Mardi | 2014-11-18
Sully 5, 16h00-17h20

Abdoul Aziz NDOYE

This study provides a Bayesian estimation method for unconditional quantile regression based on the recentered inuence function (RIF). The method consists of estimating a non linear RIF-regression model using a Gibbs-within-Metropolis Hastings sampler to perform better in the presence of heavytailed distributions. These techniques are used to evaluate the impact of changes in the distribution of covariates on the conditional quantiles of the marginal distribution of the dependent variable.Applied to a nationally representative household survey, the Senegal Poverty Monitoring Report (2005), the results show that the change in the rate of returns to education across quantiles is substantially lower at the primary level of education compared with secondary, and tertiary levels of education. The results also demonstrate that the high rural-urban inequality in Senegal is attributed in particular to the di erence in returns to various covariates, even for lower quantiles.