with Andrea Carriero (Queen Mary University of London), Todd Clark (Federal Reserve Bank of Cleveland), Massimiliano Marcellino (Bocconi, IGIER and CEPR)
Abstract: Interest rate data are an important element of macroeconomic forecasting. Not only are projections of future interest rates an object of interest themselves, but also they matter for forecasting other macroeconomic and financial variables. A popular class of forecasting models is linear vector autoregressions (VARs) that include shorter- and longer-term interest rates. However, in a number of economies, shorter-term interest rates have by now been stuck at or near their effective lower bound (ELB) for years, with longer-term rates drifting toward the constraint as well. In such an environment, linear forecasting models that ignore the ELB constraint on nominal interest rates can be problematic along various dimensions. Instead, we model nominal interest rates as censored observations of a latent shadow-rate process and relate their dynamics to other economic variables in a so-called ``shadow-rate VAR.'' We consider specifications where both actual and shadow interest rates, or only shadow interest rates, matter for forecasting. Our empirical application studies the performance of point and density forecasts generated by shadow-rate VARs for US data since 2009. In comparison to a standard VAR, shadow-rate VARs generate superior predictions for short- and long-term interest rates. By ignoring the ELB, the standard VAR can generate a negative-rate outlook which influences its economic forecasts in ways that can be similar to a shadow-rate VAR. On balance, across measures of economic activity and inflation, the accuracy of forecasts from our shadow-rate specifications is on par with a standard VAR.