실시간 항만물동량 데이터를 이용한 GDP Nowcasting = GDP Nowcasting Using Real-Time Port Traffic Data

다국어 초록 (Multilingual Abstract)

Purpose This study aims to improve the timeliness and accuracy of GDP nowcasting by incorporating real-time port throughput data into a Dynamic Factor Model (DFM). Recognizing that the official GDP figures are released with a considerable time lag, this research explores whether high-frequency port activity, which reflects a trade-dependent economy, can enhance short-term GDP predictions.


Design/methodology/approach A state-space DFM was employed to estimate quarterly GDP using mixed-frequency data (monthly and quarterly) from 2010 to 2020. To address missing values and differing release schedules, the model was estimated via the EM (Expectation-Maximization) algorithm combined with Kalman filtering. Conventional macroeconomic indicators were grouped into four blocks (global, real, soft, labor), and real-time port throughput data were added as an additional block. Model performance was then evaluated by comparing the predicted GDP values with actual outcomes based on RMSE (Root Mean Squared Error) and MAE (Mean Absolute Error).


Findings Empirical results indicate that including port throughput information reduces nowcasting errors relative to a DFM using only traditional macroeconomic variables. Although the improvement in RMSE and MAE is modest, it suggests that real-time port activity data, reflecting rapid shifts in trade flows, can enhance the early detection of economic turning points and strengthen short-term policy responses.

다국어 초록 (Multilingual Abstract)

Purpose This study aims to improve the timeliness and accuracy of GDP nowcasting by incorporating real-time port throughput data into a Dynamic Factor Model (DFM). Recognizing that the official GDP figures are released with a considerable time lag, this research explores whether high-frequency port activity, which reflects a trade-dependent economy, can enhance short-term GDP predictions.




Design/methodology/approach A state-space DFM was employed to estimate quarterly GDP using mixed-frequency data (monthly and quarterly) from 2010 to 2020. To address missing values and differing release schedules, the model was estimated via the EM (Expectation-Maximization) algorithm combined with Kalman filtering. Conventional macroeconomic indicators were grouped into four blocks (global, real, soft, labor), and real-time port throughput data were added as an additional block. Model performance was then evaluated by comparing the predicted GDP values with actual outcomes based on RMSE (Root Mean Squared Error) and MAE (Mean Absolute Error).




Findings Empirical results indicate that including port throughput information reduces nowcasting errors relative to a DFM using only traditional macroeconomic variables. Although the improvement in RMSE and MAE is modest, it suggests that real-time port activity data, reflecting rapid shifts in trade flows, can enhance the early detection of economic turning points and strengthen short-term policy responses.