Impacts of the Virginia Initiative for Employment Not Welfare

Transcription

1 Contract No.: CR MPR Reference No.: Impacts of the Virginia Initiative for Employment Not Welfare Final Report January 2002 Anne Gordon Susanne James-Burdumy Submitted to: Submitted by: Virginia Department of Social Services Mathematica Policy Research, Inc. 730 E. Broad Street P.O. Box 2393 Richmond, VA Princeton, NJ (609) Under Subcontract to: Center for Public Administration and Policy Virginia Polytechnic Institute and State University 7054 Haycock Road Falls Church, VA Project Director: Anne Gordon

2 a

3 ACKNOWLEDGMENTS Many people at the Virginia Department of Social Services (VDSS), Virginia Tech, and MPR contributed to this report. At VDSS, Carol Baron, Director of Research and Evaluation, provided overall leadership and direction. Mike Theis, Molly Sheahan, and Carol Baron developed the administrative data extracts needed for this study, provided documentation for the files, and found time in their busy schedules to respond to our questions. Carol also provided thoughtful comments on the draft report. At Virginia Tech, Renee Loeffler and Carole Kuhns provided valuable input into the design of this study, and Renee also reviewed the draft report. At MPR, Walter Corson reviewed the report and provided helpful comments. Amy Zambrowski led the development of analysis files from the administrative data files received from VDSS. This responsibility included ensuring updated files were appropriately merged with those from the earlier period covered by the previous impact report. Michelle VanNoy, an MPR programmer, worked with selected administrative data files and also prepared all of the tabulations and regressions in the report. Roberto Agodini and West Addison conducted much of the data analysis for the previous impact report and thus provided a critical foundation for the report to build upon. Walt Brower edited the report and Monica Capizzi developed the figures. Cathy Harper and Jane Nelson led the production team. We are grateful for this assistance but take full responsibility for any errors that may remain. iii

4 a

5 CONTENTS Chapter Page EXECUTIVE SUMMARY... xv I INTRODUCTION...1 A. OVERVIEW OF VIP Eligibility Reforms VIEW Transition to TANF...7 B. THE VIP IMPACT EVALUATION Why Do an Experimental Impact Evaluation? Research Questions Evaluation Design Limitations of Early Impact Results...13 C. PLAN OF THE REPORT...15 II IMPACTS ON EMPLOYMENT-PREPARATION ACTIVITIES AND SANCTIONS...17 A. DID VIEW AFFECT PARTICIPATION IN EMPLOYMENT-PREPARATION ACTIVITIES? Research Base: Activity Data and Outcome Measures Participation Overall: Higher Under VIEW Job Search: Greatest Impact CWEP: Used Surprisingly Little Job Readiness: Use Varied by Site Education and Training: Less Change than Expected...25 B. DID VIEW LEAD TO MORE SANCTIONS? Research Base: Sanction Data and Outcome Measures Employment-Services Sanctions: More Under VIEW...27 v

6 CONTENTS (continued) Chapter Page III IMPACTS ON EMPLOYMENT AND EARNINGS...29 A. RESEARCH BASE: EMPLOYMENT DATA AND OUTCOME MEASURES...30 B. DID VIEW AFFECT EMPLOYMENT RATES AND EARNINGS? Employment Rates: Higher Under VIEW Earnings: Also Higher Employment and Earnings Impacts by Education Level: Similar for Those with and Without a High School Credential...42 IV IMPACTS ON PUBLIC ASSISTANCE RECEIPT AND BENEFITS...43 A. HOW DID VIEW AFFECT TANF PARTICIPATION AND BENEFITS? Research Base: TANF Data and Outcome Measures TANF Participation and Benefits in VIEW Sites: Impacts Varied by Site...45 B. HOW DID VIEW AFFECT FOOD STAMP PROGRAM PARTICIPATION AND BENEFITS? Research Base: Food Stamp Data and Outcome Measures Food Stamp Program Participation and Benefits in VIEW Sites: Impacts Varied by Site...54 V IMPACTS ON COMBINING WORK AND TANF AND ON TOTAL INCOME...61 A. DID VIEW AFFECT HOW FAMILIES COMBINED WORK AND TANF? Combining Work and TANF: Sometimes Higher Under VIEW On TANF and Not Working: Lower Under VIEW Working and Off TANF: Higher in Petersburg Not Working and Off TANF: Little Difference...72 vi

7 CONTENTS (continued) Chapter Page V B. DID VIEW LEAD TO HIGHER INCOMES?...72 (continued) 1. Research Base: Income Data and Measure Total Income: Little Impact...75 VI IMPACTS ON CHILD CARE SUBSIDIES...79 A. RESEARCH BASE: CHILD CARE SUBSIDY DATA AND OUTCOME MEASURES...80 B. HOW DID VIEW AFFECT CHILD CARE SUBSIDIES? Subsidy Receipt Overall: Much Higher Under VIEW Dollar Value of Subsidies: Even Higher TANF Child Care Subsidies: Higher Under VIEW Post-TANF Child Care Subsidies: Mixed Impacts...92 VII SUMMARY OF FINDINGS A. OVERVIEW OF FINDINGS VIEW s Impacts Before the Time Limit Effects of the Time Limit Differences in VIEW Implementation and Impacts Among the VIEW Sites B. RESULTS IN CONTEXT REFERENCES APPENDIX A: CHARACTERISTICS OF RECIPIENT CASES...A.1 APPENDIX B: IMPACTS ON EMPLOYMENT-PREPARATION ACTIVITIES AND SANCTIONS OF RECIPIENTS: TABLES...B.1 APPENDIX C: IMPACTS ON EMPLOYMENT AND EARNINGS OF RECIPIENTS: TABLES... C.1 vii

8 CONTENTS (continued) Page APPENDIX D: IMPACTS ON PUBLIC ASSISTANCE RECEIPT AND BENEFITS OF RECIPIENTS: TABLES...D.1 APPENDIX E: IMPACTS ON COMBINING WORK AND TANF AND ON TOTAL INCOME OF RECIPIENTS: TABLES... E.1 APPENDIX F: IMPACTS ON CHILD CARE SUBSIDIES TO RECIPIENTS: TABLES... F.1 APPENDIX G: IMPACTS ON APPLICANT CASES...G.1 viii

9 TABLES Table Page V.1 OVERVIEW OF VIEW S IMPACTS ON WORK/TANF COMBINATIONS...61 ix

10 a

11 FIGURES Figure Page I.1 VIP PROGRAM AND EVALUATION MILESTONES...2 II.1 II.2 VIEW LED TO HIGHER PARTICIPATION IN EMPLOYMENT-PREPARATION ACTIVITIES...21 JOB SEARCH ACCOUNTED FOR MOST OF THE HIGHER ACTIVITY PARTICIPATION RATE UNDER VIEW: CWEP PARTICIPATION RATES WERE LOW...23 II.3 MORE EMPLOYMENT-SERVICES SANCTIONS UNDER VIEW...28 III.1 III.2 III.3 III.4 IV.1 VIEW LED TO HIGHER AVERAGE EMPLOYMENT RATES IN THE FULL VIEW SAMPLE...33 VIEW LED TO HIGHER AVERAGE EMPLOYMENT RATES IN ALL SITES...35 VIEW INCREASED AVERAGE EARNINGS IN THE FULL VIEW SAMPLE...39 VIEW INCREASED AVERAGE EARNINGS IN SOME SITES IN SOME QUARTERS...40 VIEW HAD NO EFFECT ON TANF PARTICIPATION UNTIL TIME LIMITS REACHED...46 IV.2 VIEW HAD NO EFFECT ON AVERAGE TANF BENEFITS...47 IV.3 VIEW S MIXED INCENTIVES FOR LEAVING TANF HAD VARYING EFFECTS BY SITE...48 IV.4 VIEW HAD NO EFFECTS ON AVERAGE TANF BENEFITS BY SITE...49 IV.5 IV.6 IV.7 VIEW DID NOT AFFECT FOOD STAMP PROGRAM PARTICIPATION OVERALL...55 VIEW LED TO LOWER FOOD STAMP BENEFITS FOR THE FULL VIEW SAMPLE...56 VIEW DID NOT AFFECT FOOD STAMP PROGRAM PARTICIPATION BY SITE...58 xi

12 FIGURES (continued) Figure Page IV.8 VIEW LED TO LOWER FOOD STAMP BENEFITS IN SOME SITES...59 V.1 VIEW INCREASED THE PROPORTION OF CASES COMBINING WORK AND TANF...63 V.2 VIEW INCREASED THE PROPORTION OF CASES COMBINING TANF AND WORK IN LYNCHBURG AND PRINCE WILLIAM...64 V.3 VIEW SUBSTANTIALLY DECREASED THE PROPORTION OF CASES ON TANF AND NOT WORKING...67 V.4 VIEW SUBSTANTIALLY DECREASED THE PROPORTION OF CASES ON TANF AND NOT WORKING IN EACH SITE...68 V.5 VIEW DID NOT INCREASE THE PROPORTION OF CASES WORKING AND OFF TANF OVERALL UNTIL THE TIME LIMIT APPLIED...70 V.6 VIEW INCREASED THE PROPORTION OF CASES WORKING AND OFF TANF IN PETERSBURG ONLY...71 V.7 VIEW DID NOT AFFECT THE PROPORTION OF CASES OFF TANF AND NOT WORKING...73 V.8 VIEW DID NOT AFFECT THE PROPORTION OF CASES OFF TANF AND NOT WORKING IN EACH SITE...74 V.9 VIEW HAD MODEST EFFECTS ON TOTAL INCOME...76 V.10 VIEW DID NOT LEAD TO HIGHER INCOMES AT THE SITE LEVEL...77 VI.1 MORE CASES RECEIVED CHILD CARE SUBSIDIES UNDER VIEW...82 VI.2 MORE CASES RECEIVED CHILD CARE SUBSIDIES UNDER VIEW IN EACH SITE...83 VI.3 VIEW CASES RECEIVED HIGHER CHILD CARE SUBSIDIES...85 VI.4 VIEW CASES RECEIVED HIGHER CHILD CARE SUBSIDIES IN EACH SITE...86 xii

13 FIGURES (continued) Figure Page VI.5 VIEW LED TO MORE CASES RECEIVING TANF CHILD CARE SUBSIDIES...89 VI.6 VIEW LED TO HIGHER TANF CHILD CARE SUBSIDIES OVERALL...90 VI.7 VI.8 VI.9 VI.10 VI.11 VI.12 VIEW LED TO MORE CASES RECEIVING TANF CHILD CARE SUBSIDIES IN LYNCHBURG AND PRINCE WILLIAM...91 TANF CHILD CARE SUBSIDY AMOUNTS WERE SUBSTANTIALLY HIGHER UNDER VIEW IN LYNCHBURG AND PRINCE WILLIAM...93 VIEW DID NOT AFFECT POST-TANF CHILD CARE RECEIPT ON AVERAGE...94 IMPACTS ON POST-TANF CHILD CARE SUBSIDY RECEIPT VARIED AMONG THE VIEW SITES...95 VIEW HAD MIXED EFFECTS ON POST-TANF CHILD CARE SUBSIDY AMOUNTS...97 IMPACTS ON POST-TANF CHILD CARE SUBSIDY AMOUNTS VARIED AMONG THE VIEW SITES...98 xiii

14 a

15 EXECUTIVE SUMMARY Virginia s welfare reform program the Virginia Independence Program (VIP) focuses on promoting work and family responsibility. VIP has two distinct components. The first involves eligibility requirements that differ from those under Aid to Families with Dependent Children (AFDC). These requirements were intended to encourage family responsibility. The second is the Virginia Initiative for Employment not Welfare (VIEW). VIEW is one of the nation s strongest examples of a work first program, emphasizing rapid movement of public assistance clients into jobs. Overall, VIP has involved a substantial commitment to changing the culture of welfare both for program staff and for clients. VIP eligibility requirements were implemented on July 1, 1995, after the program received waivers of federal AFDC program rules. VIEW was phased in by locality between July 1995 and October In 1996, the federal Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) replaced AFDC with a block grant for Temporary Assistance for Needy Families (TANF). Having already shifted its AFDC program to a temporary assistance program with employment as its central focus, Virginia implemented TANF in February 1997 with minimal modifications to VIP. 2 The VIP eligibility requirements include: C Stronger requirements for cooperation with child support enforcement C Cap on benefits for children born more than 10 months after assistance is authorized C Age-appropriate immunizations for children C Compliance with school attendance laws C Determination of benefits for two-parent families using the same standards as those for single-parent families 1 During most of this period, one or two Economic Development Districts (EDDs)--contiguous groups of 2 to 13 localities--implemented VIEW each quarter. Based on early success, implementation was speeded up, with roughly half the state s localities implementing VIEW in either July or October For simplicity, this report uses the term TANF to refer to Virginia s cash assistance program, unless pre-vip policies are explicitly being discussed. xv

16 Key provisions of VIEW, which applies to able-bodied parents with no child under the age of 18 months, include: C Signing of an Agreement of Personal Responsibility as a condition for receiving benefits C Required job search for 90 days or until employed, followed by mandatory work either through regular employment or participation in the Community Work Experience Program (CWEP), which involves work in a nonprofit or public setting in exchange for benefits C Full family sanctions (complete loss of benefits) for noncompliance 3 C A 24-month time limit on TANF benefits C A generous earned income disregard, which allows families to continue to receive their full TANF grant as long as their net earned income plus TANF benefits remains below the federal poverty line C Supportive services, including subsidized child care, transportation assistance, and Medicaid, while on TANF and for one year after the TANF case closes GOALS OF THE REPORT This report describes the early and midterm impacts of VIP and VIEW. To study the impacts of VIP, the Virginia Department of Social Services (VDSS) ran VIP as an experiment in five localities: the cities of Lynchburg, Petersburg, and Portsmouth, and the counties of Prince William and Wise. As soon as VIP was implemented in July 1995, VDSS began randomly assigning both new and existing cases in those localities to either an experimental group or a control group. All experimental cases were covered by VIP eligibility policies. They were enrolled in the VIEW work component (unless exempt) when VIEW started in their area. Control group cases remained subject to the old AFDC policies and received employmentrelated services only under the old Job Opportunities and Basic Skills (JOBS) program. The purpose of random assignment was to create a situation in which people in the experimental group were similar to those in the control group, on average, in all respects except for whether they were covered by VIP policies. The random-assignment evaluation design allows us to attribute differences in outcomes between the experimental and control groups to the VIP reforms with the maximum level of statistical confidence. 3 Failure to sign the Agreement of Personal Responsibility results in case closure. Failure to comply with the job search or work requirements after signing the agreement results in a 100 percent sanction for a minimum period. During the sanction period, the sanction months count toward the 24-month time limit unless the client chooses to close the case. xvi

17 This report addresses four questions: 1. How does the VIEW component of VIP affect program activities and services used? How do client experiences in VIEW differ from client experiences in JOBS, in terms of their overall participation in activities, the types of activities they engaged in, the frequency with which they were sanctioned, and their use of child care assistance? 2. What economic outcomes are observed for VIEW cases? What are the trends over time in key outcomes employment, earnings, TANF participation, TANF benefits, Food Stamp Program participation, food stamp benefits, and total income for VIEW cases? 3. What is the impact of VIEW on economic outcomes? How do the outcomes for those in VIEW differ from what they would have been had these clients remained subject to the old AFDC/JOBS program? 4. How does the impact of VIEW vary among the demonstration sites? Why do impacts vary? These questions focus on the impacts of VIEW, because the previous impact report (Gordon and Agodini 1999) showed that the VIP eligibility reforms by themselves had no impacts on these outcomes. Thus, we refer to the impacts of VIEW rather than the impacts of VIP as a whole. In October 1997, VDSS started bringing the control group under VIP/VIEW policies, as the state no longer needed to operate VIP under a federal waiver. This transition took about six months. Thus, comparisons of the experimental and control groups for the period starting in October 1997 no longer indicate the full effects of VIEW. Instead, they reveal the effects of longer exposure to VIEW versus recent enrollment. Almost certainly, such estimates understate the true impacts of VIEW relative to earlier policies. Nonetheless, the results from these quarters provide useful information. In particular, it is only during this period that we can examine the effects of the 24-month time limit on TANF benefits for VIEW enrollees. EVALUATION DESIGN Because the previous report showed that VIP eligibility policies without VIEW had no impacts on the outcomes of interest, the sample for this report includes only cases in the three sites that implemented VIEW during the experimental period Lynchburg, Petersburg, and Prince William. In addition, the sample for this report includes only cases in the VIEW sites likely to be mandatory VIEW participants, based on their characteristics (such as the age of the youngest child) at the time of random assignment. The expectation was that VIEW impacts were likely to be concentrated in this group. This imputed VIEWmandatory status was computed the same way for experimental and control cases. xvii

18 Thus, this report focuses on the experiences of 2,444 mandatory cases that were receiving AFDC on July 1, 1995, in Lynchburg, Petersburg, or Prince William. At that time, half were randomly assigned to the experimental group, and half were assigned to the control group. Outcome data were available on these cases for 14 quarters, from July 1, 1995, to December 31, These data are from the administrative records of VDSS or the Virginia Employment Commission (VEC). KEY FINDINGS The following are the key findings during the period before any cases had reached the two-year time limit: C VIEW quickly increased employment, largely by increasing the proportion of clients who both worked and received TANF. C VIEW had no impact on TANF participation and benefits when averaged across the three sites. When the time limit began to affect experimental cases who remained on TANF, the following was observed: C TANF participation and benefits were generally lower for experimental cases than for control cases after the time limit took effect or even just before. C Although control cases were phased into VIEW during this period, employment effects persisted, and the time limit may be one of the reasons for this. IMPACTS ON PARTICIPATION IN EMPLOYMENT-PREPARATION ACTIVITIES Among cases likely to be mandatory for VIEW, the program s impact on participation rates in employment-preparation activities and on sanction rates indicates that VIEW was implemented as intended and that it represented real change from JOBS. C VIEW led to a much higher participation rate in employment-preparation activities. All VIEW-mandatory experimental clients had to participate in an employment-preparation activity right away if they were not already employed. JOBS-mandatory control clients could remain on a waiting list for employment services indefinitely. xviii

19 C Job search accounted for most of the higher activity participation rate under VIEW. When they entered VIEW, all VIEW-mandatory clients not yet employed had to participate in job search or face sanction. JOBS-mandatory clients either had other options or were on a waiting list. C CWEP participation rates were low for both the experimental and the control groups. VIEW clients who do not find work within 90 days must participate in CWEP, and this requirement has been enforced. Nonetheless, because most VIEW clients found unsubsidized employment, nine percent, at most, of mandatory clients were enrolled in CWEP annually. During the experimental period, VIEW significantly increased CWEP placements in only one site. C Although VIEW was expected to lead to lower levels of participation in education and training, the difference was less than expected. Among the three VIEW sites, control cases in Lynchburg had a higher rate of participation in education activities than experimental cases. However, in Petersburg, the experimental group had a higher rate of participation in education and training, and in Prince William, the two groups had similarly low levels of participation. C VIEW led to more sanctions for not participating in employment services. In all three VIEW sites, there was a significant difference between experimental and control cases in sanction rates for failure to participate in employment services. VIEW s strong job search and work requirements were major reasons. Sanctions were more common under VIEW despite their more severe consequences. IMPACTS ON EMPLOYMENT AND EARNINGS As expected, VIEW led to higher employment rates and earnings for mandatory VIEW cases. Several aspects of VIEW were expected to increase employment. These include the expanded earnings disregard, the 90-day work requirement, the two-year time limit on receipt of TANF benefits, and enhanced transitional child care eligibility. The key findings on employment and earnings are: C VIEW led to higher average employment rates. In all three VIEW sites, more experimental cases than control cases worked in most quarters. In the quarters after VIEW was implemented and before the discontinuation of control group policies in each VIEW site, employment rates were from 5 to 13 percentage points higher among experimental cases than among control cases. After control policies ended, impacts for the full VIEW sample declined slightly in magnitude but remained significantly higher than those of the control group in three of the five quarters. xix

20 C VIEW also increased average earnings. In many VIEW experimental quarters, average quarterly earnings were higher for experimental cases than for control cases. However, the impact on earnings in each site was less consistent than the employment impact and was not always statistically significant. Nonetheless, even after control policies ended, the experimental group as a whole earned significantly more than the control group in two of the five quarters. C Impacts persist after the discontinuation of control group policies. This persistence is noteworthy, since after control policies ended, members of the control group who remained on TANF were subject to VIEW policies. There are a number of potential explanations for the persistence of the impacts. Adults in experimental cases may have already found employment, while controls who were newly subject to the VIEW policies may have just begun moving into employment. Because they started working sooner, the experimental cases who were working could be earning more than the control cases who were working. In addition, reaching the time limit may have led more experimental cases to work or to work more hours. C Impacts may be understated because of data limitations. Data used for this analysis are from VEC earnings records. These data do not capture self-employment, federal employment, or employment outside Virginia. As a result, impacts may be understated, particularly in Prince William, since Prince William is near the District of Columbia, Maryland, and several military bases. IMPACTS ON TANF AND FOOD STAMPS In the short term, VIEW reduced TANF participation, TANF benefits, and Food Stamp Program benefits among mandatory cases in some quarters, in some sites. C During the VIEW experimental period, VIEW s mixed incentives for leaving TANF had no effect in two sites but reduced participation and benefits in one site. For the full VIEW sample the sample comprising mandatory cases from Lynchburg, Petersburg, and Prince William no significant differences in TANF benefits or participation existed during the VIEW experimental period. Although TANF participation and benefits were significantly lower in Petersburg in several quarters, they were about the same or insignificantly higher in Lynchburg and Prince William, leading to no effects on average. The cross-site difference in impacts may be due to Petersburg VIEW staff placing more emphasis on the time limit in discussions with clients than staff in other sites did. C The 24-month time limit reduced TANF participation overall and especially in Lynchburg, the first site to implement VIEW. Experimental cases in Lynchburg, who began to reach the 24-month time limit in the last five quarters of the follow-up period, had significantly lower TANF participation rates and benefits during those quarters than control cases, who would not be subject to the time limit for another two years. xx

21 C VIEW led to lower Food Stamp Program benefits in the full VIEW sample, since 4 VIEW led to higher earnings but did not, on average, reduce TANF benefits. These effects did not persist after the end of control group policies. In addition, VIEW did not significantly affect Food Stamp Program participation in the full VIEW sample or in any of the sites. IMPACTS ON WORK/TANF COMBINATIONS AND TOTAL INCOME During the VIEW experimental period the period after the implementation of VIEW and before the end of control group policies mandatory experimental and control cases combined work and TANF participation in significantly different ways. Nonetheless, VIEW appears to have had little effect on total incomes, as declines in food stamp benefits partly offset increased earnings. The major findings concerning VIEW s impacts on work/tanf combinations and total income are: C In the period before the time limit applied, in the full VIEW sample and in Lynchburg, VIEW increased the proportion of clients who worked and received TANF. The larger earnings disregard offered through VIEW allowed most experimental clients who worked to continue to receive a TANF benefit. Thus, in the period before the two-year time limit took effect, more experimental cases than controls combined work and welfare. After the time limit took effect, this pattern reversed in Lynchburg and in the three sites combined. C VIEW reduced the percentage of clients on TANF and not working. Fewer experimental cases than control cases received TANF and did not work in each of the three sites and in the full VIEW sample. C Before the time limit applied, VIEW increased the percentage of clients both working and off TANF in Petersburg. This is because Petersburg was the only site in which VIEW led to lower TANF participation before the time limit. No significant effects were found in the full VIEW sample. However, after the time limit began to apply, the percentage of experimental clients working and off TANF was higher in both Lynchburg and Petersburg. C VIEW had little impact on total income from work, TANF, child support, and food stamps. For the full VIEW sample, incomes of experimental and control cases were significantly different during two early quarters. Experimental cases generally had slightly higher incomes than control cases, but effects were not statistically significant after VIEW 4 Thus, experimental cases had larger cash incomes than control cases, which automatically reduced their Food Stamp Program benefits. xxi

22 implementation in any site. Higher earnings for experimental cases were partially offset by lower food stamp benefits. As cases began to reach the time limit, this result did not change. IMPACTS ON CHILD CARE SUBSIDIES VIEW child care policies were designed to make child care subsidies available to all clients who need them. Most TANF clients are single parents with young children, so they need to find adequate and affordable child care so that they can work. Under the JOBS program, participation in work or workpreparation activities was limited in part by lack of child care funding. To make VIEW s work requirement viable for all mandatory cases, Virginia substantially increased funding for child care (Pavetti et al. 1999). 5 In addition, VIEW made it easier for those who left TANF to qualify for transitional child care. At the same time, child care funding for working parents not on welfare was also increasing, and these funds were available to both experimental and control group families who took jobs. Key findings on the use of child care subsidies include: C Under VIEW, more families received some type of child care subsidy. Experimental cases also received higher average subsidy amounts than control cases. Subsidies counted include child care subsidies for families on TANF, transitional child care for those whose TANF cases had closed within the past 12 months, and other state-administered subsidies. C Under VIEW, child care subsidies for cases still on TANF increased substantially. More experimental cases than control cases combined work and TANF. A larger proportion of experimental cases thus received TANF child care subsidies to support their movement into work. TANF subsidy amounts were also higher for experimental cases than for control cases. C Impacts on post-tanf child care subsidies were scattered and inconsistent over time and across sites. In general, experimental and control cases were equally likely to receive a subsidy. The amounts received by experimental cases were sometimes higher and sometimes lower than amounts received by control cases. 5 Before VIEW, transitional child care was restricted to cases closed because of earnings. VIEW expanded eligibility to those who closed their cases voluntarily. In addition, VIEW ended the requirement that a client had to have received benefits in three of the six months before the case closed in order to receive transitional child care. xxii

23 CONCLUSION As discussed further in the full report, the findings from the VIEW impact study accord reasonably well with those from other studies of programs that combine time limits with generous earnings disregards. All these programs have been successful in increasing employment and earnings of participants, a primary goal. The policies implemented unambiguously encouraged work. All these studies suggest that, in the short term, before the time limit is reached, programs that combine time limits and generous disregards do not usually decrease TANF participation and benefits. The extra benefits available to participants when they go to work seem to be at least as attractive as the months of eligibility they could be saving. The results for Petersburg, however, indicate that caseworkers who put a strong emphasis on the time limit may be able to persuade clients to leave TANF more quickly. After the time limit applies, the effect of these programs is generally to reduce TANF participation and benefits, as some who would otherwise have remained on TANF have their cases closed. Further studies of the role of caseworkers in affecting how clients balance time limits and earnings disregards could be of interest. xxiii

24 a

25 I. INTRODUCTION Virginia s multifaceted welfare reform program the Virginia Independence Program (VIP) is a prime example of the new philosophy of welfare reform that focuses on promoting work and family responsibility. VIP has two distinct components. The first is changes in policies regarding eligibility for Aid to Families with Dependent Children (AFDC). These changes were intended to encourage family responsibility. The second is the Virginia Initiative for Employment not Welfare (VIEW). VIEW is one of the nation s strongest examples of a work first program, emphasizing rapid movement of public assistance clients into jobs. Overall, VIP represents a substantial commitment to changing the culture of welfare both for program staff and for clients. VIP passed the state general assembly with widespread bipartisan support and was signed into law by Governor George Allen on March 20, VIP eligibility changes were implemented on July 1, 1995, after the program received waivers of federal AFDC program rules. VIEW was phased in by locality 1 between July 1995 and October In 1996, the federal Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) replaced AFDC with a block grant for Temporary Assistance for Needy Families (TANF). Virginia implemented TANF in February 1997 with minimal modifications 2 to VIP. Figure I.1 (left side) illustrates the timeline of VIP implementation. 1 During most of this period, one or two Economic Development Districts (EDDs)--contiguous groups of 2 to 13 localities--implemented VIEW each quarter. Based on early success, implementation was speeded up, with roughly half the state s localities implementing VIEW in July or October For simplicity, this report uses the term TANF to refer to Virginia s cash assistance program, unless pre-vip policies are explicitly being discussed. 1

26 Figure I.1 VIP Program and Evaluation Milestones July 1995 October 1995 Program VIP eligibility reforms implemented statewide. Lynchburg implements VIEW. Evaluation Random assignment begins. January 1996 April 1996 Prince William implements VIEW. July 1996 October 1996 January 1997 February 1997 Petersburg implements VIEW. TANF is implemented. April 1997 July 1997 October 1997 Portsmouth and Wise implement VIEW. Control Policies end. December 1988 Follow-up period ends. 2

27 This report describes the early and midterm impacts of VIP on various outcomes, including TANF clients employment and earnings, total income, receipt of public assistance, and use of child care subsidies. To study the impacts of VIP, the Virginia Department of Social Services (VDSS) ran VIP as an experiment in five localities: the cities of Lynchburg, Petersburg, and Portsmouth, and the counties of Prince William and Wise. To set the experiment in motion, as soon as VIP was implemented in July 1995, VDSS began randomly assigning both new and existing cases in those localities to either an experimental group or a control group. All experimental cases were covered by VIP eligibility policies. They were enrolled in the VIEW work component (unless exempt) when VIEW started in their area. Control group cases remained subject to the old AFDC policies and received employment-related services only under the old Job Opportunities and Basic Skills (JOBS) program. The purpose of random assignment was to create a situation in which people in the experimental group were similar to those in the control group, on average, in all respects except for whether they were covered by VIP policies. The random-assignment evaluation design allows us to attribute differences in outcomes between the experimental and control groups to the VIP reforms with the maximum level of statistical confidence. This report focuses on the impacts of VIP from July 1995 to December During the period from July 1995 to September 1997 (referred to throughout this report as the experimental period), the VIEW component of VIP was still being phased in. At the end of this period, two of the five research sites-- Portsmouth and Wise--had not implemented VIEW. Experimental cases in those two sites were not subject to the reforms likely to have the largest effects on outcomes such as employment and TANF receipt. In fact, a previous report found no impacts in these sites during the experimental period (Gordon and Agodini 1999). Thus, this report covers only the three sites that implemented VIEW during the 3

28 experimental period--lynchburg, Prince William, and Petersburg. These three localities will be referred to as the VIEW sites. Even in the sites that had implemented VIEW, the program was not in effect for the full experimental period. Thus the periods available for full evaluation of the impacts of VIP (with VIEW included) are relatively brief. For the period starting in October 1997, the control group becomes contaminated (see Figure I.1, right side). VDSS decided to start bringing the control group under VIP/VIEW policies in October 1997, two years and three months after the experiment began, as the state no longer needed to operate VIP under a federal waiver. This transition took about six months. Comparisons of the experimental and control groups for the period starting in October 1997 no longer indicate the full effects of VIP. Instead, they reveal the effects of longer exposure to VIP and VIEW versus recent enrollment. Almost certainly, such estimates understate the true impacts of VIP and VIEW relative to earlier policies. Nonetheless, the results from these quarters provide useful information. In particular, it is only during this period that we can examine the effects of the 24-month time limit on TANF benefits for VIEW enrollees. The rest of this chapter describes VIP and the evaluation in more detail. A. OVERVIEW OF VIP VIP reforms fall into two distinct groups: (1) eligibility reforms, and (2) VIEW work requirements and incentives. This section describes each in turn. It also describes how Virginia s transition to TANF affected VIP. 4

29 1. Eligibility Reforms The VIP eligibility provisions primarily encourage adults to be responsible parents. The VIP provisions that were implemented statewide in July 1995 include: C Diversionary Assistance. Working families who face a temporary loss of income are able to receive a one-time cash payment as long as they agree to forgo welfare benefits for up to 160 days. C Paternity Establishment. Mothers are required to name the fathers of their children and provide three additional pieces of information to help locate the noncustodial parent. Mothers who do not comply face an immediate sanction of their portion of the TANF grant and have their TANF case closed if they have not complied within six months. C Family Cap. A child born more than 10 months after a family begins to receive assistance is not eligible to have his or her needs included in the TANF grant (that is, benefits do not increase). All child support received for a child subject to the family cap is passed through directly to the family and is not considered income when determining a family s eligibility for TANF. C Compulsory School Attendance. Compulsory school attendance laws apply to all schoolage children receiving assistance. Children who do not comply with the laws are removed from the TANF grant. C Immunization. TANF applicants must provide verification that all children are immunized or face a monthly reduction of TANF benefits ($50 for the first child who is not immunized and $25 for each additional child who is not immunized). C Minor Parent Restrictions. Minor parents are required to live with a parent or other adult in loco parentis to be eligible for assistance. C Two-Parent Expansions. Eligibility for benefits for two-parent families is determined using the same eligibility criteria as for one-parent families. C Savings Incentive. Families are allowed to accumulate $5,000 in savings, provided that the money will be used for education, home ownership, or starting a business. 2. VIEW VIEW encourages the moving of clients quickly into the workforce with appropriate supports, and emphasizes the temporary nature of cash assistance. Unlike the VIP eligibility provisions, which apply to 5

30 all cases, VIEW applies only to able-bodied parents with no child less than 18 months old. VIEW includes: C Agreement of Personal Responsibility. VIEW participants must sign an Agreement of Personal Responsibility (APR) to receive benefits. Families who fail to sign the APR have their TANF cases closed. C Work Requirement. VIEW participants must begin looking for unsubsidized employment immediately. They are required to work in a subsidized or unsubsidized job or participate in a community work experience program (CWEP) within 90 days of signing the APR. Families who do not comply with the work requirement face the loss of all TANF cash benefits. They remain subject to the time limit as long as their TANF case remains open. C Two-Year Time Limit. Receipt of cash assistance is limited to 24 months followed by a one-year period of eligibility for transitional benefits only (child care, transportation, and medical assistance). A two-year period of ineligibility for all TANF benefits (including transitional benefits) starts when transmittal benefits end. C Generous Earned Income Disregard. VIEW families are eligible to receive their full grant as long as the total of their TANF benefit plus net earned income does not exceed the federal poverty level (and they have received assistance for 24 months or less). C Enhanced Support Services. VIEW extends eligibility for transitional Medicaid and child care benefits (previously available to individuals who lost eligibility as a result of increased earnings) to practically anyone who closes his or her TANF case. VIEW also provides enhanced funding for child care, transportation, and work-related expenses when clients work but continue to receive TANF. C Fewer Exemptions. Under JOBS, the employment program that VIEW replaced, parents with a child under three years old were exempt. Under VIEW, this age cutoff was lowered to 18 months. Requirements for temporary medical exemptions also became stricter. Implementing VIEW required major changes in the way that local welfare agencies delivered employment-related services. These changes included a shift from an education and training focus to a work first focus. Workers now must emphasize helping clients find jobs rapidly. They must also communicate to clients the importance of personal responsibility and the consequences of noncompliance. 6

31 3. Transition to TANF On December 6, 1996, Virginia submitted its state plan for the TANF program. According to this plan, Virginia will continue to operate the VIP program as approved under the 1995 federal waivers. The only program changes were the addition of the mandatory TANF provisions, which were implemented on February 1, These provisions include a five-year lifetime limit on federally funded cash benefits. However, because a VIEW case is limited to 24 months of cash benefits followed by 24 to 36 months of ineligibility, no VIEW cases will reach the federal limit until at least nine years have passed. B. THE VIP IMPACT EVALUATION In 1995, VDSS planned a comprehensive evaluation of VIP, including an experimental impact study. The passage of federal welfare reform led to delays in the evaluation and to some changes in its design. In late 1997, VDSS contracted with the Virginia Polytechnic Institute and State University (Virginia Tech) and its subcontractor, Mathematica Policy Research, Inc. (MPR), to conduct a series of studies to examine the implementation, outcomes, and impacts of welfare reform in Virginia. Funding was provided by the U.S. Department of Health and Human Services, Administration for Children and Families, and by VDSS. The evaluation consists of five studies: (1) an early impact and outcomes analysis, based on the original experimental design; (2) an implementation study; (3) a longitudinal study of what happens to cases after they reach the two-year time limit; (4) an impact and implementation study of VIEW-PLUS, a new job retention program; and (5) a study of VIEW-exempt cases, focusing on child-only cases. This is the second of two reports that MPR is preparing on the early impact and outcomes analysis. The first report (Gordon and Agodini 1999) examined the impacts of VIP during the experimental period for the full sample in the five research sites. A key finding is that only the implementation of VIEW had significant impacts on employment. This report focuses more narrowly on cases likely to be mandatory for 7

32 VIEW in the three sites that implemented VIEW during the experimental period. It also extends the analysis for five additional quarters, in order to provide information on the effects of the 24-month time limit. 1. Why Do an Experimental Impact Evaluation? The outcomes of a program such as VIP are both programmatic and economic. Programmatic outcomes include how many clients are enrolled in VIEW, how many are sanctioned, and how many receive child care subsidies. Economic outcomes include employment rates, average earnings, and TANF participation rates. Describing these outcomes for those in VIP provides useful information, such as how often cases are closed because clients failed to sign the APR and what percentage of VIEW cases close within a year of entering the program. However, some level of sanctioning would occur under most welfare programs, and cases close for a variety of reasons, regardless of the specific policies applied. Thus, information on outcomes must be compared to some benchmark or counterfactual to tell policymakers whether the program is effective. This impact analysis was designed to examine whether VIP/VIEW policies lead to outcomes different from those of the old AFDC/JOBS policies. The impacts or net effects of the program are measured as the differences in outcomes for cases under VIP/VIEW from what they would have been under AFDC/JOBS. An experimental evaluation design was used because it provides the most rigorous foundation for an evaluation of program impacts. This design involved random assignment of cases to an experimental group subject to VIP/VIEW policies or to a control group that remained covered by AFDC/JOBS policies. Random assignment ensured that the experimental and control groups were alike, on average, in all respects 8

33 except for the welfare programs and policies available to them. Thus, differences in average outcomes for the two groups represent the impacts of VIP/VIEW, rather than the effects of other factors. 3 The implementation study report (Pavetti et al. 1999) examined the implementation of VIP/VIEW in the five research sites. That report focused on the operational challenges of making the new policies work in the varied settings of the five sites and on the lessons learned that may apply to implementing similar policies elsewhere. Information from the implementation study is also used in this report to interpret differences in impacts across the research sites. 2. Research Questions This report focuses on four research questions: 1. How does the VIEW component of VIP affect program activities and services used? How do client experiences in VIEW differ from client experiences in JOBS, in terms of participation overall and in terms of types of activities engaged in, frequency of sanctioning, and use of child care assistance? 2. What economic outcomes are observed for VIEW cases? What are the trends over time in key outcomes--employment, earnings, TANF participation, TANF benefits, food stamp participation, food stamp benefits, and total income--for cases subject to VIEW? 3. What are the impacts of VIEW on economic outcomes? How do outcomes for those in VIEW differ from what would have been observed under the old AFDC/JOBS program? 4. How do the impacts of VIEW vary among the demonstration sites? Among the VIEW sites, how and why do impacts vary? 3 Because results are based on a sample of the entire welfare caseload, some differences may reflect sampling variability rather than true program impacts. Thus, we focus on differences that are statistically significant, which implies that they are very unlikely to be due to chance. 9

34 These questions focus on the impacts of VIEW, because Gordon and Agodini (1999) showed that the VIP eligibility reforms by themselves had no impacts on these outcomes. Thus, in the rest of the report, we refer to the impacts of VIEW rather than the impacts of VIP as a whole. The VIEW initiatives provide clients with strong incentives to find work and stay employed, and with important supports to help them achieve this goal. Thus, VIEW was expected to increase employment. However, the implications of VIEW policies for TANF participation during the first two years in VIEW were not certain. The time limit and expanded transitional benefits create incentives for clients to close their cases and bank their months of eligibility. In other words, some people may elect to close their TANF cases even if they are eligible to keep them open, because they want to ensure that they will be eligible to receive TANF again in the future, if they need it. The work requirement could also have a smoke-out effect, if it leads some who were working and not reporting their jobs to close their cases rather than comply. On the other hand, the expanded earnings disregard creates an incentive to combine TANF participation and work, as it is possible to remain eligible for TANF benefits (for 24 months) at much higher levels of earnings than in the past. 4 4 Each of these policies, however, creates these incentives only if clients understand the policy and its implications. 10

35 3. Evaluation Design The specifics of the evaluation design for the impact study are as follows: C Research Sites. Random assignment occurred in five localities selected to be approximately representative of the state geographically and demographically. VIP was implemented for all cases in the rest of the state (with VIEW phased in by EDD, as discussed). Only the three sites that implemented VIEW during the experimental period--lynchburg, Prince William, and Petersburg--are included in the analysis in this report. See Gordon and Agodini (1999) and Pavetti et al. (1999) for detailed descriptions of the research sites. C Recipient and Applicant Samples. Recipients are cases that were active for AFDC in early July 1995 and were randomly assigned at that time. This report focuses on the recipient cases. Applicants are cases that were randomly assigned as they were approved for AFDC/TANF from July 1995 to August Results for these cases are omitted from the main body of this report, because few of these cases were exposed to VIEW for very long during the experimental period, and because the previous analysis found no impacts on applicants. Results for applicant cases are presented in Appendix G. 5 C Mandatory Subsample. Because the previous report showed VIP eligibility policies without VIEW had no impacts on the outcomes of interest, the sample for this report includes only cases in the VIEW sites likely to be mandatory VIEW participants, based on their characteristics (such as the age of the youngest child) at the time of random assignment. The expectation was that VIEW impacts were likely to be concentrated in this group. This imputed VIEW-mandatory status was computed the same way for both experimental and control cases. 6 C Random-Assignment Methodology. Cases were assigned based on the value of a digit in their TANF case number. VDSS found this digit to be approximately uniformly and randomly distributed. For recipients, the assignments were made by VDSS centrally. For applicants, cases were assigned their previous case number if they had received benefits before. 5 Appendix G presents impacts for the 966 mandatory applicant cases who were randomly assigned between July 1995 and September The follow-up period for applicants is the nine calendar quarters after the case entered the sample. This period ranges from October 1995 to December 1997 for the earliest applicants to October 1996 to December 1998 for the latest group included. The quarter of random assignment is omitted, because many applicant cases did not receive TANF until late in the quarter. 6 Some of the non-mandatory experimental cases would have become VIEW-mandatory later in the experimental period, for example, by having their youngest child reach 18 months of age. These cases are omitted from the sample for this report because it is not always possible to identify them based on prerandom-assignment characteristics. 11