Rising correlation and computer-driven trading

July 13, 2010

Correlation among the individual equities in the Standard and Poor’s 500 index  has increased to the highest level since 1987, according to research by Birinyi Associates reviewed in the Wall Street Journal.

Birinyi Research Director Jeffrey Yale Rubin blames heightened correlation on the increasing popularity of indexing strategies, which he claims have transformed the way the whole market behaves, reducing the dispersion of returns.

Investors increasingly trade in and out of the whole market using broad indices (including exposure offered by exchange-traded funds) rather than picking individual stocks (see list of references below).

Barclays Capital’s head of quantitative strategies Matthew Rothman highlighted the same issue in a recent research note, noting “the last two months have seen historically low levels of dispersion in stock returns”. Both conclude it has been a tough time to be a traditional stock picker.

Increased correlation is not restricted to components of the big equity indices. Significantly heightened correlations are also evident among different commodity futures markets, and between commodities and equities, for example.

There is less consensus on what is causing correlations and whether they will prove an enduring feature of the financial landscape or if traditional diversification will be restored.

THE RISE OF HFT TRADING
While the correlation research focuses on the impact of equity indices (and the same could be said for commodity indices) it is intersecting with another debate on the impact of high-frequency, algorithmic and computer-driven trading strategies on price formation, liquidity and volatility in equities and derivatives markets.

The U.S. Commodity Futures Trading Commission (CFTC) has recently published proposed rules to govern the co-location of HFT servers with exchange servers to ensure equal access and equitable, uniform and non-discriminatory fee structures.

In its Notice of Proposed Rulemaking (NPR), published in the Federal Register on June 11, the CFTC cited research by Rosenblatt Securities estimating HFT amounts to approximately 35 percent of U.S. futures markets volumes. It also notes research by Tabb Group forecasting HFT will account for close to 60 percent of trading by the end of 2010 (75 Fed Reg 33198, footnote 5).

“In its review of co-location and proximity hosting services undertaken prior to this Proposal, the Commission learned that volumes from market participants that utilize co-location and/or proximity hosting services varied a great deal. Some regulated trading markets have little or no volume generated thru (sic) the use of such services, while others have significant volume”, according to the CFTC.

“One regulated trading market reported that 29 percent of its traders utilized such services, representing 68 percent of its trading volume, while another reported that well over 100 market participants utilized the service, representing 39 percent of its trading volume”.

IMPACT ON PRICE FORMATION

For supporters, algo/HFT traders are simply displacing traditional market makers as the principal providers of liquidity to financial markets.

Supporters claim HFT traders provide liquidity with narrower spreads than used to be available from traditional market makers in the pits, cutting trading costs for other market participants, including commercial hedgers as well as institutional and retail investors.

While traditional market makers tried to hold up their margins by avoiding “odd eighths”, the rise of HFT has accompanied and driven a narrowing of spreads to the penny level, and there are pressures to revise SEC regulations to allow sub-penny quotes.

For opponents, the liquidity provided by computer-driven trading strategies disappears just when it is most needed, when market conditions get rough, worsening volatility. The similar trading strategies employed by many HFT programmes (such as statistical arbitrage) heighten correlation, ensure trades become crowded, and create a self-destructive liquidation when they go wrong.

Critics focus on unusual market volatility experienced in August 2007 and again during the “flash crash” of May 2010 to show why HFT has increased rather than reduced volatility. The essential similarity among many quant strategies results in trades becoming crowded, and the application of HFT techniques ensures that when they go wrong it results in a disorderly rush for the exits.

The August 2007 and May 2010 episodes were the only ones involving high-frequency computer-driven trading. But the same basic problem (automated quant-based strategies and crowded trades causing liquidity to disappear in a crisis) can be traced back to previous market crises in 1998 (the failure of Long-Term Capital Management) and October 1987 (portfolio insurance and the stock market plunge), according to critics.

LETTING LIGHT IN UPON THE MAGIC
The Securities and Exchange Commission (SEC) and exchanges are already implementing circuit breakers that halt trading when prices move by more than +/- 10 percent within five minutes, in a bid to reduce wild price swings accentuated by automated trading programmes.

The SEC is also eyeing rules that would require market makers to post quotes within this +/- 10 percent band. It would eliminate the “stub quotes” that caused some equity prices to plunge to a single penny during the flash crash.
It is not clear whether rules would be extended to major HFT participants besides the designated market makers. But it would help end the current complaint HFT participants pull back from the market at the first sign that conditions are getting rough.

The SEC and CFTC are investigating what role HFT played in the flash crash of May 6. No one is really sure what triggered and propagated the crash, though there is a strong suspicion that the increased prevalence of HFT technologies accelerated price movements. But the issue of HFT is now firmly on the regulatory agenda.

The CFTC’s newly re-established Technology Advisory Committee (TAC) meets for the first time tomorrow to start examining how HFT and algorithmic trading is affecting the futures markets, and how the Commission should update its own technology and data-gathering to ensure supervision catches up after lagging behind for the last decade.

The CFTC has specifically requested papers on “Algo and HFT Market Impact — What are the positive or negative impacts of Algo and HFT on the futures markets and market structure (eg in terms of liquidity or volatility, the impact on fundamentals, commercials or hedgers; and other issues)?”

There has been little thought given to the impact of HFT outside the specialist community, apart from a few lurid stories mostly focused on the issue of flash orders and front-running, and almost no comprehensive statistics and visibility. The CFTC’s investigation should shed some light on a poorly understood but increasingly crucial part of the markets.

For supporters and critics alike, the next six months will be dominated by a fierce battle to understand and define HFT’s impact of on price formation and volatility, and shape the regulatory response.

No comments so far

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/