If markets are efficient, they reflect all information, so there is no profit to be had from trading on information. If there is no profit to be had, traders with information won’t trade, so markets won’t reflect it, and will not be efficient. This is the Grossman-Stiglitz paradox in a nutshell. Indeed, if there is no profit to be had from trading on information, then why would anyone expend resources to acquire the information upon which process are based in the first place?
Indeed, a visit to the “hall of fame” of equity market efficiencies popular with quantitative traders reveals a potpourri of sources of alpha (i.e., active investment ideas for out-performing a passive benchmark), e.g., earnings forecast analysis, earnings surprises, insider trading disclosure, stock splits, secondary equity offerings and stock buybacks, mergers and acquisitions, sector analysis, common factor analysis, message board counts, Twitter sentiments, web traffic analysis, etc. Financial markets in general are far from perfect; many sources of inefficiencies can be found at different times if one has the right tools and knows where to look.
An investment strategy that is quite popular with hedge funds is what is called the “market neutral long-short portfolio”. In a typical setup starting with, say $100 million, for example, a long-short, market neutral portfolio consists of $100 million in long positions and $100 million in short positions. After receiving $100 million from the short sale and spending $100 million on the long side, there is still $100 million in cash (which is the amount that the fund starts with). There is no net capital requirement to put on a position such as this. What the hedge fund manager typical does then is to use the cash to put on an unleveraged futures position, e.g., the S&P 500 index futures, so as to capture market return. This is because an ongoing market index futures position, reinvested at the contract expiration dates, closely tracks the index return.
So when a long-short portfolio and an index futures position are put together, what results is a total return equal to the return on the index (i.e., beta) plus whatever return captured from the long-short portfolios (i.e., alpha). This is called an “equitized” portfolio, named for the market return captured through putting on an equity market index futures position. Notice that all the money in the fund is working twice: once on the long side of the portfolio and once on the short side. And this alpha return comes on top of the market return. It is no wonder that David Leinweber dubbed this the “James Brown of quant stock strategies, the hardest working portfolio in the equity business.”
The general plan of quantitative strategies, such as the popular market neutral long-short portfolio aforementioned, is no mystery. After all, quantitative strategies are really just mathematical expressions of fundamental investment ideas, if one look inside the process. Quantitative methodology allows many disparate concepts to come together in a single forecast. Because the process is automated, it can be applied to many financial securities, thus spreading little bets across many active positions and limiting risk in the process. So in many ways, quantitative investing is really not that much different from traditional investing, although it may sound quite dissimilar.
Some quantitative strategies work by pure arbitrage, essentially finding the three-and-a-half-cent pennies in the market before anyone else does. Arbitrage opportunities are sweet if you can find them; statistical arbitrage works just as well. But in an increasingly wired world where the global financial markets are fully electronic, such arbitrage trading opportunities are rare, and available only to those with bleeding-edge infrastructure, or scale of capital, or both. For the rest of the trading masses, strategies based on prediction of financial markets (adjusted for risks) is far more common place. The objective here is now two-fold: increasing predictability increases investment return; while improving the consistency and downside error of predictive models reduces risk.
A useful perspective on maximizing predictability in financial markets is depicted above. The perspective is attributed to Andrew Lo, but the picture is adapted from an illustration found in David Leinweber’s Nerds on Wall Street. When viewed from a high level, there are only three key decisions to make in any financial market prediction:
- What to predict: One can choose to predict returns to an asset class, e.g., a broad market or an industry group, an exchange rate, interest rates, or returns to individual securities of many types. One can also choose to predict spreads (i.e., return differences) between individual securities or groups of securities. Predictions of volatility are useful for options-based strategies.
- How to predict: One can choose from a wide variety of statistical and mathematical methods of prediction. Many use simple windowed regression methods, which are popular. Some choose more advanced regression methods, such as moving or expanding windows, kernel estimation, auto-regressive integrated moving average (ARIMA) time series models, or even neural networks.
- What to predict with: This is the raw materials that feed into the prediction methods. Technical traders use only past prices to predict future prices; but this is quite rare in institutional trading. A wide selection of financial and economic data, e.g., commodity prices, foreign exchange rates, GDP announcements, analysts opinions, messages on bulletin boards or even measures of Twitter sentiments could find their respective predictive powers within the right context.
In an uncertain world, a stochastic world view and associated methodologies for conducting experiments, interpreting outcomes and take-away results might be important. Last but not least, the time horizon of how everything interacts together (i.e., long or short), plays just as important a role in determining success or failure, especially for electronic markets participated by high-frequency traders.
Now, here is an interesting meta-level question: Can hedge fund returns be predicted? Can hedge fund returns, assuming they are good, be replicated?
- Grossman, Sanford J. and Stiglitz, Joseph E. (1980, June). On the Impossibility of Informationally Efficient Markets. The American Economic Review. Retrieved from: https://assets.aeaweb.org/assets/production/journals/aer/top20/70.3.393-408.pdf
- Lo, Andrew, and MacKinlay, Craig (1995, February). Maximizing Predictability in the Stocks and Bond Markets. NBER Working Paper No. 5027. Retrieved from: http://www.nber.org/papers/w5027.pdf
- Litterman, Bob (2005). Active Alpha Investing. Goldman Sachs, Open Letters to Investors. Retrieved from: http://www.goldmansachs.com/gsam/pdfs/USI/education/aa_beyond_alpha.pdf
- Geer, Carolyn (2011, February 7). Index Funds Get a Makeover. Wall Street Journal. Retrieved from: http://online.wsj.com/articles/SB10001424052748703555804576101812395730494
- Hulbert, M. (2008, July 3). The Prescient Are Few. New York Times. Retrieved from: http://www.nytimes.com/2008/07/13/business/13stra.html
- Barras, Laurent, Scaillet, Olivier, and Wermers, Russ (2010). False Discoveries in Mutual Fund Performance: Measuring Luck in Estimated Alphas. The Journal of Finances, LXV(1), 179-216. Retrieved from: https://gates.comm.virginia.edu/uvafinanceseminar/2008Wermers.pdf
- Zweig, Jason (2014, August 22). The Decline and Fall of Fund Managers. Wall Street Journal. Retrieved from: http://blogs.wsj.com/moneybeat/2014/08/22/the-decline-and-fall-of-fund-managers
- Leinweber, David (2009). Nerds on Wall Street: Math, Machines and Wired Markets. John Wiley and Sons.