Last week, Professor Bo Rothstein, one of the most influential senior researchers in the anticorruption field, published a blog post entitled “Three Reasons Anti-Corruption Programs Fail.” The post (which draws from Professor Rothstein’s earlier writings and his new book) sets out to explain why the anticorruption efforts sponsored by a combination of domestic reformers and the international development community have been a “huge policy failure.” The three reasons for this purported failure laid out in the post are (1) use of the wrong definition of corruption (2) use of the wrong social science theory to frame and analyze corruption, and (3) locating corruption “in the wrong social spaces.”

I am disappointed to report that I find little in the post that is correct. Professor Rothstein’s post does illustrate some important and ongoing failures in anticorruption thinking—just not in the way that he intended. Rather, the post inadvertently illustrates certain tendencies that afflict a certain strain of academic work on the corruption topic—tendencies that render scholarship on corruption far less useful to the world than it could or should be. I’m particularly troubled when I encounter bright young up-and-coming researchers who appear to be misled by these tendencies. So with all due respect to Professor Rothstein, I will use his post as a framing device to highlight the problems I see and to urge the new generation of anticorruption researchers to be mindful of them.

Before proceeding, an important note: Despite what I just said, and what I’m going to say in the remainder of the post, I like and respect Professor Rothstein. We have met on several occasions, and he has always treated me graciously. He was the driving force behind the founding and development of the University of Gothenburg’s Quality of Government Institute, which consistently produces excellent research and researchers. He is a prolific writer, and by all accounts a generous and supportive mentor, coauthor, and teacher. My objective in this post is most definitely not to entertain readers with the gratuitous academic blood-sport that is unfortunately too popular in some quarters. Yet at the same time, precisely because Professor Rothstein is such an influential figure in the field, his writings ought to be subjected to rigorous critical scrutiny, especially given the importance of the topic. This isn’t a game, and we must hold one another to very high standards, even when this means assessing harshly the work of people we generally like and respect. I suspect Professor Rothstein would agree with that last sentiment, though probably not with anything else in this post.

With that important note out of the way, let me highlight the three common tendencies in academic writing on corruption that Professor Rothstein’s post illustrates: (1) an unhealthy and unhelpful obsession with definitions; (2) misunderstanding and misuse of social science concepts, particularly a fixation on capital-T “Theories”; and (3) sweeping and uncharitable dismissiveness of prior work and thinking on the topic.

#1: An Unhealthy Obsession with Definitions

Corruption is hard to define precisely. This is a commonplace observation, and probably doesn’t need to be repeated or developed at length. The term has different meanings in different contexts, but as used in the relevant community (modern-day international development practitioners, domestic good-governance reformers, the activist community represented by Transparency International and similar organizations), “corruption” means something like “the abuse of entrusted power for private gain.” While that definition has several built-in ambiguities (What counts as “abuse”? What is “private gain”?), there’s a rough general consensus on the sorts of things people have in mind when they talk about “corruption” in this context: bribery, embezzlement, nepotism, etc. The concept has fuzzy boundaries, and there are edge cases, but that’s hardly unique to the concept of corruption. Sometimes a greater degree of definitional clarity is necessary—for example, when classifying events for quantitative research, or determining whether a given act falls within a given anticorruption entity’s jurisdiction—but in general most folks who care about this topic have a rough-and-ready sense of what “corruption” is, even if there are sometimes some debates at the margins.

But as I’ve remarked before, when academics run out of ideas, they argue about definitions. And when we do, we try to puff up the importance of these esoteric and largely semantic debates by asserting (usually without anything resembling evidence) that a failure to define (or, if you prefer, to “conceptualize”) corruption is a big reason for the (alleged) failure of anticorruption interventions. Professor Rothstein’s post is typical: An “important reason behind [anticorruption] policy failure,” he writes, “is that the corruption problem has been poorly conceptualized[.]” How so? Well, he explains, because the term “abuse” in the standard corruption definition “is not defined,” the entire definition is “an empty signifier,” meaning that “we don’t know what norm is transgressed when a public official is deemed ‘corrupt.’”

Um… what??? There’s certainly a kernel of truth to part of Professor Rothstein’s statement—the term “abuse” entails an implicit normative judgment concerning what exercises of power are proper. This is well known in the anticorruption community, and has been for at least forty years. The mainstream literature notes that there at least three benchmarks that we could use to determine whether a use of entrusted power is improper: (1) formal rules (including but not limited to criminal laws); (2) public opinion and social expectations; and (3) an “objective” standard of just or appropriate governance. All of these benchmarks have their advantages and drawbacks. But in practice—as has also been well-known in the anticorruption community for decades—this definitional ambiguity doesn’t matter all that much, because there’s broad overlapping consensus, under all these various definitions, regarding the “core” corrupt acts (bribery, embezzlement, etc.). Surely the claim that the standard definition is “empty” flies in the face of such broad consensus on what the term means. (Indeed, as Professor Rothstein later notes in his blog post, there’s much less cultural variation in the understanding of what counts as corrupt than earlier writings had presumed).

But the biggest problem here is that the claim that the alleged inadequacies of the conceptualization of the definition of corruption (even though we all have a pretty good sense of the kinds of behaviors we’re talking about) is the cause of the failure of anticorruption policies. Professor Rothstein does not explain this assertion. He just states it as if it’s self-evident. But far from being self-evident, it’s kind of ridiculous. I mean, just think about translating this claim from airy language into a concrete context. Suppose that the IMF and EU pressure a central European country to establish a new anticorruption court to hear cases involving a specified list of criminal offenses deemed to involve “corruption” (bribery, embezzlement, influence peddling, money laundering, etc.). Suppose that a few years later, the court doesn’t seem to be very effective—convictions are rare even when the prosecution’s evidence is strong, and cases take very long to process. International funders and domestic reformers want to understand why this institutional reform doesn’t seem to have worked. What could it possibly mean to tell these people that the ineffectiveness of the court is because “the corruption problem has been poorly conceptualized,” and that the list of corruption crimes subject to the court’s jurisdiction was generated without knowledge of “what norm is transgressed when a public official is deemed ‘corrupt.’” If you said that to a group of serious people trying to diagnose an anticorruption policy failure, you’d be met with puzzled looks, and rightly so. I have not seen anyone ever provide any concrete evidence convincingly linking—as a causal matter—the ineffectiveness of anticorruption policies to the inadequacies of corruption’s definition.

And one more thing: The ambiguities and normative judgment calls in the traditional definition of corruption are baked in and largely unavoidable. Indeed, attempts to “re-conceptualize” corruption in response to these difficulties generally end up recapitulating them or making them worse. Again, Professor Rothstein’s blog post is illustrative. What’s his solution to the alleged failure to define corruption in a manner that is not empty? First he declares (incorrectly) that “what should count as the opposite of corruption has been excluded from the discussion.” (In fact, that’s been part of the discussion for decades, but put that aside.) Next, he offers up as the opposite of corruption the concept of “quality of government,” which means (he says) a system where the basic norm is the impartiality of public officials. So, the opposite of “corruption” is “impartiality” of a particular kind. What deviations from impartiality are unjust (and so, presumably, inconsistent with “quality of government”)? Those that involve “favoritism, which is a central ingredient in corrupt exchanges[.]” OK, so to define corruption, we need to define its opposite, which we’ll call impartiality. But then we’re told that the opposite of impartiality is favoritism. Does that mean that favoritism is corruption? Apparently not, because we’re also told that favoritism is a “central ingredient” of corrupt exchanges, which implies that corruption involves more than favoritism. Also, favoritism might not necessarily be corrupt. But that means that impartiality (and “quality of government”) are not simply the absence of corruption, and we still haven’t defined corruption. And how do we determine what deviations from pure equality are improper “favoritism”? As with the traditional definition of corruption (“abuse of entrusted power for private gain”), we need some normative benchmark against which to evaluate a given exercise of public power that benefits or burdens an identifiable individual or group, in order to determine whether it is improper favoritism or not. And again, our options for selecting a normative benchmark are the same—formal rules, social norms, and a general theory of justice and good governance.

I said it before and I’ll say it again: Obsessing over definitional and conceptual debates is a sign that a scholar has nothing original and worthwhile to say. Don’t do it.

#2: Misunderstanding and Misuse of Social Science Concepts

Social scientists have developed a range of theories (or, perhaps more accurately, tools, methods, and frameworks) for analyzing complex social problems. Familiarity with these tools—both their utility and their limitations—can be extremely helpful. But watch out whenever an academic asserts that the key to understanding and solving some complicated problem is to use Social Science Theory X rather than Social Science Theory Y. And academics should be extremely careful about deploying terminology that they don’t fully understand, especially if they imagine themselves to be scoring points by pointing out seemingly obvious flaws with the target they’ve chosen to shoot at. Very often the academics in question don’t really understand the terms they’re using, and end up making erroneous assertions. The larger problem, though, is the tendency to frame what should be a nuanced discussion of which analytical tools are helpful in shedding light on a given problem—along the lines of, “is this a nail or a screw, because I need to figure out whether to use my hammer or my screwdriver”—as a Clash of Big-Think Theories.

Professor Rothstein’s post illustrates this problem in spades, and does it by regurgitating a trope that I had thought was dead and buried at this point: the notion that anticorruption reforms have not been more successful because of the use of “principal-agent theory.” As Professor Rothstein puts in (in boldface in the original no less): “[T]he principal-agent theory that has dominated anticorruption efforts from the World Bank to development organizations represents a serious misspecification of the problem,” and is, in his view, the most important reason for anticorruption policy failure.

OK, a quick primer on what principal-agent theory actually is. As Professor Rothstein rightly notes, economists developed this terminology mainly in the context of employment relations and industrial organization, though the basic concepts are much older. A principal-agent problem is one in which an actor (the “agent”) is supposed to make decisions on behalf of another person, entity, or group (the “principal”), but the agent has her own interests, which may not align with those of the principal. The principal would like to control the agent’s behavior, but may not be able to do so for one or both of two reasons: The principal may not be able to observe the agent’s decisions (a “hidden action” problem), and the principal may not know as much as the agent about which actions are actually in the principal’s best interest (a “hidden information” problem). The principal would like to design her relations with the agent so as to maximize the principal’s welfare (broadly defined), but her ability to do so may be constrained by these hidden action and hidden information problems. A large body of literature considers how various entities in various contexts would or could optimally respond to these sorts of problems.

Economists and others recognized early on that corruption was a manifestation of a principal-agent problem. Indeed, a principal-agent problem is baked into the standard definition of corruption—the idea that an official could “abuse entrusted power for private gain” entails the notion that this official is supposed to act on behalf of someone else (the power is “entrusted,” after all), but may be tempted to use that power for her own benefit. Speaking at the broadest and most general level, the “principal” (or principals, plural) in this framing are members of society at large, and the agents are those who are supposed to serve the public interest. But the framework is flexible enough to characterize more specific relationships within, say, firms or bureaucracies. Principal-agent “theory” does not, on its own, prescribe specific solutions to agency problems. The optimal responses will depend on context. But this perspective can help (1) call attention to a fundamental problem, (2) highlight why that problem can be so difficult to solve—because, for example, greater control of agents can reduce their effectiveness, (3) explain the development of particular institutional structures that might otherwise seem troubling, (4) help to predict how behavior might change in response to certain interventions, (5) suggest novel interventions that might mitigate agency problems more effectively than more familiar measures.

That all seems pretty sensible to me, especially since principal-agent “theory” (and I keep putting “theory” in quotes because it’s not really a “theory” so much as it is a term for a particular kind of problem) is so flexible, and because corruption so obviously entails a classic principal-agent problem. Does that mean that focusing on classic agency problems—hidden information and hidden action—and drawing on the literature on responses to those problems is the only way to understand corruption? No, of course not. It’s one tool in the toolbox, though frankly the basic insight, as sketched above, is obviously correct.

But in Professor Rothstein’s telling, principal-agent “theory” is the main reason that anticorruption interventions have failed to transform Somalia into Denmark. What are his beefs with thinking about corruption as a kind of principal-agent problem? He asserts that “[t]he principal-agent theory posits that corruption can be remedied if an honest ‘principal’ (e.g., a mayor) changes the incentives for his/her dishonest ‘agents’ (e.g., a city’s civil servants) so that the latter find it more in line with their rational self-interest to avoid corruption.” That is simply not correct. Yes, economists (and others) generally think that incentives matter. But no serious person (least of all economists who study agency problems) believe that corruption “can be remedied” through simple measures by the principal to “change the incentives.” As noted above, one of the main reasons that economists are interested in agency problems is precisely because they can be difficult to solve, due to hidden action and hidden information.

Professor Rothstein betrays this misunderstanding later in the post when he writes that the big problem with “principal-agent theory” is that “if erasing corruption was just a matter of changing incentives, the problem would have been fixed long ago. (There is no lack of knowledge of how to structure an incentive system.) Simply put, if the principal-agent theory was correct, eradicating corruption should have been a ‘piece of cake’.” This baffling passage simply overlooks a long line of thinking and research. Nobody thinks that erasing corruption is “just” a matter of changing incentives—but does anybody seriously think that changing incentives isn’t a big part of it? And as I noted above, one of the insights of the analysis of principal-agent problems is that aligning the agent’s incentives with those of the principal is not easy, and may often be impossible, when hidden action and/or hidden information problems are severe. The claim that there “is no lack of knowledge of how to structure an incentive system” in the context of politics or public bureaucracies is just false—there’s a ton that we don’t know, although it is certainly true that the main obstacles to effective anticorruption reform are often more political than technical.

But maybe that last point highlights Professor Rothstein’s most egregious misunderstanding of principal-agent problems: He seems to think that because in one framing the bureaucrat is the agent of the mayor, it cannot be the case that the mayor is also a self-interested agent whose incentives are not perfectly alighted with that of the members of the public (the “principal” in that framing of the relationship). Indeed, Professor Rothstein’s biggest beef with conceiving of corruption as a principal-agent problem is the misguided notion that the political leaders in any given jurisdiction must be thought of as the “principals,” whose lack of honesty means that anticorruption reforms are doomed to fail. I’ve already explained why framing this observation as a critique of principal-agent theory is misguided: The theory still applies, even if the identity of the principals and the agents shift. (Professor Rothstein also asserts—again displaying his misunderstanding of the theory he is critiquing, that principal agent theory is “internally incoherent” because it assumes that the “principal” is “not a rational self-interested utility maximizer.” Not so. An economic analysis of a principal-agent problem typically entails figuring out how a rational principal would design a contract with the agent so as to maximize her utility, given that the agent will also act in a self-interested way.) But there are two other problems:

First, Professor Rothstein vastly understates the extent to which, even within corrupt systems, leaders at various levels may be interested in, or at least open to, genuine anticorruption reforms. Not every country with a serious corruption problem is an Equatorial Guinea. Places like Ukraine, Brazil, India, Indonesia, China, South Africa, South Korea, and Nigeria have seen, over the past generation, political moments when at least some people in government appear open to meaningful anticorruption reforms. This is not meant to downplay the difficulties of building political support for reforms that would threaten the influence of the powerful, or to exaggerate the extent to which anticorruption efforts in these or other countries have been successful. But Professor Rothstein’s blanket assertion that politicians in corrupt societies will “have little appetite to change the incentives for their opportunistic agents who are engaged in corruption” seems just wrong–and ungenerous to numerous dedicated public servants–when applied to many, many countries.

Of course, sometimes it is indeed the case that the leaders at the top are straight-up kleptocrats who are only interested in maximizing the rents they can extract, with no real interest in the public good. But that observation leads into the next big problem with Professor Rothstein’s assertion that the fight against corruption will become much more successful if one replaces “principal-agent theory” with his preferred alternative, which he calls “social contract theory.” If we posit that the big problem is that the people with the ability to improve the society have no interest in doing so (which, again, is basically to say that there’s a principal-agent problem between the citizens and the leaders), then it’s hard to see why these leaders would support other kinds of interventions that would constrain their ability to pillage their societies, or redirect resources from their own pockets to valuable social programs. In Professor Rothstein’s account, ditching the principal-agent theory in favor of the social contract theory would mean moving away from reforms focused on supervision and control to focusing on building “a functioning state that actually delivers what citizens expect of it, namely, security, infrastructure, meritocracy and education.” OK, fine, I can get behind all that (with a possible caveat on “meritocracy,” which has a range of meanings, not all of which are appealing). But Professor Rothstein just told us that the fatal flaw with the principal-agent framework is that in a high-corruption society, the society’s leaders are themselves thoroughly corrupt, and only care about lining their own pockets. So why should we expect these leaders to support substantial investments in infrastructure, education, and other social services? We’re not told. Oh, and suppose these leaders do somehow decide that they care about promoting all of these things that citizens demand, but due to bureaucratic corruption, the infrastructure isn’t getting built, the education is substandard, money for social services is being misappropriated, etc.? Presumably leaders who cared about upholding their end of the social contract would want to stamp out this sort of corruption. How would they do it? Well, presumably with all of the techniques for addressing agency problems within the public service that Professor Rothstein says we should ignore.

So Professor Rothstein’s discussion of principal-agent “theory” and its alternatives is inaccurate and unhelpful. Which is a shame, because he’s actually got some important (if not entirely original) ideas buried in his exposition. The idea that the big problem in anticorruption reform is less technical (“what policies work?”) and more political (“how can reformers get those in power to adopt policies that work?”) is, in my view, spot-on. We’ve got a bajillion academic papers documenting the harms of corruption, and a smaller but growing body of literature that assesses the impact of various targeted interventions, but our understanding of the politics of anticorruption reform is frustratingly limited. But where Professor Rothstein goes wrong, in my view, is by seeking to frame the big problems in the field as a clash between capital-T “Theories” of corruption, where one has to prevail and the other needs to be vanquished. And that’s the broader tendency that I see all too often in younger academics: This idea that an important Big Picture Scholarly Contribution is one that says either, “We should use [Blank] Theory to understand corruption,” or “We should abandon [Blank] Theory,” or both. So, please, young scholars, stop doing this! Talking in terms of these grand theories is not a sign of intellectual sophistication. And if you do deploy or critique technical social science concepts, be sure you understand what you’re talking about.

#3: Sweeping and Uncharitable Dismissiveness of Prior Work and Thought

Academics understandably and correctly like to frame their interventions in reaction to prior work, often highlighting shortcomings or oversights of that work. I do that quite a bit, as do many other scholars. It can be helpful to do so, because it sharpens the focus and makes clear what the intended contribution of any given book or article is supposed to be. One of the best pieces of advice I got when I was a young professor was from a mentor who remarked, “When you’re writing a paper, it’s always helpful if you can show that someone important is wrong about something important.”

But we need to be careful, when doing this, to avoid sweeping generalizations that are unduly dismissive of other perspectives. I’ve already highlighted some parts of Professor Rothstein’s blog post that I think fall into this trap, such as the suggestion that those who focus on principal-agent problems in the corruption context assume that government leaders are benevolent, and that these principals are not rational utility-maximizers. Here’s another example, also from his discussion of the alleged deficiencies of principal-agent theory: He asserts that a principal-agent perspective “points to increased supervision and control, more stringent laws and punishment, less discretion for people working in the public sector and public service cutbacks.” Um, no, that’s not right, or at least it’s a vast oversimplification. The claim that those who think about corruption as a principal-agent problem inevitably (or even usually) favor limiting discretion and cutting public services is especially egregious. Sure, some people think this, but not all—these conclusions certainly don’t flow from the logic of the principal-agent problem as such. There is, in fact, a robust debate within the academic literature on how things like government size, bureaucratic discretion, and monitoring systems affect bureaucratic corruption. It’s much more convenient to go after the straw man idea that those espousing “principal-agent theory” inherently favor a kind of neo-liberal shrink/constrain-the-state agenda, even though that isn’t really true, and hasn’t been for at least a decade.

Another example where the desire to simplify prior thinking so as to clear space for one’s preferred ideas leads to unfortunate analysis comes when Professor Rothstein address the question whether corruption may persist in part because of deficiencies in the legal system. Professor Rothstein says no, reasoning that “[i]t is not a legal problem because most deeply corrupt countries have very good laws against corruption.” Wait, what? Sure, most (all?) countries have laws against various forms of corruption. But where does he get the idea that all of these laws are “very good”? There are, in fact, quite substantial legal differences across jurisdictions. Some countries have laws against illicit enrichment; others do not. Some countries have strict conflict-of-interest rules for public officials; others do not. Some countries allow private parties to initiate lawsuits or prosecutions targeting corrupt actors; others do not. Some countries require extensive asset disclosures for a broad range of public officials; others do not. Some countries have procedural rules that facilitate settlement of corruption cases; others do not. Some countries allow whistleblowers who reveal corruption scandals to receive a portion of the resulting monetary recovery; others do not. I could go on, but I hope the point is clear. And even when the laws on the books are similar, enforcement institutions and practices can be wildly different—with potentially important implications for how effectively the laws on the books are enforced. Professor Rothstein’s rhetorical strategy in this passage seems to be to say that laws (and legal reforms more generally) aren’t that important in fighting corruption, because every country has laws against bribery and embezzlement yet corruption persists. That reasoning is obviously fallacious when stated explicitly, but this sort of passage is not uncommon in anticorruption discourse. If anyone were to take this kind of rhetoric seriously, the implications would be to ignore or discount vitally important questions regarding legal reform. Or at least I think they’re vitally important – Professor Rothstein may disagree. But he makes no real argument to that effect–there’s just a back-of-the-hand dismissal of the idea that the quality of law and legal institutions might be relevant to the fight against corruption, simply because existing laws against corrupt acts haven’t eradicated corruption altogether.

* * *

This has been quite a long post, I realize. Probably too long. But Professor Rothstein’s post provoked me to get off of my chest much that has bothered me about ongoing trends among scholars who write about anticorruption and related issues. So, if any up-and-coming scholars are still reading at this point, allow me to sum up my plea to you thusly: (1) Don’t get bogged down in debates about definitions, or think that those debates are more important on practical level than they are. (2) By all means make use of social scientific tools and techniques when they are useful (and feel free to critique their limits or misapplications), but avoid framing debates over effective anticorruption policy as clashes between Big Think capital-T “Theories,” and if you must invoke technical theories, be sure you really understand them—especially if you’re inclined to dismiss them. (3) By all means critique prior work and its limitations, but generally try to address the strongest and most sophisticated version of your targets, without mischaracterizing or oversimplifying them, and without casually dismissing the relevance of large and nuanced bodies of thought to clear space for your preferred intervention. If all of us (me very much included) make more of an effort along these lines, our work will be better, more useful, and more likely to enlighten rather than mislead.