LAUSR

ABSTRACT In this paper, we propose two types of robust estimates for the multiplicative error model, M-estimates and BM-estimates, and analyse their estimation effects for different loss functions. According to the Monte Carlo simulation, we find that the M-estimates perform well regardless of whether the data contain outliers. The BM-estimates based on the loss function L are less affected than the other estimation methods for additive outliers. The M-estimates based on the loss function L present better robustness than the other estimation methods for innovational outliers. The empirical application shows that the out-of-sample forecasting capacity of the M-estimates based on the loss function L is better.