The estimation of standard measures of intergenerational mobility ideally requires the complete income history for two generations to determine their lifetime incomes. However, empirical applications are typically based on snapshots of income over a limited number of observations in the life cycle. If those snapshots do not mimic lifetime outcomes, the estimates are subject to attenuation and life-cycle bias. The literature has followed two different strategies to address this problem. The first strand models the income processes across subgroups, based on a set of observable characteristics such as education and occupation (Creedy, 1987; Vogel, 2006; Hertz, 2007). The second strand of the literature proposes instead an errors-in-variables model, to capture how the relation between annual and lifetime incomes changes over the life cycle (Haider and Solon, 2006; Lee and Solon, 2009). In this paper, we use uniquely long series of Swedish data to study how well these methods approximate the intergenerational elasticity of income. We show that all methods are biased to some degree. While accounting for important aspects of the income process, they each fail to account for one of its key components – transitory noise, income growth explained by observable characteristics, and unexplained income growth that nevertheless correlates within families. We propose an alternative method that addresses all three components and that remains feasible across a wide range of settings.