Large-N studies reveal a strong, negative relationship between income-based measures of the middle class and perceptions of grand corruption. Yet significant cross-country variation exists, even when accounting for regime type. Why are some countries with large middle classes nonetheless perceived to be persistently corrupt? This paper argues that the effect of middle-class size on grand corruption is moderated by the strength of horizontal accountability institutions. Middle-income citizens are more likely than their poor and vulnerable counterparts to learn about corruption, to care about its prevalence, and to punish wrongdoing through the ballot box and non-electoral sanctions—but their potential to bring about change depends on their ability to form virtuous cycles with institutions tasked with constraining the abuse of entrusted power for private gain. The theory is tested on a cross-national dataset of 132 countries. The results lend empirical support to the existence of a virtuous cycle between socio-economic change (and in particular the expansion of the middle-income segment of the population) and institutional responses to prevailing levels of corruption.