The efficient market hypothesis in its semi-strong form posits that asset prices reflect all publicly available information (Fama, 1970). Therefore, to achieve outperforming trading profits, one would need to obtain information unavailable to the public. Recently, there has been growing evidence of such information flows through social connections, education networks (Cohen, Frazzini and Malloy, 2008, 2010), geographical proximity (Coval and Moskowitz, 2001), expert networks, as well as through direct contact of agents involved in financial management processes with corporate insiders.
In recent years, the US has been ramping up enforcement on insider trading both at federal and state levels. Notable examples include Galleon, a hedge fund led by billionaire Raj Rajaratnam that relied on inside information illegally supplied through expert networks; the related conviction of Rajat Gupta, a former McKinsey managing director and board member of Goldman Sachs who shared restricted information with Rajaratnam; and the recent civil and criminal cases brought against SAC Capital and staff at affiliate firms such as Diamondback and Level Global.
As the flow of information comes under greater scrutiny, hedge funds are embracing alternatives including systematic trading strategies to maintain an edge. For such strategies, trades are determined by quantitative processes and are often executed automatically. From an information-flow perspective, these programs are less controversial because once algorithms are coded and trading commences, individuals no longer play a significant role in determining investment decisions. Therefore, funds engaging in such trading strategies are shielded against insider-trading scrutiny.
Another means by which funds have striven to maintain an edge is by exploiting technological advancements in trading. For example, high-frequency trading (HFT) can process and execute new information in less than one second. There are two outstanding issues, however, with HFT. First, it has been criticised by regulators and other market participants as possible market manipulation and therefore funds engaged in HFT implicitly assume some form of regulatory, legal and headline risk. Second, while trading volume has significantly increased over the past decade, most HFT trading occurs between HFT systems, making the environment more competitive, and thus less profitable. The capacity and scalability of such strategies is also unclear and likely limited (Baron, Brogaard and Kirilenko, 2012).
Adding to the controversy around HFT, some data providers have been selling preferential access to market-moving information to HFT trading firms. Earlier this summer, Eric Schneiderman, the attorney general of the state of New York, ordered Thomson Reuters to discontinue its practice of selling early access to University of Michigan’s Consumer Survey results. Until then, some funds had obtained the results two seconds ahead of their official release to the public, which provided an edge over other subscribers.
Systematic strategies as an alternative edge
Despite the controversy around HFT, systematic trading strategies across a variety of frequencies seem to be the most logical investment style for fund managers looking for an alternative edge in the current regulatory environment.
Strong performance of systematic strategies during the period leading to – and during the first part of – the financial crisis especially contributed to the popularity of this investment style. CTA/managed futures funds (otherwise known as trend followers), which typically rely on computer models and automatic execution to trade futures and over-the-counter contracts globally, have grown substantially, with hedge funds such as Man AHL, Winton Capital, Canlab and BlueCrest Capital seeing their assets under management (AUM) increase dramatically since 2008. Recent reports suggest that this industry segment peaked at $330 billion in capital1, with the largest fund, Winton Capital, attracting approximately 10% of that figure2.
The graph below displays the time series of the aggregated AUM of US dollar-denominated funds that report to the Lipper/TASS database under the systematic diversified category. The figure demonstrates tremendous growth over the past few years until 2011.
With the increased attraction of significant capital, one would expect returns to diminish (Fung, Hsieh, Naik, and Ramadorai, 2008). “My view has always been that the raw trend-following strategy will become less efficacious over time as more money has come into it,” says David Harding, founder of two of the largest systematic strategy funds, Winton Capital, and AHL3.
Indeed, despite stellar performance early on, in which the investment style recorded negative annual returns in only one year during the 19-year period 1990–2008, these strategies have since performed quite poorly, recording, on average, only one year of positive returns over the past five years since 2009 as shown by the graphs below.
In addition to the drop in the ability to generate abnormal returns of a given strategy upon reaching substantial AUM, another risk is presented – correlation risk. Once many funds use similar information and engage in similar strategies, their returns become more correlated, which makes them a less attractive allocation on an individual basis from a diversification perspective. An example of such risk is the quant crisis of August 2007 (Khandani and Lo, 2007), which can be partly explained by a market-wide liquidity event (Sadka, 2010, 2012).
One explanation for that event is that over time many quant funds have adopted similar trading strategies, such as momentum and value investing. Once one large such fund suffers an idiosyncratic shock that forces liquidation of its positions, all funds experience losses, which further leads to depressed asset prices such as a liquidity spiral as in Brunnermeier and Pedersen, 2009.
To mitigate such correlation risk, funds set out to look for alpha elsewhere, for example, by acquiring access to new datasets. One such example is access to media data, such as tapping into the Thomson Reuters news feed. This feed became quite popular amongst HFT funds due to the ease with which its data can be processed. However, upon becoming popular, returns are rumoured to have dropped significantly. Another medium provided anecdotal evidence a few months ago, when a false tweet from the official Twitter account of the Associated Press about a possible terror attack on the White House caused an immediate impact on prices, only to be reversed after several minutes once the rumour had been proved false.4 The fundamental concern with using such new data is that data providers have the incentive to make it generic, yet once it is, profitability drops while correlation risk increases.
Therefore, given the limitations of popular datasets, to obtain uncorrelated returns one should branch out to uncharted territories and use data that will generate uncrowded trades.
Portfolio managers and financial analysts have recently begun to harness the power of unstructured data. It has been claimed that hedge fund analysts have been known to hire consultants to count cars in retailers’ car parks in order to project revenues during intense shopping seasons. A UBS analyst was reported to have purchased satellite images of Walmart car parks to estimate its business activity ahead of the release of its quarterly earnings.5
With growth in the availability of big data – capturing anything from consumer/investor behavior and point of sales transactions to the price dynamics of products and services offered online – there is a tremendous opportunity for innovative fund managers to achieve an edge by systematically obtaining and analysing unstructured data to generate uncorrelated, uncrowded returns.
Examples of the systematic collection of novel datasets and their use for financial applications are provided in two previous articles (Ozik and Sadka, 2010, 2013). The motivation for these studies was an anecdotal observation that superstar fund managers tend to underperform. To formally test this phenomenon, we set out to collect information about the media coverage of hundreds of equity hedge funds. This information, however, is not readily available. With the help of a programmer and using some advanced web-searching techniques, we were able to obtain information about the media coverage of roughly 1,000 equity hedge funds over the period 1999–2008. Indeed, as predicted, formal statistical tests provide systematic evidence of the underperformance of funds with extensive media coverage. For example, the figure below shows that, regardless of context, media-covered funds underperform uncovered funds by 4% within two years post-coverage. The monthly return spreads are statistically significant for over one year post-coverage.
Further tests show that the underperformance of media-covered funds is not due to some usual omitted variable suspects, such as past performance, past flow and fund size.
In Ozik and Sadka (2013), the authors found that not all media sources produce the same type of information. Corporate-communication covered funds tended to outperform, while those covered by general media such as national newspapers tended to underperform. Coverage in specialised magazines, meanwhile, did not seem to significantly predict hedge fund returns. Notwithstanding the fact that media coverage attracts investor flow (e.g., Jain and Wu (2000)), the source analysis of Ozik and Sadka shows that despite the ability of different types of media to separate outperforming and underperforming funds, they do not seem to differentially attract flow.
Furthermore, processing the unstructured media data to determine the tone allowed the authors to document biases in media coverage and measure investor reaction to news. Using that same dataset, we provide new evidence demonstrating how careful processing of big data may contribute to investment performance. In the table below, we show that the underperformance of star funds (covered funds) stems mainly from funds experiencing negative coverage.
In conclusion, despite the regulatory, correlation, and performance challenges faced by fund managers, the availability of new data presents a unique opportunity for innovative managers willing to explore new frontiers. The combination of unique sets of big data and technology driven investment processes is a potential new source of uncorrelated excess returns.
Gideon Ozik is affiliate professor of finance at Edhec Business School and president of investment analysis firm Alphaness SAS. Ronnie Sadka is professor of finance at Boston College.
1 Burghardt Galen, Ewan Kirk, and Lianyan Liu. Capacity of the managed futures industry, Newedge AlternativeEdge Note, July 31, 2013.
2 Newedge research shows CTA capacity concerns overblown. Hedge Funds Review, August 27, 2013.
3 Jones Sam, Hedge funds battered in quant arms race. Financial Times, June 11, 2013.
4 False Rumor of Explosion at White House Causes Stocks to Briefly Plunge; AP Confirms Its Twitter Feed Was Hacked.
5 New Big Brother: Market-Moving Satellite Images.
Baron, M, Brogaard, J and Kirilenko, A. 2012. The trading profits of high frequency traders. Working paper.
Brunnermeier, MK and Pedersen, LH. 2009. Market liquidity and funding liquidity, Review of Financial Studies 22, 2201-2238.
Cohen, L, Frazzini, A, and Malloy, CJ. 2010. Sell-side school ties, Journal of Finance 65, 1409–1437.
Cohen, L, Frazzini, A, and Malloy, CJ. 2008. The small world of investing: Board connections and mutual fund returns, Journal of Political Economy 116, 951–979.
Coval, J and Moskowitz, T. 2001. The geography of investment: Informed trading and asset prices, Journal of Political Economy 109, 811-841.
Fama, E. 1970. Efficient capital markets: A review of theory and empirical work, Journal of Finance 25, Papers and Proceedings, 383-417.
Fung, W, Hsieh, DA, Naik, NY and Ramadorai, T. 2008. Hedge funds: performance, risk and capital formation, Journal of Finance 63 (4), 1777-1803.
Khandani, AE and Lo, AW. 2007. What happened to the quants in August 2007? Journal of Investment Management 5, 5-54.
Ozik, G and Sadka, R. 2010. Does recognition explain the media-coverage discount? Contrary evidence from hedge funds. Working paper.
Ozik, G and Sadka, R. 2013. Media coverage and hedge fund returns, Financial Analysts Journal 69, 57-75.
Sadka, R. 2010. Liquidity risk and the cross-section of hedge fund returns, Journal of Financial Economics 98, 54-71.
Sadka, R. 2012. Hedge fund performance and liquidity risk, Journal of Investment Management 10, 60-72.