Abstract: | It is crucial for social policy in Less Developed Countriesto identify correlates of poverty at the household level. Thishas been done in the literature by estimating household povertyequations typically with Tobit and Probit models. However, whenthe errors in these equations are non-normal and heteroscedastic,which is usually expected, these models deliver biased estimates.Using quarterly data from Rwanda in 1983, we reject the normalityand homoscedasticity assumptions for household chronic and transientlatent poverty equations. We treat this problem by estimatingcensored quantile regressions. Our results of censored quantileregressions and of inconsistent Tobit regressions are substantiallydifferent. However, in the case of chronic poverty the signsof the apparently significant coefficients are generally inagreement, while for seasonal transient poverty different variableshave significant effects for the two estimation methods. Oursecond contribution is to study, for the first time, correlatesof poverty indicators based on quarterly consumptions. Our resultsshow that in Rwanda different correlates are significant forchronic poverty and for transient seasonal poverty. The effectsof the main inputs (land and labour) are more important forthe chronic component of poverty than for the transient one.Household location and socio-demographic characteristics playimportant roles that are consistent with usual explanationsof poverty in the literature. |