Regression results of impacts of treatment and control variables on household income levels (log of cash and in-kind income).

aVariables are defined in Table 1. bthe number of observations for the quantile model is smaller than for the combined model because the quantile model includes the series of %change variables. We lose 134 observations due to calculating changes (as opposed to values associated with each data collection visit. *significant at 10%; **significant at 5%; ***significant at 1%.