Utilizing Univariate and Multivariate Econometric Techniques to Forecast United States Corn Prices
Lindholm, Christopher (2023)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20231207152132
https://urn.fi/URN:NBN:fi-fe20231207152132
Tiivistelmä
For several hundred years, corn has been an important global commodity with a diverse range of uses. As a result, a multitude of stakeholders rely on accurate price forecasts to make informed decisions. This thesis investigates the efficiency and accuracy of two varying econometric forecasting techniques when applied to U.S. corn prices. A univariate Seasonal Autoregressive Integrated Moving Average (SARIMA) model, and a multivariate Vector Error Correction Model (VECM). This study conducts a comparative analysis of the differing econometric techniques in order to further the current understanding of U.S. corn price forecasting.
The methodology involves the collection of historical U.S. corn prices, which are utilized in both models. In addition to monthly U.S. corn price data, the VECM is constructed using monthly U.S. soybean prices, U.S. wheat prices, West Texas Intermediate (WTI) crude oil prices, and U.S corn production data. Following data collection and data preprocessing, model selection is conducted. Various pre- and post-estimation diagnostic tests are undertaken in order to ensure the validity and stability of the chosen models. An out-of-sample forecast, ranging from January 2021 through June 2023, is conducted by both models and the subsequent results are analyzed using various metrics, including Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and Theil’s U1 statistic. Finally, a future forecast is produced by both models, ranging from July 2023 to July 2026. The results of the future forecasts are compared to U.S. corn price forecasts released by the United States Department of Agriculture (USDA).
The preliminary results indicate that the SARIMA model performs better than the VECM in the out-of-sample forecast. However, both models fail to accurately forecast the price development that occurred in the U.S. corn market during the out-of-sample forecast horizon. The relatively poor performance of both models is likely due to the drastic increases observed as a result of a multitude of factors, including the effects of the COVID-19 pandemic, unprecedented export demand from China, and adverse growing conditions. The future forecast provides different results, as the VECM appears to outperform the SARIMA model, when the outputs are compared to the forecasts released by the USDA.
This research has implications for relevant stakeholders in the agricultural sector, including farmers, traders, and policymakers, providing a more comprehensive understanding of corn price forecasting. Additionally, the comparative analysis of the VECM and SARIMA model contributes to the current literature of time series analysis, and the ongoing discussion surrounding univariate and multivariate forecasting techniques.
The methodology involves the collection of historical U.S. corn prices, which are utilized in both models. In addition to monthly U.S. corn price data, the VECM is constructed using monthly U.S. soybean prices, U.S. wheat prices, West Texas Intermediate (WTI) crude oil prices, and U.S corn production data. Following data collection and data preprocessing, model selection is conducted. Various pre- and post-estimation diagnostic tests are undertaken in order to ensure the validity and stability of the chosen models. An out-of-sample forecast, ranging from January 2021 through June 2023, is conducted by both models and the subsequent results are analyzed using various metrics, including Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and Theil’s U1 statistic. Finally, a future forecast is produced by both models, ranging from July 2023 to July 2026. The results of the future forecasts are compared to U.S. corn price forecasts released by the United States Department of Agriculture (USDA).
The preliminary results indicate that the SARIMA model performs better than the VECM in the out-of-sample forecast. However, both models fail to accurately forecast the price development that occurred in the U.S. corn market during the out-of-sample forecast horizon. The relatively poor performance of both models is likely due to the drastic increases observed as a result of a multitude of factors, including the effects of the COVID-19 pandemic, unprecedented export demand from China, and adverse growing conditions. The future forecast provides different results, as the VECM appears to outperform the SARIMA model, when the outputs are compared to the forecasts released by the USDA.
This research has implications for relevant stakeholders in the agricultural sector, including farmers, traders, and policymakers, providing a more comprehensive understanding of corn price forecasting. Additionally, the comparative analysis of the VECM and SARIMA model contributes to the current literature of time series analysis, and the ongoing discussion surrounding univariate and multivariate forecasting techniques.