`Toxic' Orders Can Predict Likelihood of Stock Market Crashes, Study Says

A formula for measuring how fast the best-informed traders increase their share of market volume may help regulators prevent crashes such as the May 6 plunge, according to a study from researchers at hedge fund Tudor Investment Corp. and Cornell University.

The team developed a gauge that identifies the likelihood of market makers curtailing their activity, which can cause a rapid drop in prices. Marcos Lopez de Prado of Tudor and Cornell’s David Easley and Maureen O’Hara say the metric, similar to risk-management techniques used at some trading firms, may help regulators monitor for potential catastrophic shifts in liquidity.

Regulators are trying to prevent another rout like May 6, when a 20-minute drop briefly erased $862 billion from the value of U.S. equities. A transaction in Standard & Poor’s 500 Index futures overwhelmed computer-driven firms that bought the contracts, producing a flurry of “hot potato” trading followed by an evaporation of buy orders as investors lost faith in data, the Securities and Exchange Commission and Commodity Futures Trading Commission said in an Oct. 1 report.

“The measure would have been able to anticipate two hours in advance there was a high probability of a liquidity-induced event on May 6,” said Lopez de Prado, head of high-frequency futures at Tudor. “It measures order toxicity, or the probability the market is going to have persistent order imbalances that are going to damage market makers.”

Statistical Models

One of the statistical measures Greenwich, Connecticut- based Tudor uses to monitor risks in its high-frequency trading is based on O’Hara and Easley’s research, Lopez de Prado said. The new gauge the three developed after May 6 is called volume- synchronized probability of informed trading, or VPIN.

While no system can predict stock crashes, any research by O’Hara, whose specialty is market microstructure, “must be taken seriously,” said Paul Wilmott, owner of Wilmott.com, a website for quantitative finance that has 80,000 members.

“You can create lots of toy models, meaning simple models, to explain phenomena such as price movements, but they aren’t necessarily predictive,” Wilmott said. “Think of these models as being like an early warning indicator. Just like predicting a hurricane, you may know something’s on the way, but you don’t know where or when it will hit.”

Market Structure

The married Cornell couple are experts in market structure who have spent 20 years studying how changes in the make-up of orders on exchanges affects prices. Easley is chairman of the economics department and O’Hara is a finance professor at Cornell in Ithaca, New York. Before joining Tudor in 2008, Lopez de Prado was a partner at PEAK6 Investments LP in Chicago and head of quantitative equity research at UBS AG in Zurich.

“This could help market makers by measuring the probability they’re receiving toxic flow,” or orders from traders who are more informed about short-term price movements, Lopez de Prado said. “By providing a metric, they can regulate the way they operate and avoid dramatic decisions,” such as exiting the market. A futures contract on the gauge could enable them to manage their risk the way traders use the Chicago Board Options Exchange Volatility Index, or VIX, to hedge against volatility, he said.

Tudor filed for a patent on VPIN on Oct. 15, according to Steve Evans, a partner and director of the hedge fund’s 46- person systems trading group, where Lopez de Prado works. Evans, who also manages the $1.4 billion Tudor Tensor managed futures strategy and a portion of the firm’s flagship Tudor BVI Global Funds, declined to say how VPIN influenced Tudor’s activity or profitability on May 6. He said that while Tudor uses some high- frequency trading techniques, that’s not one of the firm’s primary investment strategies.

Informed Trading

VPIN offers a measure of informed trading “suited for a hyper-volume informational world,” said O’Hara, who is also chairman of New York-based Investment Technology Group Inc. Since informed traders usually buy or sell at the same time more often than other participants, market makers tend to lose more money when order imbalances increase, causing them to curb their activity, she said. The paper about the May 6 crash is one of three the authors wrote related to VPIN.

“These are good papers that present interesting, precise ideas that can add to our understanding” of how informed trading affects liquidity providers, said Aaron Brown, chief risk officer of AQR Capital Management LLC, a Greenwich, Connecticut-based hedge fund. “What’s missing is an empirical study linking elevated levels of toxic trading to the withdrawal of market liquidity.”

Forecasting Selloffs

Brown also questioned whether the tool was capable of forecasting selloffs such as the one on May 6.

“Liquidity providers monitor order imbalances in much more sophisticated ways than the authors, and use the analysis in their strategies,” Brown said. “It’s hard to believe you can predict flash crashes using a measure less sophisticated than the market participants have, given that the participants are trying to avoid being victimized by crashes.”

The VPIN gauge measures the imbalance of buy and sell orders to overall volume over periods that vary based on how much trading occurs, Easley said. Tracking imbalances relative to the intensity of trading instead of set periods -- what the paper calls trade time as opposed to clock time -- is a variant of existing models adjusted for high-frequency strategies.

“Informed flow is very hard to define,” said D. Keith Ross Jr., chief executive officer of PDQ Enterprises LLC, a Glenview, Illinois-based firm that operates a dark pool, or private network that matches orders without publishing quotes. “It can only be defined after the fact, based on its impact.”

Ross, a former chief executive officer of Chicago-based market maker Getco LLC, added that the paper presents an “elegant idea, but it might not be practicable.”

VIX Swings

VPIN has shown it can predict changes in the VIX, which reflects investor expectations about near-term price moves based on trading in options on the Standard & Poor’s 500 Index, O’Hara said. The VIX doesn’t predict VPIN, making the measure a tool that provides new information to the market, she said.

Algorithms, or strategies that break larger orders into smaller pieces and execute them gradually, could also be tailored to trade differently as the VPIN rises, avoiding risks caused by diminishing liquidity, O’Hara said. Regulators for their part could use the measure to identify when markets are at risk of a liquidity-induced crisis and slow down trading to forestall further problems, Lopez de Prado said.

“This would be much more effective than a circuit breaker,” he said. U.S. stock markets and the Securities and Exchange Commission instituted curbs in June that briefly halt trading in a security when its price moves rapidly. “That stops the infection after the infection is already widespread,” he said. “VPIN is more dynamic and gradual.” It allows firms to track problems and respond to changes more quickly, he said.

High-frequency strategies now help facilitate the majority of trading volume in U.S. stocks and are proliferating in futures markets.

In this environment, “a period of illiquidity could pose real challenges when markets are fragmented and difficult,” O’Hara said. “Liquidity risks are non-trivial.”

To contact the reporter on this story: Nina Mehta in New York at nmehta24@bloomberg.net.

To contact the editor responsible for this story: Nick Baker at nbaker7@bloomberg.net.

Bloomberg reserves the right to edit or remove comments but is under no obligation to do so, or to explain individual moderation decisions.

Please enable JavaScript to view the comments powered by Disqus.