230 likes | 319 Views
Food Stamp Participation and Employment among Adult-Only Households. David C. Ribar – The George Washington University Marilyn Edelhoch – South Carolina Department of Social Services Qiduan Liu – South Carolina Department of Social Services. Two gaps in the FSP literature.
E N D
Food Stamp Participation and Employment among Adult-Only Households David C. Ribar – The George Washington University Marilyn Edelhoch – South Carolina Department of Social Services Qiduan Liu – South Carolina Department of Social Services
Two gaps in the FSP literature • Relatively little FSP research on adult-only households • Most previous research has either examined • families with children, especially single-parent families, or • the caseload as a whole • Participation behavior among adult-only HHs is different • low take-up among elderly • long spells for those who do participate • Adult-only HHs are also a focus of policy • simplified application procedures for elderly & disabled • new work requirements and time limits for able-bodied adults without dependents (ABAWDs)
Gaps in FSP research (cont.) • Little research on FSP policies and procedures • Benefits follow a national formula, only variation in formula is in Alaska and Hawaii • Administration, however, is left to the states • Policies are difficult to measure and usually only apply to select groups • Hard to examine with most national data sets • Would also seem to be difficult to measure within a single state; however, there are two important aspects of policy in South Carolina that can be readily measured
ABAWD policies • From PRWORA, ABAWDs • had to work or participate in work-related activities • otherwise could only receive benefits for 3 months in any three years • At their discretion, states could exempt ABAWDs living in high unemployment areas • States later given discretion to exempt up to 15 percent of their ABAWD caseload • South Carolina used its discretion to exempt entire counties (mostly) • mix of counties changed somewhat over time • all counties exempt after Oct. 2002
South Carolina Counties Exempt from ABAWD Work Requirements 1996 – 2003
FSP recertification • FSP eligibility is based on monthly criteria; however, these are hard to check • Instead, clients need to “recertify” their eligibility either in-person or by mail regularly; states vary in their recertification intervals • South Carolina distinguishes between HHs with fixed and fluctuating incomes • HHs with fixed incomes need to recertify annually • Prior to Oct. 2002, HHs with fluctuating incomes needed to recertify quarterly • Since Oct. 2002, HHs with fluctuating incomes need to recertify semi-annually • Elderly and disabled HHs with fixed incomes have easier recertification requirements than other HHs
Identification of policy effects • Typically, it is difficult to identify policy effects within a single state • For ABAWD policies, • county variation in exemption status • variation in the expected timing of effects (should appear early in a spell) • can examine a psuedo-control group of older adult-only HHs • For FSP recertification, • effects occur relative to the start of a participation spell and thus can be seen in the spell duration pattern • change in policy in Oct. 2002 • also, differences by initial income status
Data • We examine administrative records for FSP spells that began after Oct. 1996; data extend through Dec. 2003 • Construct spells of participation or non-participation • data measured to the day • eliminate short breaks and short spells • drop spells that were on-going as of Oct. 1996 • participation & non-participation spells may be repeated • Covered employment • construct indicator for whether case head earned more than $250 in a quarter • misses some jobs – agriculture, out of state • lack detailed timing information, just use quarterly indicators
Data (continued) • Other personal and family controls include race, age, gender, education, marital status • County measures: unemployment, population density, border county, ABAWD exemption • Only examine cases without dependent children • To reduce sample size, examine 1 out of every 11 cases • Separate analyses conducted for HHs with and without members under age 50 (with and without potential ABAWDs) • 9,264 households with members under age 50 • 4,550 households with no members under age 50
FSP exit hazards for different cohorts & types of households
FSP survival functions for different cohorts & types of households Spells began after June 2002 Spells began before 2000
Multivariate analyses • Jointly estimate multivariate models of • food stamp participation spells • food stamp non-participation spells • employment outcomes • Models control for problems from omitted variables and endogenous explanatory variables • Models estimated separately for HHs with and without members under age 50
Food stamp exits • Hazard model for food stamp exits ln hFS(t) = AFS′TFS(t) + δFSE(t) + BFS′XFS(t) + η (1) • proportional hazard specification • TFS(t) is a vector of duration variables, including spell duration, calendar time and recertification indicators; spell duration controls interacted with ABAWD exemption status • E(t) is an indicator for employment • XFS(t) is a vector of other observed explanatory variables • η is an unobserved, time invariant variable; η ~ N(0, ση2) • AFS, δFS and BFS are coefficients to be estimated • Employment is endogenous; assumed to be correlated with η
Food stamp re-entry • Hazard model for food stamp re-entry ln hNF(t) = ANF′TNF(t) + δNFE(t) + BNF′XNF(t) + μ (2) • proportional hazard specification with E(t) and XNF(t) defined as before • vector of duration variables, TNF(t), only includes controls for spell duration and calendar time • μ is an unobserved, time invariant variable; μ ~ N(0, σμ2); correlated with η (ρημ) • ANF, δNF and BNF are coefficients to be estimated • Employment is again endogenous • Estimation procedure allows for multiple, alternating spells of food stamp participation and non-participation
Employment • Longitudinal earnings/employment model E*(t) = BE′XE(t) + ν + ε(t), E(t) = 1 if E*(t) > 0 (3) • random effects probit model • ν is an unobserved, time invariant variable; ν ~ N(0, σν2); correlated with η and μ (ρην and ρμν) • Equations (1)-(3) estimated jointly; because of repeated observations, numerous outcomes examined per case • Gaussian quadrature used to evaluate η, μ and ν; 10 evaluations in each dimension (1000 evaluations total)
Estimation results – simulated change in ABAWD exemption status, HHs with members < age 50
Estimation results – simulated change in employment, HHs with members under age 50 Spell begins in January 1997 Spell begins in July 2002
Estimation results – simulated change in emp. & age comp., HHs with no members under age 50 Spell begins in January 1997 Spell begins in July 2002
Conclusions • Recertification is important: more frequent recertification reduces FS participation and shortens spells • ABAWD restrictions • shorten FSP participation spells • reduce FSP re-entry • BUT have only modest effects on employment • Dogs that don’t bark: policy effects appear where they are supposed to but not where they shouldn’t • Employment reduces FS spells