هزینه های معاملات، تجربه صنعت و تصمیم گیری درباره خرید یا ساخت در جمعیتی از شرکت های خودرو سازی ایالات متحده

We study the determinants of make-or-buy decisions for engines made by every recorded U.S. auto firm during 1917–1933. Most make-or-buy studies testing predictions of transaction cost economics exclude smaller firms, even though their make-or-buy behavior might be different from large firms’ due to capital constraints or other factors. We find that transaction cost effects on make-or-buy choices were nevertheless important at the population level. We also find that firm experience in the industry, as well as pre-entry experience, significantly affected make-buy choices. These experience effects may reflect the interaction of mechanisms emphasized in evolutionary and transaction cost theories.

Many empirical studies of make-or-buy decisions testing transaction cost economics (TCE) have found support for the theory's main hypotheses. Transactions characterized by asset specificity, for example, tend to be internalized, whereas transactions lacking this characteristic tend to be carried out through market contracts (Williamson, 1975 and Williamson, 1985). TCE-inspired empirical studies of phenomena other than make-or-buy choices per se have similarly found an impressive amount of evidence supporting its main prescriptions (for surveys, see Shelanski and Klein, 1995, Klein, 2005 and Macher and Richman, 2006).

This paper tests transaction cost hypotheses about the determinants of make-or-buy decisions on data drawn from the entire population of U.S. auto firms during 1917–1933, as well as a hypothesis about the effect of industry experience on such decisions. We examine the factors that affected firms’ decisions to make or buy the engines for each of their car models and find that firms tended to produce unique engines internally while sourcing standard engines from independent suppliers. Consistent with TCE, we ascribe this finding to the greater asset specificity associated with manufacturing unique engines. We also find that, again consistent with TCE, component subsystems that shared more complex mechanical interfaces, presumably because they require more intensive coordination and therefore asset specificity, also tended to be produced internally, whereas those with less complex interfaces tended to be sourced externally.