FORECASTING INTEREST RATE OF CHINA USING UNIVARIATE, MULTIVARIATE AND COMBINATION TECHNIQUE

Abstract

Accurate interest rate forecasting is essential for monetary policy formulation, financial stability, and investment decision-making, particularly in emerging economies such as China, where market dynamics are shaped by both structural reforms and global economic shifts. This study evaluates the predictive performance of univariate models (ARIMA, naïve), multivariate models (Nonlinear Autoregressive Distributed Lag (NARDL), Vector Error Correction Model (VECM), and hybrid forecast combination techniques in modeling China’s real interest rate. Using macroeconomic variables including inflation, GDP growth, exchange rate, trade indicators, oil prices, and money supply, the research applies rigorous econometric diagnostics to ensure model validity. Findings indicate that while individual multivariate approaches, particularly NARDL, effectively capture nonlinearities and long-run relationships, forecast combination strategies consistently enhance predictive accuracy and stability (Timmermann, 2022; Yang et al., 2023). Notably, model integrations such as ARIMA + NARDL and NARDL + VECM outperform standalone models by leveraging complementary strengths, thereby reducing model-specific biases (Ai et al., 2023; Li & Chen, 2024). This supports the growing consensus that hybrid methodologies are more resilient in volatile and policy-sensitive economic environments (Wang et al., 2023). The study contributes to the literature on interest rate forecasting in emerging markets by demonstrating that model combinations yield superior performance under uncertainty, offering valuable insights for policymakers, investors, and financial institutions seeking robust forecasting frameworks in dynamic macroeconomic contexts.

Keywords: Univariate, Multivariate, combination forecasting technique, ARIMA, NARDL, VECM