One problem that people with a lot of time on their hands like to get worked up about is that academic economists sometimes write papers advocating positions that benefit organizations that give them money, while being coy about that relationship. On the other hand this newish paper about dark pools, which compete for stock trading orders with exchanges like NYSE and Nasdaq, has a first author whose affiliation is listed as “The NASDAQ OMX Group, Inc.,” so that’s fine then. Guess what he thinks? No, kidding, you don’t get to guess, he thinks dark pools are bad, duh.
The study, by Dr. Frank Hatheway, Nasdaq OMX Group; Dr. Hui Zheng, the University of Sydney; and Dr. Amy Kwan, the University of New South Wales, looks at US trading venues with restricted access and without displayed orders – generically referred to as “dark pools” – which increasingly segment order flow in the US. … The authors show that the effects of order segmentation by dark venues are damaging overall price discovery and market quality.
I’m a sucker for market microstructure papers because I like the Hobbesian world they imagine, where everyone is trying to rip everyone else’s face off, and keep their own face on, every nanosecond. The basic idea here is that liquidity providers – call them dealers – charge some price (call it the spread1) for trading shares with you. That price in part pays them for, like, hiring traders and buying computers and stuff, and in part it pays them for taking the risk that the shares they buy from you will go down and the shares they sell to you will go up. That risk should in some abstract aggregated sense be manageable – stocks go down about as often as they go up and everything should cancel out – but becomes scarier if you know what you’re doing. I guess some people do? It’s a particular skill; here it means being informed about the very short-term future path of stock prices – like, the time it takes the dealer to work out of the position he just took with you – and so tends to be the province of high-speed-type prop traders.
Anyway it is an unusual skill not generally found among big “real money” institutional investors, who tend to think in terms longer than five minutes. Now it turns out that if you can take the big people who don’t know what they’re doing – the Fidelities and Vanguards of the world who are mostly putting vast pools of mutual-fund-ish money to work in index-ish ways – and segregate them in one box, and then take the little people who don’t know what they’re doing – retail – and segregate them into another box, then that leaves you with a third box, which is disproportionately full of traders who have a good idea of what stock prices will do in the next five minutes, and the trick, for dealers, is just to stay the hell out of that box. That’s loosely speaking how the U.S. stock market operates: big institutional investors trade in dark pools, retail investors trade with internalizing dealers who pay brokers for order flow, and everyone else trades on the stock exchanges. Which are unpleasant places, for market makers, because they face the adverse selection risk of disproportionately buying from and selling to people who are selling and buying for a good reason:
Conditional on quoted spreads, liquidity providers on average incur a cost of 3.24 to 3.83 bps for bearing the adverse selection risk … if they trade on lit markets, but only 0.64 to 1.30 bps for such risk [on] dark venues. Liquidity providers on lit markets receive a mean spread of -0.06 to 2.40 bps, depending on the size of the quoted spread at the time they trade. On the other hand, liquidity providers on dark venues receive a mean spread of 1.18 to 4.12 bps.
So you want to get less informed people into dark pools, which is pretty simple: because it’s more profitable to trade with them, you can just charge them less. About 0.4 or 0.5bps less per trade:2
This is made easier because dark pools, unlike exchanges, can trade in increments of less than a penny, so you can always improve on the exchange price by a fraction of a cent. This makes dark pools attractive for certain types of traders, but the flip side is that you have to keep the guys who are trying to rip your face off out of your pool. You do this in part by making dark pools less attractive to face-rippers (no displayed orders, etc.), but mostly by just not picking up the phone when they call, or whatever the robot equivalent is:
[U]nder the current regulatory environment dark venues are not subject to the fair access rule and thus can prohibit or limit access to their services …. Our results in the previous section show that the adverse selection risk is significantly lower on dark venues, suggesting that dark venues preferentially screen for less informed orders.
This leaves the public markets – which, among other functions, effectively set the price on which dark pools trade – full of “informed” traders, which I guess sounds good, but maybe isn’t:
[F]ewer uninformed investors trading on lit markets is associated with significantly lower returns for liquidity provision on lit markets and higher transaction costs on both lit and dark markets. … [I]nformation fragmentation leads to higher volatility, wider spreads and less efficient mid-quote prices. …
Because market-makers on the lit venues have higher costs, they need to trade at wider spreads, and since lit-venue prices affect dark-pool prices, the market as a whole pays higher prices. Which is bad for everyone:
[A] 10% rise in absolute market share by dark venues is associated with a 0.16 bps rise in effective spreads or about 0.32 bps in the round trip costs of a trade market wide. Given that the average round trip cost of a trade is 7.08 bps in our sample across all market centers (not reported), this result reflects an economically significant increase of 4.5% in transaction costs.
Now, some math. Around 20% of trading in the authors’ sample is in dark pools, give or take. So by their math doubling the amount of dark-pool trading in the average stock would increase the average spread charged to customers by about 0.3bps; eliminating dark pools entirely would reduce costs by about 0.3bps. You’ll notice from the table above that the average stock trader in the average stock can save about 0.4 to 0.5bps by going to a dark pool rather than a lit market. So for Vanguard or whoever, “the market as a whole is 0.3bps worse off” is not a great argument. They’re saving 0.4bps over the market price by going to a dark pool, so on net they still come out ahead.
So the argument isn’t made to them. The exchanges have argued that the SEC should crack down on dark pools, and this paper provides ammunition. Its conclusions are explicitly regulatory:
Our study has important policy implications for the regulation of equity markets. By creating an environment which has come to emphasize information based segmentation of order flow the U.S. equity markets are increasingly characterized by intense competition within the distinct informational classes of order flow. But, there is little effective competition between lit markets and dark venues for low information content order flow. Consequently, the benchmark prices set for all classes of order flow represent the characteristics of just one class, the informed trader on lit markets. When only informed traders remain, the market may eventually collapse.
Nasdaq’s point is basically that its main constituents, high-frequency traders looking to take a few pennies off big investors, should have more access to uninformed traders, so they can take their money. Otherwise the market will collapse!
That might be true? There’s a pick-your-poison element; the high-speed trading of public markets also may be bad, or so say other academics. (Here’s a paper published yesterday called “High frequency trading tactic lowers investor profits,” which I’m sure is true; zero sum etc.) The argument here is basically that the big institutions that invest people’s retirement funds shouldn’t get to pick their poison. Instead of being able to trade in dark pools3 where they can’t be front-run or picked off by informed high-frequency traders, they should be forced to trade in centralized lit markets where market-makers can conveniently prey on them to subsidize their other, losing trades with informed traders.
Because market-making is a precious service, and regulation needs to carefully foster and subsidize those who provide it. Which is a funny thing to think? Compare the gloriously named “Tick Size Flexibility Act of 2013,” which would allow small-cap stocks to trade in 5 or 10-cent increments so as to make more money for market makers. Because then something something IPOs something something jobs!
Backers of the measure argue that wider trading increments – a 5 cent or 10 cent spread — will help increase the visibility of small publicly traded companies by giving small boutique investment banks the incentive to trade them more.
Presumably, small investment banks would also have an incentive to invest in hiring stock salesmen and analysts that would shine a spotlight on otherwise ignored small public companies. All of this also would convince many private companies to take the initial public offering plunge, knowing that their stock will not be ignored in the public markets.
You can’t buy that share of iPets for $20.53 because that small broker needs an extra two cents to pay its stock salesmen so there can be jobs. For the salesmen?
I dunno. These concerns are not, like, made up: liquidity provision really is a valuable service, adverse selection really does impact it, and market fragmentation really is a thing worth worrying about. But mostly, in a world where financial regulation is conceived of largely as a “war” against “Wall Street,” I’m tickled to see that some people think the opposite: that financial-intermediation services are so valuable that we need new regulation to protect them.
Dark Pools Are Costing US Investors Billions [Tabb Forum]
An Empirical Analysis of Market Segmentation on U.S. Equities Markets [CMCRC]
Wider tick spreads for small company stocks may be on the way [MarketWatch]
1. Half the spread, whatever.
2. This sorts stocks into buckets with small, medium, and large quoted spreads. Intuitively most people should probably care more about small/medium quoted spreads, which means bigger/more liquid stocks. On those, the effective spread (what the customer pays) is about 0.4 to 0.5 bps tighter in dark markets than in lit ones.
3. By the way, whoever named dark pools “dark pools” is an idiot on par with the idiot who named E&Y’s tax strategies unit “Viper.” “A safe, sheltered space for big real-money investors to buy stocks without being picked off by high-speed traders” sounds so soothing and nice. “Dark pools,” though, sounds terrifying. Don’t go to the dark pool, it’s a bad place. We better regulate it! It’s worth noting that Scott Patterson’s popular book about high-frequency trading and the general wildness of the lit equity markets is called “Dark Pools,” for no reason at all except that it sounds scarier than, like, “Electronic Communications Networks.”