View article

This paper presents an empirical study on the production of open source software, based on a panel of 10,553 projects registered on SourceForge over a period of 28 months (February 2005 until May 2007). We use a flexible Translog specification to estimate a production function relating the number of program updates with the number of corporate and voluntary contributors, taking into account the spillovers flowing from other projects. We find that corporate developers are the more productive ones, but that associating them with other developers in a project entails inefficiencies. We also find evidence of non-decreasing returns to scale in open source projects, thus suggesting substantial efficiency gains of division of labor in large open source projects. Our empirical analysis finally highlights a substantial impact of spillovers on productivity. Spillovers mainly benefit mature projects. They especially flow between projects with the same topic and/or programming language, and have a stronger impact in projects involving corporate developers.