Gregory C. Chow

Overview

Gregory C. Chow (born 1929) is a Princeton econometrician best known for the Chow test for structural breaks (1960) and the Chow-Lin temporal disaggregation method (1971). His 1973 paper established the theoretical foundations for Bayesian multiperiod forecasting from autoregressive models, showing that optimal multi-step predictors require integrating over the posterior of the coefficient matrix rather than plugging in a point estimate.

Key Contributions

• Chow test (1960): F-test for parameter stability across two subsamples; one of the most-cited diagnostic tests in econometrics.
• Chow-Lin interpolation (1971): GLS method for temporal disaggregation and distribution of time series using related higher-frequency indicators.
• Bayesian multiperiod prediction (1973): Proved that $E[A^k \mid \text{data}] \neq (E[A \mid \text{data}])^k$ for $k \geq 2$; derived closed-form Bayesian $k$-step predictors under normal-gamma and normal-Wishart priors. See Bayesian Multiperiod Prediction.
• Random and changing coefficient models (1984): Chapter 21 of the Handbook of Econometrics. Derives two equivalent estimation methods (Kalman filter and GLS regression) for the TVP regression $y_t = x_t'\beta_t + \varepsilon_t$, $\beta_t = M\beta_{t-1} + \eta_t$; develops MLE of hyperparameters ($\sigma^2$, $V$, $M$) via the Kalman innovation likelihood; extends to SUR systems and simultaneous equations; surveys classical constancy tests. See Random Coefficient Model.

Related

• Bayesian Multiperiod Prediction
• Vector Autoregression
• Minnesota Prior