This dissertation examines the effectiveness of the Quantile Factor Model (QFM) in pre dicting exchange rate returns, specifically examining the U.S. dollar against a basket of major currencies. Traditional econometric models often fail to accurately forecast exchange rate movements due to market complexity, especially during periods of high volatility. To ad dress these limitations, the thesis applies the QFM framework, which provides a flexible alternative by capturing non-linear dependencies and tail risks in exchange rate dynamics. Using the FRED-MD database of macroeconomic indicators and U.S. dollar index data, this study applies QFM to identify and forecast exchange rate returns across quantiles, provid ing insights into the distributional properties that are critical in high-volatility conditions. The robustness of QFM is compared to simpler models, including the Random Walk (RW) model, with predictive performance evaluated using Mean Squared Error (MSE) and Density Forecast scores. Results show that while the QFM provides valuable distributional insights across the distribution, the RW model ultimately outperforms the QFM in predictive accu racy, underscoring the persistent challenge of forecasting exchange rate movements even with advanced quantile-based methods, highlighting the unpredictability of exchange rates and confirming the validity of Efficient Market Hypothesis. This study contributes to financial econometrics by demonstrating the practical limita tions of complex factor models and emphasizing the continued effectiveness of simpler, basic forecasting models such as Random Walk. Future research should focus on better understanding the relationship between exchange rates and macroeconomic variables and which are the main drivers for movements, especially in the short to medium term forecast ing horizon
This dissertation examines the effectiveness of the Quantile Factor Model (QFM) in pre dicting exchange rate returns, specifically examining the U.S. dollar against a basket of major currencies. Traditional econometric models often fail to accurately forecast exchange rate movements due to market complexity, especially during periods of high volatility. To ad dress these limitations, the thesis applies the QFM framework, which provides a flexible alternative by capturing non-linear dependencies and tail risks in exchange rate dynamics. Using the FRED-MD database of macroeconomic indicators and U.S. dollar index data, this study applies QFM to identify and forecast exchange rate returns across quantiles, provid ing insights into the distributional properties that are critical in high-volatility conditions. The robustness of QFM is compared to simpler models, including the Random Walk (RW) model, with predictive performance evaluated using Mean Squared Error (MSE) and Density Forecast scores. Results show that while the QFM provides valuable distributional insights across the distribution, the RW model ultimately outperforms the QFM in predictive accu racy, underscoring the persistent challenge of forecasting exchange rate movements even with advanced quantile-based methods, highlighting the unpredictability of exchange rates and confirming the validity of Efficient Market Hypothesis. This study contributes to financial econometrics by demonstrating the practical limita tions of complex factor models and emphasizing the continued effectiveness of simpler, basic forecasting models such as Random Walk. Future research should focus on better understanding the relationship between exchange rates and macroeconomic variables and which are the main drivers for movements, especially in the short to medium term forecast ing horizon.
Forecasting Exchange Rates: a Quantile Factor approach
STORTONI, ALESSANDRO
2023/2024
Abstract
This dissertation examines the effectiveness of the Quantile Factor Model (QFM) in pre dicting exchange rate returns, specifically examining the U.S. dollar against a basket of major currencies. Traditional econometric models often fail to accurately forecast exchange rate movements due to market complexity, especially during periods of high volatility. To ad dress these limitations, the thesis applies the QFM framework, which provides a flexible alternative by capturing non-linear dependencies and tail risks in exchange rate dynamics. Using the FRED-MD database of macroeconomic indicators and U.S. dollar index data, this study applies QFM to identify and forecast exchange rate returns across quantiles, provid ing insights into the distributional properties that are critical in high-volatility conditions. The robustness of QFM is compared to simpler models, including the Random Walk (RW) model, with predictive performance evaluated using Mean Squared Error (MSE) and Density Forecast scores. Results show that while the QFM provides valuable distributional insights across the distribution, the RW model ultimately outperforms the QFM in predictive accu racy, underscoring the persistent challenge of forecasting exchange rate movements even with advanced quantile-based methods, highlighting the unpredictability of exchange rates and confirming the validity of Efficient Market Hypothesis. This study contributes to financial econometrics by demonstrating the practical limita tions of complex factor models and emphasizing the continued effectiveness of simpler, basic forecasting models such as Random Walk. Future research should focus on better understanding the relationship between exchange rates and macroeconomic variables and which are the main drivers for movements, especially in the short to medium term forecast ing horizonFile | Dimensione | Formato | |
---|---|---|---|
Stortoni_Alessandro.pdf
accesso riservato
Dimensione
3.12 MB
Formato
Adobe PDF
|
3.12 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/78450