Friday, May 07, 2010

Algorithmic Trading and Price Volatility

Yesterday's dramatic decline and rapid recovery in stock prices may have been triggered by an erroneous trade, but could not have occurred on this scale if it were not for the increasingly widespread use of high frequency algorithmic trading.

Algorithmic trading can be based on a variety of different strategies but they all share one common feature: by using market data as an input, they seek to exploit failures of (weak form) market efficiency. Such strategies are necessarily technical and, for reasons discussed in an earlier post, are most effective when they are rare. But they have become increasingly common recently, and now account for three-fifths of total volume in US equities:
Algorithms have become a common feature of trading, not only in shares but in derivatives such as options and futures. Essentially software programs, they decide when, how and where to trade certain financial instruments without the need for any human intervention... markets have come to be dominated by “high-frequency traders” who rely on the perfect marriage of technology and speed. They use algorithms to trade at ultra-fast speeds, seeking to profit from fleeting opportunities presented by minute price changes in markets. According to Tabb Group, a consultancy, algorithmic and high-frequency trading accounts for more than 60 per cent of activity in US equity markets.
This is a recipe for disaster:
[In] a market dominated by technical analysis, changes in prices and other market data will be less reliable indicators of changes in information regarding underlying asset values. The possibility then arises of market instability, as individuals respond to price changes as if they were informative when in fact they arise from mutually amplifying responses to noise. 
Under such conditions, algorithmic strategies can suffer heavy losses. They do so not because of "computer error" but because of the faithful execution of programs that are responding mechanically to market data. The decision by Nasdaq to "cancel trades of 286 securities that fell or rose more than 60 percent from their prices at 2:40 p.m." might therefore be a mistake: it protects such strategies from their own flaws and allows them to proliferate further. Canceling trades can be justified in response to genuine human or machine error, but not in response to the implementation of flawed algorithms.

I don't know how the losses and gains from yesterday's turmoil were distributed among algorithmic traders and other market participants, but it is conceivable that part of the bounce back was driven by individuals who were alert to fundamental values and recognized a buying opportunity. The following clip of Jim Cramer urging viewers to buy Proctor and Gamble just moments before a sharp recovery in its price is suggestive:

I would be very interested to know whether the transfer of wealth that took place yesterday as prices plunged and then recovered resulted in major losses or gains for the funds using algorithmic trading strategies. I expect that those engaged in cross-market or spot-futures arbitrage would have profited handsomely, at the expense of those relying on some form of momentum based strategies. If so, then the cancellation of trades will simply set the stage for a recurrence of these events sooner rather than later.
I thank Charles Davi for alerting me to the Financial Times piece on algorithmic trading, and Jens Kayenburg (a student in my Financial Economics course this semester) for sending me a link to the Cramer clip.


Update (5/7). David Merkel is also opposed to the cancellation of trades:
[My] sense of the day is that some algorithmic trading programs went wild, and made trades that no sane human would... NASDAQ should not have canceled the trades.  It ruins the incentives of market actors during a panic. Set your programs so that they don’t so stupid things. Don’t give them the idea that if they do something really stupid, there will be a do-over. In the absence of fraud, trades should not be canceled.
And here's Yves Smith's take on the events of yesterday:
The idea that a fat-fingered trade out of Citi was the cause has been denied by the bank. The downdraft did have the look of a monster sell order, but the more credible explanation is that it was either a sudden rise in yen or the euro hitting the magic number 1.225 to the dollar that set off algorithmic traders. And enough of them look to similar indicators and technical levels that it isn’t hard to see this as the son of program trading, mindless computer-driven selling when the right triggers are hit.
But another side effect of today’s equity market gyrations is further distrust in the markets, particularly by retail buyers. I am told that various retail trading platforms were simply not operating during the acute downdraft and rebound. I couldn’t access hoi polloi Bloomberg news or data pages then either. The idea that the pros could trade (even if a lot of those trades are cancelled) while the little guy was shut out reinforces the perception that the markets are treacherous and the odds are stacked in favor of the big players (even though we all understand that, it isn’t supposed to be this blatant).

Update (5/8). Here's an important point about liquidity by Paul Kedrosky:
Largely unnoticed... the provision of liquidity has changed immensely in recent years. It is more fickle, less predictable, and more prone to disappearing suddenly, like snow sublimating straight to vapor during a spring heat wave. Why? Because traditional providers of liquidity, market-makers and other participants, are not standing so ready to make the other side of the market. There are fewer traders prepared to make a market for the sake of market health. This is partly because they can, but mostly because of what has happened with high-frequency trading, algorithms, and the like, which increasingly jump into the trading queue in front of and around orders, creating some liquidity, but also peeling pennies for themselves, frustrating market participants and heretofore liquidity providers, but in the course of normal business generally accepted as a price that gets paid to the market's battle bots.
But all of this changes market microstructure in insidiously destabilizing ways. For the first time we have large providers of this shadow liquidity, algorithms and high-frequency sorts, that individually account for large percentages of daily trading activity, and, at the same time, that can be turned off with a switch, or at an algorithmic whim. As a result, in market crises, when liquidity was always hardest to find, it now doesn't just become hard to find, it disappears altogether, like water rushing out sight via a trapdoor to hell. Old-style market-makers are standing aside as panicky orders pour in, and they look straight at shadow liquidity providers and say, "No thanks. You battle bots take it". And, they don't.
David Murphy, who is always worth reading, thinks that its time to put some sand in the algorithmic wheels:
It is time to... throw some sand in the cogs of the algos. If every trade executed in the same, say, five second interval got the same price, instability would be greatly reduced, yet ordinary investors would not notice the effect. And if every trade were executed on the NYSE, or at least using the same market conventions, then officials could actually stop everything when things get out of hand.
I spoke with Zack Goldfarb of the Washington Post yesterday for an interesting article that ran today. The point I was trying to make is this: the problem lies not so much with the method of trading (algorithmic or otherwise) but with the underlying strategies that are being implemented. Algorithmic trading allows technical strategies to profit and proliferate, and markets dominated by technical analysis will tend to be unstable. If destabilizing strategies are prevented from taking losses when they misfire, the result will be more frequent and significant departures of prices from fundamentals. Hence my concern over the cancellation of trades.


Update (5/9). CBS Evening News had a report on this yesterday, including a couple of clips from a conversation I had with Tony Guida earlier in the day. As with most media reports on the topic, the focus is on automation, scale and speed rather than the kinds of trading strategies that these methods allow speculators to implement. We did cover this ground in the interview but (understandably, I suppose) it didn't make it into the broadcast. 


  1. Rhetorical question:

    If free markets are always optimal and rational, why would canceling trades make sense, or be necessary?

    The omniscient nature of markets suggests erroneous trades shouldn't even exist.

  2. The literature on computability due to people like Velupillai and Axtell and Markose, not to mention Mirowski at a more philosophical level, are now becoming relevant. We are approaching a point where we could have a massive collapse due to roundoff error.

  3. Golden Networking ( is organizing the High-Frequency Trading Experts Workshop 2010 (, “Practical Implementation of High-Frequency Trading Strategies," on December 9th and 10th in New York City. It has been recommended for executives in finance and investments who work at Investment Banks, Hedge Funds, Pension Funds, Broker Dealers, Consultancy Groups, Prime Brokers, Solution Providers and Exchanges and wish to gain a thorough understanding and practical knowledge of high-frequency trading.

  4. "If free markets are always optimal and rational, why would canceling trades make sense, or be necessary?"

    free markets aren't always optimal or rational. also, there is no such thing as a free market in modern society. doesn't exist outside the classroom