Do GAP Models Still have a Role to Play in Forecasting Inflation?

Since the end of Great Recession, researchers have turned their attention to studies on economic recovery, and the speed of correction in the United States. While the economy is recovering, researchers have begun to expect the possibility of inflation in the future. A recent article from the Federal Reserve Bank of Cleveland found that simple models of inflation tend to forecast inflation better than large statistical models. This paper re-visits the price gap model where the central idea is that the price level is determined by the money stock, output and velocity. A horse race is then run whereby a price gap model is tested against atheoretic models based simply on past information, and a structural output gap model. The overall results indicate that the price gap model does in fact display the lowest forecast error over the shorter term forecast horizons, and thus has the most usefulness for inflation forecasting. Robustness checks are then run – the models are re-estimated with a different measure of inflation (CPI less food and energy prices), and the forecasting horizon is extended to 12 quarters. The price gap model is sensitive to the measurement of inflation and loses some of its forecasting power when core CPI (CPI less food and energy prices) is used. Naïve forecasts tend to perform better when forecasting inflation series that are less volatile. However, from the policymaker’s standpoint, it would be more appropriate to have better forecasting power over a more volatile series and so the price gap model would be the ideal choice out of the four models tested here. When the forecasting horizon is extended, the price gap model continues to have the lowest forecast errors.