Is the Buy Side Getting Screwed in the Dark Shadows?
Dark pools are under a regulatory and legislative assault, in Europe, Canada, and the US. Scott Patterson wrote a long piece in the WSJ about FINRA’s scrutiny of Dark Pools:
The Financial Industry Regulatory Authority, Wall Street’s self-regulatory body, last month sent 15 examination letters to operators of “dark pools”—lightly regulated, off-exchange trading venues that have been a rising concern for regulators and some investors as more activity shifts away from exchanges.
Finra is seeking details about how the increasingly popular venues operate, what they disclose to clients and whether they adequately police trades. It could bring enforcement action against dark-pool operators or issue recommendations for tighter oversight, depending on the answers it receives and additional examinations, said John Malitzis, executive vice president of market regulation at Finra. The letters are a follow-up to an initial round of questions the regulator circulated last fall.
“We want to understand whether [dark pools] are disclosing to their customers how their orders work [and] whether customers are informed who their orders will interact with,” Mr. Malitzis said in an interview. “A big part of this is to get an understanding of practices that may or may not be problematic.”
Here’s an interesting bit:
Dark pools don’t disclose traders’ buy or sell orders and only publish trade data after transactions occur. About 13% of all stock-market action takes place on dark pools, up from about 4% five years ago, according to Tabb Group, a market-research firm. Most dark pools are run by broker-dealers—firms overseen by Finra that buy and sell assets on behalf of customers as well as trade for their own accounts.
While dark pools command an expanding slice of trading volume, regulators still have little idea about how they operate since, unlike exchanges, they aren’t required to regularly disclose detailed information about their trading systems.
I see. “more activity shifts away from exchanges”; dark pools “command an expanding slice of trading volume”; their volume has more than tripled in 5 years. Growing volume! Growing market share! That sounds sinister, doesn’t it?
Note that the possible nefarious practices under investigation, if they occur, would result in the users of dark pools are getting screwed. If that’s true, why are they flocking to the pools in increasing numbers? The most plausible explanation for a growing market share is that dark pools offer better value-not worse-to those who have the opportunity to choose where to trade.
I am particularly skeptical about the idea that dark pools are hugely exploitive because sophisticated buy side firms are among the most extensive and loyal users of dark pools. Buy side firms analyze their execution costs with considerable detail. They evaluate execution costs in lit and dark markets, and slice and dice the data in sophisticated ways, uncovering the circumstances under which it is cheaper to execute in one venue or another, and allocate their trades accordingly.
In other words, if they decide to execute more in dark markets, it’s because they have decided after thorough analysis that they are more likely to get screwed in broad daylight on exchanges.
This isn’t rocket science.
There is a potentially more intellectually reputable case against off-exchange trading. Dark pools utilize a variety of techniques that are designed to screen out informed, or opportunistic, traders. Uninformed traders-firms or individuals trading for portfolio balance or cash flow or hedging reasons-suffer losses when dealing with the informed. They can reduce execution costs by trading in venues where there is a lower likelihood of dealing with a better-informed trader.
But this “cream skimming” of uninformed order flow by dark pools that screen out the informed increases the trading costs of those uninformed traders who for whatever reason cannot avail themselves of dark markets: those left behind have a higher likelihood of dealing with someone with better information, and incur higher trading costs (due to adverse selection).
Dark pools free ride off the price discovery in lit markets, and cream skimming impairs price discovery. Thus it is possible-possible-that this externality reduces welfare.
Possible, but not necessarily so. The free riding externality reduces welfare if there is no other inefficiency in the lit markets. But that is not necessarily the case. As I pointed out about 15 years ago, there are other sources of inefficiency on exchanges. For instance, exchanges may exercise market power. In this case, second best considerations imply that the free riding externality may improve welfare because the benefit of increased competition for order flow from the dark market may more than offset the cost of degraded price discovery. I have produced formal models that show that this can occur under conditions that have existed in securities and derivatives markets.
As another example, much informed trading can be a form of rent seeking. Market participants expend real resources to produce information because this allows them to profit at the expense of the uninformed. This profit is a transfer from the uninformed, and unless there is some external benefit arising from more informative securities prices, the real resources expended to gain this transfer are a pure waste. By reducing the profitability of informed trading, dark pools reduce this rent seeking and improve welfare.
Both of these analyses are applications of the theory of the second best, which says if there are multiple departures from “perfect” conditions (perfect competition, no externalities), eliminating one imperfection may actually make things worse, not better. In the dark pool case specifically, if there are other imperfections (imperfect competition on and between lit markets, rent seeking informed trading), dark pools’ ability to free ride on exchange price discovery may improve welfare because it mitigates these other inefficiencies.
Determining the welfare impacts of dark pools in the presence of these various departures from first-best conditions is devilish hard. For instance, it is arguably impossible to determine the costs and benefits of informed trading. It is also difficult to determine the degree of imperfect competition in lit markets, and the costs of this market power.
But what can be said is that the kinds of things FINRA is looking at are totally irrelevant to addressing these issues. FINRA is focusing on activities that would be agency problems, or fraud. But it is hard to square the proposition that these problems are widespread with the observation that dark markets are growing relative to lit ones, especially given that most of the patrons of dark markets are quite sophisticated, and devote considerable attention to evaluating the cost and quality of the services that they obtain.
So what is going on here? Quite likely political economy, capture, and rent seeking. For the past couple of months exchanges have been importuning regulators to do something about dark pools, precisely because they are bleeding business. This headline from Bloomberg says it all:
Stock exchanges seek curbs on dark pools to curb exodus.
In other words: exchanges are losing a competitive battle, and are recruiting regulators and legislatures to hamstring the competitors that are eating their lunch. A very, very, very old story.
I have to give the exchanges credit. They have done a marvelous job of shaping the narrative, playing skillfully on the quite different connotations of light and dark. The exchanges are like a Scout master telling a ghost story beside a campfire, his face scarily illuminated by a flashlight. They have proven masterful at scaring regulators into action against the evils lurking in the dark
My bottom line? It is possible that dark pools are, on net, inefficient, but the case has not been made. Moreover, their commercial success should be considered as evidence that they actually enhance welfare. Indeed, their commercial success is almost impossible to reconcile with claims that dark pools are somehow deceiving or defrauding those who trade on them, especially given that these traders are often quite sophisticated, and quite on guard against getting screwed.
So don’t be afraid of the dark. Be especially skeptical about stories detailing the horrors allegedly stemming from the activities of entities that are gaining market share. And always remember that incumbents are past masters at inducing regulators to crack down on inconvenient competitors. Be afraid of what regulators do in the light, not what goes on in the dark.
Great topic, Prof! This truly is a fascinating subject. I have a bit of a different take than your’s re rent-seeking and its consequences.
There’s an interesting tributary in the market microstructure literature re broker-dealers as market-makers vis-a-vis the aggregation of information. See, e.g., Chae and Wang (2003, prior to the financial crisis, IMO), “Who Makes Markets? Do Dealers Provide or Take Liquidity?” The authors note, and I quote: “Standard models of market maker trading imply a negative contemporaneous correlation between market maker order flow and stock returns. We test this relation with a unique dataset containing trades of all dealers in a well-developed, liquid market. The correlation is strongly positive, implying that dealers take liquidity. Furthermore, dealers earn significant excess returns, in aggregate driven by information returns rather than market-making returns.” (Here’s the whole paper: http://web.mit.edu/finlunch/Fall03/AlbertWang.pdf . I don’t think it made it to a journal, but it is very insightful. They use Taiwan equities and position data in their research, but, homo economicus is consistent in his behavior, as Albert Kyle (cited in the paper) demonstrates, so this is generalizable.) Among other things, this paper points out the obvious: “providing dealers with low transaction costs, high transaction speed, and access to order flow information, is clearly valuable.” (p. 31)
Having spent most of my life in markets, I can tell you the only sensible strategy for a trading house operating in reasonably efficient markets is to aggregate as much information as possible at the lowest possible cost, then allocate your own capital in light of what that information tells you. The best information available in trading markets is to know who’s on the bid, who’s on the offer, what size does each side have, and what information are they trading on. That’s why trading houses have sales forces: they are the pointy end of the spear in terms of gathering real-time information from customers who are transacting. Even knowing why customers are not transacting is valuable information.
The second-best source of information is to provide the platform on which your “customers” trade — think Enron online. When you literally can see — and mine — your customer order flow you have the best information on the street. Over time, you can even figure out which way your customers’ books are leaning and fade your bids and offers on that basis (in real-time of course … you’d never leave a bid or offer sitting out there for an informed trader to hit or lift at their discretion). Operating a dark pool provides this opportunity.
If you’re able to have a customer interface via a sales desk AND operate the trading platform, you likely will have the best information available on which to allocate your own capital when trading against “customers” and counterparties.
Then it gets really interesting. Consider the incentives when a broker-dealer has really good information that the preponderance of order flow wants to sell, which would put the broker-dealer on the offer too, but a customer has more to do than the dealer. The dealer clearly has an incentive to front-run this “customer,” and may absorb all of the available liquidity on the bid. Who’d want to catch a falling knife when the dealer exhausts the bid side? What if that customer is a pension fund and comes to the dark-pool operator to re-balance its portfolio? Is the broker-dealer(s) in the dark pool incentivized to provide liquidity to the pension fund or to capture the profit opportunity for its own book? Chae and Wang — and common sense — would suggest the latter, since it would allow the broker-dealer to cover his short at an advantageous price.
Running a dark pool is the most efficient way for a broker-dealer to increase his signal-to-noise ratio, which can greatly improve his risk-adjusted returns in a low-leverage operating environment. This is, effectively, substituting technology for leverage, and minimizing your risk of loss (i.e., maximizing your probability of gain) on high trading volumes when you do decide to put your own capital at risk.
This is fairly obvious stuff to the trading markets, but it may not be as clear to the “customers” of the broker-dealers, be they in dark pools or well-lit markets. At the end of the day, it is pure rent-seeking by the broker-dealer, albeit elegant. They’ve paid to improve the quality of information on which they trade, now they’re in a position to realize excess returns at the expense of less-informed “customers” and counterparties. “Customers” will, over time, determine whether the benefits of higher liquidity compensate them for being the source of steady rent payments to the dark-pool operator. At some point, they’ll have to determine whether it pays them to attempt to capture that rent, which would require investment, or take their volumes to well-lit markets.
Quickly absorbing available liquidity — either on the bid or the offer — continuously, such that the next entity to transact has to cross a wider bid-ask, can be problematic for market efficiency, and, in the limit, deleterious for capital allocation. And, as Enron online demonstrated, it can concentrate information as to which way customers’ portfolios are leaning (i.e., long or short), which can be a huge incentive for the broker-dealer to show bids and offers designed to extract additional rent from customers’ order flows. Together, these effects can induce price volatility as broker-dealers acting as market-makers monetize their information. Chae and Wang note: “While dealers may be meant to perform the socially beneficial function of liquidity provision, the institutional advantages granted to them also give the ability to act as super-efficient proprietary traders if they choose to.” (p. 2)
Cool topic. Fascinating stuff.
Comment by markets.aurelius — June 9, 2013 @ 10:42 am
@markets-yes, fascinating. Thanks for your thoughtful and provocative comment. A more detailed response later . . . but a couple of initial thoughts.
1. I wonder about how this works out in the long run equilibrium. The information rent induces entry (new firms, the commitment of additional capital by firms). The net effect on transactions costs may be a wash. Relatedly, competing broker-dealers competing to get the valuable order flow cut fees. In equilibrium, the lower fees and the information rent add up to about the same fee as if there is no information rent. Total compensation is the same: the breakdown is different.
2. If customers are being exploited by broker dealers operating dark pools, dark pools not operated by broker dealers would have a competitive advantage. Again, broker-dealers who can profit from reducing the signal to noise ratio would be expected to offer lower fees to customers than stand-alone dark pools would.
In a nutshell-in equilibrium, where all prices can adjust, the effects of a sort of externality like one you’ve identified (servicing order flow generates information that allows profitable trading) can be complex. My rough intuition is that the effect on customer trading costs is pretty much a wash. The information rent is competed away in lower prices of other services that BDs provide.
In these sorts of situations I think of an interesting story in Donald (now Dierdre) McCloskey’s Micro textbook. In the 18th and 19th centuries, British shipyard workers could take home and sell wood shavings. The conventional story was that this was inefficient, and enriched the workers: they had no incentive to shape wood carefully (because they could sell any excess shavings), and got a windfall from the sales. McCloskey argued, and provided empirical evidence in support, that the main effect was to depress wages of shipwrights. There was a compensating wage differential: shipwrights were willing to accept a lower wage because they made up for it by selling the shavings. In equilibrium, the workers earned their reservation wage. Some of that was the formal wage, some in the form of sales of shavings.
This is quite plausible, and I think the like would apply in financial markets.
It’s not clear that this compensation structure would provide an appropriate incentive to conserve wood. But once one takes transactions costs into account, it is likely to be the efficient outcome. Monitoring to induce conservation of wood is expensive. It was probably more efficient for the shipyards to tolerate some waste of wood because the cost of reducing waste exceeded the benefit.
Getting a full understanding of what is going on in financial markets would require a similar equilibrium analysis, which would take transactions costs into account.
It gets complicated, no?
Complicated, yes.
I have a different notion of equilibrium: I do not view equilibrium as an evolution toward some long-term stable realization in which the supply of and demand for assets traded from time t = now to time T = some time in the longer-term future — S[t(i)], D[t(i)] — resolve in a price — P[t(i)] — that ultimately reflects some Pareto-optimal configuration. Markets are too dynamic for any such evolution.
The markets these firms operate in — trading markets — move continually from instantaneous equilibrium to instantaneous equilibrium that is entirely a function of the information set available at time = t* = I(t*). An equilibrium occurs at time t* when the market clears: i.e., the instantaneous price realization = P[t*(i) | I(t*)], in which S[t*(i)| I(t*)] = D[t*(i) | I(t*)]for 0 < t < t* < T. The set {I(t)} evolves randomly, since information affecting S[t(i)] and D[t(i)] is transformed randomly (e.g., by weather, war, etc.). Therefore, for all intents and purposes, prices, supply and demand follow random walks.
The only way to earn returns in excess of the risk-free rate in such environments is to aggregate information in a manner superior to one's rivals. The market has to provide a return for the information-gathering function, or it will not occur, given it is a costly pursuit (see Grossman and Stiglitz (1980) at http://www.aeaweb.org/aer/top20/70.3.393-408.pdf ).
It is not clear that the information-gathering process per se is a zero-sum game. For example, if I excel at that task and am able to provide liquidity when no one else can (because I have superior information and am willing to take the price risk providing liquidity entails), am I not providing an essential service to other market participants that is valuable to them? Clearly the answer is yes.
So far, so good. However, if I now set up and control the market in which these risk transfers occur — that is to say, I advantage myself by controlling the information flow vis-a-vis bids and offers (prices and size); I set the trading rules; I front-run the "customers" and counterparties that trade in my dark pool — I have monopolized the provision of liquidity.
Markets in the U.S. historically have been operated as a public trust — for the better part of our history, we have tried to ensure they are disinterested arenas in which numerous buyers and sellers can participate in a "fair game" (the economic meaning of which is buyers' and sellers' expected value of a zero-risk rock-solid certain profit is zero when that trade is entered into unless one is trading with inside information. See http://www.econlib.org/library/Enc/EfficientCapitalMarkets.html ). If that arena now is dominated by a dark-pool operator, forget about fair games. One could easily argue the public-trust of the market has been corrupted. How long can the capital-formation function of markets survive in such a world?
Complicated … yes 🙂
Comment by markets.aurelius — June 9, 2013 @ 5:03 pm
I believe that dark pools fragment the market. I’d rather see bids and offers congregate around one bid ask-with all the competition there. As I understand it, the reason dark pools started were two fold. The operators wanted a nice playground to trade against their customers-they could charge them a far cheaper price to entice them. The parallel is that in commodity markets I traded, if I could have stood on the top step, and traded against customer order flow; I would have paid to get order flow because of the advantage. The other problem was the big orders were being front run by specialists on the NYSE and NASDAQ. Pension funds etc wanted to trade anonymously. Dark pools allowed for that.
Dark pools are an outgrowth from a structural/regulatory/ethics problem. If they ended payment for order flow, ended the ability to trade against the customer and forced operators to choose to trade-or earn money filling orders-dark pools would end. Customers would look like they are paying a higher price-but I bet their “all in” price when you figured in slippage and the thicker bid/ask spread that would result-would actually be cheaper than costs the way the market is set up today.
Information is good for markets. What we are seeing because of regulation (including Frankendodd) is more fragmentation, and less current information on bid/ask spreads.
Comment by jeff — June 10, 2013 @ 4:50 am
[…] Is the Buy Side Getting Screwed in the Dark Shadows? Dark pools are under a regulatory and legislative assault, in Europe, Canada, and the US. https://streetwiseprofessor.com/?p=7321 […]
Pingback by The week that was (aka Dazzling Derivatives; issue of 10.6.2013) | The OTC Space — June 10, 2013 @ 10:19 am