Inference on inequality measures: A Monte Carlo experiment

Two broad types of method tend to be used to estimate the sampling distribution of inequality measure estimators: analytical asymptotic approximations and resampling-based procedures (e.g. the bootstrap). The present paper attempts to check the coverage performance, in large samples, of a series of standard estimators of both types so as to provide a yardstick to choose among competing alternatives. Two sampling schemes are considered: simple random sampling and clustered sampling. The comparison is made using a Monte Carlo experiment and an application to Belgian data. It turns out that neither basic bootstrap procedures nor asymptotic approximations significantly outperform its competitors. Both yield acceptable estimates (especially in random samples) provided that sampling design is taken into account.