Purpose– Most intergenerational mobility studies use data on two generations to estimate the elasticity between son's and father's earnings. The purpose of this paper is to use a data set spanning three generations to estimate additional relationships between a person's earnings and family background yielded by intergenerational mobility models such as Becker-Tomes (1979) model and modified versions of it.Design/methodology/approach– The paper uses data from the 1996 PNAD – a nationally representative household survey in Brazil. The author builds a data set consisting of 5,125 grandfather-father-son triplets by taking advantage of two characteristics of Brazil. First, commonly in Brazil, individuals live with their parents until they marry. Second, individuals tended to quit school and begin working at an early age. As a result, there are many households with adult sons who are not at the very beginning of their working careers. Since the sample is limited to households with adult sons, the author applies Heckman (1979) estimation procedure to address selection bias.Findings– Estimation results contradict some predictions of simple versions of the Becker and Tomes model. The paper proposes a modified version of the Becker and Tomes model that allows for a skipping generation effect, and finds that family background explains 34.9 percent of the variation in earnings among males aged 16-27 in Brazil. If there were no differences in endowments (talent, IQ, health, physical appearance, attitudes toward work, family connections, etc.), the variation in earnings would fall by no less than 26 percent. If it were possible to eliminate differences in investment in human capital, the variation in earnings would fall by at most 21.1 percent.Research limitations/implications– The paper has two main data limitations. First, the 1996 earnings of the fathers and sons are used as proxies for lifetime earnings although the transitory component of one-year earnings may be quite large, particularly at young ages. Second, in spite of the efforts to deal with the sample selection bias, the paper shows that the intergenerational elasticity in earnings for the sons aged 22-27 is about 14.6 percent lower for the subsample of households with adult sons than for the full sample.Practical implications– The paper finds evidence supporting the existence of a direct effect of the grandparents on the grandchildren beyond their influence on the parents, and reinforces consideration of this factor in intergenerational mobility studies.Social implications– The findings in this paper may suggest a room for improvements in economic outcomes of children in less privileged families through investment in formal education as well as policies that considers other aspects of a person's life. For instance, Bolsa Família – a Brazilian government program that provide cash allowances to poor families conditional on children school attendance – may improve the economic outcomes of poor children by enforcing formal education and by lessening the children hardships at home.Originality/value– The paper proposes a modified version of the Becker and Tomes model which allows for a skipping generation effect. Under the assumptions of the modified model and in hand with a three-generations data set from Brazil, the paper estimates a lowerbound for the variation in earnings explained by differences in endowments across families, and an upperbound for the variation in earnings explained by differences in human capital.