Search

The effect of welfare programme characteristics on the decision to enter the welfare system is modelled as a binary discrete-choice latent variable model dependent upon an underlying index function. The two chief classes of welfare beneficiaries in the US welfare system are analysed: two-parent families and single female-headed families. The data set is a large, recent, non-experimental multiyear cross-sectional pooled data set derived from the US Census Bureau's Current Population Survey. Behavioural impacts from a range of economic variables are consistent in sign with predictions made by economic theory. Newly estimated comprehensive descriptive statistics of the entire welfare system for both single female household heads and married couples with children are employed to characterize the welfare system. Statistically significant estimates of substantial magnitude are obtained for the impacts on welfare participation for the welfare guarantee level, the effective tax rate on earned income, and the effective tax rate on unearned income. Distinct estimates are obtained for each sample consistent with empirical studies in the labour supply literature.

Welfare grants in the United States vary from year to year and from state to state. Actual grants are very different from stated nominal levels of support for two reasons. First, grants are reduced for income. Second, not all qualified families receive welfare. Some do not apply. Others are discouraged, wrongly denied, or poorly counselled. The actual level of welfare support per child is estimated herein at the state level for single female household heads for each year from 1979 to 1990. A tobit estimation technique is employed to account for the probability of not receiving welfare when otherwise qualified.