Is enforcement of the U.S. Foreign Corrupt Practices Act (FCPA) improperly politicized? The notion that it is has gained traction in some circles, particularly in countries with multinational firms that have been sanctioned by U.S. authorities for FCPA violations, such as France and Brazil. The usual claim by those who assert that FCPA enforcement is politicized is that the US Department of Justice (DOJ) deploys the FCPA as a kind of protectionist weapon against foreign multinationals that compete with US firms. But a recent working paper by two business school professors (one American and one Chinese) claims to have found evidence for a different sort or FCPA politicization. According to this paper, individual U.S. Senators exert behind-the-scenes influence over the DOJ to manipulate the timing of FCPA enforcement actions against foreign corporations. More specifically, the paper argues that when a Senator is up for reelection, he or she will influence the DOJ to announce an enforcement action against a foreign company before, rather than after, the election. Doing so, the authors suggest, helps the Senator’s reelection chances by imposing a cost on a foreign company that competes with domestic firms in the Senator’s state.
I confess that when I first saw this paper a few weeks ago, I didn’t take it too seriously, because the central argument seemed so obviously detached from reality. (I also didn’t have time to dig into the details of the empirical methods, which are somewhat involved.) But the paper seems to generated a bit of buzz—including a Tweet from one of the best and most respected economists who works on corruption-related issues, which specifically asked me and a few others for our reactions to some of the “provocative” evidence presented in the paper. So I took a closer look.
Let me say at the outset that I very much respect what the authors of this paper are trying to do. It’s difficult to answer many vital empirical questions about the FCPA, and practitioners and scholars often find themselves relying on anecdote and intuition. Quantitative evidence can therefore be extremely useful. We always need to be open to the possibility that crunching the numbers will show that our intuitions are inaccurate, or that important things are happening below the surface, which are hard to detect without rigorous analysis of quantitative patterns. This sort of research is hard to do well, it takes time and training, and those who do it should be respected for their efforts.
But I don’t buy the central claim of this paper. I don’t buy it in the slightest. It is, in fact, ludicrous—and obviously so to anyone who knows anything about the FCPA or US politics and government institutions. Treating the quantitative findings of this paper as conclusively establishing that individual Senators manipulate the timing of FCPA enforcement actions against foreign firms would be the equivalent of listening to your GPS when it tells you to drive into a lake. Yes, sometimes quantitative analysis can show us that our “common sense” intuitions are wrong. But sometimes applying just a dollop of common sense can show us that the results of the quantitative analysis are flawed or fluky and not to be taken seriously.
Given the mostly non-technical audience of this blog, and the limits of my own capabilities, I’m not going to try to critique the statistical analysis in the paper—though in a moment I’ll have a bit to say about why I think the authors might be getting the statistical results that they’re getting. Instead, I’m going to focus on the implications of the author’s central claim, and why that claim is so wildly implausible.
Again, the principal claim goes like this: Individual United States Senators (especially but not exclusively those on the Judiciary Committee) can exert pressure on the DOJ to manipulate the timing of the announcement of FCPA enforcement actions. The paper then argues that US Senators exercise this influence in a particular way: to move up the announcement of enforcement actions against foreign firms to before rather than after the Senator’s election.
To do a basic gut check on the plausibility of this notion, let’s start with the “case study” that the authors present as an illustration of the phenomenon in action. They focus on the resolution of the case against Total, S.A., a French oil company with operations in Texas (among other places). On May 29, 2013, the DOJ and SEC announced that Total had agreed to pay $398 million (the paper incorrectly says $245 million, possibly because it omitted SEC penalties) in fines for FCPA violations related to bribes Total had paid in Iran. The timing of this announcement, the authors insinuate, had to do with the fact the Senator John Cornyn, a Republican Senator from Texas, was up for re-election on November 4, 2014. Total competes with Exxon Mobil, a U.S. multinational headquartered in Texas. So, the authors seem to suggest, Senator Cornyn likely got the DOJ to announce the enforcement action against Total in 2013 in order to benefit Exxon, thereby improving is own re-election chances.
Right off the bat, there are some things about this story that don’t add up. For one thing, in 2013, Barack Obama was President and Eric Holder was the Attorney General. So, how does this work, exactly? What’s the scenario the authors are implying took place? Senator Cornyn—a bitter partisan adversary of the Obama Administration and implacable critic of AG Holder who back in 2012 told Holder that he should resign—calls up Holder or one of his DOJ Deputies and says, “So I’m up for re-election next year, and you guys would be doing me a big favor if you could file FCPA charges against one of Exxon Mobil’s foreign rivals before then. Can you help me out?” And Holder or his deputy says, “Sure, happy to help. Especially since you’re on the Senate Judiciary Committee, we’d better do what you ask.” I mean, come on. I love number-crunchers, I love quantitative work that challenges conventional wisdom—but believing that anything like this took place is driving your car into the lake because your GPS told you to turn right. It’s just silly. And if that’s not what the authors are suggesting took place… then what are they suggesting, exactly? That there’s some mystical, diffuse process by which DOJ line attorneys time their enforcement announcements to help opposition-party Senators on the Judiciary Committee? The quantitative results are only meaningful if they substantiate a plausible story about real people doing real things in the real world. And I don’t see such a story in this case.
There are a few other problems with the Total-Cornyn example as alleged evidence for the paper’s central thesis. For one thing, though this may seem like a minor point, the empirical analysis focuses on enforcement actions brought within a year of the election. The Total settlement was announced over 17 months before the 2014 Senate election–so it wouldn’t actually count as evidence in favor of the author’s hypothesis in their full quantitative analysis. Perhaps more importantly, if the timing of the Total announcement were really driven by Senator Cornyn’s electoral interests, there are a few other things we should expect to observe.
- First, one might imagine that at some point, in some venue, Senator Cornyn would have mentioned the Total case. Yet I haven’t been able to find a single instance in which Senator Cornyn said a word about this case. It’s a bit odd that what’s supposed to be an event that provides a major boost to his campaign is something he never mentions. I haven’t even been able to find any evidence of Senator Cornyn mentioning the FCPA during his 2014 campaign. He did occasionally reference the FCPA more generally in other contexts, but not on the campaign trail (or even, so it seems, during the year leading up the the 2014 election). The paper asserts, without citation, that the Total announcement was accompanied by an “explosion of media coverage,” but I think what the authors really mean is that the settlement was covered in the business press—which is unsurprising and, for the most part, politically irrelevant.
- Additionally, if Senator Cornyn actually did try to intercede in this individual case, especially during a Democratic administration, it’s inconceivable that this would not have been reported—and probably would have been a major scandal. After all, there are extremely strong norms against politicians interceding in individual criminal cases. (This is not so true for regulatory actions, a potentially important distinction here.) Indeed, when politicians have attempted to pressure the Department of Justice to bring, move up, or refrain from taking criminal enforcement action for purposes of improper electoral advantage, this has been front-page news, all of it bad for the politicians involved. If Cornyn had tried to pressure the Obama DOJ to move up the Total action, and this had been reported, Cornyn would have had to answer questions about it for months. Would he have taken such a risk to get a news event that was so unimportant to his campaign that he seems never to have mentioned it?
Now, you might say, I’m being unfair in spending so much time on this one example. After all, it’s a quantitative paper that purports to derive evidence from thousands of data points. But the main points I’m making about the Total-Cornyn example apply more generally:
- The claim that the timing of FCPA enforcement actions is meant to help individual Senators up for re-election really only makes sense for Senators from the same party as the President. But nothing in the empirical analysis in the paper attempts to control for partisan alignment. If the alleged effect appears regardless of the partisan configuration, this strongly suggests that the results are not due to the Senator’s electoral interests.
- If the announcement of an FCPA enforcement action against a foreign corporation gives a significant political boost to an incumbent Senator up for re-election, it’s reasonable to expect that the Senator, or his or her campaign staff, would talk about the case. But to the best of my knowledge, that never happens. And if it doesn’t happen, this is strong evidence against the notion that Senators are pressuring the DOJ to move these announcements up to before the election. (The paper claims, implausibly, that the announcement of an FCPA action against a foreign corporation before the election increases the incumbent Senator’s vote share by 2.2% relative to the baseline—an effect that is too large to be credible. If a reported effect is too large to be believable, that should be taken as a reason to question whether the effect exists at all.) More generally, it’s not clear that these cases are politically significant. The paper asserts that FCPA case announcements cause a “spike in public media attention,” but what the paper actually shows is that FCPA enforcement action are mentioned in the Wall Street Journal business section. Well, sure. The WSJ will report on major legal cases involving major companies. But the fact that the WSJ business section mentions a settlement of an FCPA case does not mean that this event will have a major impact, or any impact, on a Senate election, or even that the case will generate much mainstream attention outside the business press. (The paper also includes an appendix with examples of Senators mentioning the FCPA, discussing concerns about the statute, discussing possible reforms or amendments, etc. Sure, but that’s largely irrelevant. The question isn’t whether Senators occasionally opine on the FCPA writ large—they do. The question is whether Senators attempt, in their individual capacities, to influence the timing of particular FCPA cases, and talk about that. They don’t.)
- If individual Senators were regularly and frequently pressuring the DOJ to manipulate the timing of FCPA enforcement announcements or resolutions, we would expect at least some of these cases to be reported—especially given that such pressure would violate longstanding norms against political interference with individual criminal enforcement actions. But that has never, ever happened. If the author’s story is to be believed, this nefarious political influence has been occurring for decades, but nobody has noticed. (And, to combine this point with the previous point, Senators have been courting the risk of a major scandal—“Extra! Extra! U.S. Senator Presses Justice Department to Charge Acme Corporation with Bribery!”—for electoral benefits that seem trivial or non-existent, in light of the general political insignificance of FCPA cases against foreign firms. Are we really supposed to believe that politically savvy incumbent Senators, many of whom—like Senator Cornyn—won re-election in landslides, would risk a major scandal by interceding in a criminal investigation for little more than an early announcement of an enforcement decision that’s unlikely to be covered outside of the business press?)
There’s another, perhaps even bigger problem here. The author’s main thesis, again, is that individual U.S. Senators have enough sway over the DOJ to affect the timing of FCPA enforcement action announcements, and will do so to advance their own electoral interests. But if that were so, why on earth would Senators exercise this influence only with respect to enforcement actions against foreign firms? One of the striking things about the paper’s empirical findings is that there is no statistically meaningful correlation between whether a Senator is up for re-election and whether the DOJ announces an enforcement action against a domestic firm in that Senator’s state. But why not, if the authors’ claims about Senatorial influence on the DOJ are accurate? If you were a Senator, and it is bad (political) news if one of your state’s flagship firms is charged with FCPA violations, then why wouldn’t you lean on the DOJ not to announce such an action until after the election, if you could do so without being detected? The absence of any evidence that Senators delay (or block) enforcement actions against domestic firms in the year before an election is a classic example of a dog that didn’t bark. If the fact that enforcement actions against foreign firms with substantial in-state activities are more common before elections were due to politically-motivated intervention by the incumbent Senator, then we should also expect to see evidence that enforcement actions against in-state domestic firms are less common before elections than at other times. But we don’t see that—which casts doubt on the inference we should draw from the correlation that we do see.
“But, but, but,” I hear some of you say, “but the data! The differences-in-differences in estimators! The 141,495 firm-state-year observations! The statistically-significant-at-the-1%-level correlations? How can you explain that?” Easy: Random chance. Yes, I know that the odds of getting a statistically significant correlation if there’s no real effect are, by definition, low. But it’s also the case that if you look at enough correlations—if you crunch enough numbers, mine enough data—you’ll find some flukes. There are whole websites devoted to satirical documentation of this fact. With lots of researchers running lots of regressions, and usually not bothering to report the ones that don’t find anything interesting, it’s not surprising that occasionally you’ll get a fluke. As I stressed above, the way to figure out whether what you’re looking at is a fluke or evidence of something real is to ask yourself, “If the causal story I want to tell to explain this unlikely correlation is actually true, what other observable implications would follow? And do we see evidence supporting those implications?” And if the answer is no, you need to rethink whether your story is the right one. Now, sometimes we can’t do this, because there aren’t any other observable implications that we can test. But sometimes there are.
But I think we can go a bit further than just saying that the correlation is probably a fluke, even without doing more rigorous analysis of the data. For one thing, it’s important to remember that even though the number of observations in the regressions appears very large, we’re actually dealing with a relatively small number of FCPA cases against foreign firms. (I can’t find the exact number from the paper—it isn’t clearly stated in the section on descriptive statistics—but another paper has 104 enforcement actions against foreign firms between 2004 and 2018.) There’s also a lot of year-to-year fluctuation in the number of cases against foreign firms (ranging, in the other paper I found, between 2 and 20, with no discernible pattern since 2010.) With a small number of cases, and a fair amount of random noise, it’s much more likely that random chance—a bunch of cases involving foreign firms active in a given state, that just fall in the year before an election—could produce what looks like a meaningful result. It’s also worth keeping in mind that these cases aren’t all independent of one another: often the same transaction produces several cases, which are sometimes (though not always) resolved together, even though they would likely show up in the author’s data as separate cases. It’s also worth noting that the states where foreign firms are likely to be most active are likely large states—and as it happens, three of the largest states (New York, Florida, and California) are on the same Senate election cycle. Another possibility, which I admit I haven’t investigated, is that what we’re actually seeing in the data is a Texas effect, since so many of the foreign FCPA cases involve oil companies, and many of these foreign companies have their US operations centered in Texas. If there happened to be a disproportionate number of oil company cases in years that happened to be Texas Senate election years, that might produce the result the authors report.
I don’t want to lean to heavily into any of those possibilities, since all of them would require much closer examination of the actual data than I’ve done, or that I have time to do. But what I want to stress is that just because there are a lot of fancy stats and tables and multiple asterisks and so forth, doesn’t necessarily mean we should just say, well, I guess the numbers must be right—if the numbers are telling us something that doesn’t seem to make much sense.
There are a lot more nits that I could pick with this paper. For example, I don’t understand why a Senator would care about foreign competitors only if those foreign firms have a substantial presence in that Senator’s state, given that we’re mainly talking about competition in foreign markets. I don’t know why the authors assume that the announcement of the FCPA enforcement action is necessarily bad news for the company (and good news for its competitors), given that in many of these cases the announcement of the enforcement action occurs simultaneously with the announcement of a settlement agreement (as was the case with Total SA), and in many cases this final resolution is good for the firm in question, as it removes the cloud of uncertainty. The paper asserts in passing that FCPA fines are “an important source of revenue for the DOJ,” apparently labeling under the common misapprehension that FCPA fines go into the DOJ’s budget rather than the general treasury. The paper also suggests resolution via plea bargain or other settlement is a sign of leniency by the DOJ, which is slightly baffling given that virtually all FCPA actions against corporations settle. I could go on, but these are mostly side issues. The big problem with the paper is the one I emphasized above: The authors find a strange and surprising correlation in the data, and proceed to tell a story about it, but that story (1) on its face doesn’t make any sense in light of what we know about incentives and behavior of the relevant actors, and (2) would, if true, imply that we should see a bunch of other things—none of which we actually see in the real world.
Does any of this matter? Well, yes, in the sense that—as I mentioned in the introductory paragraph—there are a bunch of folks out there who would like to discredit the FCPA and the DOJ by alleging that in fact FCPA enforcement is all about politics. Those folks will, I predict, seize on this study (and simplify its results in the process) to advance that political agenda. Now, I do not believe that academics should pull their punches or modify their findings out of a concern about how their research will be used by advocates or politicians. That way madness lies. But I do think that when academics weigh in on a matter of political controversy, they (or I should say we, because I’m one of them) have an obligation to be especially careful—before we make some hugely provocative claim that will be used to boost one side of a fraught controversy, we had better be damn sure that we’re right (or, since we never can be certain that we’re right, I should say that we’d better be damn sure that we’re as right as we can be).
The paper I critique in this post is still a working paper, and the authors are still revising it. I hope they revise it further. I will keep an open mind, and perhaps I will be more convinced by an updated version. But I hope they keep an open mind as well, and will at least consider, to paraphrase Oliver Cromwell, the possibility that they might be mistaken.