This guest post was contributed by Ilya Podolyako, a recent graduate of the Yale Law School, where he was co-chair of the Progressive Law and Economic Policy reading group with James Kwak.
The development of the news coverage of high-frequency trading has been quite interesting. The story started out with a criminal complaint that Goldman Sachs lodged against Sergey Aleynikov, a former employee who allegedly stole some secret computer code from the Goldman network before departing for a new job in Chicago. Incidentally, Mr. Aleynikov appeared to be headed to Teza Technologies, a company recently started by Mikhail Malyshev, who had previously been in charge of high frequency and algorithmic trading at Citadel, a Chicago-based hybrid fund. Immediately after the report leaked, Citadel began investigating Mr. Malyshev’s departure and filed a lawsuit to prevent him from getting his nascent business off the ground. From these facts, some reporters inferred that the surprisingly public maneuvers of two notoriously secretive finance giants vis-à-vis seemingly routine personnel matters showed that Aleynikov had tapped into the gold mine of precious proprietary trading software.
That was two weeks ago. At this point, the story has crescendoed. The New York Times ran a report on high frequency trading. The Economist published a piece on the same topic. Senator Schumer (D-NY) requested that the SEC investigate the matter and the agency acquiesced.
The cynical perspective on these events is that both Schumer’s and Mary Schapiro’s moves with respect to algorithmic trading show that the issue is a red herring. As the argument goes, neither of these actors would touch the practice if it actually underpinned Goldman’s record profits or Citadel’s outstanding performance in 2004-2006. If, however, banning the practice would eliminate a few small hedge funds and create the appearance of revising market frameworks without threatening the big players (a regulatory brush fire of sorts), high-frequency trading would form the perfect political target.
However, I am not that cynical. I am friends with a few traders (yes, I know, I am tainted), and conversations with these guys always prove pretty interesting. At this point, the securities infrastructure is completely dependent on and permeated by high-speed computers carrying out various tasks. Some of these tasks are clearly benevolent and straightforward – matching buy orders with sell orders, reporting pricing – while others are probably benevolent but too opaque for us to be certain – concealing large orders by splitting them into chunks. Of course, certain practices made possible by advanced technology appear downright abusive.
To reliably distinguish the bad strategies from the good ones, regulators must have a clear idea of whom they work to protect. As the financial rescue has demonstrated, life-saving policies for some entities turn out to be a bullet to the head of others. For this reason, market rules should not be guided by some abstract idea of “fairness.” The concept is alien to trading, and attempts to squeeze it into policy fail quickly because everyone claims that the word means something entirely different. Rather, the SEC and CFTC should seek to remedy specific problems with narrow solutions. If investors of all political orientations can agree that fake prices on public exchanges are a bad thing, the relevant agency should end the practice and move on to the next obvious flaw in the trading process (there is no shortage of these). By contrast, a roving regulator on a mission to end all fraud before it starts will likely overlook actual problem areas in favor of broad banner initiatives that conceal problems instead of solving them. In this sense, regulatory policy should seek to fight actual enemies, not wage infinite war for infinite peace.
What does all of this have to do with high-frequency trading? Well, a longstanding theory of securities regulation divides investors in capital markets into three classes. Passive investors are individuals who use their savings to buy stocks for long-term gain. Passivity here is a virtue, not a sin: regular folks do not have the time to research individual companies in great detail, build their own projections, or actively manage a portfolio. They lack the incentive to do so too, since picking individual stocks is a near-certain way to underperform a diversified portfolio. Thus, the idealized individual investor passively provides capital for functional enterprises throughout the economy and gets a steady annualized return for their services over a long period of time. Moreover, diversified portfolios make passive investors largely indifferent to whether a particular company fails or pushes a competitor out of business, because the increased earnings of the winner in the fight will displace the losses in the investor’s coffers.
Institutional investors like mutual funds form a second group. These enterprises hold large fractions of stock in several companies, either for their own account or on behalf of passive investors. They have the resources to conduct thorough investigations of various corporations and can push the corporations to improve their business practices by electing their own directors to a set of boards. These moves are impossible for dispersed individuals to carry out because the costs of coordinating and persuading millions of shareholders to do something far outweigh the potential reward from a good decision. Institutional investors, however, immediately profit from good corporate governance because the value of their shares goes up, leading either to capital gain or performance-based fees. In a well-functioning market, institutional investors are responsible for discovering information and maintaining the competitiveness of individual companies. If one institutional investor uncovers key information that other institutional investors lack and acts on it, the less informed institutional investor loses but all diversified, passive investors win. Moves by institutional investors will usually focus on medium- or long-term profit because such bets allow the firm to amortize the significant cost of information discovery.
The third class of investors consists of speculators. These are individuals or companies making a series of short-term bets on a variety of indicators. Speculators are indifferent to whether a particular asset goes up or down, as long as they were correct in predicting the direction of the price change. Speculators rarely uncover significant new information about an asset on their own, because doing so costs a lot and because short-term price changes often depend on substantive information about what will happen to a business or a product far less than they do on what other individuals think will happen to the business or product. In other words, in the realm of speculators, rumors reign supreme. Some of these rumors turn out to be accurate, others turn out to be quite false, but the medium-term pricing of an asset sorts the two out. The role of speculators, on the aggregate, is to provide liquidity and supplement high-quality data held by institutional investors with a stream of additional information that may be worth looking at. Speculative trading activity by itself, however, neither adds nor subtracts value, since rumors (by definition ungrounded in previously known material fact) are equally likely to be true or false, and speculators sit on both sides of the transaction.
Within this framework, market regulators should strive to help each group fulfill its objectives without interfering with the roles of the other group. The government should not try to turn passive investors into speculators or vice versa. Instead, it should seek to prevent speculators from spreading intentionally false information, keep institutional investors from cornering markets and acting like monopolists, and protect passive investors from blatant theft a la Bernie Madoff.
In this light, high-frequency trading in itself is not the problem. Speculators can and should be able to buy or sell things as often as they like, since the number of times you make a bet on a roll of dice does not change the probability of a six coming up in any given situation. Passive investors, who provides the overwhelming bulk of capital to US markets, should not be hurt by high frequency trading because they should be buying assets at the prevailing ask price in 2009 to sell at the prevailing bid price in 2029. Over this holding period, their profit on the investment will be the same regardless of whether the buy and sell orders are executed by humans in yellow coats running around with paper tickets or extremely advanced machines. Indeed, since people generally cost more to maintain, passive investors should profit from any fancy games that speculators play against each other, provided that the SEC keeps the two camps separate. That is, a well-functioning regulator should explain to the public that the most speculators lose money because of transaction costs and the cost of capital, and that most individuals can get a far better return on their time from reading a novel than by feverishly checking Yahoo! Finance. The last thing we as a society would want is to have the SEC announce that, with the elimination of some algorithms, active trading has become a safe and reasonable activity for working families.
Flash orders, however, are an entirely different beast. By probing the market without the intent to complete a purchase or sale, these transactions erode pricing quality by eating up the distance between the publicly known order price and the limit price, a secret rightfully held only by the investor and his servant, the broker. This pattern of activity means that instead of listing information, providing liquidity, and helping set prices, speculators are actively destroying the ability of a quoted market price to represent the availability of contracts in the market. If I understand the practice correctly, flash orders are nothing more than a simple bait-and-switch fraud enabled by really expensive computers. They should be made as illegal as other forms of lies about financial products.
By Ilya Podolyako
The Baseline Scenario 
Makes a ton of sense to me. Information that isn’t fair or transparent only impairs the market.
Okay – speculators can take care of themselves, and passive investors will be okay in the long run.
But what about the institutional investors, whose investment decisions also affect many passive investors? They are not speculators (at least, they shouldn’t be), but they’re not necessarily long-term investors either. Seems as if they’re going to be harmed by HFT, and by extension, so will their passive shareholders…
The problem with the regulators responding to specific problems with narrow solutions is that the regulators are then always responding to an issue with a rule. This increases the number and complexity of regulations, in ways that often obfuscate the origins and reasons for the regulations themselves. It also ends up making the regulators play whack-a-mole — as each new problem comes up, a new solution must be crafted. You’re just trading one infinite war for another, one in which the regulators are always fighting a losing war because they never have the initiative on the regulated.
It would be better to have a wide umbrella of principles under which various practices could be judged and regulated beforehand, where the regulated would know that a contemplated “innovation” would be against the principles articulated, and therefore banned.
However, the largest problem I have with your suggestion is that these different market segments aren’t each trading in their own sandbox. Speculators are trading with passive investors and with institutional investors, and entry and exit into the market is often provided by speculators. When those passive investors are buying, they are buying at the inflated price in 2009, inflated by the speculators, and when they are selling, they are selling at the depressed sell price in 2009, depressed by the speculators.
The problem isn’t with the speculators playing games with each other — it is with speculators taking advantage of those entering and exiting the market. The SEC can’t keep these camps separate. The market is one unitary whole – everyone trades in the same pool, so when one group pollutes it, everyone has to deal with it.
High-frequency trading allows speculators to take advantage of the low transaction costs to slightly tilt the odds in their favor. Even if the tilt is only a tenth of one percent, when you’re dealing in the volumes that move in the market, you’re talking about real money. High-frequency trading is a practice that has very little to redeem it. And I think if you asked the passive investors and the mutual funds whether they would rather trade in a market with high-frequency trading or without, they would choose the market without it.
Passive, or long term, investors need not be diversified. I think the “longstanding theory of securities” you describe conflates two different aspects of investing. The “longstanding theory” fails to truly partition the collection of investors because it omits long term, non-diversified stock pickers.
The algorithmic trading tools used by HGT in the equities markets are clearly speculative activity. The appropriate regulatory response should be driven by the goals relevant to that activity. Organizing the issue into this structural framework is very useful
The focus of the recent reporting has been on the potential abuses resulting from unfair advantages in the equities markets that HFTs are able to exploit The prevailing counterargument is that there are legitimate and beneficial HGT strategies that exist and should not be curtailed. Although it’s arcane, most investors probably grasp the implications of the flash issue, but think the HFT issue ends there
However, the broader issue, which people haven’t been able to articulate well enough yet, is that the algorithmic trading isn’t limited to the equities markets. It is used simultaneously across all asset classes by sophisticated speculators.
The recent extraordinarily public reaction by Citadel and Goldman alerts us to the fact that the code competition is fierce. If it’s so fierce the profit opportunities must be significant. If the opportunities exist due to market flaws, an appropriate public policy response is to identify those flaws, eliminate the ones caused by bad policy, and let the market players arbitrage the rest. This is what seems to be playing out right now with Flash trading.
We’ll have to have faith that the remaining flaws are being addressed and resolved in the regulatory overhaul package currently underway. But, that’s unlikely, if the huge sums HFT powerhouses are investing in infrastructure and development.
Almost ignored, or at least, overlooked and underplayed in the code theft stories is that the algorithms at issue were for equities and COMMODITIES trading strategies. The HFT issue is not limited to equities trading. The algorithmic trading in the much less effficient, variously regulated, and in some cases unregulated contract markets is extremely lucrative.
Since HFT algorithmic trading strategies cross asset classes and simultaneously transact in all markets the debate and coverage of HFT needs to be expanded (and linked) considerably.
Hedge funds have been in this space for a long time. With the recent domination by gov’t guaranteed hedge funds (i.e GS, Phibro) in this space its time to shed some sunlight into the whole HFT space.
The Sergey story is the best vehicle for this. Sergey may have done us all a great favor, especially if he’s innocent, and Goldman overplayed their hand.
“can push the corporations to improve their business practices by electing their own directors to a set of boards. ”
that’s not what “they” did in Germany when “they” discovered we needed to be taught about shareholder value.
“They” sent suit wearers fresh from universities into the offices of seasoned company bosses with the explicit task of making them comply with the latest management fad/business theories. The smart kids had only their theories and were not interested in listening to any deviating point.
I hope the rest of your interesting to read post is not guided by a similarly romantic view of behaviour in the market.
“By probing the market without the intent to complete a purchase or sale…”
Fine post but not sure you have clue what flash orders are about, they are most definitely meant to be filled.
More info: http://tinyurl.com/nud3x8
Why don’t we all voluntarily give Goldman Sachs $50 per year?
The result will be the same except that they won’t have to run this HFT circus and we can put their software engineers to a better use.
Podolyako goes to great length to explain the different types of investors, apparently aiming at a general audience. But then he talks about flash orders as though everybody knew what he was talking about.
On another note, he said. “{Institutional investors} have the resources to conduct thorough investigations of various corporations and can push the corporations to improve their business practices by electing their own directors to a set of boards. These moves are impossible for dispersed individuals to carry out because the costs of coordinating and persuading millions of shareholders to do something far outweigh the potential reward from a good decision. Institutional investors, however, immediately profit from good corporate governance because the value of their shares goes up, leading either to capital gain or performance-based fees.”
From what i hear, they have not been doing a good job in the governance of individual corporations. Boards are prone to rubberstamping management decisions. And when an institution has stakes in hundreds or thousands of corporations, how much time, energy, and expense can it devote to the governance of any one of them? Unless something seems badly wrong, how much incentive does it have to intervene? And if things have gone wrong, does their duty to their own investors extend any further than selling their stock in that corporation? Besides, cognitive dissonance applies. We bought this stock when it had the current management, and we should support management, because we did the right thing. Arguably, institutional investment may be an overall negative for corporate governance.
Highly creative thinking bungalowbill. You are absolutely right. Just give the ‘best of the best’ their compensation. And the little man does need to suffer from their ‘puppet theater’.
Patrick Earnest: “The problem with the regulators responding to specific problems with narrow solutions is that the regulators are then always responding to an issue with a rule. This increases the number and complexity of regulations, in ways that often obfuscate the origins and reasons for the regulations themselves. It also ends up making the regulators play whack-a-mole — as each new problem comes up, a new solution must be crafted. You’re just trading one infinite war for another, one in which the regulators are always fighting a losing war because they never have the initiative on the regulated.
“It would be better to have a wide umbrella of principles under which various practices could be judged and regulated beforehand, where the regulated would know that a contemplated “innovation” would be against the principles articulated, and therefore banned.”
Well put. Hear, hear!
Patrick: “The problem isn’t with the speculators playing games with each other — it is with speculators taking advantage of those entering and exiting the market. The SEC can’t keep these camps separate. The market is one unitary whole – everyone trades in the same pool, so when one group pollutes it, everyone has to deal with it.
“High-frequency trading allows speculators to take advantage of the low transaction costs to slightly tilt the odds in their favor. Even if the tilt is only a tenth of one percent, when you’re dealing in the volumes that move in the market, you’re talking about real money. High-frequency trading is a practice that has very little to redeem it. And I think if you asked the passive investors and the mutual funds whether they would rather trade in a market with high-frequency trading or without, they would choose the market without it.”
Problems with high speed trading were evident in the 1980s, with programmed trading. In theory there should have been no problem, because the success of high speed trading depends upon exploiting inefficiencies in the the markets. You can even regard maintaining the efficiency of the markets is a public good. In practice however, such high speed trading increases the volatility of markets, and increased volatility means increased risk, especially for slow moving traders. A momentary imbalance between stock index futures (on one exchange) and the underlying index (on another) triggers massive trading on both exchanges. This trading can induce other speculative trading, with the result that market volatility is amplified. Leverage is important here, because both the futures markets and the high speed trading are highly leveraged. Leverage amplifies both volume and volatility. Perhaps the vulnerability of markets to manipulation is also increased. Because of leverage and the linkage provided by high speed cross market arbitrage, a market manipulator might be able to induce moves in stocks by a relatively small investment in futures. The footprint of the manipulator would also be smaller and harder to detect than in the past.
The author asserts that
“high-frequency trading in itself is not the problem.”
Because:
“Speculators can and should be able to buy or sell things as often as they like, since the number of times you make a bet on a roll of dice does not change the probability of a six coming up in any given situation. Passive investors, who provides the overwhelming bulk of capital to US markets, should not be hurt by high frequency trading because they should be buying assets at the prevailing ask price in 2009 to sell at the prevailing bid price in 2029.”
Whether or not the basic assertion is correct, the justifications are absurd.
The dice analogy is extremely poor, because dice are truly random, while markets respond to buying and selling signals, including those from speculators. As long as speculators make up a relatively small part of the market, their high-frequency inputs will be damped out. But when speculator money makes up a significant fraction of the total money in the market, then their erratic, high-frequency signals will drive the market into oscillations and greatly increase volatility.
This is basic control theory: if you apply high-frequency inputs of sufficient magnitude to any feedback system, it will oscillate out of control.
The idea that passive investors will not be hurt by the instability created by this activity is also ridiculous, and obviously so. Anyone who buys during an speculation-fuel runup of price is going to suffer, since they’re buying assets for more than they are worth in the long term. (And by definition, these investors can’t caveat emptor — they’re not sophisticated enough.) Likewise (and contemporaneously), anyone who is forced to sell during a post-speculative-bubble collapse (because, say, they’re old) will likewise suffer. In the meantime, the speculators are laughing all the way to the bank — even the losers — because they can still pay their mortgage and buy food, and they know they’ll be another bubble to surf along some day soon.
In the described groups of traders, unsophisticated and institutional investors add to the system and get some of the value that is created in return. Speculators on the other hand play a zero-sum game. By definition any gains they make are paid by inputs from other speculators or the other two groups (I will call them investors) of investors. If we insist that investors are protected from speculators then speculators must make their profits solely at the expense of other speculators. And while I am sure that Goldman Sachs would claim that they are the best of the best and can make their money that way, I doubt that it is really the case. There just isn’t enough money in the speculators pool if it doesn’t constantly take from the investors pool. If investors are protected, then speculation will necessarily decline drastically. On the other hand, if speculation does not decline drastically we will know that speculators are continuing to make most of their profits off of investors.
I’m curious about the Sergey Aleynikov story. Maybe he was just trying to take the code he had written for GS with him because it would be useful at his new job.
What’s the deal? If you write code for a company, does the code belong to you or the company?
Get ready for the short raid. See “Stock Shock” stockshockmovie.com to learn more about this…they’ve got one more shot in mind.
When speculators or market makers earn profit from “traditional investors” it doesn’t imply that the traditional investor “lost” something. When mutual fund buys a lot of stock there should be someone on the other side. In a lot of cases it would be arbitrageurs, market-makers and speculators which earn spread by providing liquidity. Basically in this case spread investors pay is a price for liquidity they demand.
And since all evidences point to spreads decreased in latest years, there is no indication that “traditional” investors pay more than before.
And please, guys, don’t mention flash orders. When I enter flash order I _want_ this order to be pre-exposed to selected participant, otherwise I would submit regular limit which is still available.
About flash orders. First, please have a look at http://www.nasdaqtrader.com/content/ProductsServices/Trading/Flash_factsheet.pdf
To me (from viewpoint of person in a algo trading systems business) it looks like a strange way of doing “discretionary orders” allowing a participant to post a limit order saying “I want to buy 100 @ 10, but if there will be an opportunity I’m OK with buying @ 10.1”. Such order types sometimes available on different MTF (Multilateral Trading Facilities), for example Turquoise.
Reasons for parties listening to Flash order via ITCH interface:
* utilize discretionary order functionality, allowing to take incoming orders at prices slightly worser than the ones you would want to expose in the book.
Reasons for parties to ENTER Flash orders:
* receive additional fills from participants who don’t want to fully expose their trading intentions. This fills might improve my price, especially if my orders takes topmost levels of book on the other side. By entering such an order I expect to move prices less and receive better execution.
Again, if I just enter regular order, no flashing is done, so this is an additional opportunity for the sending party.
Thanks for the link. I am not an expert on trading technology, so I was writing based on information I received from various relatively reliable sources.
What are the BOLT ONLY orders used for?
BOLT is their name for flash orders, BOLT ONLY is used to avoid moving the market, in case of a buy order, it will first lift offers at NBO (best offer) if any, then “flash” the balance or part of, for a short period to attract liquidity at that offer to that venue, it’s either filled, or cancelled if no liquidity shows up during the exposure period or becomes a firmn limit bid if the market moves higher. In any case, these orders are meant to be filled.
I should add that the reason these “flash” orders have a short life is that it is against the (SEC) rules to show a crossed market i.e. bid=offer.
Normally the company.
Ilya writes:
Surely in a bubb;e that’s not true? Just because a statement is a lie doesn’t mean that all lies are equally possible. Rumors may not be grounded in truth, but they are always grounded in interest.
Brent Eubanks: “As long as speculators make up a relatively small part of the market, their high-frequency inputs will be damped out. But when speculator money makes up a significant fraction of the total money in the market, then their erratic, high-frequency signals will drive the market into oscillations and greatly increase volatility.
“This is basic control theory: if you apply high-frequency inputs of sufficient magnitude to any feedback system, it will oscillate out of control.
“The idea that passive investors will not be hurt by the instability created by this activity is also ridiculous, and obviously so.”
Well said! :)
The only reason I can see that computerized trading has not been (more) catastrophic is that the vast majority of such trades are arbitrage, and provide negative feedback. But if they provide negative feedback, why are they not stabilizers? Because of leverage and volume. The interactions are complex, but the observed result is an increase in volatility.
Our current regulators think in terms of market efficiency, not in terms of control theory. Regulatory capture aside, one problem is the shared world view of regulators and financial professionals. If we do get a systemic regulator, better not have a banker, financier, or economist. Bring in an engineer, someone who understands complex systems and their stability. :)
I think this argument would be helped by stating why, precisely, the use of micro-orders (or – the worst case scenario which is actually what the SEC is proposing to ban – early private access to order information) to discern a secretly held reservation price is “bad”.
First, there’s the issue of “peeking” at information before anyone else (which supposedly was ony about 2% of trades, a sliver of all rapid trading). This let’s the trader see the bids/asks earlier than anyone. They can then issue a tiny order meeting the bid or the ask price, which becomes the last trade price for the stock. In this way (among others), they can influence the price up or down. They can also probe ceilings and floors with repetive small orders (all those 100 to 400 shares you see flying around cheaply priced stocks).
The markets (e.g. Nasdaq itself) like this because it increases liquidity. The problem has to do with incentives for investors to discover information about an asset or company. Large mutual funds move huge piles of money around, so large that liquidity actually becomes an issue. Little specks like me can by and sell 500 stock here and there without touching the price of AT&T, or the S&P 500. Not so mutual funds – thus when they invest they need it to be based on solid information. Information like that is expensive to obtain, but is one of the “socially useful” functions that investors perform.
If a pricing-discovery algorithm can effectively discover that information with little cost, then it can parasitize the investor who obtained the information. This decreases that investor’s reward, and thus decreases the incentive to actually invest in discovering _real_ information to feed into Mr Market.
IMHO, a Tobin Tax would be better than a ban, however, it would decrease this activity, PLUS it would raise some revenue.
Is brevity still a virtue? I can state the same thing in about half as many words as Ilya.
“The interactions are complex, but the observed result is an increase in volatility.”
What is the evidence of increased volatility due to automated trading? People talk about this as a proven fact, but I still haven’t seen any evidence presented in econoblogs/press. Can you reference anything?
Please note, that algo trading is used not only by arbitrageurs, it’s also used by funds (mutual, hedge etc) or their brokers. When you need to execute a huge order in a stock or in a basket, you can’t just send it to the [single] market (if it’s not dark pool) because of market impact.
kosmik: “What is the evidence of increased volatility due to automated trading? People talk about this as a proven fact, but I still haven’t seen any evidence presented in econoblogs/press. Can you reference anything?”
I do not know if anybody has done a study of the relation between high speed, high volume trading and volatility. It would be interesting to do. However, if you compare daily charts of the Dow Jones Industrials in the days before such trading with daily charts afterwards, you can see the difference. The later charts look much more like commodities futures charts.
“Please note, that algo trading is used not only by arbitrageurs,”
Yes. But my guess is that without the arbitrage the effect on volatility would be significantly greater.
If a songwriter records a tune with a company’s money I assume this specific recording will belong to the company. But I suspect the songwriter retains the rights to the tune itself so that he/she can make alternative versions of the song.
Every time Paul McCartney sings Yesterday during a concert I don’t think he has to pay Emi or whatever company he first recorded the song with.
Anyway, for a programmer it wouldn’t matter. Rewriting your own code from scratch would just be an incredible waste of time.
I assume $50 per year is just for HFT. For gas the GS tax would be much higher (remember summer of 08).
If I were an economist I would try to figure out the cost of things if speculation were not allowed. What percentage of our income is in fact given to unproductive financial institutions through speculation and other hidden taxes?
Bring in one with experience from balancing the power grid. The nice thing with the power grid is that it is not possible to fool it. You cannot sweet talk the grid into believing that there is not a peak in voltage. You can not fool nature.
As the black outs have learned you, they also have experience in having limited resources available.