An Analysis of Demand in the Korean Automobile Industry

Empirical analysis of market demand inherits several challenging issues. Researchers are frequently under faced with the problem of deficiency in data, and undertaking estimation of market demand requires a precise theoretical model that can represent consumers’ behavior which is, in many cases, rather unknown. Based on previous literature that models an oligopolistic market with differentiated products, this paper takes advantage of the data provided by manufacturers in Korean automobile industry rather than using data of registered cars provided by Korea Automobile Manufacturers Associations (KAMA), to deepen understanding of demand structure of Korean automobile industry. Information of demand structure is a main key element of understanding any market. Firms establish their own strategies for marketing, production, profit maximization, and cost minimization based on information about the expected consumer behaviors induced by firms’ actions. Government, before enforcing any regulations such as tax and bonuses, has to consider the range and limit of their policies based on information about how elastic targeted agents response. Precise expectation on changes in consumer behavior due to firm’s or government’s actions requires utilizing complete information of each consumer’s preference, but this is, in most time, impossible. Hence, economists have tried to make inference based on statistical analysis by using data constituting main features of a market such as price, quantity, income, and others. In this sense, this paper aims to contribute to identification of demand structure of Korean automobile market by a rigorous empirical analysis. This paper follows the seminal paper of Berry, Levinsohn, and Pakes (1995), and is also closely related to Park and Cho (2005) and Park (2005) that have set a foundation of empirical analysis of Korean automobile industry based on discrete choice model. Section 2 describes the model this paper adopts. In section 3, we introduce basic statistical properties of the dataset, and then estimate the discrete choice model of our focus. In section 4, we treat various alternative specifications to test the robustness of our main model. Section 5 concludes.