Dallas Fed research seeks to improve real-time inflation forecasts

The Federal Reserve Bank of Dallas. Photo: Andreas Praefcke

It is possible to improve real-time forecast performance by stripping noise out of data on inflation and inflation expectations, research published by the Federal Reserve Bank of Dallas has found.

The working paper by N Kundan Kishor and Evan Koenig begins by breaking survey data on expectations and inflation data into trend, cyclical and noise components. This allows them to filter the noise component out when using the data for real-time forecasting.

Forecasts based on their filtered inflation