Social Capital and Contract Duration in Buyer-Supplier Networks for Information Technology Outsourcing

This paper presents new evidence on the role of embeddedness in predicting contract duration in the context of Information Technology (IT) Outsourcing. Contract duration is a strategic decision that aligns interests of clients and vendors, providing the benefits of business continuity to clients and incentives to undertake relationship specific investments for vendors. Considering the salience of this phenomenon, there has been limited empirical scrutiny into how contract duration is awarded. We posit that clients and vendors obtain two benefits from being embedded in an inter-organizational network. First, the learning and experience accumulated from being embedded in client-vendor network could mitigate the challenges in managing longer-term contracts. Second, the network serves as a reputation system that can stratify vendors according to their trustworthiness and reliability, which is important in longer-term arrangements. In particular, we attempt to make a substantive contribution in theorizing about embeddedness at four distinct levels: structural embeddedness at the node level, relational embeddedness at the dyad level, contractual embeddedness at the level of a neighborhood of contracts and finally, positional embeddedness at the level of the entire network. We analyze a dataset of 22039 outsourcing contracts implemented between 1989 and 2008. We find that contract duration is indeed associated with structural and positional embeddedness of participant firms, with the relational embeddedness of the buyer-seller dyad and with the duration of other contracts to which it is connected through common firms. Given the nature of our data, identification using traditional OLS based approaches is difficult given the unobserved errors being clustered along two non-nested dimensions and the autocorrelation in a firm’s decision (here the contract) with those of contracts in its reference group. We employ a multi-way cluster robust estimation and a network auto-regressive estimation to address these issues. Implications for literature and practice are discussed.