Title: Forecasting regional inflation using spatial correlation models
Authors: Taisiia Gorshkova - Russian Presidential Academy of National Economy and Public Administration (Russia) [presenting]
Abstract: The purpose is to discuss the need to integrate the spatial relationship in regional data for forecasting inflation. There is a comparative analysis of the models that take into account only temporal correlation between the data and models that take into account temporal and spatial correlation. Three weighting matrices are used to consider spatial correlation in data and six different models with each weight matrix, such as individual models for each region, panels with fixed and random effects, spatial lag models and spatial error models. The individual forecasts based on the models were then combined with different weights into a single forecast. The weights were chosen on the basis of five methods, including discounting and shrinkage methods and method of principal component analysis.