7 Short-Term Forecasting
The physicist Nils Bohr once quipped, “Prediction is very difficult, especially about the future.” While prediction, or forecasting, is always difficult, some kinds are more difficult than others. Long-term forecasting is the most difficult because information is limited, competing trends exist, and many variables can influence future outcomes. An example would be forecasting the development of new technologies. A decade ago, could we have developed good forecasts for the sales of cholesterol-reducing drugs? Or the demand for downloadable music? Or the uses for geographic information systems? Forecasting those phenomena at that early stage would certainly have been challenging.
Short-term forecasting is another matter. Predicting the volume of activity in an established market is relatively manageable when we have access to good records based on recent observations. For example, what if we want to predict how much gas we'll use in our car next month? Or the interest rate on 6-month CDs next week at our local bank? Or the number of callers requesting maintenance service on their kitchen appliances tomorrow? When quantities such as these play the role of input parameters in our models, the task becomes one of making intelligent extrapolations from the history of observations. Practical techniques are available for this kind of short-term forecasting.
Regression analysis can sometimes be useful in short-term forecasting because it represents ...