Evidence on economies of scale in software development

Researchers and practitioners have found it useful for cost estimation and productivity evaluation purposes to think of software development as an economic production process, whereby inputs, most notably the effort of systems development professionals, are converted into outputs (systems deliverables), often measured as the size of the delivered system. One central issue in developing such models is how to describe the production relationship between the inputs and outputs. In particular, there has been much discussion about the existence of either increasing or decreasing returns to scale. The presence or absence of scale economies at a given size are important to commercial practice in that they influence productivity. A project manager can use this knowledge to scale future projects so as to maximiT~ the productivity of software development effort. The question of whether the software development production process should be modelled with a non-linear model is the subject of some recent controversy. This paper examines the issue of non-linearities through the analysis of 11 datasets using, in addition to standard parametric tests, new statistical tests with the non-parametric Data Envelopment Analysis (DEA) methodology. Results of this analysis support the hypothesis of significant non-linearities, and the existence of both economies and diseconomies of scale in software development.