Estimated impacts of PSA on changes in welfare.

a Full sample, with N treated = 50 and N controls = 152.
b Statistical significance evaluated with a two-sided t-test of the difference in means between treated and control sub-samples.
c Calipers restrict matches to units within two standard deviations of each covariate, resulting in matched sample of N treated = 40 treatment and N controls = 40.
d Ordinary least squares regression on change in consumer durables, with all variables used in matching as covariates.
e Ordered logit regression on change in quality of life, with all variables used in matching as covariates.
Robust standard errors in parenthesis.