Streetwise Professor

September 20, 2019

The Simple (and Very Old) Economics of the Stock Market Data Pricing Controversy

Filed under: Economics,Exchanges,Regulation — cpirrong @ 4:20 pm

The most contentious battle in American securities markets right now is being waged over exchange pricing of data, in particular over proprietary order book feeds. The battle pits the exchanges against market users (e.g., HFT firms, institutional traders) with the latter claiming that the prices charged by the former border on the extortionate.

The critics actually have a very good point. The economics of the situation imply that the prices the competing exchanges charge are ABOVE the price a monopoly would charge.

No. Really.

So how could competing firms charge a supra-monopoly, let alone supra-competitive, price? The answer to this question is something pointed out by the first true mathematical economist, Augustin Cournot, in his Principes Mathematique, published in 1838. In that book, Cournot laid out “the problem of complements.” Cournot showed that imperfectly competitive firms overprice complementary goods. (Cournot’s example involved zinc and copper in the production of brass. They are complements used in fixed proportion.)

The basic issue is that when goods are complements, if firm A raises its price firm B that produces a complement to A’s good cannot steal sales from A by cutting its price (as would be the case of A’s and B’s goods were substitutes). This reduces the incentive to cut price, and actually provides an incentive to increase prices in order to get a bigger piece of the surplus that is generated when the consumer buys both goods.

This situation fits the stock market case perfectly. Execution services on US exchanges (e.g., NYSE, BATS) are substitutes, but data services are complements.

Consider an HFT firm. One source of profits for this firm is to exploit price discrepancies across exchanges. This requires having near immediate and simultaneous access to prices across all exchanges. Or consider a buyside firm that is trying to minimize execution costs by a clever order routing strategy. Optimizing the allocation of orders across exchanges requires knowing the order book on all the exchanges.

In other words, there are many market participants who have to collect the entire set (of exchange data). This makes the data provided by competing exchanges complements, which by the Cournot logic, forces prices above the competitive level, and indeed, above the monopoly level.

Furthermore, the problem becomes worse, the larger number of exchanges. This is a situation in which lower concentration leads to less competitive outcomes. (Robin Hansen made a similar point recently.)

This is yet another example of the only law that is never repealed: the law of unintended consequences. The intent of RegNMS was to increase competition in the execution of stock trades, and it has done a marvelous job of that. However, the unintended effect of this “fragmentation” (i.e., the increase in the number of execution venues and decline in concentration across exchanges) has been to create and exacerbate a complements problem in data.

A couple of final points. Perhaps one could make a second-best argument here: low execution fees and high data fees may be a good way of covering the fixed costs of operating exchanges (a la Ramsey pricing). Perhaps, but unproven.

What is the right regulatory response? Not clear. I addressed similar conundrums in my 2002 Market Macrostructure article. Natural monopoly-style/pricing regulation could mitigate the overpricing problem, but entails its own costs (e.g., undermining incentives to innovate). The issue is particularly challenging here because efficiency-enhancing competition on one dimension (execution) leads to inefficient problems-of-complements competition on another (data).

As I argued in Market Macrostructure, it really comes down to an issue of property rights. Should exchanges have exclusive ownership of their data? Should this ownership be attenuated in some way, such as limitation on prices, or a required pooling of data that would be sold by a monopolist, with revenues shared by the exchanges? Here is a case where a monopoly would actually improve outcomes.

Maybe that is the way to split the baby, politically. Exchanges would get rents, but efficiency would be improved. Not a first-best solution, but maybe a second best one, and one that could represent a Coasean bargain between exchanges and their customers. And perhaps the regulator–the SEC–could help facilitate and coordinate that deal.

Print Friendly, PDF & Email



    That right there is a rent scheme by ICE designed so they they can control who has their data. It’s a smaller market but a more egregious abuse of the monopoly power ice has in the energy markets.

    Comment by clf28264 — September 20, 2019 @ 6:17 pm

  2. One of the commenters on the Hansen article beat me to it, but this situation comes up in technical standards committees quite often when various patents are required to implement a spec. It is obvious to all involved that the standard will die ignominiously without a pooled patent scheme, but whose patents get into the standard and how the licensing is allocated is often a contentious topic (and beyond tedious to technical people like me who get roped into such proceedings). The conflict can lead to a split and competing standards.

    Comment by M. Rad. — September 22, 2019 @ 11:40 pm

  3. Any thoughts?

    Comment by Mario Gómez — September 23, 2019 @ 5:21 pm

  4. If you need orderbook data, you are either contributing to the orderbook or ripping it off. Perhaps a mix of both depending on the security/market. Why not give-get for those who contribute flow?

    Comment by cjc — September 24, 2019 @ 7:55 am

  5. “…or a required pooling of data that would be sold by a monopolist, with revenues shared by the exchanges? Here is a case where a monopoly would actually improve outcomes.”

    It is called a SIP – the idea to nationalize the orderbooks is not new. But does it make sense?
    Market participants want fragmentation and want to reference prices provided by exchanges at no cost at the same time. Explain how this makes sense?

    Comment by Joe — September 24, 2019 @ 8:00 am

  6. pardon me for having no sympathy for HFT firms etc. The exchanges also sold out the top step via co-location so HFT had a clear competitive advantage over the rest of the schmucks in the market. The exchanges forgot about retail, and ran roughshod over them.

    It is what it is. The exchanges should price to maximize their producer surplus. If you want to be in the business, you have no choice but to buy the data so the demand curve has a pretty vertical slope.

    Comment by Jeff Carter — September 24, 2019 @ 7:07 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress