Laws of Forecasting
 

Forecast Bias

Obvious examples of forecast bias are the sales person wanting to make sure their quota is as low as possible, the development manager trying to gain approval for a new project, and the industry trade group economist creating an industry forecast. All are likely to nudge the forecast in a direction that is favorable to their goals.

These intentional biases are pretty obvious, but there are many ways that unintentional bias can enter a forecast.

Every time a forecast is built, the forecaster has to decide which factors to include and which seem safe to exclude. If the forecaster is not careful to provide equal balance, it is possible for these decisions to be more heavily weighted on one side or the other, possibly biasing the overall results.

Another possible source for bias is the way the data is used. Do I look at that unusually high or low value last month as an anomaly, or do I treat it as important information? Do I use all of the data in the forecast, or just the last 12 months? We all have a tendency to see in data what we want to believe, even data that is factual and precise. And if the data is at all subjective or incomplete, this tendency is amplified.

A third source of unintentional bias can be the interpretation of trends. As noted in the laws, all trends end, but it is possible for a forecaster to forget that rule if the trend happens to be favorable. In this case, they might be less inclined to challenge it than if it were an unfavorable trend.

By asking the right questions, it is frequently easy to tell what a forecaster's preference might be, and that can lead you to possible biases in the forecast.

The most difficult situation is when you are the forecaster. The best question to ask here is "what do I really want the answer to be?" The answer to that question will tell you the direction that you are likely to be biasing the forecast. If you have just bought 100 shares of XYZ stock and are creating a forecast of their stock price, you are likely to be biasing that forecast on the upside.

Once you have established your likely bias, try looking at the forecast from the opposite point of view. If you have a colleague that likely has an opposite bias (e.g., maybe they just shorted XYZ), ask them for their views and then listen intently. If you don't want to ask them, at least ask yourself what they might say. Then modify the forecast to consider these points. The results of this exercise will frequently be illuminating.

Return to The Laws of Forecasting


footer