Overview
Mark W. Watson is a macroeconometrician, Professor of Economics at Princeton University and Research Associate at the NBER. His research spans dynamic factor models, structural VARs, and empirical macroeconomics. His 1981 JASA paper with Robert Engle introduced the scoring algorithm for ML estimation of single-factor state-space models from sectoral wage data. Subsequent work with James Stock on large-panel factor models and diffusion indexes has been foundational in modern macroeconomic forecasting and business cycle measurement.
Key Contributions / Features
- Dynamic factor model for wages (Engle-Watson 1981): Single-factor state-space model extracting an unobserved metropolitan wage component from five sectoral series; scoring algorithm computing the Fisher information matrix via K Kalman filter passes without second derivatives; LM diagnostics for serial correlation under null. Journal of the American Statistical Association 76(376): 774–781.
- Testing for common trends (Stock-Watson 1988): Cointegration recast as common I(1) stochastic trends; GLS-based test for the number of common trends. Journal of the American Statistical Association 83: 1097–1107.
- Stochastic Trends and Economic Fluctuations (King-Plosser-Stock-Watson 1991): Applied balanced-growth cointegration restrictions to identify common stochastic trends in postwar U.S. data; found balanced-growth shock explains less than half of output variability once nominal variables are included. American Economic Review 81(4): 819–840.
- Inference in VARs with unit roots (Sims-Stock-Watson 1990): Conditions for standard asymptotic inference in models with some unit roots. Econometrica 58: 113–44.
- Dynamic OLS (Stock-Watson 1993): The DOLS estimator of cointegrating vectors — augment the levels regression with leads and lags of the regressors' first differences to obtain asymptotic efficiency and standard χ2 Wald inference; applied to stable 1900–1989 U.S. M1 money demand. See Stock-Watson (1993). Econometrica 61(4): 783–820.
- Diffusion indexes (Stock-Watson 2002): Principal-components estimation of latent factors from large panels of macroeconomic time series (approximate dynamic factor model); diffusion indexes as forecasting variables via direct h-step projection; outperformed AR/VAR/leading-indicator benchmarks. See Stock-Watson (2002). Journal of Business & Economic Statistics 20(2): 147–162.
- External-instrument identification of dynamic causal effects (Stock-Watson 2018): Sargan Lecture (with Stock) unifying the proxy-SVAR / external-instruments literature; the LP-IV vs. SVAR-IV trade-off (invertibility-free vs. efficient), a Hausman-type test of invertibility, and the "no free lunch" theorem. The Economic Journal 128(610): 917–948.
Related