Using Propensity Score Matching to Evaluate PSNP Capacity Development Impact

January 15, 2025 | Impact Evaluation | PSNP | R

The Productive Safety Net Programme (PSNP) is Ethiopia's flagship social protection program. Evaluating its capacity development investments requires rigorous quasi-experimental methods when randomization isn't feasible.

Methodology

We employed Propensity Score Matching (PSM) to address selection bias in our evaluation of PSNP IV capacity development initiatives. The approach involved:

• Constructing propensity scores using logistic regression with pre-treatment covariates
• Matching treated and control woredas using nearest neighbor matching
• Assessing balance through standardized mean differences
• Estimating average treatment effects on the treated (ATT)

R Implementation

library(MatchIt)
ps_model <- glm(treatment ~ covariate1 + covariate2, data = df, family = binomial)
matched <- matchit(ps_model, method = "nearest", ratio = 1)
summary(matched)

Key Findings

Government implementation capacity improved by 34% (p < 0.01), beneficiary targeting accuracy improved by 18%, and timely public work delivery delays reduced by 25%.