Intended and Unintended Effects of the War on Poverty: What Research Tells Us and Implications for Policy. Marianne Bitler, UC Davis & NBER

Transcription

1 Intended and Unintended Effects of the War on Poverty: What Research Tells Us and Implications for Policy Marianne Bitler, UC Davis & NBER

2 Funding acknowledged/disclaimer FUNDING ACKNOWLEDGMENT: Research reported in this talk was supported by the Eunice Kennedy Shriver National Institute Of Child Health & Human Development of the National Institutes of Health under award number P01HD and by the National Science Foundation under award SMA DISCLAIMER: The findings, conclusions, views, and opinions are solely those of the authors and do not necessarily represent the views of the US Department of the Treasury, the National Institutes of Health, the National Science Foundation, or the United States government.

3 War on Poverty In his 1964 State of the Union address, President Johnson declared a War on Poverty. Congress passed the Economic Opportunity act, creating the Office of Economic Opportunity to administer local funds (8/64). Community Action Program (set up Comm. Action Agencies-local non-profits to promote self-sufficiency and run grants, Head Starts, etc. led by community members) Project Head Start (school readiness for low-income kids) Job Corps (training for job readiness) Others (Section 8 housing, Legal Services, Community Health Centers, Family Planning Grants, VISTA, Upward Bound )

4 War on Poverty Food Stamp Act of 1964 (8/64) (allowed counties to join what had been a pilot program ) Elementary and Secondary Education Act (4/65) (Title 1) Social Security Act of 1965 (7/65) (created Medicaid and Medicare) School Breakfast Program

5 1970s Earned Income Tax Credit (1975) (for families, small) Supplemental Security Income (1974) (means tested cash benefits for disabled and aged) WIC program (1974) (food vouchers for low income pregnant women and children under 5)

6 Spending Great Society: Real 2014 $45 Billion 65, real 2014 $140 Billion 1972 (Haveman et al., JPAM) Poverty: %, % Participation in food stamps: ,000 participants, ,000,000 participants

7 Summary of talk This talk is organized around programs aimed at the non-elderly and children. It is also organized around the type of benefits: Cash, in-kind, and investments in human capital. Focus on long run impacts as well. (Emerging economics literature)

8 Types of programs Timing Cash In-Kind Human Capital Original War on Poverty or earlier AFDC/TANF Food Stamps (SNAP) School Meals Housing Assistance Medicaid Head Start Title I Higher Education Student Aid Job Corps Later program SSI EITC WIC CHIP CCDF Early Head Start

9 The social safety net for nonelderly families: Eligibility 1. Cash welfare (AFDC/TANF) [means tested, categorical] 2. Food Stamps [means tested, nearly universal] 3. The EITC [means tested, requires employment, categorical] 4. SSI [means tested, categorical] 5. Medicaid [means tested, categorical until ACA] 6. WIC [means tested, categorical]

10 The safety net for nonelderly families: How big in 2010? 1. In 2010, 35.2 million families with kids, 79 million families 2. Programs (not all targeted just to children 1. Medicaid non disabled children: 27 million children, $82 billion, $2363 per child per year 2. TANF cash: 1.9 million families, $11 billion, $402/month per family on the program 3. Food Stamps: 18.6 million families, $65 billion, $285/month per family on the program 4. The EITC: 26.2 million tax units, $59 billion, $187/month per tax unit 5. SSI (non-aged): 6.7 million persons, $46 billion, $518/month per recipient 6. WIC: 10 million persons, $7 billion, $5 billion food, $41.43/month per recipient 7. Head Start: 904,000 persons, $7 billion, $8000/year per recipient, if annualized, $667 per month

11 Assessment I will review the causal literature about the largest of these programs. I focus on intended effects such as poverty reduction and self-sufficiency. I also consider unintended effects on employment, earnings, income, & poverty. I also look at longer-run effects on family formation, investments in children, etc.

12 Ongoing work I will summarize some ongoing work of mine aimed at looking at long run effects of the Food Stamp program on earnings and the food environment

13 Stated intended effects of programs Cash aid: Provide a floor for consumption Food programs: Avoid hunger, malnutrition Health programs: Improve health Human capital: Increase attainment, raise wages Intergenerational goals

14 Other key features of last several decades in the US Transition from out-of-work assistance to in-work assistance in the US post welfare reforms of the late 1990s Cash welfare entitlement (AFDC) changed to block grant (TANF), time limits, other tightening EITC expanded Other factors that are relevant Business cycle Wage and income inequality Changing social norms

15 Effects of programs: Unintended Standard economic model of a cash transfer program with utility maximizing recipient and means testing Benefit reductions with increased earnings lead to work disincentives, asset tests lead to savings reductions Iron triangle: More work incentives (lower tax rate) means more people on the program unless generosity is reduced Categorical benefits in theory can cause changes in categories (single mother programs lead to more single motherhood) Intergenerational effects Take-up, stigma

16 In-kind programs In-kind means you do not get cash. Economists think cash is more efficient (you pick) Basic models predict biggest effects for those who would spend less on the good than the value of the in-kind benefit. Those who would spend at least that much should only increase use a little. Behavioral science, dynamics predict otherwise. Give in-kind to constrain people to consume the good (food stamps, medical benefits)

17 Why can t we just compare recipients to non-recipients or look at trends? People who use these programs look systematically different than those who don t (education, age, disability, single parenthood) Might worry they are different on other dimensions we can t see If so, the comparison might overstate the effects of the program

18 Why can t we just compare recipients to non-recipients or look at trends? Trends can be driven by other variables

19 Causal estimates Experiments are the gold standard, randomize who gets an offer of a program, ensure their characteristics can t be correlated with the offer Quasi-experimental variation-arbitrary timing or imposition of policies, next best

20 Variation for obtaining causal impacts RCT Rollout Federal Policy St/Local Policy Exog. Formula AFDC/TANF X X X X SSI X X X EITC X X X Food Stamps/SNAP X X X Child Nutrition X X X X WIC X X Medicaid X X X X Housing X X CCDF X Head Start/Early HS X X X Title I Higher Ed Student Aid X X Training X X Other

21 AFDC/TANF AFDC was an entitlement, changed to block grant TANF program in 1996 State variation in benefits/eligibility High tax on earnings No time limit until kids age out Came with Medicaid, Food Stamps, other benefits Concerns about rising welfare caseloads lead to waivers, culminated in PRWORA

22 AFDC vs. TANF Lifetime time limit with TANF More work emphasis More state flexibility to spend $ on non-cash assistance aimed at higher income folks (some states have basically dissolved their AFDC programs), block grant not entitlement More state flexibility with sticks (sanctions) and carrots (reduced tax rate on earnings)

23 TANF No Longer Acts as a Stabilizer

24 TANF Reaches Fewer Poor Families

25 A small share of the TANF block grant is used for cash assistance Only 25% of the TANF block grant is cash assistance Less assistance is going to the poorest families

26 Effects of welfare reform More work, less welfare use in immediate wake of PWRORA (booming economy, EITC expanded), some evidence from experiments states had to do to get waivers Little evidence state choices about earnings reductions affected income Little evidence of any impacts on family structure Children do better if income goes up, not otherwise

27 Long Run Effects of the Food Stamp Program Rollout: Food Stamps, series of papers by Hoynes, Schanzenbach, and Almond Food Stamps began in 1961, pilot Food Stamp act 1964,, was voluntary at the county level Mandatory by 1975 County rollout design, compare changes before and after my county gets Food Stamps, counterfactual is what other counties not getting it experience

28 Rollout is a good quasi-experiment Exogeneity: Is adoption plausibly random? Variation across time and place in where roll out occurred (Maps, time series) Existing papers look at time to adoption as a function of other welfare programs/county characteristics pre adoption, no relationship Event study graphs

29 Figure 2: Food Stamp Program Start Date, By County ( ) Note: Authors tabulations of food stamp administrative data (U.S. Department of Agriculture, various years). The shading corresponds to the county FSP start date, where darker shading indicated later county implementation. 45

30 Share of counties with Food Stamps by year SNAP adoption (from Hoynes and Schanzenbach, 2009): Share of counties adopting by year Counties participating in FSP (weig h ted percent) : Pilot prog rams inititated 1964 FSA: counties can start FSP 1973 Amend: mandatory FSP b y : PSID b eg ins Figure 1. Cumulative Percent of Counties with Food Stamp Program, Source: Authors tabulations of county FSP start dates. Counties are weighted by their 1960 population. 7

31 Effects of Rollout Existing papers show that roll out leads to less out of pocket food spending, more food at home (Hoynes & Schanzenbach, 2009) Some work disincentive effect (similar to cash) (Hoynes and Schanzenbach, 2012) Increases in birthweight, less LBW from natality data (Almond, Hoynes, and Schanzenbach, 2011)

32 Long run effects on adult health This paper uses the PSID (a panel data set). You know where kids were when they were 0 to 5, and can see when they were first exposed to Food Stamps. Hoynes, Schanzenbach, and Almond find large effects on reducing metabolic syndrome and increasing self-sufficiency for women.

33 Table 2 Metabolic Syndrome Index for High Participation Sample Metabolic syndrome (index) Components of metabolic syndrome index Heart disease Heart attack Diabetes High blood pressure Obesity FS share IU *** * (0.107) (0.048) (0.086) (0.086) (0.027) (0.019) Mean of dep var Observations 8,246 8,431 8,430 9,217 8,430 8,432 R-squared Table 3 Additional Health Outcomes for the High Participation Sample Other health outcomes Health behaviors In good health Disabled Height below 5th perc. Ever smoked Drink 3+ day now FS share IU ** (0.074) (0.039) (0.026) (0.131) (0.052) Y-mean Observations 25,738 25,731 9,398 8,430 8,413 R-squared Notes to Tables 2 and 3: Each parameter is from a separate regression of the outcome variable on FSP exposure (share of months between conception and age 5 that FSP is in the county). The sample comes from the PSID and includes heads and wives born between who are between 18 and 53 (or for economic outcomes). The high participation sample includes those born into families where the head had less than a high school education. Estimates are weighted using the PSID weights and clustered on county of birth. The models control for individual demographics, family background, and fixed effects for year of birth, year of interview, county, state specific linear cohort, and 1960 county characters interacted with linear cohort. Standard errors are in parentheses and ***, **, and * indicate that the estimates are significant at the 1%, 5% and 10% levels.

34 Other papers Barr finds exposure to Food Stamps while young reduces the probability of incarceration My own work with Tim Beatty and Cynthia Van der Werf shows that there was an effect of rollout on the food environment. We see an increase in employment in grocery and food stores.

35 Bitler and Figinski (I) Look at long run effects of exposure to Food Stamps while 0-5 Use tax records for a 1 percent sample of Social Security Numbers Linked to place of birth from Social Security Card applications Sample: Born in US 1955 to 1980 and could match to place of birth

36 Bitler and Figinski (II) Outcomes at age 32. Control for county, year of birth, gender, and white/non-white. Also control for average spending in the county of birth through age 18 on spending on public medical care the EITC unemployment insurance payments SSI

37 Bitler and Figinski (III) Results for FS exposure women at age 32, cohorts (1) (2) (3) (4) (5) (6) FS thru (311) (302) (312) (300) (342) (311) Controls for: White Y Y Y Y Y Y YOB FE Y Y Y Y Y Y COB FE Y Y Y Y Y Y SOB trend N Y Y Y N N COB trend N N N N Y Y REIS birth N N Y N Y N REIS to 18 N N N Y N Y 34

38 Bitler and Figinski (IV) No effects for men For women, also find a reduction in ever having had SSI and SSDI

39 EITC The EITC is a refundable tax credit, quite large for families with children, small otherwise. It offers a wage subsidy, then a flat amount, then is taxed away. Take-up is high. Bitler, Hoynes, and Kuka (2016) find that it is not very countercyclical for low-skill/single parents, when times are bad they do not get more EITC. Higher skill/married parents claim more when times are bad.

40 EITC EITC expanded considerably in mid 1990s, with a much bigger increase in benefits for those with 2 or more versus 1 kid Series of papers by Hoynes and Eissa, Simon and Miller, Patel, DD of 1 versus 2 kids post expansion on labor supply, infant health, income; Evans and Garthwaite on mother s health: All find positive effects Dahl and Lochner use own predicted EITC funding to look at effects on kids test scores: More predicted EITC leads to higher text scores

41 Head Start Non-profits get matching grants from the Federal government to provide low income children with pre-school Based on Perry Pre-School experiment Started in 1965 Lots of evidence of short run cognitive gains using family comparisons

42 Head Start Head Start Rollout: poorest 300 counties got aid in applying Ludwig and Miller QJE: Find some long run positive effects on mortality, other outcomes Variation in rollout: Barr and Gibbs find large effects of your mother being exposed to HS

43 Other evidence using the Head Start Impact Study (experiment) Bitler, Hoynes, and Domina: Distributional approach Positive effects at the bottom of the PPVT distribution, and some achievement measures Little difference across groups once condition on where in distribution the control group is Kline and Walters and Page, Feller and Miratrix use other approaches, find evidence the counterfactual may matter Gelber and Isen: Positive effects on parenting

44 Medicaid Rollout Goodman-Bacon looks at rollout, places with larger AFDC use pre-rollout see a bigger increase in coverage. He finds a significant decline in non-white child mortality. Boudreaux et al. find long run positive effects of exposure on health in adulthood.

45 Medicaid Big literature using late 1980s/early 1990s expansions Simulated eligibility: Project eligibility using national data Currie/Gruber/Cutler/Shore-Sheppard/Yelowitz Experiment in Oregon: Baicker, Finkelstein, others More coverage, more health care use, better self-rated health, less depression, little on objective health, less bankruptcy RD of Medicaid expansion Those born on or after September 1983 (expansion happened when they were young kids) Card and Shore-Sheppard coverage went up Wherry, Miller, Kaestner, and Meyer 2015: Less hospitalizations Miller and Meyer: Less mortality for teens East, Miller, Page, and Wherry: Large positive effects of exposure in utero on the next generation s birth weight

46 Broader lessons for policy (I) Individuals respond to incentives Effects on labor supply are somewhat mitigated by a shift to in-work welfare Effects largely on labor supply, incentives for altering family structure and investment not large enough?

47 Broader lessons for policy (II) Some positive long term effects, can think of these programs as investments Tradeoffs for policy makers (Iron Triangle, more generosity vs. better work incentives) Program complexity is challenging, can attenuate program impacts

48 Broader lessons for policy (III) In some settings, need attention to quality as well as quantity Coordination across levels of government important (e.g., Medicaid and ACA) Continue agenda, counterfactual of 1980s to now not the counterfactual about which we are learning long-run effects