Is One Simple Poverty Scorecard Enough for India?

This paper explores whether one simple poverty scorecard is enough for India. It uses out-of-sample bootstrap tests to compares accuracy of an all-India scorecard with urban- and rural-specific scorecards. The bootstrap tests examines differences between predicted and true poverty likelihoods, differences between estimated and true overall poverty rates and rank-ordering of households by poverty scores. The study finds no evidence that segmentation provides large increases in accuracy results. Findings include:

• All-India scorecard is as accurate for targeting purposes as the rural- and urban-specific scorecards, in general;
• All-India and rural scorecards are in an even competition, while the urban scorecard is slightly better than All-India in terms of the overall poverty rate;
• Differences are smaller than several other potential sources of error;
• There is little evidence of large returns to incurring the cost and complexity of constructing and using segmented scorecards.

The study suggests that simple poverty scorecards offer development programs a quick, easy, inexpensive and accurate way to identify the poor, improve their targeting, track progress out of poverty through time and report to external stakeholders the share of their participants who are poor.