This paper aims to answer how best to model education attainment, which is an individual-level variable, in household-level income functions. The accepted practice in the literature is to use the education level of the household head. This paper compares the head-of-household model to three competing models and concludes that the maximum or average level of education in the household is a better explanatory variable of household income. Least absolute deviations (LAD) estimators and censored least absolute deviations (CLAD) estimators are used to predict income. Standard errors, which are robust to violations of homoscedasticity and independence, are generated by a boot-strap method that replicates the two-stage sample design.