Forecasting Exchange Rates Using the Kalman Filter with Oil Price as Exogenous Input: Evidence from CNY/SEK and USD/NOK
(2025) NEKN01 20251Department of Economics
- Abstract (Swedish)
- In this paper, we investigate the prediction performance of several models across different
forecasts horizons for the CNY/SEK and USD/SEK exchange rates, with a focus on
incorporating macroeconomic inputs. Over the course of time, the Kalman filter method has
been widely applied in the exchange rate forecasting literature, while the predictive ability of
the Box-Jenkins Transfer Function (BJ) model remains largely unexplored in this area. We
propose incorporating oil prices into the BJ model and further refining the model with the
Kalman filter approach. For comparison, we include the random walk model as the
benchmark and ARMA model to analyze the dynamics of the exchange rate itself. We then
introduce oil prices as an exogenous... (More) - In this paper, we investigate the prediction performance of several models across different
forecasts horizons for the CNY/SEK and USD/SEK exchange rates, with a focus on
incorporating macroeconomic inputs. Over the course of time, the Kalman filter method has
been widely applied in the exchange rate forecasting literature, while the predictive ability of
the Box-Jenkins Transfer Function (BJ) model remains largely unexplored in this area. We
propose incorporating oil prices into the BJ model and further refining the model with the
Kalman filter approach. For comparison, we include the random walk model as the
benchmark and ARMA model to analyze the dynamics of the exchange rate itself. We then
introduce oil prices as an exogenous variable to extend the model into the BJ framework and
further enhance its forecasts performance using the Kalman filter. Through out-of-sample
forecasts evaluation of two different exchange rates (across both major oil exporting
countries and not), we demonstrate that the Kalman filter approach consistently outperforms
all other models in this study, including the random walk model that is hard to beat. This
finding is applicable even for the high volatile currency pair such as USD/NOK, where the
Kalman filter significantly improved the forecasts accuracy and archives dramatically lower
RMSE and MAE compared to the random walk model. Our study fills the gap in the existing
literature by incorporating oil prices as an input to the BJ model and refining it with the
Kalman filter to forecast exchange rates. The proposed approach offers practical value for
both exchange rate risk management and macroeconomic policy formulation. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9201742
- author
- Liu, Tiangang LU and Huang, Yuanyuan LU
- supervisor
- organization
- course
- NEKN01 20251
- year
- 2025
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Exchange rate forecasting, Random walk, ARMA, BJ model, Kalman filter.
- language
- English
- id
- 9201742
- date added to LUP
- 2025-09-12 10:00:02
- date last changed
- 2025-09-12 10:00:02
@misc{9201742, abstract = {{In this paper, we investigate the prediction performance of several models across different forecasts horizons for the CNY/SEK and USD/SEK exchange rates, with a focus on incorporating macroeconomic inputs. Over the course of time, the Kalman filter method has been widely applied in the exchange rate forecasting literature, while the predictive ability of the Box-Jenkins Transfer Function (BJ) model remains largely unexplored in this area. We propose incorporating oil prices into the BJ model and further refining the model with the Kalman filter approach. For comparison, we include the random walk model as the benchmark and ARMA model to analyze the dynamics of the exchange rate itself. We then introduce oil prices as an exogenous variable to extend the model into the BJ framework and further enhance its forecasts performance using the Kalman filter. Through out-of-sample forecasts evaluation of two different exchange rates (across both major oil exporting countries and not), we demonstrate that the Kalman filter approach consistently outperforms all other models in this study, including the random walk model that is hard to beat. This finding is applicable even for the high volatile currency pair such as USD/NOK, where the Kalman filter significantly improved the forecasts accuracy and archives dramatically lower RMSE and MAE compared to the random walk model. Our study fills the gap in the existing literature by incorporating oil prices as an input to the BJ model and refining it with the Kalman filter to forecast exchange rates. The proposed approach offers practical value for both exchange rate risk management and macroeconomic policy formulation.}}, author = {{Liu, Tiangang and Huang, Yuanyuan}}, language = {{eng}}, note = {{Student Paper}}, title = {{Forecasting Exchange Rates Using the Kalman Filter with Oil Price as Exogenous Input: Evidence from CNY/SEK and USD/NOK}}, year = {{2025}}, }