Welfare recipient work choice and in-kind benefits in Washington state

Abstract

We analyze the work choice of welfare recipients. Potential welfare recipients compare their on and off welfare utility from after-tax income and in-kind benefits via employment or welfare, and choose whether to work. Our null hypothesis, which we reject, is that benefits affect only the decision to work or not, not the hours worked, which will depend on wages. Using Temporary Assistance for Needy Families (TANF) administrative data from Washington state, we find that employer provided health insurance and child care subsidies significantly raise exit rates of TANF recipients and induce greater work effort. Other work inducing factors include wages and the Earned Income Tax Credit, while increased levels of Medicaid, Food Stamps and the income guarantee increase welfare dependency.

Notes

¹ A related tax credit, the Earned Income Tax Credit (EITC), also affects the labour force behaviour of low-income individuals.

² These exact policies can differ between states. According to DHHS (2000), by August 2000, 38 states imposed 60-month lifetime time limit while seven other states established even stricter time limits that range between 21 and 48 months of lifetime assistance. In addition to the lifetime limit, 13 states have intermittent limits that either restrict the number of months a household can accumulate during a given period of time or make it ineligible for a consecutive set of months, once it has accrued a fixed number of months on welfare.

³ The term government refers to the state government only. This is consistent throughout the manuscript. When we refer to the Federal government, we will explicitly say ‘Federal government’.

⁴ For convenience and to be consistent with our data, we usually refer to TANF rather than more generically ‘welfare’, but use the terms interchangeably in the article.

⁵Other separable functional forms should give qualitatively similar results.

⁶ By default, being on welfare means income is below EITC income eligibility. So all individuals on welfare are in fact eligible for the EITC. We assume, for simplification, that an individual receives the EITC while on welfare, paying no taxes. When an individual exits the welfare system she can still qualify for the EITC depending on income, however, not through the welfare recipient budget constraint. The tax reduction at the higher income level can be seen through a lower income tax rate modelled in the budget constraint.

⁷Second order conditions can be found in the article's first appendix.

⁸ This is in fact a testable conjecture which is supported in our empirical analysis.

⁹ We provide the results from a representative nonseparable utility function in the second appendix.

¹⁰ An example of a ‘trivial solution’ in a nonseparable utility function that would not depend on in-kind benefits would be if all of the individual's utility weights (the ψs) were restricted such that they take the same value. But in doing so one would restrict the model to take on a form that was separable in its arguments.

¹¹ In this case, we believe it is a stronger statistical statement to formulate our null hypothesis so that rejecting it gives a clear result. Stating it so not rejecting the null hypothesis as a test of the conjecture of seperability gives a more ambiguous result.

¹² Washington State called their TANF design Work First.

¹³ The Automated Client Eligibility System (ACES) is maintained by the Washington State Department of Social and Health Services (DSHS). Dollar values were converted to real values, with a base year of 2000 using the consumer price index.

¹⁴ JOBS Automated System (JAS), Washington State wage records, and ES202 Files are maintained by the Washington State Employment Security Department (ESD). Dollar values were converted to real values, with a base year of 2000 using the consumer price index.

¹⁵ In the 2002 fiscal year, 88% of all clients on TANF in Washington State eligible to file an income tax return did so. The IRS provided this percentage.

¹⁶ We refer to per capita UI claims and per capita employment as UI claims and Employment from here forward.

¹⁷ Average monthly earnings are $442.52 over the five-year panel. This average includes observations corresponding to the decision not to work. When individuals that chose to work zero hours during the month are removed, average labour market earnings increases to $1034.31 per month.

¹⁸ TANF participation is empirically determined by whether an individual received her guarantee in a given month.

¹⁹ When conducting the empricial analysis, we also ran several different model specifications to check the appropriateness of our random effects specification. First, we estimated Equation 17 via OLS. However, White's test (1980) indicated that the results were heteroskedastic. We also ran several different specifications (based on how the group effects were defined) of the fixed effect model and found only minor differences in the signs and significance of the coefficient estimates between these fixed effect models and our random effect model. Further information on this issue is available from the authors upon request.

²⁰ Not one individual in the panel has an individual insurance policy. Individual policies are very expensive, and this suggests that current as well as potential welfare recipients only subscribe to health insurance directly provided to them. Additionally, we also considered adding a dichotomous variable indicating participation in Medicaid. However, this variable was dropped because it was nearly perfectly multicollinear with the employer-sponsored health, for reasons discussed in the first part of this endnote.

²¹ An individual could have no insurance, thus having zeros for both these variables.

²² We are, of course, moving well beyond a marginal change, so this gross change is offered only as an example.