Forecasting economic outcomes is a challenging and time-consuming endeavor. Nonetheless, the ability to predict future economic activity is vital for many purposes. Among them, it allows governments to make policy decisions that affect the economy and business, and it helps households and companies plan for the future.
As one might expect, there are a wide variety of methodologies employed in creating economic forecasts. In broad terms, the various methods can be categorized into two polar opposites:
On the one hand are statistical time series models that require only a statistical characterization of the behavior of a particular economic variable over time. Using these models does not necessarily require any knowledge about why the variable behaves in a certain way, but it does generally require immense amounts of data and faith that the pattern of that behavior will persist into the future.
A more complex group of forecasting methods combines statistics with economic knowledge. Examples of these include factor models, mixed frequency models, and non-linear models. They may be used in conjunction with more traditional linear time series models or as a replacement. These techniques generally involve estimating relationships between different variables and then using those relationships to predict the behavior of a third variable.
Regardless of the type of model, it’s important to remember that a good economic forecast requires not just good data and a model but also some understanding of how the data is collected and revised. For example, a model that forecasts quarterly GDP growth will suffer from revisions of the data on a year-over-year basis. It will also be affected by the fact that the horizon for forecasting GDP growth is short.
About the Author
