Firms' use of external labor arrangements (ELAs), such as temporary, contract and seasonal workers, has become increasingly prevalent over the last two decades. Despite the increasing importance of this phenomenon, little is known about firms' reasons for using ELAs. Most research to date has been exploratory, using qualitative methods or archival data not well suited to the constructs. The result of this research has been a long and often contradictory list of proposed antecedents of ELA use. In this study, I tested the ability of the transaction costs theory to predict when firms will fill a given job using an ELA rather that a permanent employment relationship. According to this theory, three characteristics of the job will determine whether the job will be filled using an ELA: transaction-specific investment, likelihood of repetition, and uncertainty of performance. Firms will be less likely to staff a given job using an ELA when the job requires investment in idiosyncratic skills, when the firm is likely to require a person with that set of skills regularly, and when performance in that job is difficult to measure.