Recent scholarly and expert attention has called for more scrutiny of the failure of infrastructure public private partnerships (PPPs). Yet, there is still uncertainty about how we can measure failure and very little systematic evidence about its occurrence. We employ a mixedmethods design to explore the notion of failure and to identify the conditions under which PPPs fail. A first phase, based on documentary analysis and semi-structured interviews with PPP experts, sought to understand failure as project cancellation. A second phase, based on observational data analysis, assessed whether PPPs are less likely to be cancelled in countries with more veto points, which check the ability of politicians to intervene in the implementation of public policy. We find robust support for this claim using regression models and a unique World Bank dataset capturing the provisions of over 4,000 infrastructure PPPs funded between 1990-2015 in 89 countries.