Advances in Economic Forecasting

1 - Advances in Economic Forecasting

1 1 Advances in Economic Forecasting Matthew L. Higgins Western Michigan University The six chapters that follow this introduction are based on lectures the authors gave at Western Michigan University as part of the 2009– 2010 Werner Sichel Lecture Seminar Series, organized under the same title as the present volume. The lectures were given over an academic year during a time when the U.S. economy was just beginning to recover from the “Great Recession.” The economics profession’s inability to predict this catastrophe may have seemed like an inauspicious background for a lecture series on economic forecasting. However, the economic distress and surrounding uncertainty actually benefited the series because they heightened interest in the topic of economic forecasting. The question-and-answer sessions following each lecture revealed that some audience members had a healthy skepticism of economists’ ability to ever predict the future course of the economy. I think, however, that each speaker’s candid and realistic assessment of opportunities to improve economic forecasting left most attendees with some sense of optimism. Several common recommendations emerge from the following six chapters for improving the reliability of economic forecasts. Authors Dean Croushore, Kajal Lahiri, and H.O. Stekler all emphasize that improvements in forecasting will require proper evaluation of the performance of forecasting methods, focusing particularly on the ability of methods to forecast in real time and predict turning points in major macro aggregates. David E. Rapach and Tae-Hwy Lee, in their chapters , argue that the abundance of economic data can be more efficiently exploited through model and forecast combination. Rapach, Lee, and coauthors Michael D. Bradley and Dennis W. Jansen each advocate using models that are adaptive and perform well in the presence of nonlinearity and structural change. Below, I briefly summarize each author’s chapter to help direct the reader to these specific themes. Higgins.indb 1 Higgins.indb 1 11/3/2011 10:22:07 AM 11/3/2011 10:22:07 AM 2 Higgins In the book’s second chapter, Croushore addresses the complications that data revisions have on economic forecasts produced in real time. He begins by advocating the use of forecasts provided by the Survey of Professional Forecasters (SPF). He argues that these forecasts are unbiased and efficient over long time periods. This survey is made publicly available by the Federal Reserve Bank of Philadelphia in an easily accessible format. (Croushore provides the Web address for this.) Furthermore, because this survey can be easily matched to the real-time macroeconomic data also maintained by the Philadelphia Fed, the survey provides a good data set for studying the role of data revisions in the forecasting process. Croushore asserts that many forecast evaluation studies are flawed because real-time forecasts are compared to ex post forecasts that are based on revised data that actual forecasters did not have access to. To the real-time forecaster, recognition of the possible magnitude of data revisions causes uncertainty about model inputs, structure, and coefficient values. Croushore’s research indicates that all three factors can degrade the quality of forecasts. Based on his own work, and the work of others, Croushore suggests that attempts to explicitly incorporate the process of data revision into model construction have so far had limited success in improving the quality of economic forecasts. To illustrate these issues, Croushore examines forecasts in the 1990s from the SPF. When compared to the 2001 vintage actual values, the forecasts of gross domestic product (GDP) growth were consistently too low, and forecasts of inflation and unemployment were consistently too high. Forecasters appeared to be slow to recognize the effect of high productivity growth in that decade. When the forecasts are compared to actual values observed in real time for the variables, the forecasts appear much better. This may explain why forecasters were slow to adapt their models to the surge in potential output. Croushore’s chapter demonstrates that research on the role of data revisions in economic forecasts is at a very early stage and should prove to be productive in the coming years. In the third chapter, Lahiri addresses the intriguing question of how far into the future forecasters can provide information about the growth of GDP. He uses a survey of forecasts of the annual growth rate of GDP for 18 countries in the Organisation for Economic Co-operation and Development (OECD), obtained from Consensus Economics Inc. The Higgins.indb 2 Higgins.indb 2 11/3/2011 10:22:07 AM 11/3/2011 10:22...