Theme 5: Seven extremely simple poverty scorecards

How poor are participants in rural finance programs? Using data from national household surveys, this paper presents easy-to-use, objective poverty scorecards for seven countries: Bangladesh, Bolivia, Haiti, India, Mexico, Pakistan, and the Philippines. Each scorecard uses five simple indicators to estimate a person’s poverty likelihood, that is, the probability that income or expenditure is below the national poverty line or below $1/day. Field workers can compute scores by hand. Estimates of individuals’ poverty likelihoods are typically accurate to 5–14 percentage points. Estimates of groups’ overall poverty rates are accurate to 1–2 percentage points with 90-percent confidence. The poverty scorecards can help programs target services, report on poverty rates, and track changes in poverty rates over time. This helps rural finance programs make their management of poverty outreach more explicit, quantitative, and intentional.

The poverty scorecards provide an inexpensive, simple, quick way to measure poverty outreach, without the cost of exhaustive (and essentially infeasible) income/expenditure surveys. The scorecards are built using Logit regression on national survey data, but the regression results are transformed to make the actual scorecard understandable to field workers. Accuracy is measured using out-of-sample bootstrap tests.

The main conclusion is that simple, inexpensive poverty scorecards are remarkably accurate, even though their main design consideration was ensuring organizational buy-in and facilitating routinization. Tests in India and Mexico also show that a single scorecard provides essentially the same performance as do two distinct scorecards, one designed for urban areas and one designed for rural areas.

Related Resources