Bootstrapping in Applied Welfare Analysis

Bootstrapping procedures are used to estimate the statistical properties of common empirical welfare measures. Results from a Monte Carlo experiment indicate that welfare estimates such as Marshallian consumer surplus often exhibit significant bias. Standard errors of welfare estimates are found to often exceed the magnitude of the point estimate for typical cross‐section data sets and are generally larger than the difference between comparable Hicksian and Marshallian measures. Precision of welfare estimates can be markedly enhanced through generating larger data sets, obtaining better model fits, and through imposition of innocuous inequality restrictions on the demand function parameters.