A Tale of Two Papers, or, Humpty Dumpty Writes About Exchanges
| Craig Pirrong |
The American Economic Association/American Finance Association Meetings are just about over. I made a quick trip there to comment on a paper. Upon returning home, I downloaded a couple of the papers presented that seemed of interest. Good call on one, bad call on the other.
The bad one is “Centralized versus Over The Counter Markets” by Viral Acharya of LBS and NYU, and Alberto Bisin of NYU. Although the motivation of the paper is admirable, the execution is execrable, and is representative of a lot of what is wrong in the profession.
The motivation is to compare the efficiencies of alternative ways of organizing derivatives trades: centralized exchanges and over-the-counter (OTC) markets. Great. Big question. I’ve written a lot about it, and would be very interested in seeing other takes thoughtful on the subject.
The paper concludes that organized exchanges are (constrained) first best efficient, and more efficient than OTC markets. A quick review of the paper makes it clear, however, that they’ve rigged the game to produce that result.
Specifically, Acharya and Bisin assume that there are different “types” of traders; they differ based on the characteristics of their endowments. Due to their finite wealth, and the risks of derivatives positions, traders are sometimes unable to meet their derivative contract obligations. So far, so good: all that makes sense.
Then they go off the rails. Specifically, they assume that centralized exchanges set price schedules for a single derivative contract, and specifically, set a price for every trader that varies depending on (a) the trader’s type, and (b) the size of the trader’s positions. They motivate (b) by arguing that exchanges can view a trader’s entire position. That is, every trader pays a different price depending on his position and his type. In contrast, in OTC markets, they assume that no counterparty can observe any trader’s entire position, and hence although prices can be conditioned on type they cannot be conditioned on position.
It’s no surprise, then, that they find exchanges to be a more efficient way of trading. Exchange prices are conditioned on more information that is relevant in determining the likelihood of default, and the cost thereof, and hence can be set to control default and risk taking more efficiently than is possible in the OTC market.
Perhaps there is a planet somewhere in the universe in which exchanges can do what Acharya and Bisin claim they can do. All I can say is that on earth they can’t and don’t. In fact, the situation here on the planet in the Milky Way with the blue sky is almost an inversion of the what they assume.
Specifically, exchanges don’t condition trade prices on anything. Indeed, the whole idea of clearing on an exchange is to create fungible contracts in which prices are the same for all traders at a given point in time, and don’t depend on the traders’ endowments or positions. That is, exchanges facilitate anonymous trading among consenting adults by taking actions that make trader types and positions irrelevant to the price determination process.
In reality, collateral depends on position size, but not trade prices. Moreover, exchanges quite specifically do not condition collateral terms on trader type (e.g., on traders’ balance sheets). So, even if you argue that collateral variations are effectively price variations, the Acharya-Bisin assumptions are unwarranted. Exchanges are not, as Acharya and Bisin assume, central planners implementing a centrally imposed price schedule conditional on fine information on trader type and trades.
Put differently, clearing on exchanges creates a potential moral hazard that exchanges control at cost. The cost can be sufficiently high to make it inefficient to adopt clearing. Acharya-Bisin assume that exchanges have and can use costlessly the information required to control this moral hazard. Again, an inversion of the truth.
Also in reality, OTC market participants do condition trading prices on counterparty type, and explicitly take balance sheet risk into account. Although OTC traders don’t view their counterparties’ entire exposures/positions, they do have noisy signals on this and take it into account when setting trading terms. Moreover, as I’ve argued before, there are a variety of channels (e.g., the repricing of short term debt) by which diffuse information on the entire risk exposure of a financial institution is aggregated, and affects the incentives of said institution to take on risk.
So, the tradeoffs between exchanges and OTC markets are more complex: exchanges/CCPs have better information on some dimensions, worse on others. It’s not altogether clear a priori which dominates. Moreover, there are institutional innovations other than exchange trading and clearing that can mitigate the OTC markets possible information disadvantage: for instance, data warehouses that collect and disseminate information on positions could remedy the supposed weakness of the OTC markets.
And here’s where the Coase question comes into play: If a particular institution is inefficient, what transaction costs precludes agents from adopting it? Or, to put things differently, if you believe Acharya and Bisin, agents are leaving a lot of money on the table by trading in the OTC market: there are gains to trade to be captured by trading on exchanges. Why isn’t this happening? Why does a putatively inefficient mechanism survive, and indeed, thrive and grow relative to the putatively efficient one? This can happen, and policy can perhaps improve efficiency, but someone who claims that a given arrangement is inefficient, as Acharya and Bisin do, should identify the friction or frictions that precludes a change. They, suffice it to say, don’t do this.
This model wouldn’t be objectionable if it were a purely theoretical exercise. But Acharya and Bisin make broad policy recommendations based on their analysis. The problem is, they are like Humpty Dumpty, who declared: “When I use a word,” Humpty Dumpty said, in rather a scornful tone, “it means just what I choose it to mean — neither more nor less.”
Humpty Dumpty-like, they use the word “exchange” to mean just what they choose it to mean, even though it bears no relation to exchanges in the real world, or the kinds of exchanges that would be the beneficiaries of exchange trading mandates, or efforts to raise the cost of OTC trading. Their recommendation is a sort of bait and switch: sell the concept of exchange trading on the basis of an idealized model that assumes exchanges have decisive information advantages, and then deliver real world exchanges that are much more limited.
This is not uncommon, sad to say. But it is pernicious. And if this is the best that can be done to support exchange trading mandates, well, there you go.
The other, much better paper is “Danger on the Exchange: Counterparty Risk on the Paris Exchange in the 19th Century” by Angelo Riva of EBS and U-Paris X, and Eugene White of Rutgers. Rather than construct otherworldy exchanges and decree that their visions should be implemented here on earth, Riva and White do something radical: they actually take a detailed look at the evolution of counterparty risk sharing mechanisms and defaults on a real world exchange–the Paris Bourse. They carefully trace the evolution of mutualization of countparty risk on the PB, and pay close attention to a very real-world problem: namely, that mutualization creates the potential for moral hazard that is costly to control.
The empirical analysis could be improved (and when I have some time I’ll pass on some suggestions to them), but this paper undertakes a detailed analysis of actual institutions, and actual agents grappling with hard problems. The Paris Bourse was a real exchange. The problems it faced are inherent to any efforts to manage performance risk. We can learn a lot more from these sorts of analyses, than we can from constructing fanciful exchanges in a galaxy far away. Formalism is valuable — I even commit formalism from time to time, myself — but it is important that if formal models are intended to be the basis for far reaching policy conclusions, that they bear some relation to reality. Acharya and Bisin would put their modeling talents to much better use by trying to construct theories that reflect the gritty realities that Riva and White quite clearly document, instead of building models that bring to mind all those economist jokes with punch lines like “I assume the existence of a 100-story ladder.”