Bias-Variance Trade-offs in Demand Forecasting

Konstantinos Katsikopoulos and Aris Syntetos

Foresight: The International Journal of Applied Forecasting, 2016, issue 40, 12-19

Abstract: In supply-chain forecasting, we have traditionally used point forecasts to predict the mean level of demand per time period. In doing so, we place emphasis on finding forecast methods that minimize bias in the forecasts, because forecast bias ultimately leads to either excessive or inadequate inventory levels. This emphasis on avoiding bias, however, can neglect variability in the forecasts. Konstantinos and Aris explain in this article that the neglect of forecast variability is a mistake, because forecast variance is a factor - along with, and of the same importance as, bias - that determines the Mean Squared Error (MSE) of a forecast method, which is often (and possibly too often) used to set safety stocks. The authors show that, in selecting a forecast method, there is in fact a trade-off between forecast bias and forecast variance. Their provocative takeaway is that simple methods tend to have large bias but low variance, while complexity reduces bias at the expense of increasing variance. Hence, simple forecasting methods, even if their resulting forecasts are biased, might be preferable to complex methods for minimizing forecast error. Most generally, the right amount of complexity should be sought. Copyright International Institute of Forecasters, 2016

Date: 2016
References: Add references at CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
https://foresight.forecasters.org/shop/

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:for:ijafaa:y:2016:i:40:p:12-19

Access Statistics for this article

More articles in Foresight: The International Journal of Applied Forecasting from International Institute of Forecasters Contact information at EDIRC.
Bibliographic data for series maintained by Michael Gilliland ().