.0620 0.00973 0.202*** 0.235*** 0.179** 0.216*** 0.132 -0.00406 0.243*** 0.120*** SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the coefficient from a separate model. The dependent variable is a bivariate on whether a student whose initial credential was a short-term certificate in one of our scanned colleges earned a related credential in the same program (ICN7) within 3 years. All models include covariates listed in Table B4 Model 5, which includes student and program characteristics as well as year, college, and 2-digit TOP code fixed effects. Column headings refer to the subsample used in the regression model. Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1. The sample size varies across each model, and are provided in the next table. Full results are available upon request. All models only include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first career education credential, in addition restrictions on the comparison programs or subsamples noted in the table. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 14 TABLE B6 Sample size for main fixed effects models of program features Overall Latino Asian African American White Male Female Business IT Engineering Family/ Public/ Consumer Protective Explicit pathways compared to… All others 22,429 8,520 3,026 1,774 8,173 11,566 10,863 4,262 1,237 5,332 6,713 4,885 Programs with no pathways Programs with light pathway features Programs with pathway features (but not explicit) 6,531 11,181 19,761 2,330 3,949 7,686 1,043 1,473 2,699 451 782 1,549 2,464 4,536 7,002 3,647 6,793 9,404 2,884 4,388 10,357 964 1,356 4,218 972 1,010 382 1,348 2,600 4,838 1,651 2,479 6,645 1,596 3,736 3,678 Other comparisons within pathway type Explicit launchpad/ lattice programs vs programs with no explicit pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificateAA only) Launchpad/lattice programs vs programs with no pathway feature 20,557 20,687 13,171 17,602 22,044 7,915 7,684 5,388 6,729 8,409 2,509 2,893 1,710 2,523 2,983 1,708 1,631 1,101 1,382 1,756 7,557 10,697 9,860 3,623 7,598 11,059 9,628 4,126 4,416 5,016 8,155 3,154 6,208 8,799 8,803 3,723 7,970 11,500 10,544 3,903 1,191 1,174 230 319 1,237 4,827 5,052 1,861 4,538 5,344 6,147 5,697 5,183 5,604 6,688 4,769 4,638 2,743 3,418 4,872 SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the sample size of a separate model, corresponding with Table B5. Column headings refer to the subsample used in the regression model. Full results are available upon request. All models only include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first CTE credential, in addition to the comparison programs in the first column and subsamples noted in the first row. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 15 Table B7 reports the results from a number of additional tests of our main models. As above, each cell in this table is the main coefficient of interest from a separate regression model that also includes all covariates in Table B4 Model (5). The first three columns test the importance of pathway design on a variety of alternative student samples: (1) only students who re-enroll in a community college after their first short-term certificate (perhaps indicating an intent to complete a pathway), (2) adding students whose first award is a long-term certificate to the main sample, and (3) including students who transfer. Estimates on the relationship between pathway design and stacking are very similar across these student samples—they are also very similar to our main estimates above. The last three columns of Table B7 test alternative dependent variables. We remove the restriction that a student must stack within 3 years, we require the student to stack on a slightly more narrowly defined pathway (ICN8), and we test stacking to higher-level credentials (from short-term to long- or associate). Once again, the results are quite similar. TABLE B7 Additional tests on main models of program features on earning a stacked credential Dependent variable: Stack credential within program Students Students Students who whose first whose first re-enroll after award is award is a S first short- either a short-term term short- or certificate, certificate long-term regardless of certificate transfer Strong pathway compared to… Testing alternative dependent variables Stack within ICN7 but no restriction on how long it takes Stack within ICN8 within 3 years Stack within ICN7 to a higher level credential All others 0.0550*** 0.0452*** 0.0422*** 0.0383*** 0.0479*** 0.0346*** Programs with no pathways 0.169*** 0.164*** 0.162*** 0.167*** 0.165*** 0.0969*** Programs with light pathway features Programs with pathway features (but not explicit) 0.0973*** 0.0797*** 0.113*** 0.0527*** 0.101*** 0.0489*** 0.0925*** 0.0468*** 0.108*** 0.0589*** 0.0663*** 0.0230*** Other comparisons within pathway type: Explicit lattice programs vs programs with no explicit pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificate-AA only) Lattice programs vs programs with no pathway feature 0.0732*** 0.0373** 0.0950*** 0.150*** 0.159*** 0.0427*** 0.0553*** 0.117*** 0.175*** 0.162*** 0.0489*** 0.0251** 0.0918*** 0.149*** 0.142*** 0.0387*** 0.0303** 0.103*** 0.163*** 0.159*** 0.0411*** 0.0427*** 0.0879*** 0.146*** 0.142*** 0.0163 0.0471*** 0.0964*** 0.115*** 0.114*** SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the coefficient from a separate model. The dependent variable is a bivariate that changes across models as indicated in column headings. All models include covariates listed in Table B4 Model 5 as well as year, college, and 2-digit TOP code fixed effects. Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1. Only main effects are shown here, but full results are available upon request. Baseline models include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first CTE credential, as well as sample restrictions noted in column headings. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 16 The Public Policy Institute of California is dedicated to informing and improving public policy in California through independent, objective, nonpartisan research. Public Policy Institute of California 500 Washington Street, Suite 600 San Francisco, CA 94111 T: 415.291.4400 F: 415.291.4401 PPIC.ORG PPIC Sacramento Center Senator Office Building 1121 L Street, Suite 801 Sacramento, CA 95814 T: 916.440.1120 F: 916.440.1121" } ["___content":protected]=> string(196) "

Stackable Credentials in Career Education at California Community Colleges, Technical Appendix

" ["_permalink":protected]=> string(125) "https://www.ppic.org/publication/stackable-credentials-in-career-education-at-california-community-colleges/1018sbr-appendix/" ["_next":protected]=> array(0) { } ["_prev":protected]=> array(0) { } ["_css_class":protected]=> NULL ["id"]=> int(16788) ["ID"]=> int(16788) ["post_author"]=> string(1) "4" ["post_content"]=> string(0) "" ["post_date"]=> string(19) "2018-10-22 20:54:40" ["post_excerpt"]=> string(0) "" ["post_parent"]=> int(16518) ["post_status"]=> string(7) "inherit" ["post_title"]=> string(94) "Stackable Credentials in Career Education at California Community Colleges, Technical Appendix" ["post_type"]=> string(10) "attachment" ["slug"]=> string(16) "1018sbr-appendix" ["__type":protected]=> NULL ["_wp_attached_file"]=> string(20) "1018sbr-appendix.pdf" ["wpmf_size"]=> string(6) "429818" ["wpmf_filetype"]=> string(3) "pdf" ["wpmf_order"]=> string(1) "0" ["searchwp_content"]=> string(43266) "Stackable Credentials in Career Education at California Community Colleges Technical Appendices CONTENTS Appendix A. Data and Sample Construction Appendix B. Analysis of Student Outcomes Sarah Bohn and Shannon McConville With research support from Bonnie Brooks and Sophie McGuinness Supported with funding from the ECMC Foundation, the James Irvine Foundation, and the Sutton Family Fund Appendix A. Data and Sample Construction Data sources The data used in this report come from the California Community College Chancellor’s Office (CCCCO). The Chancellor’s Office Management Information System (COMIS) is a longitudinal dataset that includes information on all students enrolled in colleges across the California Community College system. This dataset includes detailed information on student characteristics, including demographics, measures of economic and academic disadvantage, and disability, along with course enrollment, financial aid receipt, and award completion. We have these records for all students between the fall term of 1993 and the spring term of 2017. The data system also contains information on the awards or credentials that a student completes.1We classify all awards that students earn in the system into three categories—short-term certificates, long-term certificates, and associate degrees. Short-term and long-term certificates are defined based on the length of time, measured in terms of units and assuming full-time course loads, it takes to complete the degrees. Short-term certificates take less than 1 year to complete and include certificates requiring less than 30 units. Long-term certificates are defined as those requiring between 1 and 2 years to complete and include certificates requiring 60 or more units and those requiring 30-59 units. Associate degrees can be either associate of art or science and typically take 2 or more years to complete. All courses and awards include information that designates a specific field of study called a Taxonomy of Program (TOP) code. The TOP system of numerical codes is used to collect and report information on programs and courses in different colleges throughout the state that have similar outcomes. We use TOP codes to identify career education awards. The CCCCO designates all career education or vocational programs based on the 6-digit TOP code. The TOP codes were designed to aggregate information about programs and all courses and awards are coded with a 6-digit TOP code. The first two digits of the six-digit TOP denote the discipline and is used to define our broad career education areas which include Business and Management (05), Information Technology (07), Engineering (09), Family and Consumer Sciences (13) and Public and Protective Services (21). The first four digits are intended to denote a sub-discipline (e.g. 1305: Child Development/Early Childhood Education), and the entire six digits denote a specific field of study (e.g. 130580: Child Development Administration and Management). Defining Stackable Credential Features In order to systematically examine stackable credentials, we developed a strategy that would allow us to identify career education programs based on interrelated credentials they offer and to classify those credential features. Although programs are intuitive, akin to college departments like culinary arts, automotive technology, information technology, etc., no statewide data source tracks programs in this way. The TOP code contained in COMIS is both too broad and too detailed. A sub-discipline (4-digit TOP) typically contains multiple programs and sets of credentials that have no overlapping course requirements. The field of study (6-digit TOP) is too detailed, with credentials in different fields of study actually closely related in terms of course requirements and 1 One caveat in our research pertains to the potential underreporting of short-term certificates that are not approved by the Chancellor’s office. These ‘local’ certificates are college- and department-specific certificates less than 12 units. Because the CCCCO does not require colleges to report local certificates in COMIS—although many do—non-reported local certificates are not included in our analysis. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 2 career options. Furthermore, there is a lot of variation across colleges and program areas in what may or may not be offered in a sub-discipline or field of study. To appropriately identify career education programs and classify students according to their program awards, we first attempted to connect information available from COMIS on the types of awards students earn (i.e. TOP codes and credential length) to the inventory of programs and courses available from the Chancellor’s office. Unfortunately that approach was not workable due to the high levels of mismatch—there were too many credentials in COMIS that did not have a match to the inventory and vice-versa. In the end, we determined the best way to identify and describe programs according to their stackable pathways would require a detailed scan of college course catalogs that would allow us to explicitly connect credentials based on overlapping course requirements and other available descriptors of programs and credentials. To employ this more qualitative approach, we also needed to identify a subset of colleges on which to focus our scan in order to make the task more manageable. At the same time, we wanted to ensure that the colleges we scanned had career education programs that had enough different types of credentials available to make stackable credentials a possibility. We selected a group of colleges to conduct in-depth catalog scans for each of the five focal career education disciplines based on the number of awards they conferred by credential length (short-term, long-term, associate). First, we sorted all schools by the number of certificates (long-term and short-term) and also associate degrees they conferred in school year 2015-16 and selected colleges with the highest number of each type of credentials that cumulatively added up to 50% of the total awards systemwide that were conferred in that career education discipline. In cases where the number of schools needed to encompass 50% of associate degrees was close to that of certificates, we chose whichever had the greater number of schools. We find that this selection method organically provided a large and diverse sample of programs with stacking potential. Using this method, across all disciplines we identified 64 community colleges overall to be scanned and between 25 and 32 colleges for each career education discipline (business, IT, engineering, family/consumer science, and public/protective services). Second, from the selection of colleges identified for the scan, we used available information from COMIS and the CCCCO program inventory to enumerate all of the specific fields of study available in the career education discipline based on 6-digit TOP code and the program names and descriptors provided in the inventory. The TOP codes were necessary so that we could link the scan information back to the student-level COMIS data and the program names assisted us in finding the programs of interest in the course catalogs. With our scan information in college catalogs to answer questions aimed to identify the differing degrees to which a program fulfills the characteristics of three types of pathways to stacking credentials: PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 3 TABLE A1 Stackable pathway definitions and features Pathway type Progressive Lattice Definition Students can upgrade from lower-level credentials to higher-level credentials (vertical) by completing additional coursework within the same program Students complete a group of courses that serve as a “Launchpad” to multiple same-level (horizontal) and/or higher-level (vertical) credentials Features Are there smaller credentials that can be upgraded to higher-level credentials (certificate or associate) by adding additional units in the CTE area? If yes, is there a sequence that involves only certificates, not associate degrees? Does the catalog explicitly state there are levels or stages within a program (i.e. using terms like beginner, intermediate, advanced OR I, II, III in program titles) Are there 3 or more degrees and certificates that share a core group of (2+) classes? Is the shared group of courses called out (identified as a “core”) in the catalog to indicate that it can lead to multiple credentials? Even if the core is not directly labeled “core,” does the group of core courses alone earn the student either a local certificate or a certificate of achievement? After coding each field of study using the guiding questions, programs that share certain features (and subsequently share stacking potential) we assign a unique code called the Internal Control Number (ICN) to programs that share coursework. The ICN is intended to identify sequences of credentials by linking programs that share coursework across six-digit TOP codes in a given college. In this way we can connect multiple fields of study within a discipline that could be part of stackable credential pathway. The definition of the ICN is listed below. ICN = College ID + Two-Digit TOP Code + First-Degree Connection + Second-Degree Connection To determine if programs share either a first-degree or second-degree connection we ask the following questions. For the first-degree connection: at least 2 course requirements come from other programs? If so, assign the same digit to programs that overlap by at least 2 courses. For the second-degree connection: are programs themselves pieces of the same program sequence (and/or have 3-4 classes that overlap)? If so, assign the same digit to programs that overlap by 3-4 classes. In the final step, we link the information on pathway features at the ICN level (8-digit code that includes the 1st and 2nd degree of connection) to the COMIS data by college and 6-digit TOP code. Table A2 provides information on the pathway features for each of the two stackable pathway types we define at the ICN 8-digit level as it provides the most granular detail on pathway characteristics. In our main analysis, we use the ICN 7digit code to determine if students stack additional credential along a program pathway to allow for a more broad definition of potential connection between programs. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 4 TABLE A2 Characteristics of programs across career education disciplines Pathway characteristics Business Progressive Potential to upgrade credentials Certificate-only paths Explicit credential sequence Lattice Common set of course requirements for multiple credentials Explicit core/launchpad Core-alone earns credential Total programs identified 106 (82%) 72 (56%) 18 (14%) 101 (79%) 14 (11%) 0 (0%) 129 Engineering 137 (87%) 82 (52%) 10 (6%) 58 (37%) 3 (2%) 7 (4%) 157 Public and protective services 66 (71%) 20 (22%) 4 (4%) 48 (52%) 4 (4%) 1 (1%) 93 Information technology 41 (56%) 26 (36%) 7 (10%) 28 (39%) 5 (7%) 4 (6%) 72 Family and consumer sciences Total 95 (90%) 64 (61%) 27 (25%) 445 (80%) 264 (47%) 66 (12%) 63 (60%) 1 (1%) 6 (6%) 105 298 (54%) 27 (6%) 18 (3%) 556 Sample Construction We build our sample universe from the COMIS data file that records the awards students earn. For the purposes of this report, we restrict our sample to those students who earned their first observed award (at least since 1993) within the community college system between the years 2000 and 2014—so that we can observe students for at least three years following the completion of their first community college award. We further restrict our sample to students who have an SSN (scrambled and de-identified in our data extract) recorded as an identifier. Each community college assigns school-level student identification numbers, but these are not unique across the entire system. Restricting our sample to students with an SSN as an identifier allows us to track students across multiple colleges so we have a complete picture of their coursework and credentials. For our examination of the broad trends in career education credentials presented in the first section of the report, we include all students age 18 to 54 who earn their first community college credential between school years 2000–01 and 2013–14. We designate students earning career education credentials based on the designation of vocational programs provided in the Taxonomy of Programs (TOP) manual 6th Edition (California Community Colleges, Academic Affairs Division, July 2013). Table A3 provides descriptive statistics across different student groups—including all students who earned a credential regardless of if it was in career education, all students who earned any career education credential as their first community college award with breakdowns by the level of their first credential. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 5 TABLE A3 Sample means by group Total Age at first award Mean 18-24 25-29 30-34 35-39 40-54 Sex Male Female Race/Ethnicity White Latino African American Asian/Pacific Islander Other/Multi/Missing Citizenship status Citizen Non-Citizen Unknown Highest level of education No high school diploma High school diploma/GED Associate degree Bachelor’s degree or higher Other/Missing Markers of disadvantage Low income (Pell or age 25+ and BOGW) Limited English proficiency Academic disadvantage Student with disabilities Ever CalWORKs Markers of ability Ever developmental math or English Ever transfer math or English GPA Financial aid Ever BOG Waiver All students 1,131,711 Any career education, 1st award 512,377 Highest level of 1st career education award Associate degree Long-term certificate Short-term certificate 222,241 95,205 194,931 27.6 52.2% 17.7% 10.0% 7.0% 13.1% 40.5% 59.5% 30.1% 15.9% 7.1% 42.9% 4.0% 86.6% 9.6% 3.8% 3.7% 78.8% 7.8% 4.3% 5.4% 51.1% 8.2% 63.1% 7.6% 4.6% 43.7% 51.4% 2.07 60.9% 30.4 36.5% 20.1% 13.3% 10.0% 20.0% 43.2% 56.8% 30.6% 16.1% 7.4% 42.3% 3.6% 84.9% 10.9% 4.1% 4.4% 74.2% 7.1% 8.3% 6.0% 53.5% 10.0% 62.1% 8.1% 6.4% 36.0% 28.8% 2.13 60.3% 29.6 38.7% 22.1% 13.4% 9.4% 16.3% 38.5% 61.5% 42.8% 28.7% 6.6% 18.5% 3.4% 83.9% 11.3% 4.8% 2.8% 72.3% 12.2% 6.9% 5.8% 56.7% 9.3% 64.1% 7.1% 5.4% 39.6% 43.9% 2.15 63.3% 31.1 32.9% 20.2% 14.2% 10.6% 22.1% 45.8% 54.2% 41.3% 30.8% 8.9% 15.0% 4.0% 84.9% 11.2% 3.9% 4.9% 78.1% 3.3% 8.5% 5.3% 54.8% 9.8% 61.8% 9.1% 7.4% 31.4% 15.1% 2.13 61.2% 31.0 35.9% 17.9% 12.8% 10.4% 23.1% 47.3% 52.7% 42.2% 32.5% 7.5% 14.0% 3.7% 86.1% 10.4% 3.5% 5.9% 74.4% 3.1% 9.9% 6.7% 49.3% 10.8% 60.0% 8.8% 7.0% 34.2% 18.4% 2.09 56.4% PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 6 All students Any career education, 1st award Highest level of 1st career education award Associate degree Long-term certificate Short-term certificate Ever Pell 42.1% 41.7% 44.3% 42.4% 38.3% Ever Cal Grant B 12.7% 12.2% 13.9% 11.8% 10.5% Ever Cal Grant C 2.2% 3.6% 3.8% 4.8% 2.9% Received Pell after first award 24.5% 26.3% 23.8% 27.8% 28.4% Post First Award Transfer to four-year university, 3 years 41.3% 18.8% 32.8% 6.4% 8.9% Source: Authors’ calculations based on COMIS data. Includes students age 18-54 at the time of their first award earned between 2001 and 2014 school years. We also exclude a small group of students (7,043) who earned both career education and non-career education credentials as their first award in the community college system and the non-career education credential was of a higher level than the career education credential. Limitations There are limitations to our data and analytic approach. First, our pathway analysis relies on the subset of colleges included in our pathways database and is not necessarily representative of the whole California community college system. While we sought to include colleges that conferred a large number of career education credentials and those with different types of credential lengths that could support stackable pathways, the colleges not included could very well have different program features. In addition, only a small percentage of programs we do include in our sample explicitly define their pathways (12% of progressive pathways and 6% of lattice pathways). While we use college catalog information to flag features within programs that signal progressive or lattice pathway design, colleges may not have intended to design a program based on the features and definitions we impose. Furthermore, the college catalog information we used for our scan to categorize programs and their stackable pathway features is limited to the 2016-2017 academic year, which we use to then retroactively apply to the programs in previous academic years. Our main models present findings from 2010 forward, but it is possible that the stackable features we identified based on 2016-17 catalog information were not in place in the earlier years. Finally, colleges are not required to report short-term certificates that are not approved by the Chancellor’s Office to COMIS—although many do. Our catalog scan and pathways database include these local certificates, but to the extent that we do not observe them in the student-level COMIS data, we are likely under-estimating the number of students who earn short-term credentials as well as those that complete a stackable pathway that starts with local certificate—sometimes referred to as Certificates of Accomplishment. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 7 Appendix B. Analysis of Student Outcomes Empirical Strategy Our analysis is primarily descriptive, with the goal of quantifying the prevalence of stacking within the community college system and identifying correlates of stacking credentials. As the report describes, we focus on students whose first award in the system is a short-term career education credential. For these students, we track whether, through the period of study, they earn:  Any other credential  Any other credential in the same career education discipline as the first  Any other credential in the same career education program (as identified with ICN7/ICN8 in our pathway database) The last constitutes pure “stacking” in this analysis, as it indicates accumulating skills/credentials within a given career pathway. TABLE B1 Students earning multiple credentials by the level of their first career education credential Highest level of 1st career education credential Short-term certificate Long-term certificate Associate degree Any other credential Within 3 years of first 27.7% 20.3% 8.2% Within 6 years of first 31.9% 24.4% 9.6% Another credential in the same discipline (TOP2) Within 3 years of first 20.1% 15.3% 4.2% Within 6 years of first 22.5% 17.8% 4.7% Total students, 3 year 194,931 95,205 222,241 Total students, 6 year 149,413 74,668 161,329 Another credential in the same pathway program (ICN7) Within 3 years of first 23.8% 18.7% 5.1% Total students in colleges in pathway database, 3 year 73,433 26,174 45,939 SOURCES: Author’s calculations from COMIS data and pathway database. NOTES: Restricted to students 18 to 54 at the time of their first career education credential. The 3-year measures include students who earned their first award between school years 2000–01 and 2013–14 and the 6-year measures include students who earned their first award between school years 2000–01 and 2010–11. Next, to describe the characteristics of students according to their likelihood of stacking, we examine a number of socioeconomic characteristics, as defined in Table B2. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 8 TABLE B2 Variable definitions Variable Description Outcomes Stack in same CTE program (ICN7) Stack to higher credential This variable is 1 if a student earned an additional award within three years of earning their first award This variable is 1 if a student earned a higher-level award than the award they first earned Student demographic attributes (COMIS) Sex Categorical variables for female and unknown sex, with male as the reference category Race/Ethnicity Citizenship status Age Categorical variables for Latino, African American, Asian-Pacific Islander, other race (includes two or more races, Native American) and unknown race. White is the reference category Categorical variable for non-citizen (permanent resident, temporary resident, refugee/asylee, F-1 or M-1 student visa, other status) and unknown citizenship. US citizen is the reference category Continuous variable for the age of a student upon receiving their first award between 2001 and 2014. Range is between 16 and 70 years old Student academic preparedness proxies (COMIS) Highest level of education at first term Disability status Limited English proficiency Academically disadvantaged Categorical variable indicating the highest level of education completed in the term that the student completes their first career education award. Categories include less than high school (includes adult ED and GED), associate degree, or bachelor degree or higher. High school graduate is the reference category This variable is 1 if a student was ever recorded as having at least one primary disability during their academic career in the community college system. Primary disabilities include mobility, visual, hearing, or speech impairment; intellectual, learning, or mental health disability; brain injury, ADHD, autism spectrum or other. This variable is 1 if a student was either enrolled at some time in their academic career in a basic skills English as a Second Language (ESL) course or was identified as needing ESL services. This is a derived variable created by the CCCCO. This variable is 1 if a student had enrollment or been flagged as needing basic skills instruction or was reported on academic probation or dismissal at least once in their academic career. This is a derived variable created by the CCCCO. Student socioeconomic status (COMIS) Low-income Ever CalWORKs This variable is 1 if a student ever received a Pell grant or is over age 25 and received a Promise Grant (formerly called a Board of Governor’s (BOG) waiver) sometime during their academic career at the community college system. This variable is 1 if a student was ever identified as receiving CalWORKs at any time during their academic career at the community college system. Markers of ability (COMIS) Ever developmental math or English Ever college math or English GPA This variable is 1 if a student ever enrolled in a developmental math or English course within two years of earning their first award This variable is 1 if a student ever enrolled in a college-level math or English course within two years of earning their first award Continuous variable indicating a student’s grade point average in the two year period prior to their earning their first career education credential. Pathway characteristics (pathway scan database) Number of credentials offered Associate degrees offered Local short-term credentials offered Explicit pathway feature Pathway feature, but not explicit Continuous variable indicating the number of credentials that a given program offers Continuous variable indicating the percent of credentials a given program offers that are associate degrees in science or arts Continuous variable indicating the percent of credentials a given program offers that are short-term credentials not approved by the CCCCO, also referred to as local certificates This variable is 1 if a program is categorized as having either an explicit progressive pathway or a defined launchpad core for lattice. This variable is 1 if a program has smaller certificates that can be upgraded to higher level certificates (certificate-only paths) for progressive pathways, or for lattice pathways has at least three credentials that share a core group (2+) of courses. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 9 Variable Description Pathway with minimal features No pathway features This variable is 1 if a program only has certificates that can be upgraded to associate degrees, often with just the completion of general education requirements or in the case of lattice pathways, no feature. This variable is 1 if a program has no identified features of either a progressive or lattice pathway. NOTES: The CCCCO MIS data element dictionary provides a more detailed description of each variable used in our study (CCCCO undated). Table B3 provides sample means for our analytic sample of short-term certificate earners in career education programs included in our pathways database. The analytic sample is restricted to students age 18 to 54 at the time of their first award, earned in 2010-2014, and excludes students who eventually transferred or who had missing values for the control variables. It is also restricted only to students who completed their first career education award at colleges that were scanned and included in our pathways database. TABLE B3 Sample means of dependent and independent variables All years Stacking credentials within an ICN Gender Male Female Age Age at first award Age^2 Race/Ethnicity White Latino Asian-PI African American Other/Missing Citizenship status Citizen LPR Other/Missing Highest level of education HS or equivalent No HS AA BA+ Other/Missing Markers of disadvantage DSPS Mean 24.1% 47.4% 52.6% 31.1 (9.8) 1060.6 (676.8) 38.3% 36.3% 14.2% 7.8% 3.4% 84.4% 11.6% 3.9% 76.1% 6.0% 2.8% 8.8% 6.3% 9.0% 2010 forward Mean 27.2% 51.6% 48.4% 30.1 (9.7) 1001.3 (668.8) 36.4% 38.0% 13.5% 7.9% 4.2% 87.0% 9.5% 3.5% 79.6% 3.8% 2.2% 9.9% 4.5% 9.7% PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 10 All years 2010 forward Mean Mean Academic disadvantage 63.8% 66.4% Limited English proficiency 13.5% 11.4% Pell 41.0% 48.7% CalWORKs 8.5% 7.3% Markers of ability Developmental Math or English 35.8% 39.7% Avg. GPA 2.0 2.02 Course characteristics Prop. work-based courses 7.0% 8.9% Prop. evening courses 33.1% 33.5% Prop. online courses 16.1% 17.8% Wage return Wage return 9.47 (9.49) 8.75 (9.20) First award: discipline Business 16.4% 19.0% Information technology (07) 4.1% 5.5% Engineering and industrial technologies (09) 19.1% 23.8% Family and consumer sciences (13) 34.4% 29.9% Public and protective services (21) 26.0% 21.8% First award: school year 2001 6.0% 2002 6.1% 2003 6.8% 2004 6.2% 2005 6.9% 2006 7.5% 2007 7.4% 2008 6.9% 2009 7.6% 2010 7.0% 18.0% 2011 6.6% 17.2% 2012 7.6% 19.7% 2013 8.8% 22.9% Number of observations 58,111 22,445 SOURCES: Author calculations from COMIS data. NOTES: Summary statistics from sample used in model (5) of Table B4, showing all years and main analytic sample years (2010-forward). Standard deviations are included for continuous variables in parentheses. In order to assess individual characteristics one by one, and to identify the effect of program features ceteris paribus, we estimate a number of linear and fixed effect regression models. These models estimate the relationship between earning a second credential in the same program (same ICN7 code) and individual PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 11 characteristics, certain non-design program characteristics, and college, period, and discipline effects. Table B4 provides detailed regression results for the full analysis sample, with alternative specifications and fixed effects. Model (1) is the simple descriptive model used to calculate odds of stacking credentials across demographic groups and across major programs (see report Figure 9). For parsimony and clarity in creating predicted odds of stacking, not all covariates are included in this initial model. Also, the sample of students is much broader— including those in programs that are in our scan database and whose first award was a short-term certificate. Models 2 through 5 present full covariates used in our main models, and test alternative specifications with regard to fixed effects and sample criteria. Model (5) is the chosen specification we utilize for all subsequent analyses. To this specification we add covariates of interest on program design, summarized in a subsequent table. For the baseline specification, we include rich student-level characteristics that account for student background and ability, factors strongly correlated with completion, broadly speaking. We also incorporate a handful of program characteristics (aside from stackable design) that may be related to completion—course delivery methods including online, evening, and work-based. In addition we utilize estimates of the wage return with in a 6-digit TOP code level, estimated in Bohn, McConville and Gibson (2016). In theory, a student who earns an initial highvalue credential may be less likely to return to stack another credential. These wage returns are themselves results from a student-longitudinal regression model, but are not refined enough to signal wage returns for the specific credential a student earns, rather the program the credential is from. Lastly, fixed effects for discipline (2-digit TOP code), college, and year, are included to account for the fact that certain colleges or programs may be better at ensuring student completion (or their local industry, labor markets, etc., may provide more or less incentive to complete). Year effects, in particular, may also address the cyclicality of enrollment. We tested alternative fixed effects—more detailed TOP codes and additional interactions, for example—but given our sample size and degrees of freedom we chose specification (5) as our baseline. TABLE B4 Main linear and fixed effects models of earning a stacked credential in a program within 3 years Dependent variable: Stack credential within program Female Age Age^2 Latino Asian African American Other/missing race Legal Permanent Resident Other status Less than high school AA BA+ Other/missing education Disability Academic disadvantage (1) -0.0101** -0.0100*** 0.0001*** -0.0128*** 0.0312*** -0.0152** -0.0058 0.0095* 0.0513*** -0.0401*** -0.0156 -0.0192*** -0.0430*** 0.0291*** 0.0266*** (2) -0.0072** -0.0108*** 0.0001*** -0.0133*** 0.0300*** -0.0107 -0.0007 0.0136** 0.0498*** -0.0432*** -0.0190* -0.0339*** -0.0473*** 0.0230*** 0.0257*** (3) -0.0145*** -0.0091*** 0.0001*** -0.0181*** 0.0297*** -0.0148** -0.0061 0.0093 0.0510*** -0.0438*** -0.0148 -0.0198*** -0.0354*** 0.0202*** 0.0277*** (4) -0.0049 -0.0021 0.0000 -0.0127*** 0.0058 -0.0135** -0.0121 0.0153*** 0.0495*** -0.0350*** -0.0155 -0.0333*** -0.0241*** 0.0207*** 0.0203*** (5) -0.00702 -0.00472** 0.0001* 0.0021 0.0128 -0.0157 0.0116 0.0034 0.0452*** -0.0342** -0.0306 -0.0332*** -0.0013 0.0309*** 0.0176** PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 12 Dependent variable: Stack credential within program (1) (2) (3) (4) (5) Limited English 0.0382*** 0.0484*** 0.0452*** 0.0415*** 0.0567*** PELL 0.0665*** 0.0666*** 0.0638*** 0.0521*** 0.0431*** CalWORKs 0.0415*** 0.0361*** 0.0361*** 0.0337*** 0.0445*** Developmental math or English 0.1065*** 0.1060*** 0.103*** 0.0990*** 0.0956*** GPA 0.0279*** 0.0319*** 0.0280*** 0.0263*** 0.0424*** Work based courses -0.0379*** -0.0526*** -0.171*** -0.189*** Evening courses 0.0336*** -0.0030 0.0323*** 0.0008 Online courses -0.1170*** -0.0685*** -0.202*** -0.241*** Average wage return for TOP code -0.00346*** -0.0030*** -0.0020*** -0.0014*** IT -0.0062434 0.0079 0.0319*** 0.0517*** Engineering 0.0816474 0.0835*** 0.0356*** 0.0959*** Family and consumer 0.0649579 0.0670*** 0.0510*** 0.0723*** Public and protective -0.0145007 -0.0020 -0.0043 0.0042 Constant 0.2412 0.327*** 0.275*** 0.139*** 0.219*** Observations 66050 65,187 65,187 65,187 22,445 R-squared 0.055 0.055 0.061 0.098 0.108 Fixed effects Year XX TOP 2 X X XX College XX Exclusions Transfers included X XX In scan only X XX Only 2010 forward X SOURCES: Authors calculations from COMIS data and pathway database. NOTES: Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1 The dependent variable is a bivariate on whether a student whose initial credential was a short-term certificate in one of our scanned colleges earned a related credential in the same program (ICN7) within 3 years. To test whether program design is related to the odds a student stacks credentials, we estimate Model 5 including variables to indicate program features, as described in the text. Table B5 presents the estimates from those primary variables of interest. Each cell in the table is derived from a separate regression model. We do this to control which programs we are comparing in each case. For example, we test programs with strong progressive or launchpad/core pathways compared to programs with pathways that are just not explicit or to programs with no pathway, in turn. For this reason, the sample sizes vary greatly across models and are shown in Table B6. Our main models examining pathway types include a host of demographic and program characteristics. However, to understand how the relationships vary across key dimensions, we re-estimate models separately by race/ethnicity, gender, and 2-digit TOP code. These models allow all of the factors related to stacking (not just program design but also age, educational background, college, etc.) to vary in their importance across groups. This specification is more flexible than a model that incorporates interactions, though in those we find qualitatively similar results. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 13 TABLE B5 Main estimates from fixed effects models of program features on earning a stacked credential Dependent variable: Stack credential within program Overall Latino Asian African American White Male Female Business IT Engineering Family/ Public/ Consumer Protective Explicit pathways compared to… All others 0.0456*** 0.0834*** 0.0712*** -0.0468 0.0304* 0.0321** 0.0360** 0.0923*** 0.0396 0.0832*** -0.0294 0.198*** Programs with no pathways Programs with light pathway features Programs with pathway features (but not explicit) 0.164*** 0.173*** 0.0994*** 0.113*** 0.186*** 0.139*** 0.167* 0.00615 0.110*** 0.150*** 0.0713*** 0.117*** 0.135*** 0.0499** 0.0225 -0.0669 0.0558*** 0.0969*** 0.0650** -0.0458 0.0524*** 0.0441*** 0.0404** 0.113*** 0.104 0.105 0.150 0.160*** 0.135*** 0.0351 0.492* -0.00935 -0.0202 0.252*** 0.245*** 0.114** Other comparisons within pathway type Explicit launchpad/lattice programs vs programs with no strong pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificateAA only) Launchpad/Lattice programs vs programs with no pathway feature 0.0423*** 0.0421*** 0.0888*** 0.148*** 0.144*** 0.0672*** 0.0931*** 0.145*** 0.192*** 0.206*** 0.108*** 0.0610* 0.170*** 0.182*** 0.211*** -0.0812 0.0422* 0.0466** 0.0174 -0.00959 -0.0126 0.00505 0.0322* 0.0415* 0.114*** 0.136*** 0.0177 0.0611*** 0.162*** -0.0396 0.199*** 0.0949*** 0.148*** 0.204*** -0.0931 0.196*** 0.0753*** 0.124*** 0.203*** -0.141** 0.0382 0.106 0.0592 0.0767* 0.0560 0.0877*** 0.114*** 0.251*** 0.283*** 0.239*** -0.0620 0.00973 0.202*** 0.235*** 0.179** 0.216*** 0.132 -0.00406 0.243*** 0.120*** SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the coefficient from a separate model. The dependent variable is a bivariate on whether a student whose initial credential was a short-term certificate in one of our scanned colleges earned a related credential in the same program (ICN7) within 3 years. All models include covariates listed in Table B4 Model 5, which includes student and program characteristics as well as year, college, and 2-digit TOP code fixed effects. Column headings refer to the subsample used in the regression model. Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1. The sample size varies across each model, and are provided in the next table. Full results are available upon request. All models only include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first career education credential, in addition restrictions on the comparison programs or subsamples noted in the table. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 14 TABLE B6 Sample size for main fixed effects models of program features Overall Latino Asian African American White Male Female Business IT Engineering Family/ Public/ Consumer Protective Explicit pathways compared to… All others 22,429 8,520 3,026 1,774 8,173 11,566 10,863 4,262 1,237 5,332 6,713 4,885 Programs with no pathways Programs with light pathway features Programs with pathway features (but not explicit) 6,531 11,181 19,761 2,330 3,949 7,686 1,043 1,473 2,699 451 782 1,549 2,464 4,536 7,002 3,647 6,793 9,404 2,884 4,388 10,357 964 1,356 4,218 972 1,010 382 1,348 2,600 4,838 1,651 2,479 6,645 1,596 3,736 3,678 Other comparisons within pathway type Explicit launchpad/ lattice programs vs programs with no explicit pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificateAA only) Launchpad/lattice programs vs programs with no pathway feature 20,557 20,687 13,171 17,602 22,044 7,915 7,684 5,388 6,729 8,409 2,509 2,893 1,710 2,523 2,983 1,708 1,631 1,101 1,382 1,756 7,557 10,697 9,860 3,623 7,598 11,059 9,628 4,126 4,416 5,016 8,155 3,154 6,208 8,799 8,803 3,723 7,970 11,500 10,544 3,903 1,191 1,174 230 319 1,237 4,827 5,052 1,861 4,538 5,344 6,147 5,697 5,183 5,604 6,688 4,769 4,638 2,743 3,418 4,872 SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the sample size of a separate model, corresponding with Table B5. Column headings refer to the subsample used in the regression model. Full results are available upon request. All models only include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first CTE credential, in addition to the comparison programs in the first column and subsamples noted in the first row. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 15 Table B7 reports the results from a number of additional tests of our main models. As above, each cell in this table is the main coefficient of interest from a separate regression model that also includes all covariates in Table B4 Model (5). The first three columns test the importance of pathway design on a variety of alternative student samples: (1) only students who re-enroll in a community college after their first short-term certificate (perhaps indicating an intent to complete a pathway), (2) adding students whose first award is a long-term certificate to the main sample, and (3) including students who transfer. Estimates on the relationship between pathway design and stacking are very similar across these student samples—they are also very similar to our main estimates above. The last three columns of Table B7 test alternative dependent variables. We remove the restriction that a student must stack within 3 years, we require the student to stack on a slightly more narrowly defined pathway (ICN8), and we test stacking to higher-level credentials (from short-term to long- or associate). Once again, the results are quite similar. TABLE B7 Additional tests on main models of program features on earning a stacked credential Dependent variable: Stack credential within program Students Students Students who whose first whose first re-enroll after award is award is a S first short- either a short-term term short- or certificate, certificate long-term regardless of certificate transfer Strong pathway compared to… Testing alternative dependent variables Stack within ICN7 but no restriction on how long it takes Stack within ICN8 within 3 years Stack within ICN7 to a higher level credential All others 0.0550*** 0.0452*** 0.0422*** 0.0383*** 0.0479*** 0.0346*** Programs with no pathways 0.169*** 0.164*** 0.162*** 0.167*** 0.165*** 0.0969*** Programs with light pathway features Programs with pathway features (but not explicit) 0.0973*** 0.0797*** 0.113*** 0.0527*** 0.101*** 0.0489*** 0.0925*** 0.0468*** 0.108*** 0.0589*** 0.0663*** 0.0230*** Other comparisons within pathway type: Explicit lattice programs vs programs with no explicit pathway Explicit progressive programs vs programs with no explicit pathway Progressive programs vs programs with no pathway feature Progressive programs vs programs with no feature (or certificate-AA only) Lattice programs vs programs with no pathway feature 0.0732*** 0.0373** 0.0950*** 0.150*** 0.159*** 0.0427*** 0.0553*** 0.117*** 0.175*** 0.162*** 0.0489*** 0.0251** 0.0918*** 0.149*** 0.142*** 0.0387*** 0.0303** 0.103*** 0.163*** 0.159*** 0.0411*** 0.0427*** 0.0879*** 0.146*** 0.142*** 0.0163 0.0471*** 0.0964*** 0.115*** 0.114*** SOURCES: Author calculations from COMIS data and pathway database. NOTES: Each cell is the coefficient from a separate model. The dependent variable is a bivariate that changes across models as indicated in column headings. All models include covariates listed in Table B4 Model 5 as well as year, college, and 2-digit TOP code fixed effects. Reference categories are: white, high school graduate/GED, citizen. *** p<0.01, ** p<0.05, * p<0.1. Only main effects are shown here, but full results are available upon request. Baseline models include those students enrolled in programs in our catalog scan and enrolled from 2010 forward and who did not transfer within 3 years of their first CTE credential, as well as sample restrictions noted in column headings. PPIC.ORG Technical Appendices Title Stackable Credentials in Career Education at California Community Colleges 16 The Public Policy Institute of California is dedicated to informing and improving public policy in California through independent, objective, nonpartisan research. Public Policy Institute of California 500 Washington Street, Suite 600 San Francisco, CA 94111 T: 415.291.4400 F: 415.291.4401 PPIC.ORG PPIC Sacramento Center Senator Office Building 1121 L Street, Suite 801 Sacramento, CA 95814 T: 916.440.1120 F: 916.440.1121" ["post_date_gmt"]=> string(19) "2018-10-23 03:54:40" ["comment_status"]=> string(6) "closed" ["ping_status"]=> string(6) "closed" ["post_password"]=> string(0) "" ["post_name"]=> string(16) "1018sbr-appendix" ["to_ping"]=> string(0) "" ["pinged"]=> string(0) "" ["post_modified"]=> string(19) "2018-10-22 20:54:57" ["post_modified_gmt"]=> string(19) "2018-10-23 03:54:57" ["post_content_filtered"]=> string(0) "" ["guid"]=> string(59) "http://www.ppic.org/wp-content/uploads/1018sbr-appendix.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) }