Identifying Targeting with Nonparametric Methods: An Application to an Indian Microfinance Program

This paper discusses non-parametric methods and statistical tests that are appropriate to assess poverty targeting in public programs. It states that:

• These methods account for the possibility that the population distribution of participants and non-participants cross;
• Crossing points provide the upper bounds on the income of those who have been excluded from the program.

The paper uses two approaches to examine targeting in public programs:

• Parametric models that estimate the probability that a household participates in the program as a function of income or other characteristics of the targeted group;
• Differences in the share of participants and non-participants in some pre-determined intervals of income distribution.

The paper applies these methods to data from a microfinance program in the state of Jharkhand in India, and analyzes:

• Household survey data to compare the composition of newly formed microcredit groups to other randomly chosen households in the area;
• The economic conditions of the members before they received any benefits from the program.

The paper finds that:

• Members of the microfinance program are predominantly poor;
• There is no evidence of stochastic dominance of the distribution of non-members over that of program members;
• The authors' estimation of the crossing point is statistically significant;
• A smaller proportion of members among the poorest households are excluded from village level activity and from government-sponsored programs aimed at poverty alleviation.