This paper studies the longevity of socio-economic inequality across generations. Descriptively, we provide first evidence on the long-run persistence of occupational and educational attainment in Germany, using well-suited data that track family links across up to four generations. We then use this data to test two recent theories of multigenerational transmission. First, we address Gregory Clark’s hypotheses that (i) the true rate of social mobility is far lower than standard parent-child estimates suggest and (ii) constant across countries and time, unaffected by the environment or policy. We argue that multigenerational data offer a direct and appealing path for identification of the latent factor model that is underlying these arguments. We do find that social mobility in Germany is substantially lower than standard estimates imply, but its rate varies over cohorts and is not as low as Clark suggests. Second, we test if an independent causal effect of grandparents may contribute to the observed persistence of inequality across generations. We show that the coefficient on grandparent status will be positive in a wide class of Markovian models and formalize a test procedure to distinguish the latent factor from the grandparental effects model. We find evidence against the latter, but also that grandparents play a larger role in those cohorts that were subject to large societal transformations. Joint with Sebastian Braun.