What drives inflation and how? Evidence from additive mixed models selected by cAIC

Philipp F. M. Baumann, Enzo Rossi and Alexander Volkmann



JEL classification
C14, C33, C52, E31

Theories of inflation, longitudinal data, additive mixed models, model-based boosting, conditional Akaike criterion


We analyze the forces that explain inflation using a large panel of 122 countries from 1997 to 2015. Models motivated by the economic theory are compared to a boosting algorithm, and non-linearities and structural breaks are explicitly considered. The boosting algorithm outperforms theory-based models. Further, we provide compelling evidence that the interaction of energy price and energy rents stand out among 37 explanatory variables. Other important determinants are demographic developments. Contrary to common belief, globalization and technology, public debt, central bank independence and transparency as well as countries’ political characteristics, are less relevant. Exchange rate arrangements are more important than inflation-targeting regimes. Moreover, GDP per capita is more relevant than the output gap and credit growth is generally superior to M2 growth. Many predictors exhibit a structural break since the financial crisis. In particular, credit growth has lost its grip on the inflation process.