Regional Inflation Forecasting: Econometric Models Versus Machine Learning Methods?

Abstract

The paper reveals the forecasts for regional inflation based on the regions of the Privol­zhskiy Federal District (PFD). The purpose of the study is to determine the model that most ac­curately predicts regional inflation. The paper compares the tools of machine learning – sup port vector machines, gradient boosting, and random forest – with econometric models of time series, autoregression and integrated autoregression-moving average – models that are often used to predict national inflation in Russia. The results of the study help to identify the key macroecono­mic indicators that most significantly affect regional inflation. The top three of them for each region include the value of inflation within last month, the average inflation, and the count number of analyzed month. We show that machine learning models are not as bad at predicting regional inflation over long time periods (21 months and 24 months), while econometric models predict quite accurately on short time horizons.