The Structure of Adjustment Costs in Mainframe Computer Investment

This paper provides new empirical insights into the source of investment adjustment costs. In contrast to prior models of investment that gave only ad hoc treatment to the concept of adjustment costs, my analysis pays serious micro-level attention to investment adjustment processes and their dynamics. I construct a structural dynamic model that predicts the lumpy and episodic patterns of mainframe computer investment in the micro-level data and estimate it using a dynamic programming technique. The model identifies the size of implicit lump-sum adjustment costs from the investment observations. Furthermore, the paper investigates how lump-sum adjustment costs differ across organizations. The empirical results show that investment adjustment costs arise due to the specificity of the complementary assets. Adjustment costs are higher when computer applications are more firm-specific. The estimates show, however, the main costs of this firm-specificity are not the engineering costs of producing in-house application software. Instead, the results point to the difficulty of changing users' work routines and coordinating tasks as the source of lump-sum adjustment costs. The idiosyncrasies in firms' internal transactions and tacit organizational knowledge constrain the ability of engineers to promote common, standardized applications. Hence, they also limit firms' ability to economize the computer development efforts by delegating tasks to external markets. The findings of this paper substantiate the importance of internal organization in understanding the source of investment adjustment costs.