Whitney Tilson Probes Deeper Into High Frequency Trading
From Whitney Tilson's letter to clients today:
Lots of feedback on the high frequency trading issue. I spoke with Joe Saluzzi of Themis Trading -- he and his partner's comments are below.
1) I've been in this business for 10+ years, know a lot of people and read a lot -- but until a few days ago, I'd never heard an explanation of what high-frequency trading is and why it's so profitable.
2) If the Themis guys are right -- and I think it's very likely they are -- then I think what the exchanges are doing is analagous to the mutual fund scandal from a few years ago, when many mutual fund companies conspired with their biggest customers to steal from all of their other customers in exchange for a share of the loot. The mutual fund scandal struck a chord because average investors were being bilked, whereas today it's mostly the Fidelitys of the world who are paying the price -- but guess who Fidelity's investors are?!?!
3) To the extent that HFT strategies are creating more volatility, that's great for investors like us because we don't look to the market for signals. If HFT strategies (and quant strategies in general, which are often momentum-based) push stocks below where they otherwise would be, then we can cover our shorts and buy longs at better prices.
4) HOWEVER, increased volatility and prices moving to extremes is HORRIBLE for most investors, our markets and financial system in general. If we buy a stock at $10 and it drops to $5 on no news, then it's an opportunity for us to buy more -- but we're unusual. Most investors will think that someone knows something they don't, panic and dump the stock. That's good for buyers like us, but excessive volatility shakes investor and consumer confidence.
Also, as we're seen, declining stock prices can lead to a vicious cycle for many types of regulated business and companies with leverage, as the declining stock prices leads to the market and/or regulators calling on them to raise more money, which they can't do -- because of the depressed stock price! It's so ironic that the heads of Wall St. firms were blaming short sellers for the collapses in their stocks, yet to the extent that exogenous factors (beyond their own leverage and losses) were behind the declines, it was more likely to have resulted from strategies employed BY THEIR OWN PROP DESKS!
In summary, the SEC and/or the NYS Attorney General need to start looking into this IMMEDIATELY. They've been spending all sorts of time and resources investigating short sellers, but I think it's likely that they've been barking up the wrong tree... A quick change such as the one Themis proposes, requiring that orders be valid for one second -- one second! -- might get rid of a lot of this nonsense and certainly wouldn't impact other investors.
1) Comment #1:
Why does this surprise you? This is a natural outcome when technological advancement hooks up with Wall Street....you get
devilish behavior. Buffett's partnership compounded at 24% from '57-'69; Renaissance and SAC have done much better; is it b/c Cohen and Simons are that much better than Buffett or......?
2) Comment #2:
A few points:
- Look at Themis Trading and what their incentives are.
- Why is it a moral issue when people exploit markets to redistribute wealth? Is that not what every hedge fund manager is trying to do?
Take advantage of its counterparty?
- There are many other investing niches which are focused on gaining an informational edge over counterparties. The distressed investing world, for example, where secured lenders with inside information trade bonds and stock like crazy, is much scummier.
- Quantbots and automated trading played a huge role in disintermediating the specialists, many of whom were truly fucking
predatory, pure and simple.
- You are correct that that Goldman dude is an HF trader. Anyway, this is not a black and white issue and there many issues with the quant arms race...but I don't like the sloganeering sometimes.
3) Comment #3:
I've read the paper by Themis and the feedback from some of your readers and thought I would provide a different perspective, as I think this is much ado about nothing. As background, I don't purport to be an expert on the electronic market, but I did build and manage for several years a trading desk at a multi-billion dollar hedge fund, so I do know something about trading.
The main argument of the paper is that high frequency traders, predatory algos and automated mkt makers are using their competitive advantage to increase trading costs and market volatility. Upon further thought, the argument just seems to fall apart under its own weight. Let's say I'm Fidelity or even Joe Sixpack and I pay a penny a share commish. Why should I care if someone who is trading millions of shares a day in rapid succession gets 1/4 penny a share rebate? I still pay my penny a share commish. In that very narrow sense, the commission component of trading costs remain unchanged. ECNs are certainly entitled to offer rebates for their best customers to attract trading volume. Nothing nefarious there.
Now, according to the paper, these HFT and predatory algos sniff out large orders and attempt to front-run them, driving prices higher. Let's think about this for a second. Implicit in the argument is the notion that somehow models are infallible. Haven't we learned by now that models can be and often times are wrong. In fact, they are usually the instruments of their own destruction. Maybe size is trading on a plus tick because the buyer thinks the company will beat earnings, or because his analyst just finished the work and loves the story. Or maybe it's a derivatives trade where someone is coming out of swap and the broker is crossing stock. In other words, it may not be because the buyer has size behind; there may not be an order to front-run in the first place.
Second of all, even if the HFT is right and front-runs the buyer and takes the stock up a nickel or a dime or whatever, the buyer can always cancel if he doesn't like the price, or he can instruct his trader to keep bidding below the mkt if he believes the stock will come back in. No one can force a buyer to pay a price he doesn't want to pay. Third, it's not just HFT and algos that front-run. It's how many trading-oriented shops operate. HFT and algos just do it using new fangled methods and technology. We don't have to like it, but it's an age-old phenomenon, one that doesn't require new legislation. If market manipulation is taking place, then there are already laws on the books to address this. Fourth, the examples the author uses
totally stack the deck in favor of his argument. When a market maker pings the mkt, he may get hit or lifted; the mkt can then move away from him instantaneously, leaving him exposed. These guys are playing for a quarter of a penny and often times employing massive leverage (in the case of HFT). This is not riskless arbitrage; there is no margin for error. And if these guys really comprise 70% of the market (a stat I doubt BTW), then they should compete away any economic rents
that may exist over time. Others will trip up and self-destruct under enormous financial leverage. How good a business can this be if it relies on co-locating a server on the exchange to confer a millisecond timing advantage to make a quarter of a penny? And, by the way, there are zero barriers to entry to co-locating an exchange. It's not expensive, I know people who do it.
Not to beat a dead horse, but another problem with the argument was raised to me by a friend earlier today. To the extent that buy-side firms who are outsourcing the trading function to algos feel like they are getting skimmed, they can always abandon the algos and work the order manually or via a sell-side trading desk. In other words, predatory algos should ultimately be revealed as predatory and cede mkt share to their non-predatory compeitors. The Themis paper is
predicated on the gullibility of a victimized public in the face of predatory super-computers.
Anyway, I thought I would lend a different perspective to a debate that was striking me as increasingly one-sided. I have no axe to grind-- I'm not a HFT nor do I invest with any-- so these are just my unbiased off-the-cuff thoughts.
4) Here's a blog post by Themis, clarifying their views and motivations:
22 Jul, 2009
We have watched in recent weeks increased scrutiny to the highly opaque world of high frequency trading. We are happy to have aided many others in shining a light on these murky waters. We are less than happy, however, to hear our name sullied by some, who have accused us of self promotion. We have a history of having our goals alligned with the buyside, and we stand by that history.
We make these observations:
1. In recent years exchanges have become for-profit, and less regulated, and more self-regulated.
2. In recent years trading technology has become increasingly sophisticated.
3. In recent years HFT has taken off. 70% of the tape’s volume is HFT. Their might and “business level” is much sought after among the for-profit exchanges. These exchanges need to show growth and increases to their shareholders. That is a higher priority for them than self-regulating, and providing a level and transparent playing field.
4. Few understand HFT.
5. HFT is highly profitable.
6. Some HFT is rebate driven.
7. The rest of the HFT is predatory (detect institutions and front run those orders).
8. Institutional trading has been decimated in recent years (courtesy of hedge fund deleveraging and a generally shocking market sell off. Their “clout” and power to have their voices and needs be addressed by for-profit exchanges is diminished.
9. Trading institutional orders means managing “wiggle and ripple” more today than ever before. Ask any buy-side trader at the next Trader Forum.
Joe and I did not write the HF computer codes. We are not proclaiming that they should be eliminated. We do not blame HFT for our budget deficit. We don’t question whether HFT falsely proclaimed a lunar landing.
However, Joe and I do question how it came about that a secretive method of highly profitable trading now accounts for 70% of the trading volume today. We question that it is being done by a little understood group, with little understood motives, and with little regulatory oversight and understanding. We question how 70% of the tape will react during a “crash” . Is today’s HFT akin to 1987’s “portfolio insurance trading”? It is good for us to question it, and others to debate it.
We want a market that is fair and transparent for all, with fair an equal access for all. Most of all we want a market that will be fair and thriving for all. Tomorrow, next week, and 5 years from now. We have a history of being agency-only no-conflict brokers for our clients since the early 1990’s. Our goals have been aligned with the buy side community back then, today, tomorrow, and also 5 years from now.
We just wanted to throw that out there. Thanks for everyone involved in the debate, no matter your point of view.
Sal and Joe
5) Another blog post -- the full article is below:
23 Jul, 2009
Pipeline’s Al Berkeley Roasts High-Frequency Traders as “Natural Enemy” of Institutions
Long time industry veterean and former president of NASDAQ, Al Berkeley had some choice comments for the high frequency trading community. Here are some excerpts from an interview that he did with Traders Magazine:
“High-frequency traders are the natural enemy of the individual investor and the large institutional investor. Most market centers are doing things that disadvantage the individual investor and institutional investor, because they’re doing things to advantage the high-frequency trader.”
“The idea that an exchange would let one set of market participants locate their computers inside their data center and not give the same speedy response to everyone is a blatant tilt in favor of high-frequency trading. There are very few people who can resist the heroin needle of trade volume and do something that optimizes the market for institutions.”
“These markets have never been better for hedge funds and high-frequency traders–they’re magic. We have optimized our market for hedge funds to make money. How do they make money? They front-run institutional trades. Traditional long-only traders have had to go into defensive mode because of the pattern recognition software that statistical-arbitrage and other funds have deployed to look for institutional orders–to look for the tracks they leave in the tape and on the bid-ask montage.”
“If you went to a movie theater and were lined up to get a ticket and somebody ran to the front of the line and bought all the tickets and scalped them to you, would you say that was fair? All this front-running is a version of scalping.”
“You must look through to the ultimate beneficiary. Do we want a handful of bright high-frequency traders front-running the
citizen-savers in the country? It’s bad public policy to have tilted our markets so far in favor of speculators. We’ve created the greatest casino the world has ever seen in our equities markets, because we’ve got so many tilts in favor of speculation and against investment”
Here is a link to the entire article:
Pipeline's Al Berkeley Roasts High-Frequency Traders as "Natural Enemy" of Institutions
Traders Magazine Online News, July 23, 2009
In this Q&A, Alfred R. Berkeley III, a 30-year industry veteran, argues that the growth of high-frequency trading has hurt the ability of traditional buyside firms to execute orders with limited market impact. Exchanges and broker-dealers, pursuing market share, cater to high-frequency firms at the expense of institutions, said Berkeley, chairman of Pipeline Trading Systems, a block-trading platform geared to institutions and broker-dealers.
Berkeley joined Pipeline in 2004. From 1996 to 2003, he was president and then vice chairman of Nasdaq Stock Market. Prior to that, he was a managing director at investment bank Alex. Brown & Sons. He co-founded the firm's technology group in 1975, three years after joining Alex. Brown. In this interview, conducted by Senior Editor Nina Mehta,
Berkeley discusses block trading, high-frequency firms, pattern-recognition software, the national best bid and offer, and the
Traders Magazine: In your view, what distinguishes market centers from one another?
Alfred Berkeley: It's about what systems are optimized to do, and who they're optimized to serve. A recent TABB Group report said that 73 percent of trading is attributed to high-frequency traders. Displayed markets and large brokerage firms, in order to have a large market share of the larger market, have to optimize their systems to attract high-frequency traders. High-frequency traders are the natural enemy of the individual investor and the large institutional investor. Most market centers are doing things that disadvantage the individual investor and institutional investor, because they're doing things to
advantage the high-frequency trader.
TM: What's an example of this?
Berkeley: A great example is co-location. The idea that an exchange would let one set of market participants locate their computers inside their data center and not give the same speedy response to everyone is a blatant tilt in favor of high-frequency trading. There are very few people who can resist the heroin needle of trade volume and do something that optimizes the market for institutions. Pipeline's business model is premised on optimizing around the institution's needs, even though they are just a quarter of the market now.
TM: Are there other ways in which the growth of high-frequency trading has led to changes in business models that disadvantage institutional investors?
Berkeley: The development of algorithmic trading favors high-frequency traders. The creation of so-called dark pools that many broker-dealers run benefits high-frequency trading firms, including a broker's own internal prop desk. That's why you see such small execution size in those pools--250 shares on average. Institutions don't trust those pools and can't put large orders in them.
TM: Does high-frequency trading have an impact on the quality of liquidity in the market?
Berkeley: Of course it does. The best way to see that is to look at the size at the inside and see how many orders are canceled when it looks like momentum is going from the buyer to the seller. And the average trade size is just lousy.
TM: The public markets provide price discovery. There's now more volume getting done than ever before. Have changes in the market had an impact on the quality of price discovery?
Berkeley: There are prices that are discovered on the basis of retail orders vs. size on wholesale orders. The issue of price discovery can't be disengaged from the issue of size discovery. Most of America's trading rules have been focused on price discovery at the expense of size discovery. Another issue is that there are informed and uninformed orders in the market. Most high-frequency trading is about picking off short-term imbalances in supply and demand, both long and short. That price discovery is perfectly legitimate.
Many people now think blocks should be part of the price discovery process. But the argument to me is a little bit specious, because blocks are not interested in prices determined by 100-share lots. Institutions don't care about the immediate short-term price. If they're moving 5 million or 10 million shares of stock, they know that the price can't be set accurately by the last 100-share trade. They're looking for a way to discover price for that size. In 1975, the average daily volume in all U.S. markets together was 25 million shares, and we were delighted with the quality of price discovery. Now 25 million shares trade before the market opens in big stocks like Intel.
Blocks were not part of the price discovery mechanism. They were bid on by upstairs desks. Blocks rarely went to the floor of exchanges, and if they did, they changed the price. They upset the pricing equilibrium because they produced big temporary imbalances in supply and demand. The U.S., Western Europe and some Asian markets have beautiful price discovery--magic price discovery. But we are vastly over-concerned with the argument that all orders should be in transparent markets to promote price discovery. It's being raised by price discoverers that are losing market share to other venues.
TM: I suppose you could look at changes in market structure in recent decades by looking at the evolution of the national best bid and offer. What was the relationship between block prices and the NBBO 10 or 20 years ago?
Berkeley: There wasn't a relationship. Blocks got discount bids. They could be a few dollars away. Those were negotiated markets, not auto-ex markets. Blocks traded where blocks would trade depending on how willing a broker-dealer was to put up capital. The NBBO became a reality in 1975 with the National Market System Amendments to the Securities Exchange Act of 1934. Before that, you had to call around to three or four places to get quotes, and you might trade with someone you called 20 minutes earlier.
TM: After 1975, did the Securities and Exchange Commission want blocks to be part of the price discovery process, or was there ambivalence about that?
Berkeley: Blocks have always been a difficult problem because size discovery and price discovery are inextricably bound up with each other. Conflict exists because price discovery and size discovery work against each other. The existence of a verbal wholesale market worked well until the Order Handling Rules and Order Display Rules, when [SEC Chairman] Arthur Levitt basically said that all Americans get to buy wholesale. We don't do that in any other market, such as automobiles,
sugar or telephones. They have separate markets.
Until then, we had a perfectly functional market in four segments. In NYSE stocks, the retail market was on the floor, and blocks were done upstairs at firms like Salomon Brothers and Goldman Sachs, and brought to the floor to be blessed. If a few retail orders were around to be price-improved, they would be included with the blocks. When electronic markets became a reality with Nasdaq in February 1971, and later in 1979 with Instinet, there was a functioning retail market on Nasdaq with wide spreads, and a functioning wholesale market at Instinet with quite tight spreads.
But the block market was a separate and distinct thing--it had a legal definition. For most stocks, it was defined as 10,000 shares or more. Everybody knew blocks were exempted from all sorts of things, including being displayed in quotes. Levitt changed all that, essentially at the request of the day-trading community, who wanted access to better prices in the wholesaling market.
TM: What about the price-fixing scandal on Nasdaq that academics Bill Christie and Paul Schultz exposed in 1994?
Berkeley: I wouldn't put too much weight on that. There's some debate about whether that was correlation without causation, and whether the results were statistically valid. I'd look at the Christie issue more as an immediate political tool for generating support for the SEC's actions than as causation.
TM: Why would Arthur Levitt want to appeal to the day-trading community? Berkeley: He thought he was helping the individual trader, helping the mom-and-pop trader by giving them access to wholesale prices. His intentions were fine, but the unintended consequences have cost citizen-savers using mutual funds and other institutions a lot of money.
TM: Regulation ATS led to the creation of a lot of new electronic markets, and Reg NMS ensured that the public markets were automated and linked. In this new world, is the market better or worse for blocks?
Berkeley: The markets are better overall, since there's more liquidity, but they're worse for institutional investors and the
citizen-savers they represent. These markets have never been better for hedge funds and high-frequency traders--they're magic. We have optimized our market for hedge funds to make money. How do they make money? They front-run institutional trades. Traditional long-only traders have had to go into defensive mode because of the pattern recognition software that statistical-arbitrage and other funds have deployed to look for institutional orders--to look for the tracks they
leave in the tape and on the bid-ask montage.
In trying to defend themselves, one of the only tools available to institutions is to cut the size of their orders, to vanish into the river of small orders. But pattern-recognition software sees that. If I have 10 million shares to trade and I slice that into 300-share pieces, I'm leaving tracks in a huge way. Someone doesn't have to see an order to know it's there. I can give you a Pipeline commercial here--that our business is about confounding that pattern-recognition software. We have some of the smartest guys in the country figuring out how to make it hard to see customer orders either in our block facility or through our outbound switching engine [for algorithmic orders].
TM: But institutions do have options. Pipeline exists, Liquidnet exists. Is that not enough?
Berkeley: We and Liquidnet are the only true wholesale markets. The average execution size on our markets is over 50,000 shares, rather than 250 shares in the public markets. We're trying to earn a living by serving the underserved institutional investor. And it works pretty well. But our market share plus Liquidnet's is just a teeny piece of the overall market share.
TM: But that shouldn't matter if most of the volume in the market comes from high-frequency trading firms, which is not your audience.
Berkeley: Right. That audience isn't relevant for us, but it's very relevant for our customers because those high-frequency firms are front-running institutions.
TM: Are the current regulatory rules sufficient to enable institutions to trade blocks easily? What rules do you think should change?
Berkeley: Rules aren't keeping most brokers optimizing their systems for high-frequency traders. It's the high volume of high-frequency trading and the profit opportunities that are causing ECNs, exchanges and broker-dealers to say, "I want to increase my share of high-frequency trading, and therefore I'll allow high-frequency traders to co-locate with me, or allow small inbound orders to probe for institutional orders, or invest in very high-speed systems which only high-frequency traders use." Large institutions, by and large, don't have a financial motive to invest in all the things high-frequency traders have the motive to invest in.
TM: What should regulators focus on with block trades that may not be high on their to-do list at the moment?
Berkeley: Public policy issues almost always have to balance more than one good. Rarely is there a bad in the equation. Vested interests will describe something as bad, but another constituent will say, "That's my good." Regulators are trying to figure out how we raise our standard of living--which is the goal of public policy--in the face of competing interests. There are vested interests in rules that exist, and in rules that are as yet undefined.
Regulators are approaching these issues with genuinely open minds. They are inventorying these goods, such as price discovery vs. size discovery, and the issue of off-exchange trading vs. on-exchange trading. A good regulator doesn't see one as better than the other. They say, "Why is this suddenly an area of contention between different factions of people? What have we done in evolving since the 1934 Securities Exchange Act to deal with exchanges and markets, and is that the right model? What's going on in other countries? How are other people dealing with trade reporting?" We now have a trade reporting discussion going on in the U.S.--should trades in dark pools be reported immediately with the name of the dark pool? Some countries allow trades to be reported with delays appropriate to keeping the existence of the block secret.
TM: Do you think blocks should always be printed to the tape?
Berkeley: In the cultural milieu of America, we print blocks to the tape, but we make it ambiguous where and when they occurred--the tape doesn't tell you where they were done, and there's a 90-second delay. If the tape did tell you where it was done, no one would print a block there again, because blocks are part of larger blocks. Would you really want to destroy the liquidity available in the wholesale market? I don't think so. If high-frequency firms could identify the venue where blocks were trading, that would give them that much more free information to aid their front-running. If you went to a movie theater and were lined up to get a ticket and somebody ran to the front of the line and bought all the tickets and scalped them to you, would you say that was fair? All this front-running is a version of scalping.
You must look through to the ultimate beneficiary. Do we want a handful of bright high-frequency traders front-running the citizen-savers in the country? It's bad public policy to have tilted our markets so far in favor of speculators. We've created the greatest casino the world has ever seen in our equities markets, because we've got so many tilts in favor of speculation and against investment.
TM: How do you fix that tilt?
Berkeley: You're looking for a regulatory answer, but I'm more interested in free-market ideas. We have a viable solution at
Pipeline. Other firms have developed viable solutions. What I don't want is one-size-fits-all regulatory rules cramping my ability to innovate. The market will tell me if I have a viable solution or not. I'll keep modifying my product until I have something institutions want.
TM: If you were starting a new company in the markets now, what type of firm would it be?
Berkeley: I would do exactly what Pipeline is doing. The great unsolved problem is optimizing a system for the traditional mutual fund, investment manager and pension fund. The traditional buyside is the neglected part of the market. I would want to build a system that solves the problem for the buyside moving large blocks in and out of the market. Almost everyone else is addicted to the volume that comes from high-frequency trading.