P.A.V.B. Swamy

Overview

Phoebus A.V.B. Swamy was an econometrician at the Board of Governors of the Federal Reserve System. He is the originator of the random coefficient regression model for panel data — now standard in panel econometrics as the "Swamy estimator" — developed in a series of papers from 1970 onward.

Key Contributions

• Random coefficient regression model (Swamy 1970, Econometrica 38): introduced the i.i.d. random coefficient model $y_t = x_t'\beta_t + \varepsilon_t$ with $\beta_t \sim (\bar\beta, \Delta)$ for time-series data; derived the GLS estimator and its asymptotic properties.
• Panel monograph (Swamy 1971, Statistical Inference in Random Coefficient Regression Models, Springer): extended the model to cross-section–time-series panels, yielding the Swamy panel estimator — feasible GLS with a MINQUE estimate of the cross-unit coefficient dispersion matrix $\Delta$.
• Full dispersion matrix and switching regression critique (Swamy-Mehta 1975, JASA 70): relaxed the diagonal restriction on $\Delta_1$ to a full $K\times K$ covariance matrix; proved that ML fails for Quandt's (1972) switching regression (likelihood is almost always unbounded as $\sigma_i^2 \to 0$); derived minimum average-risk estimator $\bar\beta_1^* = (X'\Sigma_1^{-1}X + \psi_1^{-1})^{-1}(X'\Sigma_1^{-1}y + \psi_1^{-1}r_1)$ and MINQUE for $\Delta_1$; extended to panel decomposition $\beta_{it} = \bar\beta + \alpha_i + \xi_{it}$; applied to U.S. bank deposit demand (49 states + DC, 1949–1965).
• General RC model with concomitants and direct/indirect effects (Swamy-Tavlas 1995, Journal of Economic Surveys 9(2)): with George S. Tavlas, extended the simple RC model to the general form $\boldsymbol{\beta}_t = \boldsymbol{\Pi}\mathbf{z}_t + \mathbf{L}\boldsymbol{\varepsilon}_t$, $\boldsymbol{\varepsilon}_t = \boldsymbol{\Phi}\boldsymbol{\varepsilon}_{t-1} + \mathbf{a}_t$, where concomitants $\mathbf{z}_t$ separate direct causal effects from indirect proxy effects in each coefficient; demonstrated that all 19 conventional fixed-coefficient models are special cases; established GARCH as Case 6 via the Swamy-Tavlas (1994) transformation; documented RC superiority over CFC alternatives by 11–77% RMSE across nine macroeconomic applications.

Related

• Random Coefficient Model
• Swamy-Mehta (1975)
• Swamy-Tavlas (1995)
• George S. Tavlas