Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Parameter instability due to potential structural breaks is an important problem affecting out-of-sample forecasting performance of econometric models. This paper uses four types of methods addressing parameter instability, including rolling window, regime switching model, time-varying parameter model, and the time-dependent weighted least squares. The hyperparameters in each method which control the degree of parameter variation are determined via a simple machine learning approach of cross-validation and forecast combination. Our results show significant improvement in predictability of oil prices using these methods accounting for parameter instability except the rolling window method. Forecast combination for models with different hyperparameters produces more robust results than the cross-validation selecting the ex-ante optimal hyperparameters.