Application of ARIMA Models in the Analysis of Economic Time Series

Xinyue Zhang

Bachelor's thesis

Anotácia:

This thesis explores the application of ARIMA models in forecasting key economic indicators in China, including GDP growth, inflation (CPI), and unemployment. Through empirical modeling based on historical data, the study evaluates the effectiveness, limitations, and predictive performance of various ARIMA configurations. Results indicate that while ARIMA models like ARIMA(3,1,2) and ARIMA(2,1,3) can provide reliable short-term forecasts under stable conditions, their performance weakens during structural shifts or external shocks. The study also conducts joint analysis of CPI and GDP to assess economic trends and highlights the policy implications of predicted recession and deflation risks in the coming years. By incorporating VARIMA extensions, the research further examines the interaction between variables, revealing the potential of multivariate approaches in enhancing forecast accuracy. In-sample and out-of-sample tests show that model performance varies depending on data volatility and the presence of exogenous shocks. The thesis emphasizes that although ARIMA remains valuable for transparent and interpretable forecasts, its limitations in handling nonlinearity and external influences should not be overlooked. These findings provide insight for economists and policymakers seeking to apply time series models in practice and suggest directions for further development using integrated or hybrid methods.

Abstract:

This thesis explores the application of ARIMA models in forecasting key economic indicators in China, including GDP growth, inflation (CPI), and unemployment. Through empirical modeling based on historical data, the study evaluates the effectiveness, limitations, and predictive performance of various ARIMA configurations. Results indicate that while ARIMA models like ARIMA(3,1,2) and ARIMA(2,1,3) can provide reliable short-term forecasts under stable conditions, their performance weakens during structural shifts or external shocks. The study also conducts joint analysis of CPI and GDP to assess economic trends and highlights the policy implications of predicted recession and deflation risks in the coming years. By incorporating VARIMA extensions, the research further examines the interaction between variables, revealing the potential of multivariate approaches in enhancing forecast accuracy. In-sample and out-of-sample tests show that model performance varies depending on data volatility and the presence of exogenous shocks. The thesis emphasizes that although ARIMA remains valuable for transparent and interpretable forecasts, its limitations in handling nonlinearity and external influences should not be overlooked. These findings provide insight for economists and policymakers seeking to apply time series models in practice and suggest directions for further development using integrated or hybrid methods.

Keywords

ARIMA, Time series, Economic forecasting, Model evaluation, Stationarity

Jazyk práce: English

Datum vytvoření / odevzdání či podání práce: 30. 7. 2025

Obhajoba závěrečné práce

• Obhajoba proběhla 18. 9. 2025
• Vedúci: Ing. Šárka Čížková, Ph.D.
• Oponent: doc. Mgr. Mária Králová, Ph.D.

AMBIS University

AMBIS University

Bachelor programme / odbor:
Business Economics and Management / Business Economics and Management