Forecasting with Regression Analysis

Regression analysis is a statistical method that enables us to fit a straight line that on average represents the best possible graphical relationship between sales and time. This best fit is called the regression line.

One way regression analysis can be used is to simply extrapolate future sales based on the trend in past sales. Another way of using regression analysis is to look at the relation between two measures, say, sales and capital expenditures.

While regression analysis gives us what may seem to be a precise measure of the relationship among variables, there are a number of warnings that management must heed in using it:

• Using historical data to predict the future assumes that the past relationships will continue into the future, which is not always true.
• The period over which the regression is estimated may not be representative of the future. For example, data from a recessionary period of time will not tell much about a period that is predicted to be an economic boom.
• The reliability of the estimate is important: If there is a high degree of error in the estimate, the regression estimates may not be useful.
• The time period over which the regression is estimated may be too short to provide a basis for projecting long-term trends.
• The forecast of one variable may require forecasts of other variables.
• For example, the management may be convinced that sales are affected by gross domestic product (GDP) and use regression to analyze this relationship. But to use regression to forecast sales, management must first forecast GDP. In this case, management’s forecast of sales is only as good as the forecast of GDP.