This paper proceeds from the more recent literature on corruption as a principal-agent problem whose significance for development depends on dimensions related to the nature of the corrupt transaction itself, such as distinctions based on the agents involved, scale, type of deal, predictability, industrial organization, etc., all of which affect, for better or worse, the nature of the relationship between the principal (as represented by the public interest) and the agent (politicians and bureaucrats). Corruption may thus be regarded as an incentive-design problem. The paper also argues, however, that there is a larger dimension to corruption determined by the historical and social context. Here, the ultimate factors are those affecting social cohesion (e.g., income and wealth, education, ethnic and other differences), the economic strategies pursued by the government (e.g., minimalist versus interventionist), the political system (the autonomy of the bureaucracy, the degree of centralization), the extent of market transactions (local, global) and the rate and sources of economic growth. It is these factors that determine the credibility of the formal institutional constraints (however designed) on the behavior of public officials and private agents alike.