The aim is to both develop a new theory-based framework for analysing cost overruns, and to use this for an empirical study of cost overruns in infrastructure projects in Sweden. The conceptual part is based on a literature review and the empirical part is primarily based on a questionnaire to project managers. The framework has a descriptive part comprising two dimensions: when, during the process, the cost overrun arose and what part of the cost function was responsible: change in the product, change in quantities of the inputs and change in price of inputs. The explanatory part is a development of Flyvbjerg's theories and identifies four possible explanations: political/strategic aspects, psychological aspects, competence-related aspects and bad luck. The result from the empirical study is that most cost overruns occur in the initiation and planning stages up to the final design and are related to design changes and increases in the amount of inputs needed because of technical and administrative problems. Of the explanatory factors, there is most support for lack of competence and optimism bias.