Title: A suite of inflation forecasting models
Authors: Luis J Alvarez - Bank of Spain (Spain) [presenting]
Isabel Sanchez - Bank of Spain (Spain)
Abstract: A set of econometric models is described which is used to forecast consumer price inflation, a topic of renewed interest following the Great Recession and its aftermath. First, we consider highly disaggregated univariate time series models. Second, we propose transfer function models that include explanatory variables, such as oil prices or unit labour costs and that also allow for a slowly evolving local mean. Third, we consider macro-based models, including hybrid New Keynesian Phillips curves. These models have a theoretical foundation and include forward looking elements. Our main findings are that successful inflation forecasting models have to account for a slowly evolving local mean, in order to be able to cope with changes in trend inflation, as in the period after the Great Recession. Furthermore, differences in the features of product markets suggest that it is relevant to employ some sort of disaggregation to deal with heterogeneity in price setting. Finally, transfer function models tend to show a better forecasting performance than alternative models.