Welfare Caseloads (Real Research Not to be Found at NRO Edition)

Rich Lowry wants us to believe that the 1996 welfare reforms really did cause a miracle:

Welfare caseloads have dropped 60 percent since the passage of welfare reform. Was that just the result of a strong economy? No. Caseloads didn’t decline significantly in any of the eight periods of economic expansion from the 1950s to the mid-1990s. From 1953 to 1994, the number of families on welfare dropped in only five of those years, and dropped two years in a row only once. By 2005, welfare caseloads had been declining for a stunning 11 straight years.

As I read this, I had to think – aren’t caseloads a poor measure of poverty as they also reflect the proportion of poor people in these programs as opposed to be in dire poverty without government aid? The Center on Budget and Policy Priorities noted a couple of years ago:

The U.S. Department of Health and Human Services announced today that caseloads in the Temporary Assistance for Needy Families (TANF) program fell below 2 million during the first quarter of 2004. Echoing similar past press releases, Assistant Secretary Wade Horn claimed that “more Americans are leaving welfare, entering the workforce and becoming part of the economic mainstream.” “Contrary to HHS’s assertions, falling TANF caseloads in recent years have not meant that more families are working or out of poverty,” noted Sharon Parrott, the Center’s director of welfare reform policy. “Earlier this year, shortly before the Census Bureau released data showing a marked rise in child poverty in the United States in 2003, HHS issued a similar press release that trumpeted TANF caseload declines in 2003.” (Data on poverty are only available through 2003.) “While poverty did fall in the 1990s – as a result of a growing economy, increased supports for working poor families, and welfare reform efforts – poverty and joblessness among single mothers increased significantly between 2000 and 2003, data noticeably absent from the HHS press release,” said Parrott. “Contrary to Horn’s claim, rising poverty rates in recent years show that more families have left the economic mainstream rather than joined it.” “HHS is once again touting declining caseloads absent any research or data showing that this decline is the result of improving economic prospects for poor families,” said Parrott. “Effective welfare reform should help families join the economic mainstream by providing temporary income support, help finding and keeping jobs, and child care. But to accomplish these goals, welfare programs have to respond to increases in the number of families who need help,” said Parrott. “The decline in the number of families being helped by TANF, despite the increase in the number of families living in poverty in recent years, suggests that steps need to be taken to reduce barriers that keep needy families from benefiting from the benefits and services the TANF program can provide.”

Figure 1 from the Welfare Caseload Watch documents there was a decline in caseloads during the Clinton years that predated the passage of the 1996 act. Caseloads have not materially changed since Bush took office:

For most states, welfare caseloads have remained fairly stable, despite concerns over caseload increases during the recent economic downturn. National caseloads have leveled off even in the midst of relatively high unemployment rates. Some states have seen increases and others decreases, but overall, caseload numbers have not responded strongly to the weak economy. NCSL’s Welfare Project is actively collecting welfare caseload information from state agencies to provide a national picture and analyze state trends.

Jeffrey Grogger has authored a couple of papers of interest. In Markov Forecasting Methods for Welfare Caseloads, he notes:

The Personal Responsibility and Work Opportunity Restoration Act (PRWORA) replaced AFDC with Temporary Assistance for Needy Families (TANF) and changed the funding formula used to finance welfare expenditures. Whereas AFDC expenditures were shared between federal and state governments, TANF is funded with a block grant. This means that sufficient growth in TANF cases could force states to curtail eligibility, reduce benefits, reduce spending on other state programs, or raise taxes to support higher spending. Although the threat of such welfare-driven fiscal dislocations is currently low due to the high funding level of the TANF block grant, recent proposals to reduce the federal block grant could present states with difficult choices if caseloads were to rise substantially … In the simplest terms, today’s caseload depends on yesterday’s caseload plus entries and exits. Because today’s caseload depends in part on yesterday’s caseload, the caseload exhibits inertia. Markov forecasting exploits that inertia to base forecasts of the future caseload on current entries and exits. Put differently, current entries and exits affect current caseloads, but since today’s caseload affects future caseloads, current entries and exits affect future caseloads as well. If turnover is not too great, then there is a lag between changes in entries and exits and the time that those changes are fully reflected in the caseload. Thus current trends in entries and exits help predict future caseloads. Markov theory motivates a leading indicator that is based on entries and exits … The key trends in the entry and exit series discussed above are quite prominent. The smoothed ISS seems to fit quite well, even though it was computed indirectly from the smoothed entry and exit rate series rather than estimated It achieves its maximum value in January 1994, 14 months before the caseload peak in March 1995. It achieves its minimum value in November 2002, 12 months before the recent caseload minimum in November 2003 … The forecasts achieve their minimum value in January 2004, just two months after the caseload achieved its minimum value. furthermore, the forecasts strongly suggest that the November 2003 low point represents a true turning point, rather than just a relatively minor fluctuation around a stable level. After the beginning of 2004, the forecasted caseload rises sharply.

Grogger’s data and Markov model shows that caseloads rose from 1989 to 1995 before they began their decline. We know that poverty was already high during the 1980’s with the 1990 recession increasing poverty. Groggery’s model suggests that changes in caseloads will lag changes in poverty and it suggests that caseloads will begin to rise given the recent increase in poverty. Also, note that caseloads may decline because government assistance to the poor becomes more stingy.

None of what Lowry writes has anything to do with the impact of the 1996 act on economic behavior. Grogger’s The Effects of Time Limits, the EITC, and Other Policy Changes on Welfare Use, Work, and Income Among Female-headed Families is more promising:

Of all of the welfare reforms that were implemented during the 1990s, time limits may represent the single greatest break from past policy. This paper expands on what is known about this important welfare reform measure by exploiting the predictions from Grogger and Michalopoulos (2003) to estimate the effects of time limits on welfare use, employment, labor supply, earnings, and income among female-headed families. Results based on data from the March Current Population Survey suggest that time limits have had important effects on welfare use and work, accounting for about one-eighth of the decline in welfare use and about 7% of the rise in employment since 1993. They have had no significant effect on earnings or income, however. The analysis also shows that the collective effects of other reforms have had important impacts on employment and labor supply. Furthermore, it identifies the Earned Income Tax Credit (EITC) as a particularly important contributor to both the recent decrease in welfare use and the recent increase in employment, labor supply, and earnings.