Sentiment Analysis as an Innovation in Inflation Forecasting in Romania

Abstract

Romania faced the highest inflation rate in the European Union at the beginning of 2024, but progress has been made compared to that in 2023 due to the increasing interest rate. This inflation stemmed from a combination of global and domestic factors (global factors such as the Russia-Ukraine war, supply chain disruptions caused by the COVID-19 pandemic and war, rising commodity prices, domestic factors such as wage and pension increases, tax and charge hikes, and a strategy of gradual increase in the monetary policy interest rate). The National Bank of Romania (NBR) uses a combination of monetary policy instruments to target inflation and provides quarterly forecasts. However, under uncertain conditions, numerical forecasts are less reliable, and the inclusion of sentiment analysis in forecasts might lead to innovation in the field by improving the prediction accuracy. Sentiment analysis has become increasingly important in the field of economics, offering valuable insights and potentially improving economic forecasting and decision-making due to rapid technological progress. Sentiment analysis can identify potential changes in consumer behaviour and business decisions before they are translated into actual economic data, providing an early warning system for economic trends and potential crises. The methodological background relies on natural language processing to extract sentiment indices for large amounts of texts in Inflation Reports provided by NBR. Moreover, the sentiment indices calculated by IntelliDocker are incorporated into autoregressive distributed lag (ARDL) models to provide quarterly inflation forecasts. This type of econometric model has the advantage of addressing endogeneity. Moreover, the unemployment rate is considered an inflation predictor since tensions in the labour market might impact inflation. This paper contributes to empirical forecasting by proposing sentiment forecasts that are more accurate than NBR numerical forecasts corresponding to the 2006: Q1-2023: Q4 horizon. The new forecasting method might be used to make inflation predictions for the next quarters. More accurate forecasts would be valuable for businesses, the central bank, policymakers, and the general public. However, while sentiment analysis offers valuable insights, it is important to remember that human judgment and expertise remain essential for interpreting the data and making informed economic decisions.