Comparative Analysis of Conditional Hedging Strategies for Rupiah Holders Using Forecasting Models

Research Aims: This study aims to compare the effectiveness of various conditional hedging strategies for Rupiah holders by leveraging forecasting models, including traditional statistical, machine learning, and hybrid approaches. The goal is to determine which models offer the most accurate exchange rate forecasts and how these forecasts translate into better hedging decisions and financial outcomes. Design/methodology/approach: This research adopts a quantitative experimental approach by testing fourteen forecasting models—including ARIMA, ETS, GARCH, XGBoost, and hybrid methods—within an iterative forecasting framework using monthly USD/IDR data from 2014 to 2024. The study evaluates hedging strategies (Always Hedge, Unhedge, Conditional Hedge) using out-of-sample forecasting and performance metrics such as MAE, RMSE, MAPE, cumulative return, average return, and return volatility. Research Findings: The hybrid ETS-ARIMA-XGBoost model emerged as the most accurate forecasting method, achieving the lowest MAE and MAPE values. Conditional hedging strategies based on this model produced the highest cumulative and average monthly returns with moderate volatility, outperforming both Always Hedge and Unhedge approaches. Simpler models (e.g., ARIMA, ETS) performed adequately but were less effective in dynamic hedging. Meanwhile, models like GARCH and Prophet exhibited lower forecasting precision and inferior hedging outcomes. Theoretical Contribution/Originality: This study bridges the gap between advanced forecasting techniques and real-time financial decision-making in currency risk management. It contributes to financial economics by demonstrating how model-driven conditional hedging strategies align with rational decision theory and expected utility frameworks. The use of iterative hybrid forecasting models in practical hedging contexts is a novel empirical contribution.