Abstract

We address key challenges in estimating multivariate earnings models, including selection into employment and jobs, and the nonlinear nature of outcomes. Leveraging subjective expectations data on counterfactual outcomes, we achieve identification under weak assumptions and employ straightforward estimation methods. Subjective probabilities directly inform the joint distribution of latent earnings and employment/job choices. Our model features a productivity process, a job match process, and employment dynamics involving job-to-job and unemployment-to-employment transitions. First-step fixed-effects regressions estimate risk, persistence, and transition effects, while second-step GMM estimates reveal covariance structures, such as ability, mobility, and job match effects. Using the NY Fed Survey of Consumer Expectations, we disentangle uncertainty from heterogeneity. Key findings include lower persistence of ability than prior studies, greater variability in ability and mobility effects, smaller risk estimates, and low persistence in job quality after job changes. Unobserved heterogeneity explains much observed reduced-form persistence.