ECONOMIC EFFECTS OF RUSSIA-UKRAINE WAR: A SENTIMENT ANALYSIS APPROACH

Abstract

The Russia-Ukraine war has had far-reaching economic consequences, significantly influencing both nations’ macroeconomic indicators and public sentiment. This study investigates these economic impacts by integrating multi-source sentiment analysis with macroeconomic data spanning from 2015 to 2025 for Russia and Ukraine. Economic indicators were sourced from the World Bank and the International Monetary Fund, including consumer price index (CPI), inflation rates, GDP growth, unemployment, government debt, trade volumes, foreign direct investment (FDI) inflows, and military expenditures. Missing data for select years were estimated using time series forecasting models: Vector Autoregression (VAR), Seasonal Autoregressive Integrated Moving Average (SARIMA), and Prophet, with SARIMA yielding the most reliable forecasts for Ukraine and a combination of SARIMA and VAR performing best for Russia. In parallel, sentiment data were extracted from over 500,000 social media posts, news articles, and Reddit comments relating to the war. These texts were preprocessed and analyzed using multiple lexicon-based sentiment tools including TextBlob, VADER, AFINN, and SentiWordNet. Annual sentiment scores were calculated and then correlated with economic indicators using Pearson and Spearman correlation coefficients, Granger causality tests, and visual trend analyses. Ukraine’s economy shows heightened sensitivity to inflation and trade disruptions, with sentiment reacting more rapidly to economic changes. Russia, on the other hand, exhibits a more delayed sentiment response, aligning with its relatively more controlled economic structure. Granger causality tests confirm a more immediate influence of sentiment on Ukraine’s economic variables, whereas in Russia, such impacts surface more gradually. War-related spending is a common inflation driver in both nations, though Ukraine demonstrates a heavier reliance on debt-financed defense efforts. Machine learning models were applied to assess predictive performance, with XGBoost outperforming Random Forest overall, especially in modeling Russia’s indicators. In contrast, Random Forest showed a slight edge in predicting Ukraine’s economic trends. Lagged correlation analyses reinforce sentiment’s predictive value, particularly during periods of conflict. Post-war analysis indicates that Ukraine is experiencing a faster, though more volatile, recovery compared to Russia’s steadier but slower rebound. This research highlights the importance of sentiment in economic forecasting during geopolitical crises. Future work could enhance predictive accuracy through the integration of deep learning-based natural language processing models and explore the role of sentiment-informed economic policy in managing shocks during conflicts.