RESEARCH

ABSTRACT

Idiosyncratic labor incomes are typically modeled either by stochastic processes with heterogeneous income profiles (HIPs) or restricted income profiles (RIPs). The HIP assumes that individual labor income grows deterministically at an unobserved rate and contains a persistent but stationary component, while the RIP assumes that income contains a random walk, a stationary component, and no unobserved deterministic growth-rate component. I show that if idiosyncratic labor income contains a persistent component, a deterministic household-specific trend, and a random-walk component, then all of the components can be identified in small unbalanced panels. Using data on idiosyncratic labor income growth from the Panel Study of Income Dynamics, I find that the estimated variance of deterministic income growth is zero, that is, the HIP model can be rejected. The RIP model with a permanent component cannot be rejected. This result is important for an appropriate choice of modeling the heterogeneity in individual incomes and calibrating/estimating macromodels with incomplete insurance markets and heterogeneous agents.