Forecasting at capacity: the bias of unconstrained forecasts in model evaluation

Revenue management practices require accurate forecasts for optimal rate decisions, and therefore researchers and industry are keen on identifying the most accurate methods. This study is first to discuss the challenges of forecasting evaluation when predictions exceed capacity. Specifically, evaluators face two choices that are identified and defined, leading to the studies hypotheses. The empirical investigation confirms the importance of considering how to manage these predictions in the evaluation phase and demonstrates how the choice may sway overall accuracy measures and bias the results of model performance. The findings have important implications for capacity-based forecasting research and revenue management practice, since this previously undiscussed capacity-induced bias may alter the results of forecasting studies in academia, as well as the effectiveness of revenue management practices.