The main aim of our paper is to improve the extracted information on forward inflation compared to the one provided by the standard linear term spread model. As a result, we examine three alternative formulations, capturing the information content of the term spread under regime switching effects. The first is intended to capture different states of the monetary policy stance, by separating positive and negative values of the slope of the yield curve. The second and third regime switching formulations incorporate unobserved state dependent variables that follow a Markov ergodic chain distribution, intended to capture different states of the term spread's coefficients and the constant of the underlying equation respectively.
The sample contains monthly data of long-term (benchmark) bond yields, short term Treasury yields and CPI for the period of 1973:1 to 2007:1, collected by OECD. The forecasting horizon begins from 12 months and extends to 5 years while we report results for the
Comparing the results of the three alternative formulations against the ones of the standard model we find that the incorporation of regime switching effects improves substantially the predictive power of the term spread. Additionally, a second, but not less essential, outcome is the decomposition of the information content for forward inflation. Specifically, our results indicate that states of high volatility of inflation are associated with increased significance of the term spread as a predictive variable for future inflation. Finally, the comparison of the three alternative models provides evidence on the optimality of introduction of Markov Switching effects in the constant of the underlying equation. Overall, these results, being consistent with previous empirical research findings and theoretical considerations, indicate that macroeconomic and monetary conditions should be taken into account in order to properly extract the predictive content of the term spread for inflation.