Tourism Demand for G20 Countries Under Global Uncertainty: ML Versus Dynamic Panel Models for a Comparative Analysis

In the last twenty years, as global uncertainties increased, tourism demand has become more fragile and harder to predict. This study aims to examine the impact of global uncertainty on tourism demand in G20 countries from 2001 to 2023 and to compare the forecasting performance of machine learning algorithms with traditional econometric approaches. The analysis uses international tourist arrivals as the dependent variable; per capita income, the real effective exchange rate (REER), the Global Uncertainty Index (WUI), and lagged demand are included in the model. The dynamic panel (System GMM) estimation confirms strong habit persistence in tourism demand, with the coefficient on lagged arrivals being highly significant at 0.8672 (p<0.01) for the full G20 sample, 0.8812 (p<0.01) for advanced economies, and 0.8524 (p<0.01) for emerging markets. The negative and statistically significant WUI coefficient across the full sample (−0.4568, p<0.01), advanced economies (−0.4683, p<0.01), and emerging economies (−0.4452, p<0.01) demonstrates that global uncertainty systematically suppresses tourism demand. Furthermore, income elasticities (lnGDP) exert a positive and statistically significant impact across both advanced (0.2145, p<0.01) and emerging nations (0.3412, p<0.01), whereas real exchange rate appreciation (lnREER) significantly deters demand in emerging markets (−0.1145, p<0.05). The negative and significant WUI coefficient (−0.45 to −0.47) in both developed and developing countries indicates that uncertainty systematically suppresses tourism demand. Machine learning models, especially during periods of high volatility, produced more successful results; the hybrid model achieved an average MAPE of 3.98% for the 2016–2023 period, demonstrating high accuracy. The findings highlight the importance of flexible and data-driven approaches during periods of uncertainty; they suggest that policymakers should focus on reducing uncertainty shocks and increasing demand resilience.