DOES THE TIME SHAPE OF INVENTORY DATA HOLD CLUES ON INVENTORY BEHAVIOR

This paper tries to show that there may be a serious problem in the wa y inventory data are observed and processed. Inventory cycles may be a symmetric and the asymmetries may switch directions over the business cycle. That is, inventory holdings may follow a strongly non-linear ti me path that cannot be represented by periodic observations that are a ppropriate for stationary processes. It may lead to a cumulative error , with passage of time, in the difference between the actual inventory levels and the data we think monitor them. It has been shown that tem poral aggregation leads to a downward bias in the estimated speed of a djustment of the flexible accelerator model [1, 2]. Assuming this mode l correctly describes inventory behaviour and that the objective is to recover the model in the data this evidence raises the question why i s there this estimation bias? Could it be that the bias results from i gnoring the time shape of the data? If this is true, more frequent obs ervations could reduce that bias. This paper tries to show that the ti me shape of inventory data may have a significant effect on what we ob serve with periodic readings.