In 2015, following indications that a few Canadian Senators had been using government money for disallowed personal expenses, the Canadian government launched a major investigation that cost approximately $23 million—but led to no convictions, and exposed corruption that cost the government less than $1 million. To many, me included, this seems like overkill, even if we acknowledge that corruption is a serious problem. Yet the “zero tolerance” ethos that motivated the Canadian Senate’s investigation is widely embraced. As previous posts have pointed out, zero tolerance policies fail to account for the fact that corruption might be expensive to root out, and that the extraordinary expenditures required to reduce corruption closer to zero might not, after a certain point, be justified.
This does not mean that corruption is good. But the efficient amount for the government to invest in corruption-reduction—which in turn determines the amount of corruption that will prevail—is that for which, as an economist would put it, marginal benefit equals marginal cost. Or to put this another way, “tolerance” of some corruption is efficient (and appropriate) once the costs of achieving further reductions in corruption are greater than the costs of the corruption that would be eliminated by those additional efforts.
The right amount of corruption, therefore, is probably not zero. Existing anticorruption policies neglect this point, leading to inefficient spending like that on the Canadian Senate probe. With that understanding, how can anticorruption policy be targeted to get the biggest bang for the buck?
- First, the government should invest more in anticorruption efforts in those areas where the marginal costs of corruption (and hence the marginal benefits of corruption reduction) are higher. Money stolen from orphanages is worse than money stolen from the parks departments of wealthy cities, not just because it feels worse, but because the marginal value of the dollars stolen from orphanages would have been higher. More generally, corruption in more budget-strapped programs, and in poorer countries, likely has a larger marginal impact, and so justifies more aggressive corruption-reduction efforts. For this reason, it makes sense to pursue more aggressive anticorruption measures in developing nations, not just because the potential for systemic corruption is higher in these countries, but more importantly because it is in those countries that corruption reduction is likely to have the greatest positive impact on people’s welfare. And within any given country, anticorruption efforts should be targeted toward programs that service the impoverished, because dollar-for-dollar, providing more resources to the least-well-off has the greatest social benefit.
- Second, it may make sense to invest relatively less in reducing those forms of corruption that involve a transfer of resources, but do not otherwise lead to significant misallocations, distortions of incentives, or other inefficiencies. By contrast, relatively more investment should be made to combat those forms of corruption that involve destruction of value, rather than simply transfer of value. As an illustration, consider the contrasting examples presented by Professors Banerjee, Mullainathan, and Hanna: In an infrastructure procurement context, say for a road construction project, it might be the case that the most efficient firm (that is, the firm that can build the road at lowest cost) may be able to offer the largest bribe. If that is so, then even if the procurement official cares only about maximizing the size of the bribe he can collect, he will still end up awarding the construction contract to the most efficient firm. By contrast, suppose a public health official is allocating health care resources like hospital beds, and does so on the basis of bribe payments. If those who need the beds the most are also the poorest, their inability to pay bribes will lead to an inefficient allocation of beds to wealthier, less unhealthy individuals. In this circumstance, the bribe isn’t just a transfer, but causes a serious, harmful distortion in the allocation of government resources. The second example is much more harmful than the first, and anticorruption policy should therefore prioritize those forms of corruption that are most likely to create these sorts of distortions.
- Third, an important qualification to the preceding claim is that it considers only the direct costs of corruption. Even corruption that does not entail inefficient misallocation of resources can have hard-to-quantify “spillover” costs, including reduced trust in government and lower civic engagement. But we must be careful not to automatically invoke spillover costs to justify arbitrarily high levels of investment on anticorruption efforts. Consider the Canadian Senate example, in which $26 million was spent to root out less that $1 million in misappropriated public resources. Unless we think the spillover costs of that corruption were are an order of magnitude bigger than the actual money recovered, this audit was unnecessarily expensive.
- Fourth, the government should focus on low-cost solutions to enforcement of anticorruption provisions. For example, whistleblower statutes with bounties tied to the recovery of public money, such as those used by the US Internal Revenue Service and Securities and Exchange Commission, are a very effective way to compensate individuals, as the government always ends up paying less than the amount the government recovers.
Rather than reflexively embracing expensive “zero tolerance” policies, governments should develop cost-effective anticorruption efforts that consider issues such as allocative efficiency and marginal benefit. The political will to accomplish this may be lacking, as it can appear “soft on crime” to forgo prosecution of certain types of corruption in favor of a more nuanced enforcement. But effective policy is not built on zealousness: rather, behaving in the interest of the public good requires a thoughtful, measured, and economically savvy approach.
Fascinating, Jetson. Reading through your post, I found myself considering how your third point — about the “spillover costs” — interacts with your first and second points. In particular, I imagine we could see spillover ~within~ corruption. You do mention the ostensibly greater risks of systemic corruption in developing countries. But you then conclude your first point saying, “And within any given country, anticorruption efforts should be targeted toward programs that service the impoverished, because dollar-for-dollar, providing more resources to the least-well-off has the greatest social benefit.” Keeping with this focus on the victims of corruption, I wonder if this conclusion is complicated by the fact that the least-well-off are far less capable (and thus far less likely) to engage in reciprocal corruption. To cherry-pick from your examples, the construction firm would seemingly be more likely to carry out tit-for-tat corruption (e.g. bribery on the next project) if it feels aggrieved by corruption, in contrast to the orphan or destitute hospital patient, who likely doesn’t have the means to respond to a decreased faith in institutions by turning around and making the problem worse. This makes me wonder, uncomfortably, if the risk of systemic corruption would actually suggest that in some instances we care more about protecting the wealthy/influential/powerful from experiencing corruption, given their likely reactions to it.
Jetson, thanks for this thought-provoking post. I agree with your general point, that in a world of scarce resources, government should think carefully and critically about how they allocate anticorruption resources. And I agree too that it may make sense to allocate resources to areas where the marginal costs of corruption are higher, or even where corruption involves a “destruction” rather than a “transfer” of value, assuming those areas can be identified.
However, I’m not sure that the Canadian example you open with and return to in your third point is the most forceful way to lead into these compelling arguments because it reminds me of an issue you don’t address in the piece: the prophylactic effect of enforcement. Your third point discusses the negative spillover effects of corruption, but I would be interested to hear your views on the positive spillover effects of anticorruption enforcement.
The social utility of the $26 million spent by the Canadian Senate is impossible to calculate without knowing how much additional corruption was dissuaded by the Senate’s investigation. If other potential corrupt Canadian officials see that the Senate is serious about enforcing anticorruption laws to the tune of $26 million, they may think twice about misappropriating funds themselves which may lead to overall greater savings in the long term.
I think this example illustrates a broader point about cost-benefit analysis in general, which is that it can be very difficult to calculate accurately. You’re likely right to say “we must be careful not to automatically invoke spillover costs to justify arbitrarily high levels of investment,” but I think the sentiment holds in the reverse too: we must be careful not to automatically discount spillover costs or prophylactic benefits to justify arbitrarily low levels of investment in anticorruption efforts.
Jeteson, fascinating and thought-provoking post; on my end, your overall point is very well taken. However, to follow up on the points made by Jason and echoed by Hilary—specifically vis-à-vis the Canadian government example—I think there are a few additional important considerations.
First, the $23mm cost of the investigation is misleading (as your hyperlink to a CBC article shows) because it is inclusive of costs that would have been borne by the government anyway. Holding for what the article dubs “routine costs,” the probe cost more like $12mm.
Second, it’s important to note that the entire investigation was paid for within the confines of the Auditor General’s Office, meaning those in his office would’ve been investigating something else, if not the Senate audit. In other words, the $12mm net cost is an opportunity cost, and the example presupposes that there was another investigation that was worth a greater amount of money available for the Auditor General’s Office to work on.
Third, I believe that there may be some hindsight bias here. Indeed, aren’t corruption probes often commenced precisely to investigate the extent and severity of misconduct by those in public office? It’s very likely that the Auditor General’s Office didn’t know how expansive (and economically significant) the Senators’ misconduct was until the investigation was actually completed.
Fourth, finally, and most critically, I think it’s absolutely vital to consider the fact that Canadian Senators, unlike United States Senators, don’t have traditional term limits. Instead, as you probably know, Canadian Senators serve until their 75th birthday! Mac Harb, Pamela Wallin, Patrick Brazeau, and Mike Duffy—the four Canadian Senators who were subject to the probe—collectively had 110 years of eligibility remaining to serve on the Senate! As such, the deterrence value of this probe was both general and specific. General, in that other Senators were put on notice that they must spend responsibly. But also specific, in that the four Senators who were the subject of the probe would be deterred from future corruption (the potential of which would be great over a combined 110 years).
Thank you for this interesting and well written post, Jetson. There are two questions which I wish to ask:
1) Based on the logic and rationale behind your proposed guiding principle, would you also support cost-effective punishment? Theoretically speaking, should people who were convicted of corruption offenses be excused from punishment when it is not economically cost-effective?
2) Assume that a government does officially decide to adopt a guiding principle for its anticorruption policy which is similar to the one proposed by you. What influence will such a declared official policy have, for example, on foreign direct investment in the relevant country? You mentioned “hard-to-quantify ‘spillover’ costs, including reduced trust in government and lower civic engagement”, but a declared policy of nonenforcement (of some, but definitely not few, incidents of corruption) may also have a major direct negative effect on economic growth.
Again, thank you for your fascinating post.
Jetson, I found your post extremely interesting. In regards to your second point, and to follow up on Guy’s point, how should a country go about implementing a more lenient standard? Assuming there’s no officially declared policy, would this simply be through prosecutorial discretion to decline to bring charges based on some sort of initial cost-benefit analysis on a case-by-case basis, or would this be a blanket policy/guideline that prioritizes the investigation of certain types of corruption over another in the first place?
Jetson, thank you for your though-provoking post. Following up on Signa’s question, I would be curious to know more about how you would propose to determine which cases may be worth our while. I read your post to suggest that we should prioritize cases of corruption that produce the most harm. To illustrate the inefficiencies of some of today’s anti-corruption efforts you point to the total amount of money lost by Canada in a recent corruption scandal (<$1 million), as compared to the price of the investigation ($23 million). However, I would argue that the cases that produce the most harm may have the smallest "costs", while requiring expensive investigations. Oftentimes, it is the poor that are most affected by petty corruption (as you illustrate with your hospital bed example). This type of corruption typically involves individual actors acting alone to serve their own interests. They are rarely stealing millions of dollars. And yet, the funds that they take can have large consequences for the poor families that they are stealing from. At the same time, investigating these individual actors and enforcing anti-corruption laws against all of these people can take a lot of resources. Thus, I could see a scenario in which the very anti-corruption investigations that you advocate for in your post, would not pass the efficiency metric that you use in your example from Canada. So long as we measure costs and benefits using only the value of the stolen money in question weighed against the cost of the investigation itself, governments will be less inclined to go after petty corruption, which can create some of the largest harm towards poor citizens. How would you propose adjusting cost-benefit analyses for corruption investigations to account for these other prevented "human costs" as benefits that may outweigh the price tags of these investigations?
Interesting point. I’m not sure I can really say it any better than Jason, but I was curious how you thought the spillover effects of enforcement factored into all of this. Is it ever worth paying for a front-page anti-corruption story? Even if a high profile investigation is inefficient in terms of the amount of money stolen vs. money spent prosecuting the crime, the “signal” might still be relatively cheap given that, generally speaking, most people don’t follow corruption prosecutions so as to be aware or “deterred” by them. Then again, maybe high profile corruption investigations just increase people’s perception of corruption making them more, not less, likely to take part.
I am a huge proponent and a firm believer in applying cost-benefit analysis to decision making in the public sector. I have often thought about how this applies to sometimes very costly investigations of crimes of corruption as well, so I found your post to be extremely interesting.
I do think that – just like Hilary and Jason noted above – the benefit of deterrence should be factored in. There is strong evidence suggesting that existing penalties alone do not deter corruption – but enforcement does (I was about to search for links to articles, but I found that it is nicely summarized in this post: https://globalanticorruptionblog.com/2015/04/29/anticorruption-enforcement-policy-tips-from-the-deterrence-scholarship/; there is also an interesting study into how the adoption of death penalty for corruption in China ended up not working as a deterrent due to a lack of enforcement ).
I do not, however, think that this goes entirely against your suggestion of applying cost-benefit analysis to decisions on which corruption investigations to pursue. The “welfare” elements can be (and are – as Cass Sunstein repeatedly showcases in the “The Cost-Benefit Revolution”) factored in cost-benefit analysis in public policy and deterrence, I think, is exactly such an element. When making the decision on whether to pursue a concrete investigation, the prosecutors can evaluate what effect it will have on the particular issue / sector, how many investigations have already been conducted in the field, what the observed corruption trends are, etc. All of this could be taken into account together with calculating how much the investigation will cost and how much can be “won” by the state. What I think should also be a part of such an analysis is trying to weigh in how a particular investigation will advance the strategic goals of fighting corruption in the country / state. For example, if an anti-corruption institution identifies fighting corruption in, say, the healthcare sector as its top goal for the year and invests in issues surrounding it that year, even a costly investigation with a relatively small monetary “win” can help advance that particular goal by raising awareness, helping to ensure public support, etc.- thus making sure that other state investments already made for the goal will be supported strategically. Taking all of this into account would leave the decision maker with a tentative amount of money needed for the investigation on one side – and the “welfare” benefits on the other side.
What makes me a bit more doubtful about applying cost-benefit analysis to making decisions on which investigations to pursue is the other side of the question. I am wondering if the way modern countries construct criminal policy in general can be subject for cost-benefit analysis at all, given its core founding principals. Forgive a dramatic comparison, but it would be hard to argue that investigation of a single murder always holds against a pure cost-benefit evaluation (especially when no evidence supports the possibility of recurrence). However, many countries seem to have agreed that the public interest of fighting crime holds an essential value when drafting criminal policies and making sure that even one-off crimes will always be investigated. Furthermore, hand picking which investigations to pursue purely based on a cost-benefit calculus may instigate a sense of unfairness in some groups of the society – and I will say upfront that I don’t think this would not apply to corruption as I do not believe it is a “victimless” crime as some argue now (I think that certain groups of the society might feel more hurt by particular forms of corruption then others). Thus, I am not entirely sure that the cost-benefit analysis as such is a good tool for criminal policy decisions in general as it seems to go against the logic of the way we draft criminal policies in general…
That being said… As I have already mentioned, I am a big fan of the cost-benefit analysis, so I am leaning towards dismissing this latter argument in favor of saying that cost-benefit analysis should be made (and agreeing that in some cases the “welfare” factors will trump any numerical arguments)…
Ruta, thanks for your thoughtful comment on this post. I enjoyed reading it, and always appreciate when you bring your behavioral economics knowledge to the discussion. Your “murder” analogy (which I appreciated, although the “value” of deterring a homicide–i.e., what is the value of a given life–is always an interesting debate), I think, perfectly illustrates why we don’t use cost-benefit analyses in criminal justice. Indeed, as you imply, a murder conviction typically yields the largest criminal justice penalty–much greater than, say, a conviction for a securities/wire/mail fraud scheme worth hundreds of millions of dollars. Conversely, however, our civil court system often yields the opposite result: damages (to the extent that they can be paid) for a hundred-million-dollar fraud mirror the actual losses faced by investors, whereas wrongful death suits typically worth much less than such a fraud. I guess this is a long-winded way of saying: perhaps the cost-benefit approach is better suited for individuals bringing private actions in civil courts? N.B., I haven’t fully thought through this yet . . . .