In my post last week, I discussed a recent working paper (by Cheung, Rao, and Stouraitis) that attempted to measure the economic returns firms reap from foreign bribery — and which reached the depressing conclusion that, much as we would like to believe otherwise, bribery still “pays.” In doing a bit more research along these lines, I came across another terrific working paper — by Jonathan Karpoff, D. Scott Lee, and Gerald Martin — that investigates a similar question, using a somewhat different (though complementary) method, and reaches a similar conclusion: in the authors’ words, “firms engage in bribery because it pays to bribe, on average.”
Let’s suppose — plausibly, in my view — that these papers are correct in their conclusions that, given the expected costs of foreign bribery (probability of detection times expected magnitude of direct and indirect sanctions), many firms will find it in their rational self-interest to bribe. If one believes (as I do) that foreign bribery is a social cost that we should try to deter (if we can do so at reasonable cost), then this implies that we should increase the expected cost to firms of paying bribes abroad — either by increasing the probability of detection, or the size of the penalty, or both.
One of the cool things about the Karpoff, Lee, and Martin paper is that it attempts to calculate just how much higher the penalties would have to be in order to deter foreign bribery, if the probability of detection remains constant. Their sobering answer is that the average penalties would have to by about 9.2 times larger. So, despite all the hyperbole about the enormous size of FCPA penalties (with the US government bragging about the large penalties, and the business community and defense bar griping about same), this recent research suggests that the penalties are almost an order of magnitude too small, if we really want to deter foreign bribery. (The authors also calculate how much the probability of detection would have to increase to deter foreign bribery, if the penalties remain the same. Their conclusion is that it would have to increase from the current estimated level of 6.4% to about 58.5% — clearly unrealistic.)
What to make of this? A few preliminary thoughts:
First, it’s important to understand how Karpoff, Lee & Martin go about making this estimate. I can’t do justice to the sophisticated methodology they employ in a short blog post, but here’s a simplified take: We can’t observe how many firms are actually paying bribes abroad; we can only observe how many firms are doing business abroad (of the sort that might involve FCPA violations), how many firms were actually caught, and the observable characteristics of all these firms (size, nature of business, regions where the firm operates, “aggressiveness” of corporate culture, etc.). But we can use these observable characteristics to develop a model that predicts which firms are likely to pay bribes abroad. One can then use the model to predict which firms probably violated the FCPA at least once in the sample period — the idea is that those firms with observable characteristics similar to the firms that were caught paying bribes probably also paid bribes. This, in turn, allows us to calculate the probability that a bribe-paying firm is caught and sanctioned.
The latter calculation involves the crucial assumption that firms with similar observable characteristics have similar probabilities of committing an FCPA violation — in other words, that firms caught violating the FCPA did not actually have a different probability of violating the FCPA than firms with similar observable characteristics that were not caught violating the FCPA. In reality, that’s almost certainly not true — US government enforcement actions are probably not that random. The question is how close an approximation of reality this assumption is. Imagine two poles on a continuum of possibilities: (1) the government catches 100% of the firms that violate the FCPA, so that we can assume that if a firm is not sanctioned, it in fact didn’t engage in foreign bribery; (2) all firms with identical observable characteristics either violate, or do not violate, the FCPA, so that if there’s a firm with observable characteristics identical to a sanctioned firm, we can assume the former firm violated the FCPA but wasn’t caught. Both of these extreme cases are unrealistic. The question is whether we think the truth lies closer to the former or the latter. Or, to put this another way, using the authors’ other calculations: Do we think the probability that a firm that violates the FCPA is caught is closer to 6% or 60%? I tend to think it’s closer to the former, lower number — in which case FCPA sanctions would indeed need to be several times larger in order to deter foreign bribery. But the data don’t allow us to make this calculation directly.
Second, the authors are careful to note that their analysis does not tell us whether raising FCPA sanctions to the level that would deter foreign bribery is in fact socially optimal. Even if we agree that foreign bribery is bad, and ought to be stopped, there may be significant collateral costs to dramatically increasing FCPA penalties — perhaps most significantly, deterring firms from engaging in transactions in high-risk industries or jurisdictions. This is important to keep in mind, especially in light of the fact that most corporate FCPA violations, at least in large firms, seem to be the result of bad decisions by lower-level employees, not a conscious firm-wide strategy (notwithstanding exceptional cases like Siemens and BAE).
Despite those qualifications, this is an important paper and definitely deserving of close attention.