This paper examines the effect of grandparental education on their children's and grandchildren's educative outcomes. The endogeneity of parental schooling is addressed by the use of a two-fold instrumental variable approach. A natural experimental set up from a regional war that occurred in 1926 is exploited to instrument years of schooling of the “grand-parents” generation whereas labour market indicators serve as an instrument for the education of the “parents” generation. Using a nationally representative Mexican survey that gathers retrospective information on the three generations, the paper first shows that accounting for endogeneity unveils less mobility than ignoring it. This allows documenting more persistence of family background in the older pair of parent-child link than in the younger pair in the three generations at hand. Finally, results also suggest that the influence of the grandparents' educative legacy, conditional on parental education, does not seem to reach the grandchildren's generation.