#04-16:

Very poor households may be excluded from public programs intended for their benefit for a variety of reasons as lack information, a permanent residence or membershiip in social networks. We are interested in methods of testing for such exclusion based on independently drawn samples of program participants and non-participants. We discuss three alternative nonparametric procedures; sign tests, tests for stochastic dominance and a test for distribution crossing. In the cases where there is a poverty threshold below which program participation is difficult, our simulation results suggests that the last of these test procedures is the most powrful. we apply this test to data from a microfinance program in India and find evidence that the poorest households in the area were largely outside the program.

Download Paper