Overview
James P. LeSage is an econometrician at the University of Toledo (later Texas State University) known for empirical work on Bayesian VAR forecasting and, more broadly, for contributions to spatial econometrics. His 1990 paper provided early large-scale empirical evidence on ECM versus BVAR forecasting performance across 50 industries.
Key Contributions
- "A Comparison of the Forecasting Ability of ECM and VAR Models" (1990): 50-industry Ohio labor market experiment; ECM dominates at horizons 3–12 for cointegrated series (~20% less error at month 12); BECM hybrid outperforms VAR/BVAR even without confirmed cointegration at long horizons; refuted Engle-Yoo's prediction that BECM would perform poorly.
- "A Spatial Prior for Bayesian Vector Autoregressive Models" (1999, with A. Krivelyova): Extended the Minnesota prior to differentiate cross-variable shrinkage by geographic contiguity; spatial random walk prior (0.5 own + 0.5 neighbors); 20-industry × 8-state Midwest employment forecasting; SVAR outperforms standard BVAR at horizons 4–12.
Related