Forecasting the RMB/USD Exchange Rate Using an ARIMA Model: Evidence from 2015 to 2025

Department of Statistics, University of Wisconsin–Madison, Madison, WI 53706, USA

^* Corresponding author: zwang3479@wisc.edu

Abstract

This research applies an ARIMA (2,0,0) model to examine and predict the monthly RMB/USD exchange rate trends over the period from April 2015 to April 2025, aiming to explore how fluctuations in the exchange rate influence China’s monetary strategies and international trade decision-making. The raw daily data are first integrated into monthly frequencies and processed through unit root tests and smoothing to ensure that the series meet the modeling requirements. The model selection process takes into account the goodness-of-fit and complexity, and ARIMA (2,0,0) is finally identified as the optimal solution, and the data from 2015 to 2022 are used as the training set, and 2023 to 2025 are used as the test set. The residual diagnostics show that the model error is consistent with white noise characteristics, and the Ljung-Box test shows no significant autocorrelation, indicating a reasonable fit. In the out-of-sample forecasts, the average absolute percentage error of the model is 7.53%, showing a moderate level of forecasting accuracy. Despite the slight underestimation of the upward trend of the exchange rate after 2022, the overall forecasting ability is still of practical value.