Economic Forecasting, discipline in economics concerned with the prediction of economic phenomena. Almost everybody in society needs to make some sort of economic forecast at some points in their life. Most of those made by ordinary individuals in their private lives are probably made subconsciously, or at least informally. Companies will be obliged to make more explicit forecasts of market conditions and so on in planning their production schedules. And governments will need to make forecasts before deciding their budget strategies, their monetary policies, and so on. Depending on who is making the forecast, therefore, and the nature of the forecasts being made, the methods used will vary from extremely crude extrapolations based on past experience to very detailed forecasts for the whole economy based on computer models of it described by hundreds of equations.
The fundamental element in any forecast of the future value of any “dependent” variable—such as the rate of inflation—is some explanation of how, in the past, the variable in question has been linked to some other “explanatory” variable which it is thought possible to predict or take as given. For example, it may be believed, on the basis of data for previous years, that the rate of inflation in any one year has been related in a precise manner to the rate of increase in the money supply over the previous two years. Since the latter is known at any time it is simple to predict next year’s rate of inflation. In this instance, the past increase in the money supply is given and would be regarded as an “exogenous” variable in the determination of inflation. The reliability of the prediction would depend on factors such as how strong the correlation is between past increases in the money supply and subsequent rate of inflation, and what direction of causality is implied by the correlation (if any). The former depends largely on how many other variables affect the inflation rate and the latter is mainly a matter of economic theory.
Apart from the validity of the underlying model used to predict any dependent variable X on the basis of its relationship with some explanatory variable Y, the use of data for back years—which is the economist’s main source of data—encounters certain serious technical difficulties (such as “autocorrelation”) which cannot be covered here. And insofar as any one variable, X, may be determined by a variety of “explanatory” variables, U, V,Y, and Z, more complicated statistical techniques will be necessary to derive the relationship between each of them and the “dependent variable”. This is because the “explanatory variables” may be interrelated amongst themselves.
Over the course of the past few decades, the models used to predict the evolution of the economy have become bigger and bigger and more and more complex. This has become possible on account of the vast increase in the power of computers. It seemed to be desirable because of repeated failures of most economic forecasts to be reasonably accurate. As a result, continuous attempts have been made to strengthen underlying equations of the models used for the predictions by adding more and more explanatory variables, most of which will be determined by other equations in the model, thereby accumulating the scope for an error in one part of the model to be propagated throughout.
Despite all the improvements and refinements in the models—or, as some economists believe, because of them—the record of short-term macroeconomic forecasts over the past few years has been extremely poor. It would appear safer to stick to very long-term forecasts since by the time they are shown to be wrong nobody will remember that they were ever made in the first place.