Developing and Testing Poverty Assessment Tools: Results From Accuracy Tests in Bangladesh

The purpose of this report is to present the results of the accuracy tests in Bangladesh. The report provides an overview of:

• The design of the field research for the accuracy test;
• The computation of the applicable poverty line;
• The analysis of the research;
• Results on selected poverty indicators from 9 regression models;
• Two other poverty assessment methods namely, loan size and Participatory Wealth Ranking (PWR).

The authors conclude that:

• The nine models show quite satisfactory levels of overall accuracy;
• Accuracies in prediction are systematically higher among the very poor compared to the not very poor;
• All models have especially high inaccuracies of prediction for households that fall just below or above the poverty line;
• In larger geographical areas, PWR achieves a fairly low accuracy rate;
• Loan size is a very practical and a relatively low cost tool for micro finance institutions and compares favourably to PWR.

The authors recommend:

• Introduction of a confidence interval defined by a certain X percent below and above the poverty line.
• Differentiating the subjective scales, on which the practitioners participating rate indicators from 1 to 5, into difficulty of obtaining and of verifying the information.