Forecasting inflation in Iran with two approaches of econometrics and artificial neural network; Comparison of NARDL, NARX nonlinear models

Azizi Ganzagh, Hamed; Jafari Samimi, Ahmad

doi:10.22075/jem.2022.26727.1716

Forecasting inflation in Iran with two approaches of econometrics and artificial neural network; Comparison of NARDL, NARX nonlinear models

Document Type : Original Article

Authors

¹ Ph.D. Student in Economics, Department of Economics, University of Mazandaran

² Professor in Economics, Department of Economics, University of Mazandaran

10.22075/jem.2022.26727.1716

Abstract

Inflation forecasting is one of the most important actions of economic policymakers and monetary officials in the field of decision-making. Besides, researchers always try to identify appropriate methods for predicting inflation. Considering the non-linearity of macroeconomic indicators due to the shocks caused by business cycles, it would be better for inflation rate to be estimated by nonlinear models. Accordingly, in the current research study, attempts have been to utilize theoretical and nonlinear models, such as NARDL and NARX. Apart from the mentioned models, two other models, entitled ARIMA and NAR were employed as non-theoretical models. In fact, after estimation of Iran's monthly inflation rate in the period of 4/21/2005 to 9/22/2019; the time span of 10/22/2019 to 11/21/ 2021 was examined. The findings of the study indicated that for the short-term time span and long-term span NARX and NARDL models respectively performed well based the RMSE, DM criteria.

Keywords