What helps forecast U.S. inflation? - mind the gap!

The Phillips curve, which posits a relationship between inflation and domestic economic activity, introduces a crucial trade-off between real and nominal objectives for the central bank. Atkeson and Ohanian (2001), among others, present evidence that forecasts of U.S. inflation from Phillips curve-based models tend to underperform relative to naïve forecasts. We propose that globalization can be an important factor in explaining the poor performance of forecasts under a closed-economy Phillips curve. To illustrate that, we empirically evaluate the performance of open-economy Phillips curve-based forecasts constructed with global variables, such as G7 credit growth, G7 money supply growth, terms of trade, and the real effective exchange rate. These global variables perform significantly better than domestic variables, and serve as proxies for poorly-measured indicators of global slack. Moreover, we show that forecasts using simulated data from a workhorse open-economy New Keynesian model support our empirical findings on the open economy Phillips curve and also suggest that better monetary policy and aspects of the Great Moderation have improved the forecast accuracy of open-economy models.