What is the best way to measure bias in forecasts?
The objective of bias is to determine whether forecasts that are prepared have a tendency to over- or under-forecast. In other words, no one is biasing them in one direction or the other. Since academicians define bias differently, formulas given in statistics and forecasting books are not applicable here. Practitioners calculate bias as follows:
Bias = | Sum of Errors
Sum of Actuals |
x 100 |
If the bias is positive, forecasts have a bias of under- forecasting; if negative, the bias is of over-forecasting. Incidentally, this formula is same as Mean Percentage Error (MPE). There is a fifty-fifty chance for an error to be of under- or over-forecasting. The bias is of what goes above that. The formula shown above captures that component of the error.
Related Posts
« Why don’t forecasters get the respect they deserve from upper management? How can a demand planner estimate his or her value in the market in terms of salary expectation? »