The full consumption insurance hypothesis is tested using a dataset collected in the Kagera region. The full consumption insurance hypothesis is rejected in all cases. However, all coefficients for the change (in natural logarithm) of income are small. In the CARA preference cases the coefficients for change in income were larger when we estimated them in all clusters than when we estimated them only in rural clusters. In the CRRA preference cases the coefficients for the change in natural logarithm in income are smaller when we estimate them in all clusters than when we estimate them only in rural clusters.