In 1986, Mosely first drew attention to an apparent paradox in the performance of international aid. Microeconomic data from evaluations of aid financed projects showed a majority of projects were successful, whereas macroeconomic data from regressions of aid on growth were discouraging. The paradox, if real, implied that the aggregate impact of aid was less the sum of its parts. Mosely asked whether the paradox was real of whether the “data deceived.” This question, which has come to be equated with the issue of whether aid works, has been the subject of numerous cross‐country regressions to test whether aid has an impact on growth (or related variables). But the regression results have been inconclusive, and the methodology has come under attack. Evidence from case studies offers an alternative test. One prominent case study approach is that of Picciotto (2009), which claims to find strong evidence for the existence of the paradox, namely the fact that one third of World Bank country assistance program evaluations show success at the project (micro) level but not at the country (macro) level. This paper re‐evaluates Piciotto’s claimed findings. Only about one‐third of the disconnects survive critical scrutiny, and the source of these remaining disconnects has nothing to do with negative effects of aggregate aid. Although in the Picciotto case, the data do indeed deceive, we conclude that country‐level aid studies are nevertheless a useful tool for donors to use to guard against possible, albeit uncertain, negative impacts of aid at the country level.