Let me admit right up front that this is going to be kind of a nerdy post, focused mostly (though not entirely) on questions of terminology.  But there’s a particular meme that seems to have emerged in much of the discussion of anticorruption strategy, which I think is just wrongheaded and misleading.  The meme goes something like this:

“Although some people describe corruption as a principal-agent problem, corruption is actually a collective action problem.”

(I hate to point fingers, but just so you know I’m not making this up, examples of this meme appear here, here, here, here, here, here, here, and here.  Most depressingly, the not-very-good Wikipedia entry on corruption also includes this claim.)

This statement uses sophisticated-sounding social science jargon, but to anyone who knows what the terms “principal-agent problem” and “collective action problem” actually mean, in their technical economic context, it doesn’t make much sense–at least insofar as it implies that “principal-agent problem” and “collective action problem” are mutually exclusive alternatives. In fact, corruption is both a principal-agent problem (always) and a collective action problem (often).  Recognizing the latter claim need not and should not entail rejecting the former claim; the assumption that it does not only reflects conceptual confusion, but also entails the risk that we will neglect the many insights that principal-agent theory has to offer the study and practice of anticorruption.

I am certainly not the first to make this point. Heather Marquette and Caryn Peiffer have a very nice U4 Issue Paper developing many of the same themes (in greater depth). But because this has been bugging me for a while, let me offer my take on this issue:

A principal-agent problem is a problem in which one or more actors (individuals, organizations, or what have you)–the “agents”–are entrusted with authority to act on behalf of some other individual, group, or cause (the “principal”).  Depending on the context, the principal might be some specific identifiable actor or set of actors (like the shareholders of a corporation, or a bureaucratic supervisor), or the “principal” might be characterized as “society,” to capture the idea that certain agents (like political leaders and public servants) are supposed to act in the best interests of the community.  However, the principal(s) often cannot perfectly monitor and control the agents, which creates the risk that an agent may act in his/her/its own interest, rather than that of the principal(s).  A great deal of social science research–particularly in subfields like contract theory, industrial organization, and political economy–is devoted to studying the principal-agent problem, diagnosing its sources, and suggesting ways to mitigate the problem.

Corruption is clearly a principal-agent problem.  The most widely-used definition of corruption–the abuse of entrusted power for private gain–makes this abundantly clear, as the idea of “entrusted power” implies an agency relationship, and “abuse” implies that the agent is acting in ways that are not in the principal’s interests.  (Of course, while corruption is always a principal-agent problem, not all principal-agent problems necessarily involve corruption, unless “corruption” is defined very capaciously.)  Therefore, using a principal-agent framework to understand corruption is not only appropriate, but probably essential, and the insights derived from principal-agent theory more generally have a great deal of relevance to the study and practice of anticorruption.

A collective action problem involves a situation in which some group of actors would all be better off if they could all choose some action (like citizens contributing money to build a lighthouse, or workers joining a union, or criminal suspects refusing to talk to the police, or citizens and public servants refusing to take or pay bribes), but for any one individual, that individual would be better off “defecting” (failing to contribute to the lighthouse or union, blabbing to the cops to get a lighter sentence, engaging in bribery, etc.).  This is a problem because each individual’s choice of an action that is individually rational produces an outcome which is collectively inferior for the whole group.  In some settings characterized by collective action problems, individuals always have an incentive to defect no matter what everyone else does: in these “prisoners’ dilemma” or “free rider” situations, something else–coercion or special incentives or perhaps a non-instrumental sense of solidarity–is typically required to achieve collective action.  In other settings, both low and high levels of cooperation (or defection) may be self-reinforcing: if (almost) everyone else is “defecting” (bribing, failing to contribute to the common cause, etc.), then it might be rational for each individual also to defect, but if (almost) everyone else is “cooperating” (staying honest, contributing to the cause, etc.), then in this setting cooperation is actually preferred by (most) individuals.

Corruption can clearly also be a collective action problem–often of the second variety.  Even if everyone in a society dislikes corruption, no one individual may have an incentive to avoid corruption on his or her own, particularly if many other individuals are engaged in corrupt activity. The society may be “trapped” in a bad equilibrium.  Many implications flow from this, which I may discuss in a later post, but the basic idea is straightforward.

But–and this is where the meme I identified at the outset goes badly awry–there’s no reason that recognizing that corruption often has the characteristics of a collective action problem negates the fact that corruption is also a principal-agent problem.  Indeed, in principal-agent theory, a collective action problem on the part of the agents (or sometimes the principals, in those cases where the principal is a collective entity) is sometimes identified as the cause of the principal-agent problem, or at least a contributing factor.

The claim that corruption is not in fact a principal-agent problem (but rather is only a collective action problem) often depends on one or both of two associated claims, neither of which is right:

• The first is that, when corruption is systemic, everybody in society engages in corruption, so there is no identifiable “principal.” I don’t think that’s conceptually right, but we don’t need to go there, because it’s not empirically right either: there is no society where every person engages in corruption (or all types of corruption).
• Second, proponents of the corruption-isn’t-a-principal-agent-problem view often say something like, “The principal-agent framework assumes the existence of a fully benevolent principal who can solve the problem by putting in place the right institutions and incentives.”  No, it doesn’t — that assumption is sometimes made, either as a modeling convenience or because in a particular context it may be reasonable, but the theory does not require this. Indeed, often the root of the problem is that the principal (often conceived as the citizenry in general) is not able to fully control its agents, due to a range of problems — though certain interventions may help.

So part of the issue is simple confusion over terminology, but there’s also a more substantive concern.  The “corruption is a collective action problem, not a principal-agent problem” line often appears in contexts where the intended implication seems to be that various techniques that principal-agent theory has suggested for addressing agency problems are not relevant to anticorruption.  But that’s just wrong.  Of course, if one is facing a collective action problem–if a collective action problem is one of the reasons the principal-agent problem exists–then one must correctly diagnose that fact and respond accordingly.  But it would be foolish to put aside a whole range of tools and theories that might well be useful in addressing the problem because of a misguided notion that the presence of a “collective action problem” negates the relevance of principal-agent theory.