A central issue in the debate over the inequality of wage outcomes for young workers has been the source of the increasing cross-sectional polarization over time. At one extreme, wages have simply become more volatile over time and, at the other, individual wage trajectories have diverged. In the latter case, inequality persists over the life course. We address this question by developing a class of latent curve models for longitudinal data that capture the underlying stochastic variation in a substantively interpretable manner. The approach decomposes the variation into a small number of data-adaptive latent curves, which we call proto-splines. These curves represent the population-average features of the responses and allow individual-specific differences to be expressed via random coefficients of the population-average features. By incorporating both types of features in a single modeling framework, we can make more definitive statements about the components of wage inequality than can be obtained using traditional models. Using these new models, we evaluate wage inequality in the context of the National Longitudinal Survey of Young Men from 1979-94.
Relevant pre-prints can be found at http://homepages.nyu.edu/~ms184.
