Streetwise Professor

August 1, 2006


Filed under: Derivatives,Exchanges — The Professor @ 11:49 pm

Growing up in the Midwest during the Cold War, the word “silo” evoked images of blue A.O. Smith grain silos, or Minuteman missile facilities. Now the word has come to mean “vertical integration” in MBA-speak.

The “vertical silo” issue is front and center in debates over the structure of European financial markets. Deutsche Borse owns both a trading platform and a clearing system. Trading is pretty well understood, but clearing is part of the essential “plumbing” of the financial system that most don’t understand. Clearing is the process of confirming and matching transactions, establishing obligations to complete trades, and in some cases, putting a central counterparty in between ultimate buyer and seller in order to reduce performance risk.

The DB silo has been criticized as an impediment to competition in the trading for German stocks. It has been alleged that by denying other exchanges to its clearing platform, DB protects its market dominance in trading. The recent merger discussions between DB and EuroNext stalled in part over DB’s refusal to shed its clearing operation. The proposed EuroNext-NYSE combination will supposedly eschew the silo model.

So who’s right? As a dyed-in-the-wool Chicago School guy, I look askance at anti-competitive explanations of vertical restrictions and vertical integration–it’s in my DNA. Nonetheless, I do recognize that there are circumstances in which vertical restraints can be used to impede competition. (Note that quintessential Chicago guy Les Telser’s famous article on resale price maintenance considered both pro- and anti-competitive explanations of this vertical restriction.) Hence, rather than respond reflexively, I think I should provide some analysis of the issue.

This analysis must be predicated on a recognition of the importance of network effects in both trading and clearing. It is well known that the nature of liquidity creates a network effect in trading of financial instruments; see my 2002 JLEO paper, or an earlier paper by Pagano for a formal analysis. This network effect exerts a centripetal force that tends to induce all trading to tip to a single venue. In my JLEO piece, I show that alternative trading venues (e.g., the block market, the third market) can survive only by limiting participation to the verifiably uninformed. The predictions of these models are largely supported by the data. Most trading does tend to concentrate on a single market, and off-exchange trading tends to have much less information content than on-exchange trading. Attempts of entrants to compete with incumbents, e.g., LSE’s attempt to wrest the Dutch equities trade from EuroNext, Eurex v. CBOT in Treasuries, or EuroNext.Liffe v. CME in Eurodollars, almost always end in failure.

Although it has not been analyzed as extensively, there are strong sources of network economies in clearing too. The number of matches between buyers and sellers is maximized when all trades are submitted to one clearer. If there are several clearers, and the buyer and seller direct orders to different ones, no match occurs, or the clearers must take the additional step of contacting each other to complete the trade. This is duplicative, and increases the possibility for costly errors or delays in trade processing.

The use of netting to economize on the flows of cash and securities also creates network effects. Netting economizes on liquidity needs; traders who both buy and sell in a given settlement cycle need only the cash necessary to pay for their net transactions. Netting possibilities are maximized when all orders are directed to a single clearer. [N.B. The term “liquidity” used in this context is somewhat different than its use in the trading context. In the clearing/settlement context it refers to to the amount of liquid assets (cash, readily marketable securities) needed to support/finance the clearing and settlement process.]

Network effects create a centripetal force which induces tipping to a single venue. Therefore, if clearing and trading in a particular instrument are separate, one would expect tipping to lead to the survival of a single trading platform and a single clearer. This creates the real possibility of a bilateral monopoly problem. This has two potentially adverse consequences. First, it leads to double marginalization. An exchange that possesses market power due to the network effect in trading will markup the above marginal cost clearing cost that a clearer that also possesses market power due to the network effect in clearing. Second, it raises the possibility of holdup and opportunism as the monopoly exchange and the monopoly clearer attempt to capture the quasi rents that arise due to specialized investments in clearing and trading infrastructures. (There will be an additional source of inefficiency if clearing services and trading services are not consumed in fixed proportions, i.e., they are not perfect complements. In this case, production may be inefficient due to distortions in the mix of clearing and trading inputs. However, it seems that fixed proportions is a reasonable assumption so I focus on double marginalization and opportunism.)

Integration is a well-known–and usually efficient–way to mitigate both double marginalization and holdup. Therefore, one expects to observe integration between trading and clearing. This is typically, though not always the case. Most US futures exchanges own their clearing arms. The US stock exchanges jointly own their clearing entity, the NSCC, just as US options exchanges own the Options Clearing Corporation. As already noted, DB is integrated into clearing. EuroNext is something of a hybrid. It owns a 25 percent stake in its clearing provider, LCH.Clearnet.

There are exceptions, of course. The LSE secures its clearing services in an arms length transaction with LCH.Clearnet. The LCH has long provided clearing services for numerous (and relatively small) UK futures exchanges. Nonetheless, there is considerable integration between clearing and trading functions, which is not surprising given the bilateral monopoly issues.

So what are the efficiency implications of integration, and how would mandatory dis-integration of clearing and trading affect efficiency? I am putting the finishing touches on some models that address these issues. In the basic model, there are network effects in both trading and clearing. Moreover, neither stage is contestible. These effects lead to tipping at each level. That is, competition between clearers leads to the market tipping to a single clearer, and competition between exchanges leads the market to tip to a single trading platform. If clearing and trading are separate, double marginalization impairs efficiency, and creates an incentive to integrate clearing and trading. Similarly, any holdup costs also reduce efficiency. Therefore, integration is the efficient outcome in this case even though it is not first best because there is market power due to network effects.

Therefore, when the realities of network effects are considered, the case for dis-integration, well, disintegrates. Since network effects lead to considerable market power–and arguably natural monopoly in certain segments of the market for trading–dis-integration can create serious double marginalization inefficiencies since it is highly unlikely that an independent clearer–which will likely be a monopoly or at the very least a firm with considerable market power due to network induced tipping–will sell its services at marginal cost. Moreover, dis-integration will likely raise holdup costs. If there is considerable market power in trading as one would expect due to tipping (though regulations such as RegNMS can mitigate this power), dis-integration will not appreciably reduce exchange market power, but will raise double marginalization and holdup inefficiencies. This is a lose-lose proposition.

Therefore, any case for dis-integration must be based on the presumption that trading in and of itself is reasonably competitive. I am working on other models which attempt to derive conditions in which dis-integration can enhance efficiency. In these models, there is tipping in clearing due to network effects (so the clearer is a natural monopoly) but Cournot competition among N>1 exchanges at the trading level. If (a) the monopoly clearer in independent, and for some reason sells at marginal cost, and (b) there are sufficiently high fixed costs of operating a clearinghouse (which obviously makes assumption (a) problematic unless there is some two part pricing scheme), then foreclosure and a loss of efficiency can occur. One of the Cournot competitors can buy the clearer, and the other Cournot competitor(s) find it better to go out of business rather than incur the fixed costs of establishing their own (inefficiently small) clearing operation. In this case, output and welfare go down. Welfare is higher if the clearer is separate–and is constrained in its ability to exercise market power.

Back here on earth, we should doubt that a monopoly clearer will provide services at cost (more on this below). If the monopoly clearer charges a super-marginal cost price, double marginalization arises with Cournot competition at the trading level. In this case, integration of clearing and trading into a single monopoly firm actually results in larger output and welfare, even though deadweight monopoly losses persist; integration doesn’t increase the amount of monopoly power, which the clearer possessed in any event, but does reduce the double marginalization inefficiency.

This “foreclosure” outcome can be an equilibrium if the fixed cost of operating a clearing function is sufficiently high. If fixed costs are sufficiently high, one exchange and the clearer merge, and the other exchange loses money if it creates its own integrated clearing operations (because of the fixed costs and the fact that these integrated operations are inefficiently small since they cannot exploit all the network effects). If fixed costs are sufficiently small, the model predicts that one may observe competing integrated exchanges, each with its own suboptimally small clearing operation. I have not yet pinned down the welfare comparison here, as there are offsetting effects–redundant fixed costs and suboptimally small clearing networks on the one hand and greater competition on the other. My intuition is that there is no unambiguous welfare ranking in this model.

Regardless, I don’t find this model very plausible. The Cournot model banishes network effects from trading, which is unrealistic. Moreover, it’s hard to have a great deal of confidence that a separate, monopoly clearer will charge marginal cost prices. In this case, double marginalization inefficiencies likely overwhelm the competitive implications of foreclosure; barring integration doesn’t reduce market power (because the monopoly power arises at the clearing level) but it does raise double marginalization costs. Moreover, the multiple-integrated-exchange outcome that is possible in the model is not widely observed. Indeed, it is hardly if ever observed. Thus, although there are conditions in which dis-integration is welfare enhancing, it is highly doubtful that one would observe these conditions in practice.

Standing back and reviewing the argument makes it clear that a necessary–but not sufficient–condition for forced dis-integration to make sense is that the natural monopoly clearer sell its services at marginal cost. Is this realistic? Color me skeptical.

In essence, this requires either (a) contestibility in clearing or (b) regulation of clearing prices. Don’t even get me started on the latter possibility, so let’s focus on contestibility. On a priori grounds, there is some reason to be skeptical about the complete contestibility of the clearing market. There are likely large costs to switching clearers, especially inasmuch as system users must invest in an IT infrastructure specific to a given clearer in order to connect to it. Switching clearers requires investment in a whole new infrastructure. Given the sums of money involved in the clearing process, and therefore the high costs associated with errors, technical problems, lack of connectivity, etc., this investment is not likely to be small. This clearly creates switching costs, and thereby impedes contestibility.

That said, it must be noted that there are exceptions to vertical integration that suggest some degree of contestibility at the contract renewal stage. Recall that LSE is not integrated with its clearinghouse. Moreover, when its clearing contract with LCH neared expiration, LSE let the contract out for tender. There was competition for the business, which LCH retained, but at a 25 percent savings according to LSE officials. The interesting–and unanswered question–is how the contract price compared to cost. As George Stigler said, data is not the plural of anecdote, but this episode does suggest the possibility that there is a degree of contestibility in the clearing business. On the other side of the ledger, this may be the exception that proves the rule–integration or exclusive contract is the norm, and the LSE episode is the exception.

LSE is also attempting to encourage competition between LCH.Clearnet and a Swiss clearing firm x-clear. Specifically, LSE has signed an agreement whereby x-clear can provide member firms with the ability to choose where to clear. x-clear has a “sub-CCP” account with LCH.Clearnet. If the buyer and seller in a trade both choose x-clear to clear the trade, that entity serves as the CCP. Conversely if one chooses x-clear and the other LCH.Clearnet, x-clear serves as the equivalent of a clearing member of LCH.

It is evident that there are trade-offs in this arrangement. On the one hand, it may increase competition–at least for awhile. On the other, it almost certainly will result in cost duplication and failure to exploit all netting economies. This trade-0ff raises questions. If the arrangement does not result in a relatively seamless merger of the two clearing pools, does the increase in competition more than outweigh the inefficiency that results from the simultaneous existence of two suboptimally scaled clearers that do not exploit all network effects? Will the two actually survive, or will the market tip? That is, will the increase in competition endure, or will it prove evanescent? Do customers need links with both clearing houses? How much additional cost does this involve? The experiment is new, and deserves watching.

In sum, dis-integration is not a magic spell that will make European securities markets more efficient. Indeed, it is plausible that it would make these markets even less efficient. In essence, it is necessary to acknowledge the implications of network effects in trading and clearing. These are the sources of market power and inefficiency, and dis-integration will not fix them. In my view, it is plausible that not only will dis-integration not fix these inefficiencies, it will introduce other inefficiencies (double marginalization and holdup). Therefore, dis-integration in and of itself is not equivalent to cutting some Gordian knot that constrains competition in securities markets.

Instead, improving efficiency will require separate responses to facilitate competition in both clearing and trading. Measures like RegNMS in the US–which effectively weakens the order flow/liquidity externality by opening access to the total liquidity pool–can improve competition in trading. Clearing seems to be a bit more nettlesome.

In brief, don’t expect any easy solutions. Network industries always pose competitive challenges, and virtually every conceivable regulatory scheme to address these challenges is problematic, at best.

Interestingly, integration has attracted little regulatory or legislative attention in the US futures industry. It will be interesting to see whether US futures bourses will be able to fly under the radar on this issue forever. Even if US regulators don’t express concern, the growing internationalization of the futures business might spark foreign (esp. European) regulatory/anti-trust scrutiny. Note that the EU is engaged in a long raging anti-trust battle against Microsoft. Note too that there is considerable opposition to the vertically integrated clearing model in the EU. Putting two-and-two together, this suggests that EU regulators might put CME or CBT or NYMEX in their crosshairs as the globalization of the futures business proceeds apace. This may be particularly true if a US exchange attempts an acquisition in the EU. Just speculatin’, but it should be an issue that US futures exchange executives give some thought to.

Print Friendly, PDF & Email

No Comments »

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress