Advances in Economic Forecasting

2 - Real-Time Forecasting

7 2 Real-Time Forecasting Dean Croushore University of Richmond and Federal Reserve Bank of Philadelphia This chapter will discuss real-time forecasting in a macroeconomic policy context. I will begin by talking about the Survey of Professional Forecasters (SPF), a survey of private-sector forecasters. Next, I will discuss research on real-time data analysis and its importance in forecasting . Finally, I will discuss real-time forecasting in the 1990s. In a policy environment, such as the one I faced for 14 years at the Federal Reserve Bank of Philadelphia, you have three basic choices for developing forecasts in a real-time forecasting environment. One possibility, used by many policy analysts, is simply to rely on forecasts made by others, such as the consulting firm Macroeconomic Advisors. After all, forecasting firms devote considerable resources to forecasting , so why not trust their forecasts? An alternative is to look at surveys of forecasters, such as the SPF. This gives you a range of forecasts, and you can base your decisions on the median forecast, which is usually a better forecast than the forecast provided by any individual forecaster. The third possibility is to create your own forecasting model. This gives you the ability to tweak the forecast to your own needs and to specify your own baseline underlying the forecast. You can do some simple things such as I did at the Fed—for instance, forecasting GDP for the current quarter based on the employment data that are released early in the month. Or you can run time-series models of your own specification , which often hold their own against much larger, more sophisticated models. Or, you could buy a large-model forecasting software program, such as the one provided by Macroeconomic Advisors, and then modify some of its assumptions to your own liking to produce your own forecast based on its model structure. Unless you have many resources at your disposal, however, you probably do not want to produce a large-model forecast on your own. You are unlikely to do better Higgins.indb 7 Higgins.indb 7 11/3/2011 10:22:08 AM 11/3/2011 10:22:08 AM 8 Croushore than others, and almost certain to produce much worse forecasts, unless you have a large number of economists working on it, such as the dozens that work on forecasting at the Federal Reserve Board. The major concern that you should have about all these forecasting models is the role of judgment in the outcome of the forecasting exercise . The more you study forecasting, the more you realize how much impact judgment has—there is no such thing as a pure model forecast. First, there is judgment in determining what model to use. Second, there is judgment about the underlying key parameters of the model: how do you determine the natural rate of unemployment, or the growth rate of potential GDP, or the equilibrium real interest rate, which are generally not determined within a model? Those factors tend to drive the forecast much more than you might think. THE SURVEY OF PROFESSIONAL FORECASTERS My own involvement with forecasting began in 1990 with a research paper in which I wanted to get data on inflation expectations. I used a survey that was run by the American Statistical Association and the National Bureau of Economic Research (NBER) and was known as the Economic Outlook Survey. The survey contained much useful information , and I was impressed that the survey had begun in 1968 and was the longest quarterly survey of forecasters in existence. But shortly after using the survey myself, I read an announcement that the survey was folding because of lack of interest and because there was no organization that was willing to run it. As an economist at the Federal Reserve, I thought the survey was incredibly useful—it gave great insight into the expectations of the country’s leading forecasters. I was determined that the survey should not die, and so I contacted Robert Allison of the NBER and Victor Zarnowitz of the University of Chicago about the possibility of the Federal Reserve Bank of Philadelphia taking over the survey. They were both enthusiastic about having an institution like the Fed running the survey, and so we took over, missing only one quarterly survey in the transition. After taking over the survey, I, along with my coresearcher Leonard Mills, began to rehabilitate it. We renamed it the Survey of Professional Higgins.indb 8 Higgins.indb 8 11...