Analysing inflation with semi-structural models
This chapter has been prepared for the Research Handbook of Inflation (Edward Elgar Publishing). It explores semi-structural time series models as a tool for analysing, forecasting, and nowcasting inflation. We define as semi-structural models a class of multivariate time series model, in the tradition of Harvey (1985, 1990), where minimal economic restrictions are used to identify common and idiosyncratic trend and cyclical components in the data. This chapter shows that these models are suitable to analyse price pressure at business cycle frequency and to provide real-time indicators of unobserved economic measures such as trend inflation and output gap. Moreover, it compares semi-structural models with other common models in macroeconomics, like vector autoregressions (VARs), showing that semi-structural models often outperform them in forecasting accuracy.
Comments