Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
The accuracy of inflation forecasts has important implications for macroeconomic stability and real interest rates in economies with nominal rigidities. Erroneous forecasts destabilize output, undermine the conduct of monetary policy under inflation targeting and affect the cost of both short and long-term government borrowing. We propose a new method for forecasting inflation that combines individual forecasts using time-varying-coefficient estimation along with an alternative method based on neural nets. Its application to forecast data from the US and the euro area produces superior performance relative to the standard practice of using individual or linear combinations of individual forecasts, especially during periods marked by structural changes.