Overview
Tao Zha is an econometrician at the Federal Reserve Bank of Atlanta, known primarily for contributions to Bayesian VAR methodology in collaboration with Christopher Sims and Daniel Waggoner. His work focuses on structural identification, Bayesian estimation, and time-varying parameter models for macroeconomic analysis.
Key Contributions / Features
- Zha (1997/1998) "Dynamic Multivariate Model": policy article presenting the 6-variable monthly BVAR (CPI, commodity prices, fed funds rate, real GDP, M2, unemployment; 1959:1, 13-month lags). Demonstrates OLS overfitting failure, Bayesian remedy, 2/3 error bands, joint error regions, and regime-shift robustness (dropping pre-1983 data worsens forecasting). Blue Chip GDP 1995 forecast far outside model's error band.
- Sims-Zha (1998) "Bayesian Methods for Dynamic Multivariate Models" (IER 39(4): 949–968, with Sims): unified derivation of A+∣A0 posterior; Kronecker symmetry condition H(a0)=B⊗G enabling ~400× speedup for large VARs; three-table dummy-observation taxonomy (equation-specific tightness, sums-of-coefficients μ5, initial-observation μ6); empirical comparison showing A+∣A0 (31 sec/1,000 draws, well-calibrated bands) dominates B∣A0 (2.8 hr mode + 16 min/1,000 draws, unemployment and price posteriors miss data). See Sims-Zha Prior.
- Waggoner-Zha (2003a): Gibbs sampler for identified VARs under linear restrictions, enabling efficient posterior simulation and MDD computation.
- Waggoner-Zha (2003b): likelihood-preserving normalization rule for non-recursive VAR posteriors.
- Sims-Zha (1999): joint error bands for impulse responses via eigendecomposition of posterior covariance.
- Sims-Zha (2005a, 2005b): Markov-switching VARs; empirical finding that the Great Moderation reflects variance changes in private-sector shocks, not policy shifts.
- Leeper-Zha (2003): test of independence of structural shocks; measure of Lucas-critique relevance for regime switches.
Related