인터넷 검색지수는 농식품 구매를 선행하는가?

Abstract

The purpose of this paper is to demonstrate the usefulness of Internet search data in the rapid prediction of food consumption at the retail stage. For 30 agricultural food products, which are the very topics in the KREI's Agricultural Outlook, we collected the internet search index provided by Naver Data Lab and the consumers' daily purchasing costs obtained from the Agri-Food Consumers Panel Survey conducted by the Korea RDA. We utilized time-series analysis such as VAR model, Granger Causality test and Impulse Response Function, and verified whether the Internet search index of each item preceded its purchases. According to the empirical results, 24 out of the total 30 items showed that their search index and purchase amounts were inter-influential or that the search index influenced only their costs of purchasing items. It was also found that the more they searched the Internet for information on an item, the higher the probability was of buying the item in about 5 to 6 days. For 80% of the items, past Internet search activities had positive and negative effects on its current purchases. Considering that the data in the consumption outlook was is not sufficient to identify and predict real-time food consumption trends, the Internet search index can be an effective alternative at low cost.