EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0151
Title: An empirical evaluation of some long-horizon macroeconomic forecasts (virtual) Authors:  Kenneth D West - University of Wisconsin (United States) [presenting]
Kurt Lunsford - Federal Reserve Bank of Cleveland (United States)
Abstract: Over a century of annual cross-country data are used to evaluate pseudo-out-of-sample forecasts of six mean-reverting variables and four quite persistent variables for horizons up to 50 years. The mean-reverting variables are real per capita GDP growth, CPI inflation, labor productivity growth, population growth, money growth and equity returns; the quite persistent variables are real exchange rates, the investment-to-output ratio, and long- and short-term nominal interest rates. Our models for forecasting include simple time series models and frequency domain methods recently developed. We consider both point estimation and coverage of 68\% intervals for forecasts. For the six mean-reverting variables, a simple AR(1) and a frequency domain model are best in terms of point estimation and, as well, have well-calibrated 68\% forecast intervals; calibration of forecast intervals, does, however, distinctly degrade when increasing the horizon from 25 to 50 years. For the four very persistent variables, a random walk is perhaps the best choice, though forecast intervals are not well calibrated at any horizon for any of the models. We conclude that forecasting over very long horizons is viable, at least for data that are not very persistent.