The Incentive Effects of Cash Transfers to the Poor

By Anna Aizer, Shari Eli, and Adriana Lleras-Muney

Source Link

Spending on means‐​tested anti‐​poverty programs in the United States accounted for $688 billion, or 16 percent, of total federal government expenditures in 2012; such spending is projected to increase to $877 billion by 2023. All large redistributive and social insurance programs trade off the potential benefits of transfers with their costs, which include the distortions they generate through eligibility rules. All social insurance programs today, including the Earned Income Tax Credit, Temporary Assistance for Needy Families (TANF), Social Security, and food stamps, have eligibility criteria based on some combination of income, work status, and family size. These criteria create incentives for individuals to reduce their work effort and income, remain single, and increase their fertility, raising the cost of the programs and possibly having other undesirable effects on the well‐​being of the families.

While a great amount of theoretical literature explores the negative incentive effects of means‐​tested anti‐​poverty programs, the evidence is mixed regarding the empirical importance of these incentive effects, especially regarding long‐​term effects. We estimate the short‐ and long‐​run incentive effects of the first welfare program in the United States, known as mothers’ pensions. First implemented in 1911 in Illinois, the program had been enacted in 47 states by 1930. In 1935, it was replaced by the federal Aid to Dependent Children program, the precursor to TANF, today’s welfare program. Like TANF, the objective of mothers’ pensions was to improve the long‐​term outcomes of children growing up in poverty. Before 1910, mothers who could not care for their children were forced to place their children in orphanages or training schools. But in response to evidence that children in orphanages fared poorly, states established mothers’ pensions to provide cash transfers to poor mothers with dependent children to allow them to care for their children at home. Previous research documents that boys of recipient mothers had higher schooling levels, lower rates of malnutrition, higher earnings as adults, and ultimately longer lives.

We assess whether mothers’ lifetime behaviors and outcomes were affected by receipt of the transfer. Mothers’ pensions, like many transfer programs today, had built‐​in incentive effects related to eligibility and benefit level. Mothers who remarried would lose the transfer, encouraging women to remain unmarried. Moreover, the transfer was an increasing function of the number of children, encouraging out‐​of‐​wedlock fertility. There were also residency requirements, which affected mobility: only those who had resided for a certain amount of time in the county would be eligible, and those who left would lose benefits. Finally, most states required women to stay home and care for their children, which discouraged recipient mothers from working.

To answer the question of whether this welfare program generated negative incentive effects, we construct a novel data set of about 16,000 women who applied for the program between 1911 and 1930, and we follow them from the time of application until their death. We match data from the program’s administrative records to family trees from Fam​il​y​Search​.org, federal census records, and vital statistics records. This allows us to observe how the transfers affected marriage‐​market outcomes (remarriage, duration to remarriage, and characteristics of the new husband), fertility (before and after application to the program), labor market outcomes(labor force participation, work, and earnings), and geographic mobility. These outcomes are interesting not only because they affect the cost of the program but also because they can indirectly affect the well‐​being of recipients and their children. We also directly investigate the long‐​term effects of welfare on maternal well‐​being, measured by family income and longevity.

We use a basic search model to make predictions about how the cash transfers affect marriage, fertility, and work decisions of recipients. The model predicts that welfare receipt should increase duration to remarriage and the quality of the new match. It should also increase fertility. Similarly, transfers will increase nonemployment spells but increase the wages of those who are employed. Finally, the model predicts that mobility will decline but that those who move will move to better locations. Similar to the unemployment literature based on search models, our empirical findings reject many of its predictions. We find no differences in the remarriage rates of women who received transfers and those who did not—about 47 percent of them remarried, regardless of welfare receipt. Among those who remarried, those with transfers took an average of a year longer to find a new husband. Despite longer durations and contrary to the predictions of the theory, cash recipients did not marry higher‐​quality partners. Cash recipients were not more likely to have children after the transfer, and they were not less likely to work. They were, however, more likely to remain in the county in which they applied for the transfer, but when they moved, they were not more likely to move to a location with better economic or marriage markets as theory would predict.

Our results suggest that current concerns over behavioral distortions caused by social programs may be overstated. The average cash transfer we study amounted to about 30 percent of family income and was available to recipients so long as they had children under the age of 14 or 16 (depending on the state). By modern standards this program was very generous, yet we find rather modest behavioral responses. While welfare programs and welfare recipients today are very different, our results are in fact in line with research on the incentive effects of contemporary welfare with respect to fertility, marriage, and work. Similar to our findings, research examining contemporary welfare reform in the United States has found negligible effects on fertility. While some research finds that welfare reform in the United States had large effects on marriage, we show that these large effects on marriage are only observed in the short run and for a rather small number of recipients. In the long run, marriage rates are not different between those with and without welfare. Our findings on work disincentives mirror the findings from anti‐​poverty programs in developing countries today, which also find modest effects of cash transfers on labor supply.

These findings raise the question of why basic economic predictions are not borne out in the data. Our results suggest that program incentives were likely small relative to other factors that determine maternal behaviors. First, while transfers were relatively large as compared with modern standards, they were insufficient for women to provide for their families, and so women still had to find alternative or additional sources of income through marriage, work, or family. Second, remarriage rates fall rather dramatically with age in our data, and this cost of delaying marriage likely outweighed the benefits of receiving the transfer for longer. Economic opportunities also fell with age. Third, there were important norms regulating marriage and work behavior as well as potential stigma associated with welfare receipt that might have encouraged women to remarry or remain out of the labor force. Indeed, once stigma is added to the model, the theoretical predictions become ambiguous and can be reconciled with the empirical findings. Many of these factors are still relevant today.

Finally, we find no effects on the long‐​term economic outcomes of affected mothers and positive but small and statistically insignificant effects on the longevity of the affected mothers. Thus, in the long run, maternal outcomes were neither diminished nor improved by the program. However, if there are even modest benefits to the children in terms of longevity or income, the program pays for itself. This suggests that the overall evaluation of the program depends crucially on children’s outcomes and less so on maternal incentive responses.