Interpretation problems related to the use of regression models to decide on economy of scale in software development

Many research studies report an economy of scale in software development, i.e., an increase in productivity with increasing project size. Several software practitioners seem, on the other hand, to believe in a diseconomy of scale, i.e., a decrease in productivity with increasing project size. In this paper we argue that violations of essential regression model assumptions in the research studies to a large extent may explain this disagreement. Particularly illustrating is the finding that the use of the production function (Size = a·Effortb), instead of the factor input model (Effort = a·Sizeb), would most likely have led to the opposite result, i.e., a tendency towards reporting diseconomy of scale in the research studies. We conclude that there are good reasons to warn against the use of regression analysis parameters to investigate economies of scale and to look for other analysis methods when studying economy of scale in software development contexts.

► Diverging opinions on economy or diseconomy of scale.
► Production function models show economy of scale, while factor input models show the opposite.
► Belief in economy of scale among researchers a result of models violating of regression model assumptions.
► We need other methods than regression analysis for analysis of economies of scale in software development.