Comparison of Transfer Function Model and ARIMA-GARCH on Daily Stock Data in Agribusiness and Trade Sector

Abstract

Shares are one of the long-term financial instruments traded in the capital market and are one of the popular investment alternatives for investors in Indonesia. One of the goals of investors investing in a company is to get a profit (return). The higher the stock's selling price is above the purchase price, the higher the return that investors will get. Stock data is time-series data. Therefore, time-series data modeling is needed to determine when investors will get returns shared. One of the models used for time-series data is the ARIMA model. This model assumes that the data's volatility (rate of fluctuation) is constant. In financial data, there are many cases of data with non-constant volatility. This non-constant volatility will cause heteroscedastic problems in the data. Therefore we need a model that can accommodate heteroscedastic problems in the data. One model used is the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model. This study also examines the effect of independent variables ( input series ), namely world crude oil prices, the positive number of Covid-19, the dollar exchange rate against the rupiah, and the Shanghai composite index on stock return time-series data ( output series ) in the agribusiness and trade sectors. One of the models used to analyze the effect of input series variables on time series data is the transfer function model. The data used is from March 2, 2020, to June 30, 2021. These two models are compared to find out which time series forecast is better. The results showed that the transfer function method is a better method used for forecasting the next seven days on the stock return data of PT. Sawit Sumber Mas Sarana Tbk and PT. Astra Internasional Tbk compared to the ARIMA-GARCH model.