| Standard Industrial Classification (SIC) codes are one of the most widely employed methods used to categorize firms into industry groups. In this dissertation, I criticize the prevailing assumption that assignments of SIC codes to organizations are simple reflections of coded organizations' economic activities and argue that coding outcomes are contingent on social contexts of coders and coded organizations. In order to do so, I explore three factors of SIC coding, namely, the coding scheme, coders, and coded organizations. With respect to the coding scheme, I suggest that the imperfectness of the SIC coding scheme often generates situations in which two or more SIC codes fit coded organizations to a similar degree, based on the literature review. Regarding coders, I argue that low intercoder reliability should be attributed to different social contexts faced by each coder based on the production of culture perspective. With respect to coded organizations, I draw on the neoinstitutional theory and argue that value associations of SIC codes due to institutionalized usages among various social actors stimulate coded organizations to obtain SIC codes that confer more favorable images. At last, combining the two observations on coders and coded organizations based on the resource dependence theory, I argue that absence or presence of exchange relationships between coders and coded organizations influence the coding outcomes. In particular, comparisons between two coders, the Securities and Exchange Commission (SEC) and Dun & Bradstreet (D&B), based on how they coded 3,083 public firms between 1995 and 1999 show that having exchange relationships with coded organizations is positively associated with better reflecting coded organizations' preferences on SIC codes. |
