Bayesian quantile regression: An application to the wage distribution in 1990s Britain

This paper illustrates application of Bayesian inference to quantile regression. Bayesian inference regards unknown parameters as random variables, and we describe an MCMC algorithm to estimate the posterior densities of quantile regression parameters. Parameter uncertainty is taken into account without relying on asymptotic approximations. Bayesian inference revealed effective in our application to the wage structure among working males in Britain between 1991 and 2001 using data from the British Household Panel Survey. Looking at different points along the conditional wage distribution uncovered important features of wage returns to education, experience and public sector employment that would be concealed by mean regression.