Geoffrey A. Manne is Executive Director of the International Center for Law & Economics and a Lecturer in Law at Lewis & Clark Law School.
I take to heart Jim’s claim that fraud is too-little discussed in this realm given its cost, and thus I’ll try my hand at it.
Every discussion of the industrial organization of credit card networks owes a debt to Bill Baxter. Baxter, a law professor and former Assistant Attorney General in the Antitrust Division of the DOJ, was one of the first (maybe the first?) scholars to discuss the economics of two-sided markets, in a paper, as it happens, on the economics of interchange fees in credit card networks.
In simple terms, the essence of Baxter’s analysis is that the role of the interchange fee in credit card networks is to balance and maximize demand for credit card transactions on both the consumer side and the merchant side–optimizing the system by drawing in as many consumers on the one side and merchants on the other as possible while still matching up demand for credit transactions on each side (thus maximizing network benefits). The lever of the interchange fee allows the system to re-allocate some of the costs that are otherwise born by only one side of the market in order to effect this optimization. One of these costs is the cost of fraud–and the interchange fee is, it seems to me, an essential lever for re-allocating the costs of fraud within the credit card system to where they can best be born.
Fraud costs are an important, if oft-neglected, component of payment systems’ functioning. Every payment system by its nature includes the risk of fraud, and every payment system, by design or by default, imposes the risk of fraud on one or more parties in the system. For example, a merchant that accepts cash in exchange for goods bears the risk that the cash will be counterfeit. The cost of counterfeit currency to merchants is substantial. (Although, speaking of cross-subsidies, the cost to merchants likely captures barely a fraction of the full cost of counterfeit currency—a cost born mostly by the government (and passed on to taxpayers) in policing and deterring counterfeiting). The cash system, essentially by default, imposes the residual fraud costs on the merchant. Checks present an even greater fraud problem than cash and, again, the costs are allocated essentially by default: A merchant that accepts a fraudulent check will bear the cost of the fraud.
In principle, the fraud costs of checks or cash could be allocated differently. The government could offer some sort of guarantee, or the issuing bank could agree to bear the cost. But in part because the government requires banks to clear checks at par—has, in other words, fixed the interchange fee at zero for checks—there is little opportunity for the system’s lever to operate to reallocate these costs, ensuring that the costs lie where they fall, and that redistribution is made only in the parts of the system governed by explicit contracts (thus, for example, depending on a host of factors, some of this cost may be redistributed from merchants to merchants’ banks via reductions in various fees in the agreement between merchant and bank).
In contrast, the flexible interchange fee in the credit card system allows fraud costs to be allocated differently throughout the system, presumably ensuring not that the costs lie where they fall, but rather that they are born by the party best positioned to bear the costs. In the case of credit cards, assuming the merchant complies with the network’s rules for seeking authorization of payment, the issuing bank, in fact, guarantees the payment (and thus bears the risk of non-payment). As Bill Baxter noted, “[t]his shifting of risk under the [credit card] system obviously increases [the issuing] bank’s cost, enhances [the merchant’s] demand for the system, and increases the amount of discount [the merchant] is willing to pay to [the acquiring] bank.” It is this re-allocation of costs, facilitated by the interchange fee, that helps to optimize the system.
In his comment on Todd’s last post, Bob makes some helpful points. He notes that for checks, merchants do contract with third parties to help minimize the fraud costs that they otherwise bear. And, again, this is not surprising: Since it is hard for merchants to negotiate away the risk to issuers because of the Fed’s par requirement, they must turn to self-help. I wouldn’t be too quick to draw conclusions from this for the credit card system–the risks of fraud are different–but it supports the basic point that interchange incorporates compensation for risk allocation that would (and may) be otherwise obtainable through costly self-help.
Great comments. First off, let me clarify that I do not at all think that interchange reflects only risk allocation; certainly it is also a source of profits–and that is by design and essential to the operation of the system.
With respect to large merchants–many of them negotiate separate agreements (what we’re talking about in all of these discussions is a default rate, but many merchants–especially large ones–do negotiate deviations from these rates). And I have no doubt that some of the lower interchange fees they are able to negotiate reflect precisely the dynamic you describe.
I don’t know about the existence of third party insurers, but I agree that in principle this could be more efficient–it need not be the case that any of the direct participants in card networks is in fact the least cost insurer. But this is a Coasian world, and my first-cut guess is that either such entities do exist and do affect the interchange rate in certain cases, or that they are not, in fact, the least cost alternative, taking transaction costs into account. Does anyone have more info on the presence and/or effect of third-party insurers?
Ron: Of course these rules exist–my point is that because side payments are not allowed in the system (banks are required to process checks at par), the system may be inefficient. The inability to make side payments in fact makes complicated rules more rather than less likely, as a second-best effort to compensate for the difficulty in re-allocating risk through price.
The discussion of fraud seems to me to include some odd assumptions. First, with respect to check fraud, the system includes a detailed set of rules that shift the costs of fraud among the parties. Although the rules are outdated in many respects, they do serve to place different types of fraud on the shoulders of parties that arguably are best-placed to avoid it. At a minimum, the rules are clear enough that parties understand what types of fraud they should pay to avoid. Merchants do use check-verification services to avoid certain types of fraud, but most fraud losses in the checking system fall on the banks that process checks, not on the merchants that accept them.
With respect to credit cards, it is surely not true that merchants bear the majority of direct fraud costs on credit cards. Merchants do bear those costs in card-not-present transactions, but even now that is a relatively small share of all transactions. I would love to see (or hear more about) the study that Mr. Van Dyke references.
With respect to the allocation of risk, Mr. Van Dyke suggests that merchants bear the majority of direct fraud costs on credit cards. If that is the case, then the issuing banks’ incentives may be problematic. That is, if issuing banks make the decisions about who gets credit cards and whether transactions are authorized, but the costs of an error are born primarily by someone else, then there might be too little effort to lower those costs.
Fraud can be difficult to measure, but it is clearly an important issue. All payment methods are subject to fraud, and one can debate the best method for measuring fraud. Overall, I believe that there is a general consensus that PIN debit has the lowest fraud rate of the major electronic payment methods.
On a related note, there are private mechanisms for reallocating fraud costs. For example, merchants may choose to self-insure against bad checks, or they may purchase check verification services where the risk is shifted to the insuring party. Some merchants do one, some the other. As Omri notes, if efficient insurance is the theory supporting interchange fees, the existence of self insuring and third party coexisting for other payment methods is an interesting alternative.
The insurance element of interchange fees is important, and it’s good that you raise it. You also suggest, following Baxter, that issuers are the efficient insurers. I can see why this might be the case. Issuers have the best data on the risk of fraud, so they can price it accurately. Issuers also have some of the more effective ways to reduce this risk, by following the patterns of card use (I get occasional calls from my issuer when an atypical or suspicious charge is made) and by suspending cards.
It is possible to imagine, however, that some large merchants can self insure in a more efficient manner. First, they are big enough to be risk-neutral. Second, they can uniquely take some measures to reduce the risk, in real time, at the point of transaction. If efficient insurance is the theory supporting interchange fees, it seems to suggest that merchants who want to self insure ought to be allowed to negotiate such arrangements.
Moreover, it is far from clear that interchange fees are priced to reflect merely the risk. I take it that the the basis for the regulatory concern is that the price reflects a substantial component of profit. If that is the case, then some insureds might prefer to buy risk coverage from a less efficient insurer, which is priced more competitively. I don’t know if there is direct insurance product for the fraud “peril” but, again, self insurance can turn out to be more desirable than the overpriced coverage provided through the interchange scheme.