Mihaela Simionescu (mihaela_mb1@yahoo.com, mihaela.simionescu@unibuc.ro)
Faculty of Business and Administration, University of Bucharest, Romania
Academy of Romanian Scientists, Bucharest, Romania
Institute for Economic Forecasting, Romanian Academy, Bucharest, Romania
Alexandru-Sabin Nicula (sabin.nicula@ubbcluj.ro)
Academy of Romanian Scientists, Bucharest, Romania
Romanian Academy, National Institute for Economic Research ”Costin C. Kiri!escu”, Mountain Economy Center, Vatra Dornei, Romania
Centre for Research on Settlements and Urbanism, Faculty of Geography, Babeș-Bolyai University, Cluj-Napoca, Romania
ABSTRACT
Most Eastern European countries experienced high inflation because of the war in Ukraine, which makes more difficult to get accurate inflation forecasts necessary for policymakers, central bank and business environment. The main aim of this paper is to provide accurate forecasts for inflation rate in Romania and Czech Republic by using non-parametric Bayesian models and generalized regression neural networks (GRNN) that include sentiment index determined using central banks official reports. The forecasts based on Bayesian linear regression models outperformed the ones based on Bayesian linear regression model with LASSO prior, Bayesian linear regression model with stochastic search variable selection (SSVS) and GRNN on short-term horizon 2023: Q1- 2023: Q4. The approach based on diff erence-in-diff erences estimators suggested that the strategy of the Czech National Bank (ČNB) based on proactive increased interest rates in 2021-2022 to control infl ation was eff ective and it reduced expected infl ation, but had no signifi cant impact on unexpected component. The expected inflation in Czechia decreased by 0.972 percentage points relative to the case when interest rate would have not increase sharply. These empirical findings bring contribution for providing better short-term inflation forecasts by taking into account
experts opinion on future evolution of inflation and their interventions to reduce the phenomenon.
Keywords: inflation; non-parametric Bayesian models; generalized regression neural networks; interest rate
