A recent study by Tabb Group reported that the level of confidence in the markets continues to fall with every technology-associated market crash, based on the survey of 260 market participants. The study reported that following Knight Capital’s fiasco, 26% consider the present market structure to be “very weak.” By contrast, in June 2012, only 7% of market participants deemed market structure “very weak,” and in May 2010 only 3% of the polled parties thought so.
One of the regulatory solutions to boosting market structure mentioned in the poll is the idea of slowing the markets down. In fact, according to the poll, 31% of asset managers, 20% of broker-dealers, 18% of financial services vendors, and 10% of exchanges believed that slowing down markets “would help minimize the types of events we have seen in 2012, or, more broadly, help the industry regain the trust of investors.”
While the idea of slowing down markets did not generate a majority vote among any specific group, the support for the measure is quite surprising. It is particularly surprising to find 20% of broker-dealers backing the idea, as profitability of their primary business—market-making—grows with increases in trading speed.
A study of market-making on actively traded stocks on the Stockholm Stock Exchange, for example, found that the expected profit on limit orders increased as the time duration between market orders decreased. The study, written by Sandas (Review of Financial Studies, 2001) and confirmed by Beltran, Grammig and Menkveld (working paper, 2005), shows that market-makers’ profitability does not grow as they wait their turn at the exchange; instead, market-makers’ profit grows as their order matching frequency increases. In other words, modern market-makers’ profits are directly related to the number of market orders they service; the more trades the market-maker can take on within a given fixed period of time, the higher is the market-maker’s profitability.
By contrast, market-making theories of the 1980s presumed that market-makers’ profit increased with their waiting time to order execution, identifying slower execution as a driver of higher profitability. The thought went that market makers were patient traders and were compensated for their time making markets, not for the number of trades market-makers took on. Such models reflected exchange conditions circa 1970s. Changing the details of a limit order and cancelling orders was prohibitively costly, and market makers indeed expected a tidy compensation for bearing the risk of ending up in an adverse position once their limit orders were hit. In most modern markets, however, limit order cancellations and revisions can be performed free of charge at the time this book is written, and high-frequency market-makers enjoy better profitability than their low-frequency counterparts.
Still, some brokers who argue against fast execution are under the impression that the technology required for successful trading in today’s computerized markets is prohibitively expensive. The main reason behind such thinking also happens to be brokers’ past experience: just some fifteen years ago, the cost of purchasing technology required for a modern high-frequency setup was tens of millions of dollars. Today, comparative systems cost a few thousand dollars. The drastic reduction of technology in costs has been driven by two main developments, completely independent of financial services:
1. Demand for cheap yet powerful technology by numerous video-game players lacking funds
2. Overseas’ manufacturing capabilities
The very latest iterations of video games require even more drastic technological improvements: specialized chips to quickly receive and process information, much alike the demands of high-frequency markets. These chips, called Graphical Processing Units (GPUs) and Field-Programmable Gate Arrays (FPGAs), were originally designed for ultra-fast processing of video graphics. The chips are now making rapid inroads on Wall Street, where the technology is adopted by more serious applications, like market-making algorithms processing billions of dollars in positions a day. A blank FPGA chip costs as little as $100, yet, when properly programmed and installed in a regular PC, it can deliver performance similar to a cluster of thirty to three hundred interconnected computers. Such amplification of computer power delivers further savings to financial institutions seeking to modernize their technology via savings on computer power, ventilation and physical space.
In summary, fast markets improve market-makers’ profitability without outrageous cash outlays. The biggest cost of implementing modern market-making models lies in the know-how of designing and implementing of algorithms, a discipline hardly straightforward. The last two years, however, have witnessed an explosion of research and other information on the subject, bound to convert even the most hard-crusted anti-speed market-makers to the camp of profitable trading.
Irene Aldridge is teaching courses on design and implementation of algorithms in Chicago (September 5 and 6, 2012) and New York (September 20 and 21, 2012).For more information and to register, please visit http://www.hftcourse.com today. She will also be speaking on a panel about algorithmic trading at HedgeWorld New York 2012, to be held Sept. 19 at the Metropolitan Club.
This article was originally posted on Hedgeworld.com, click here to visit article.
Comments | Post a Comment
19 Comments to "Slowing Down Markets is the Wrong Way to Go":
Anonymous
28 August 2012
I question the validity of data from reseach on trading activity on a stock exchange that has and never will see the volumes associated with exchanges here in the US. Hardware, software and network capabilities have advanced tremendously in that 11 year period. The spate of breakdowns in the systems suggest we have reached a law of dminishing returns - the volumes and speed of HF, the risk vs reward as the market gets faster, sniping a mix of real or synthetic liquidity - the actual market back in 2001 had "real trades" not "fishing lures" hidden as quotes.
kennymcb
28 August 2012
"In summary, fast markets improve market-makers’ profitability without outrageous cash outlays." - Cheap profits creates a cheap mindset and that is where the problem lies - if you wont spend the money to solidfy your own infrastructure then everbody else suffers from sub optimal technocal strategies. The market becomes as weak as the lowest common denominator. Sometimes the democratization can go too far, with failures casting shockwaves that spread wider and undermine investor confidence - sounds like a perpetual negative cycle will occur unless something is done - this is not just about the "right" for people to create trading firms to jump inbetween liquidity to make money, its ultimately about confidence in the fundamental structure of the monetary system for the country.
Comments (49)
Anonymous
28 August 2012
"Since Eugene Fama, many academics believe financial markets are too efficient to allow for repeatedly earning positive Alpha, unless by chance. To the contrary, empirical studies of mutual funds spearheaded by Russ Wermers usually confirm managers' stock-picking talent, finding positive Alpha, however this work has been criticized and having a survival bias (not adequately accounting for censored observations). However, they also show that after fees and expenses are deducted, the effective Alpha for investors is negative. These results also explain why passive investing is increasingly popular."
bobbyh
28 August 2012
in theory, the markets were originally for created for "investors" if anybody can remember that, not hi speed skimming machines, if you cant wait 1/2 second for a bid or offer to be filled, you are not a investor, perhaps we create 2 markets, one for the computers to trade with each other and another for people who actually know what a company does and would like to invest in it. I am curious if the author has ever worked on the buy or sell side.
csparrow
28 August 2012
I'm pretty sure market makers would make more profits in a slow market with high natural investor confidence than in a fast market where the natural investors have all pulled out. MMs can't make money just trading amongst themselves.
I agree with bobbyh, the raison d'etre of the markets is to allocate capital not showcase technology. Market makers exist to serve natural investors, not the other way round.
As kennymcb states, the weakness is in the linkages between systems. I don't think Stockholm had nearly the degree of fragmentation nor the proliferation of dark pools that exist in the US today, i.e. far fewer linkages and therefore much lower systemic risk.
brokers want to slow the market down so their clients, both corporates and institutional investors regain confidence and come back into the market and pay the brokers commission.
Comments (37)
crammond1964
28 August 2012
until market abuse/surveillance has caught up and we can regulate and capture the cheetahs our markets cannot progress . Sadly manipulation play a far too higher % in our markets . > 35 %
Comments (252)
Anonymous
28 August 2012
This article is preposterous. First and foremost, to write an article about the profitability of market makers without touching on spread widths, and instead only focusing on speed of execution is ridiculous. In addition, the failure to address the complexity of systems needed to interact with the marketplace right now is absurd. Yes, it’s true that a single computer is much cheaper than it used to be, but what about the sheer number of computers you need now vs. 15 years ago? And this doesn’t even take into account the spend on AI, algorithms, developers, etc. And this is only the tip of the iceberg in terms of “what is wrong with this article”…
Anonymous
28 August 2012
I agree with Anon. This article is preposterous. It's scary to think the author is holding class soon.
Anonymous
28 August 2012
"I am curious if the author has ever worked on the buy or sell side."
Good point: soon market participants will be manned by math and physics phd holders and computer geniuses. No wonder real life finance and economics have no place left in there, these guys have never been trained in that other than from a pricing algorithm perspective. Just like banks are more and more often led by front office people with a typical trader mindset.
CrossingRoss
28 August 2012
High frequency market maker profitability? The faster the market, the greater their profits? While I am a firm believer in capital commitment, I am pretty sure that the Tabb's survey responses for "slowing the market down" is not about insuring that market maker profitability is secured. That you focus on the low cost of infrastructure as the greater equalizer completely misses the point that an increasingly low-latency marketplace introduces a number of market structure concerns.
For example, over the course of a usual millisecond trading day, there are 23,400,000 milliseconds...or chances for traders to miss one another (what Prof Bob Schwartz at Baruch describes as
"Fractured" markets). It also introduces the practice of repeatedly pinging the market with small trades to extract contraside trade strategy and order information (the more trading instances in a day the more opportunity to divide and conquer...maybe this is where the MMs profits come from?). And then there is reverse engineering...with the highly sophisticated capabilities of HFTs, they can quickly ping the market and then get in front, short, or trade alongside to their own short term profit benefit. (PS this is NOT liquidity provision.) Let's not forget the fact that few can match the technology (whether it is cheap to acquire or not as you say) or the staff to implement low-latency strategies...so we now create a two-tiered market between those who have this technology and those who do not.
Your article speaks nothing to the quality of price discovery, liqudity discovery, or a level playing field for all market participants. It is not that speed is not important but it should not supersede price or liqudity discovery...and the interests of the long-term investor.
anonymole
28 August 2012
Irene, I wonder if ABLE Alpha might have a ventured interest in ensuring the continued and expanding use of HFT as a means for generating income for yours and other high speed trading focused enterprises?
"Able Alpha Trading LTD. -- a cutting edge quantitative investing and high-frequency trading RIA. "
The argument for favoring traders over investors is breaking down. And please do not confuse the two. We've explained numerous times in these forums that they are not the same practice. It's natural for you to try and curb the animosity against your company's targeted line of business. But face it, the return to an investor favored set of market regulations is just getting started. Best start shifting you business practices now.
Comments (83)
sarnuk
28 August 2012
Petterfy: http://www.npr.org/blogs/money/2012/08/27/159992076/a-father-of-high-speed-trading-thinks-we-should-slow-down
Comments (160)
lehalle
29 August 2012
Interesting viewpoint Irene. Nevertheless what you said about GPU and FPGA is not true. FGPA are more expensive that $100 each, are standalone (and not to be installed on PCs). Moreover GPU are not really useful for trading since then need to be "plugged" on motherboards (servers or PCs) and thus for computing on high frequency flows of data, their weak point in the "data bus" that has to be faster than the common ones. For military and aerospace purposes, such HF data bus exist, but they are very expensive. Hence GPUs are good for Monte-Carlo simulations and the associated derivative pricing and risk computation capabilities, and not trading. Another detail: FPGAs are not graphic cards but generic computing units, existing for long (before the raise of 3D games); they have been proposed by chip builders to lower the price of manufacturing small numbers of chips for specific clients. FPGAs are programmed in two layers: the "circuits" that can be designed a soft way (cheaper that burning a new chip each time you need a new design) and the "software" (this is what you expect to be). FPGAs are used for long in data flow analysis (for pattern matching, video analysis, biological simulations, etc), see for instance Evaluation of the CNAPS neuro-computer for the simulation of MLPS with receptive fields Biological and Artificial Computation: From Neuroscience to Technology, by Bertrand Granado and Patrick Garda. They are good hardware for fast trading, and actually used that for.
For more details about using hardware for trading you can have a look at High Frequency Trading Acceleration Using FPGAs, by Christian Leber, Benjamin Geib and Heiner Litz.
To discuss your original point ("slowing down markets"), my opinion is that we should go back to the idea of the purpose of "market design". If this is more than a void concept, it means that choices have to be made to shape the dynamics of the price formation process in order to improve the efficiency of the price dynamics. Here "efficiency" means that the double auction game confronting the offer and demand (since they are said to reflect rational choices of investors) converges towards a "fair price" for the benefit of the listed firms.
I do not see what is your proposal other than "let it be"? What are, according to you, the area of improvement? none? We would be very lucky if the actual regulations (in US and Europe), fired the perfect bullet more than 8 years ago, before the crisis, setting the basis of a market designed adapted to 2013 needs. "Slowing down the markets"? I think that the question should be restated as "modifying the price formation process dynamics putting more emphasis at liquidity providing instead of liquidity capturing". The slower the market (one way or the other, increasing the tick size could be one way), the more you need to act as a liquidity provider to obtain a transaction, the faster the market the more you can "jump the queue" and capture liquidity an opportunistic manner. Thus: the faster the market the more advantage to opportunists and the less to long term investors that have to take their investment decisions before launching their trading process.
I like your reminder of 1980ies theories. My analysis is that an important difference between these "former theories" and today is that we now know that each investor is simultaneously liquidity provider and liquidity consumer (it is contained in the papers you cited). You do not have only two cliques: the long term investors who just send market orders and market maker using limit orders only. As an example, a typical brokerage algo obtain at least 1/3 of its shares passively (to obtain a slippage to the VWAP around -1/3 of the bid-ask spread). It means that the competition takes place among liquidity providing orders, sent by all market participants. According to me this is a very important point, for more details about the implications, see Navigating Liquidity 6: A global menu for optimal trading.
Comments (2)
ialdridge
30 August 2012
Ah, I really like the point of Anonymous "Since Eugene Fama,...". Very well said, thank you.
@bobbyh, yes, I've worked on both buy and sell sides, and not at the same time (duh!).
@csparrow, lack of investor confidence is a popular buzzword, but perhaps the real issue is poor economy (hence, low stock returns and disinterest of investors?)
@lehalle, I have actually built these systems, and have to disagree with your FPGA cost assessments from practical experience. GPUs are indeed much more expensive and less robust.
@anonymole. I am not ashamed of deploying technology to help firms replace costly "fat cat" brokers with reliable computers. Contact me to learn how your company can save millions and millions of dollars (steep consulting fees apply!)
Comments (1)
crammond1964
30 August 2012
Hopefully we can weed out the "cheetahs " in HFT ; it would help if their industry admitted they had problems ; then the HFT can provide the role they have been generously given .
Comments (252)
csparrow
30 August 2012
@ialdridge, there are many factors that impact confidence. A poor economy didn't cause the flash crash or the Knight problems. A poor economy didn't cause the BATS or Facebook IPO problems. A poor economy is likely to contribute to reduced volumes, though, but I don't think we can attribute all the reduced volume to the poor economy.
Systemic risk in the equity markets would still be high even if the economy was much stronger.
Comments (37)
anonymole
30 August 2012
• HFT is here to stay. That's obvious. Those that would regulate it are in the pockets of those who benefit from it. We're fools to try and combat such a thing.
• HFT will become UHFT and eventually every nanosecond will be squeezed from transactions. Every player will be as close as they can possibly get to the exchange/darkpool/ecn endpoints. Every HFT algo will be AI enabled and automatically adjust to the behavior of the markets. Eventually the ROI on HFT investment will flatten and cease. Eventually the players will isolate themselves so well that the HFT game will be nothing more than bad dog eating bad dog.
• Perhaps this is an opportunity to build another type of exchange. An investor exchange. A non-for-profit exchange that trades on a paced time frame. An exchange that hosts corporate stock transfers that elect to have their shares traded in an investor friendly venue. Where new IPOs are handled with utter transparency. A new exchange where investors can feel safe knowing that the bad dogs are not allowed to play. Let the HFT predators consume each other over there. Let's build the investor friendly exchange over here.
Comments (83)
jimross1313
31 August 2012
There IS a new market model which neutralizes the tactics and games of HFT and predatorial algos. It is called AX. It is the FIRST ATS to promote price competition and liquidity discovery on a LEVEL PLAYING field for investors and broker-dealers. Speed to market has no consequence within the AX market because AX uses a call auction model similar to the Open and the Close which aggregates multiple orders in a single moment (ie non-continuous). What is essential is that AX enables traders to initiate their OWN 5 minute auction at any time from 8 am-5 pm EST and invite whomever they want to participate in the auction. During the ensuing 5 minutes, participants respond with priced orders that go into a sealed bid book. Neither the initiator nor the participants see one another’s orders and at the end of 5 minutes, they are all aggregated at their various price levels. At that time, the auction price is determined by where the most shares will trade...not who is the fastest but based on PRICE. Orders priced at the clearing price or better are allocated shares based on a pro-rata basis. Price competition finally meets liquidity discovery…but it has taken an ATS, the AX ATS to do it.
AX is particularly effective for liquidity that has been forced outside the market by the insanely frenetic and increasingly less relevant NBBO. Many AX traders price their orders aggressively through the NBBO (seeking liquidity) or passively outside the NBBO (providing liquidity). As a result, nearly 1/3 of all AX trades occur OUTSIDE the NBBO (and of course, honoring Reg NMS Trade-Through Rule).
I agree that the time has come for change in our market design. A market design that is built upon the fundamental tenets of price competition, liquidity discovery and a level playing field for all market participants. Exchanges have made their choice to cater to the HFT community through co-location, fee rebates and billion dollar data centers....AX instead focuses on quality over quantity and puts its market in the hands of the long-term investor and the broker dealer community that works hard to support them. Both investors and Brokers desereve a market that is not going to compete with them and/or cater to a specific market group for its own self-interests. Markets must serve investors and brokers alike…..not shareholders. It is time for us to get back to our financial market basics….and AX is a great place to start. (www.ax.com)
Comments (1)
anonymole
31 August 2012
Well damn! That is certainly timely. Ask and ye shall receive I suppose. I hope it works out.
Comments (83)