Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
The ordinary least squares (OLS) estimator inflates the estimation variance of parameters in the presence of outliers, thus providing poor out-of-sample forecasts. We forecast the real price of crude oil using a robust weighted least squares (RWLS) approach that has the potential to improve forecasting performance by dealing with outliers. This approach down-weights extreme observations using a class of kernel functions. Our results show that the RWLS improves forecasting accuracy relative to the standard OLS approach. Furthermore, the predictability of oil prices revealed by the RWLS is statistically significant for longer horizons. The predictive ability of the RWLS comes from more efficient estimates and the trade-off between forecast bias and variance. Sensitivity analyses indicate that the findings are robust to alternative validation samples, kernel functions, and combination strategies.