As I have mentioned in several earlier SWP posts, and in my academic writing, competition in financial exchange markets is likely quite imperfect. Dominant financial exchanges have a considerable competitive advantage due to the nature of liquidity. When those who want to trade have to decide where to submit their order based on the expected cost of execution, they have a compelling reason to submit their order to the exchange to which they expect most other traders to direct theirs. This creates a network effect that induces tipping to a single exchange. Any competing venue faces daunting obstacles to overcome this positive feedback mechanism.
Things would be different if traders could condition where they submit their order on the actual terms of trade offered at competing exchanges, rather than on the expected terms of trade. In this case, a smaller exchange might be quoting a better price than the larger exchange, and could attract some orders.
This pre-trade transparency is difficult to create in some trading technologies. For instance, in an open outcry futures market, current trading opportunities are expressed in the bids and offers of floor participants, and are said to be “good only as long as the breath is warm.” Much of the liquidity is latent–represented by the limit orders held in brokers’ decks, and known only to them. Although one can call to get the current bid and ask prevailing on the floor, this takes time, and frequently prices change before an order is submitted and a trade can be executed. The better-safe-than-sorry strategy is to submit the order to the bigger exchange where good execution is more likely.
Electronic trading tends to improve pre-trade transparency. In many trading systems, one can observe the current bid and offer in real time, and often a good portion of the entire order book is displayed. Theoretically, it is possible to monitor multiple markets simultaneously on the computer, and submit the order to the market displaying the best price.
There are still difficulties, however. It is not a trivial task for a human being to monitor multiple markets on a computer screen, identify which offers the best trading opportunity, and pull the trigger. The markets move quickly, prices jump around on the screen, and sometimes the human is too slow to identify and exploit the best trading opportunity.
How can this problem be mitigated, thereby weakening the order flow network effect, undermining the tendency of markets to tip, and reducing barriers to entry?
One way is to socialize liquidity pools by creation of a central limit order book, or “CLOB.” That is, a single open entry execution facility where all orders from multiple exchanges or trading systems are directed.
Proposals to create a CLOB have been around in securities markets since the 1970s. They have never gone very far, despite receiving some potent political support from Morgan Stanley, Goldman Sachs, and Merrill Lynch (the “MGM” group). As I noted in my 2002 JLEO piece, a CLOB raises thorny issues relating to pricing of access, organization, and governance. Moreover, CLOB proposals engendered vociferous opposition from the incumbent exchanges–primarily the NYSE.
In securities markets, the SEC has taken a different approach. Since 1975, it has advocated what I have termed (in my 2005 Regulation Mag article) the “information and linkages” approach. This approach mandates display of quote information, and attempts to forge linakges between exchanges that facilitate the flow of orders from the exchange to which they are submitted to other exchanges offering better prices.
The first attempts to implement this approach (in 1975) did not change substantially the competitive playing field. The impending RegNMS has the potential to have a much bigger impact. By mandating the rapid and automatic direction of orders to the exchange offering the best price RegNMS will facilitate the direction of orders to the venue offering the best current terms of trade, rather than the venue that offers the best expected terms of trade.
In essence, RegNMS requires the individual exchanges to monitor trading opportunities in other markets, and redirect orders they receive to other exchanges offering better current trading opportunities. RegNMS is having a major impact even before it is in effect. It has motivated the NYSE to expedite its movement to electronic trading. It has led to the entry of new securities exchanges, such as the ISE’s stock trading service. It has revitalized previously moribund regional exchanges, attracting investments to these entities from major brokerage firms that anticipate that the regulation will loosen the NYSE’s stranglehold on order flow.
Is something like RegNMS essential? Maybe not. Rather than forcing exchanges to monitor trading opportunities at competing exchanges, and directing orders to the venue offering the best prices, it may be possible to rely on customers to do this monitoring themselves. In particular, institutional customers (who account for progressively larger shares of equity and futures trading) can utilize computer technology and algorithmic trading tools to monitor trading opportunities on myriad markets, and direct their orders to the exchange offering the best terms. Although humans have difficulties in monitoring many markets simultaneously, this is easy for computers.
That is, rather than having centralized intelligence in the system (in the extreme form in a CLOB, or to a lesser degree via RegNMS) it may be possible to rely on distributed intelligence to monitor diverse markets and direct orders accordingly.
This would work best in electronic markets where exchanges display extensive order book information, allowing large traders to determine the optimal way to trade orders bigger than the best bid or offer displayed at any single exchange. This raises the possibility, though, that exchanges may reduce the information that they display in order to influence trading strategies and the competitive environment. This is an interesting subject for future research.
Distributed intelligence holds open the possibility of weakening the positive feedback mechanism that gives the biggest trading venue a strong competitive advantage. That is, it might weaken the strong centripetal force of liquidity.
One should not get carried away at the prospect, however. The high valuations of futures, options, and equity exchanges suggest that the equity market believes that these institutions will continue to earn substantial rents now and in the foreseeable future. If incumbent exchanges’ competitive advantages were widely believed to be under threat, we would not observe these rich stock prices.
Nonetheless, this discussion suggests that it may be possible to increase inter-exchange competition without heavy handed regulation Encouraging and facilitating distributed intelligence would be one way of doing so.
Apropos my earlier posts on clearing, in the event that distributed intelligence and algorithmic trading effectively opens the liquidity network, clearing will be the main monopoly bottleneck in the trading system.
There are other interesting competitive issues in markets. In particular, the growth of “dark liquidity” is intriguing. Large brokerage firms that receive substantial order flows are matching some customer orders internally (customer v. customer) rather than directing them to exchanges. Some are also internalizing orders and serving as the principal in the transaction. These intra-broker order flows are called “dark liquidity pools.”
These are understandable reactions to the pricing power of for-profit exchanges. Why pay super-competitive fees to exchanges if they can be avoided? From a price efficiency perspective, if one believes that a Kyle model is a reasonable characterization of price determination, the matching of offsetting customer orders has no impact on the efficiency of market prices; order imbalance is what matters in this setting, and customer v. customer matches don’t affect the order imbalance at the locus of price discovery.
Internalization is somewhat different. It is likely that internalized orders are the result of cream skimming, i.e., the market maker/broker identifying orders that are unlikely to be informed, and taking the other side. This exacerbates the adverse selection problem that market makers in the price discovery market(s) face, widening spreads and reducing depth.
Cream skimming is often criticized, but these criticisms frequently assume (incorrectly) that there is perfect competition in the exchanges where price discovery occurs. In this case, exacerbating adverse selection is inefficient.
However, in reality exchanges exercise considerable market power. In the old days of non-profit exchanges, they exercised market power by limiting entry to the exchange. Nowadays, for-profit exchanges charge supercompetitive trading fees.
In this environment, cream skimming can be a second best response to exchange market power. As I show in my paper “Third Markets and the Second Best,” cream skimming can increase total surplus, even though it makes some traders (those who are uninformed but cannot prove it) worse off.
Thus, dark liquidity and cream skimming internalization are market responses to exchange market power, and likely improve welfare (though they are not Pareto improving). They represent competition for exchanges, and increase the elasticity of demand for exchange services, thereby making it less profitable for exchanges to jack up fees. These are likely to be the main forms of competition unless and until distributed intelligence and algorithmic trading make exchange v. exchange competition more viable.