I have used an extended version of Mincer's original model to estimate the returns to schooling in rural Ethiopia. In a first step a multinomial log it model is applied to distinguish between for groups of people,(1) full-time farmers, (2) part-time farmers and part-time wage workers,(3) part-time farmers and part-time traders and (4) full-time non-farmers. In a second step, a correction for sample selectivity is made using the Lee-Heckman method and the returns are estimated. The results show that returns on schooling are high in group (4)and lower in groups(2) and (3). Entry in well-paid jobs is constrained for non-educated people. Women are particularly well represented in the third group but strongly underrepresented in the fourth group. The estimation shows overall that education is a worthwhile investment in rural Ethiopia and the fact that households under invest in education can be attributed to the lack of resources at the household level.