Inflation forecasting with inflation sentiment indicators
In this paper we analyze the forecasting power of different inflation indicators using the example of the US and the German economy. We try to construct some simple measures of inflation sentiment which basically try to capture whether inflation is broad based, i.e. caused by price acceleration in many components of the CPI basket. In addition, the oil price is evaluated as an inflation predictor, too. The investigation shows that inflation sentiment indicators can improve forecast accuracy in comparison to predictions based on a standard Phillips curve approach. Forecast accuracy is improved for both countries and the gains are in most cases highly significant according to Diebold-Mariano and modified Diebold-Mariano test statistics. Moreover, forecast breakdown tests indicate that the forecast-relevant effects are quite stable. However, because the German sample is split up into two parts we also show that the forecast accuracy of the different models applied can change over time. Also the forecasting power of oil price differences is time-varying. This result stresses scepticism about the usefulness of traditional core inflation measures. The advantage of our newly-developed inflation sentiment indicators is that they do not assign certain components of the CPI certain forecasting power.