Modeling vector autoregressive and autoregressive distributed lag of the beef and chicken meat prices during the COVID-19 pandemic in Indonesia

ABSTRACT

The impact of the COVID-19 pandemic has spread to all aspects of life. Modeling the price of beef and chicken meat is very important for the government to avoid extreme fluctuations of both commodities in the prices so that society's purchasing power can be maintained. This study has several objectives, namely building VAR and ARDL models from multiple time series data (beef and chicken meat prices), conducting variable selection with forwarding subset selection on input lag in the ARDL model, and measuring the performance of the VAR and ARDL models on the both of beef and chicken meat prices based on the value of RMSE, MAE, and R_square both in the training and testing set. The novelty in this study is to propose an identification method for the lag inputs of the ARDL model based on the criteria of both the Alkaide Information criteria (AIC) value and the adjusted R square value by visualizing both criteria for all possible amounts of lag inputs. The results of the identification of the VAR model structure using the conventional method in time series modeling are yielded the different lag inputs that are compared to the ARDL model structure with lag inputs identified by using the proposed method. The ARDL model of the beef and chicken meat prices has better performance than the VAR model both on training and testing sets. In addition, the resulting VAR model also clearly shows the occurrence of overfitting problems.