Overview
Michael Keane is a labor and applied microeconomist known for structural models of individual decision-making and earnings dynamics. He held positions at the University of Minnesota, New York University, University of New South Wales, and the University of Technology Sydney / Arizona State. His research spans earnings mobility, dynamic discrete choice models, health economics, and the methodological foundations of structural vs. experimental identification.
Key Contributions
- Geweke-Keane (1999): Mixture of normals probit model for binary discrete choice; six-block conjugate Gibbs sampler with dual augmentation (latent utilities + component indicators); Gelfand-Dey marginal likelihood on the observed-data parameter space; PSID female LFP application (T=1,555): BF ≈107 against conventional probit; best model is scale mixture of 4 normals; welfare benefits effect weaker and spouse income effect stronger than in standard probit. See Geweke-Keane (1999).
- Geweke-Keane (2000): Dynamic panel earnings model for PSID 1968–1989; 3-component mixture of normals for shocks; Bayesian Gibbs sampler with data augmentation; non-Gaussianity substantially changes lifetime earnings gaps (253Kvs.201K college premium) and poverty persistence. See Geweke-Keane (2000).
- Geweke-Keane (2007): Smoothly Mixing Regressions — full Bayesian conditional distribution modeling via covariate-dependent finite mixture; multinomial probit state probabilities; five-block Gibbs with Metropolis-within-Gibbs; PSID 1993 earnings quantile surface and S&P 500 return distribution applications (SMR beats t-GARCH by 57 log-points out-of-sample). See Geweke-Keane (2007) and Smoothly Mixing Regressions.
- Keane (2010): Methodological defense of structural econometrics against the Angrist-Pischke experimentalist critique (Journal of Economic Perspectives 24(2): 47–58). Central claim: the structural–experimental dichotomy is false — all inference requires unverifiable prior assumptions; structural models make them explicit. Leamer (1983) diagnosis was correct but the remedy is more rigorous structural modeling, not reduced form. Marketing as counter-example: Erdem-Keane (1996) brand choice model and Keane-Moffitt (1998) welfare participation model achieved external validation without holdout experiments. External holdout/backcast is the right criterion for both camps. See Keane (2010) A Structural Perspective on the Experimentalist School and Structural Identification.
Related