Allowing for asymmetry in forecast errors : Results from a Monte-Carlo study

The paper presents an heuristic to estimate the distribution of future values in a forecasting process. The heuristic is based on the split-normal density used as a proxy for the expected distribution of future values. The approach is novel in that it allows for asymmetric non-stationary distributions of future values. Results of Monte-Carlo tests in a context of inventory control indicate that the heuristic may significantly reduce costs.