Proxy means test for targeting welfare benefits in Jamaica

Motivation

The proxy means test (PMT) has become the predominant targeting mechanism for social assistance schemes in many low- and lower-middle income countries, including Jamaica. It has many powerful advocates amid claims that it can accurately and cost-effectively target the poor. However, recently, there have been concerns expressed by beneficiary groups, government stakeholders, and civil society on the selection of households for benefits under Jamaica's Programme of Advancement Through Health and Education (PATH).

Purpose

An evaluation of PATH suggests that it covers only about 71% of households ranked in the poorest quintiles, while 23% of programme beneficiaries are from households considered to be non-poor (ranked in the top two quintiles). This study introduces innovations to the current PMT model with the objective of improving targeting of the poor for social protection benefits.

Approach and methods

We draw on data collected by the Statistical Institute of Jamaica (STATIN) regarding the PATH programme to show that an application of the poverty-weighted least squares regression estimation method improves coverage of households ranked in the poorest quintiles. Poverty-weighted least squares places higher weights on the squared errors of poor households, which avoids the tendency of least squares regression to increase the predicted consumption of the poor.

Findings

Using data compiled by the STATIN through the annual Jamaica Survey of Living Conditions, we showed that the construction of a national model estimated using poverty-weighted least squares, and adjusting predicted consumption to the lower limit of the 95% confidence interval, improves coverage of the poorest households from 71% to over 85%. Achieving this high rate of coverage among the poor came at a cost, in the form of high coverage among the non-poor and, particularly, an increase in inclusion errors.

Policy implications

We believe these results demonstrate that in certain contexts it is possible to improve upon the Basic PMT model and meet the goals the PMT was intended to achieve, efficiently directing social assistance to the poor, minimizing leakage to the non-poor, and maintaining integrity in the overall social assistance mechanism.