Our findings differ from those of Chandoevwit and Ashakul (2008), who apply propensity score matching on a region-by-region basis to the households in the (rural) panel component of the 2004 Socioeconomic Survey, and also apply double differences to this group. They “find that the VF program does not have a positive impact on alleviating the country's poverty. The lack of such an effect on poverty is the product of its insignificant impacts on income and expenditure” (pp. 14–15). Our propensity score equation includes more variables than does the Chandoevwit and Ashakul model, and uses all the 2004 data, not just the (rural) panel component. And our fixed-effects panel estimators are more complete than the double differences that they use, in that we control for a wide variety of household variables. It does appear, however, that the results of the impact evaluations of the VF are sensitive to the choice of model.