The operations literature remains inconclusive about the most adequate forecasting approach - top-down or bottom-up. This paper presents analytical results for mean values and variances of forecasting errors with exponential smoothing under these approaches. The share of the series in the total aggregate data is considered stochastic. Based on the generation of random numbers, two major impacts are confirmed on the variance of forecasting error, which were presented in previous papers, namely, portfolio and anchorage effects. The impacts are also evaluated in terms of forecasting bias and safety inventory levels.
Forecasting approach; Exponential smoothing; Forecasting errors; Safety inventories
