Policies that have been developed to counteract corruption inevitably make a number of assumptions about corruption’s causes. Indeed, in many cases, these assumptions are quite clear; people are rational, self-interested creatures, who, if the projected benefits outweigh the costs, should be expected to indulge in corrupt activity. These assumptions have led rational-actor and rent-seeking models to dominate much of the thinking on corruption’s causes. Yet they clearly are not the only ways of thinking about this. Different disciplines approach the challenge of working out what precisely causes corruption in strikingly different ways. This has led to the development of a body of work that questions the dominance of these rational-choice approaches. This chapter unpacks both the dominant thinking in the field and the revisionist positions of its critics.

One thing that everyone can agree on is that understanding how to tackle something is much easier if you understand why it exists in the first place. Doing that, of course, is unfortunately not quite as straightforward as it might appear. There are often many factors that contribute to any given outcome, some much more obvious (and easily identifiable) than others. Assumptions that initially seem plausible can often turn out to be far off the mark. This is not just a problem for those looking to understand what drives corruption; pinning down causal effects in the social world more broadly is far from simple. There may well be strong statistical relationships between phenomena, but that does not mean that A is necessarily causing B. Between 1999 and 2010, for example, there was a strikingly strong correlation between the number of people in the US who drowned after falling out of a fishing boat and the ups and downs of the marriage rate in Kentucky. Between 2000 and 2009, the relationship between per capita consumption of mozzarella cheese and the number of civil engineering doctorates awarded was also remarkably strong and consistent. Yet no one has (yet) come anywhere near to explaining how the relationships between these phenomena might actually function in practice; there are, in other words, plenty of spurious correlations that exist by chance and not by design.