In many economic models a central variable of interest is lifetime or permanent income which is not observed in survey data sets and typically proxied by annual income information. To assess the quality of such approximations, we use a unique source of lifetime earnings - the German pension system - and focus on two important issues that have been largely ignored in the existing literature. The first is how to deal with zero income observations in the analysis of women. The second is whether these approximations differ between natives and guest workers. For female earners, we find that estimates of the associations between current and lifetime income are highly sensitive to the treatment of zero earnings. The reason turns out to be the highly cyclical nature of the labor supply behavior of mothers. Furthermore, immigrants' income proxies are prone to significantly larger attenuation biases over the entire life-cycle. This result is explained by the larger share of annual income variance attributable to the transitory income component for immigrants. Averaging income over up to 15 years alleviates the attenuation bias as well as the difference in biases between natives and guest workers.
Generalized errors-in-variables model life-cycle bias lifetime income guest workers