Reducing the Risk of Black Swans
Page 9
Chapter 8:
AQR Style Premia Alternative Fund
Traditional mutual funds are long-only, allowing investors to capture only a portion of the factor premium to which they are seeking. In general, research shows that about half a factor’s premium comes from its long side and half from its short side. Let’s consider an example, for illustrative purposes only. The value factor is the average annual return on the top 30 percent of stocks ranked by book-to-market ratio (value stocks) minus the average annual return on the bottom 30 percent of stocks (growth stocks). The middle 40 percent are core stocks. About half the value premium comes from the outperformance, relative to core equities, of stocks in the top 30 percent and about half of it comes from the underperformance of stocks in the bottom 30 percent. To earn the full value premium, investors would have to be long value stocks and short growth stocks.
To give investors an alternative that would provide greater exposure to factor premiums than long-only funds, as well as greater diversification benefits, AQR Capital Management created the Style Premia Alternative Fund (QSPRX). QSPRX is a long-short fund that invests across four styles (or factors), each of which has support in the academic literature. Being both long and short allows investors to achieve greater exposure to factors that have delivered premiums, without also gaining any net exposure to market beta (equity risk). Each of the four styles (value, momentum, carry and defensive) is backed by both economic theory and decades of data on long-term performance across geographies and asset groups.
Further support for factor-based investing strategies comes from Antti Ilmanen and our colleague, Jared Kizer, in their paper, “The Death of Diversification Has Been Greatly Exaggerated,” which appeared in the Spring 2012 edition of The Journal of Portfolio Management. The paper, which won the prestigious Bernstein Fabozzi/Jacobs Levy Award for best paper of the year, made the case that factor diversification has been much more effective at reducing portfolio volatility and market directionality than asset class diversification.
Let’s take a brief look at the four styles employed by the AQR Style Premia Alternative Fund.
Value: The tendency for relatively cheap assets to outperform relatively expensive ones. It is implemented across stocks, industries, bonds, interest rates, currencies and commodities.
Cross-Sectional Momentum: The tendency for an asset’s recent relative performance to continue into the near future. It is implemented across stocks, industries, bonds, interest rates, currencies and commodities. (Cross-sectional momentum is different from times-series momentum, which is based on absolute performance instead of relative performance. More on that in Chapter 9.)
Carry: The tendency for higher-yielding assets to provide larger returns than lower-yielding assets. It is implemented across bonds, interest rates, currencies and commodities. (See Appendix B for a more extensive explanation.)
Defensive: The tendency for lower-risk and higher-quality assets to generate larger risk-adjusted returns. It is implemented across stocks, industries and bonds.
The AQR fund accesses each style through long-short portfolios across multiple asset groups. These groups are:
Stocks and industries: 2,000 stocks and industry portfolios across major markets.
Country equity indices: 20 country indices from developed and emerging markets.
Bonds: 10-year futures in six developed markets.
Interest rate futures: Short-term interest rate futures in five developed markets.
Currencies: 22 currencies in developed and emerging markets.
Commodities: Eight commodities futures.
Correlations
Not only has each of the styles used in the fund provided a premium, but they all have exhibited low to negative correlations with market beta and to each other. For the period from 1990 through 2013:
The monthly correlations of the value, momentum, carry and defensive styles to market beta were approximately 0, 0, 0.3 and 0, respectively.
Relative to bonds, the monthly correlations of the value, momentum, carry and defensive styles were 0, 0.1, 0.1 and 0.2, respectively.
The monthly correlations of value to the momentum, carry and defensive styles were -0.6, -0.1 and -0.1, respectively.
The monthly correlations of momentum to the carry and defensive styles were 0.2 and 0.1, respectively.
The monthly correlation of carry to the defensive style was 0.1.
Implementation
Each strategy is implemented in a systematic manner, using a clearly defined, transparent process. The process employs liquid securities, which helps keep transaction costs low. The historical results demonstrate that, while each style within its asset class (or group) has on its own shown a premium, the diversification benefit leads to a whole that is greater than the sum of its parts. In short, composites perform better than the components. For example, the hypothetical Sharpe ratio for the carry factor from 1990 through 2013 was about 0.7 for fixed income, currencies and commodities. The composite Sharpe ratio for the carry factor across these three asset classes was about 1.2. Note that these Sharpe ratios are based on gross returns, and thus overstate the actual results. However, the important point is that the composite outcomes are superior to component results.
Target Allocations
The target risk allocations of the fund are:
30 percent equities across stocks and industries.
20 percent equity indices.
20 percent bonds.
15 percent currencies.
15 percent commodities.
This results in an implied style allocation that is:
34 percent value.
34 percent momentum.
18 percent defensive.
14 percent carry.
Use of Leverage
The fund uses leverage to target an annual volatility of 10 percent. The level of leverage employed is adjusted over time, adapting to market conditions. The expectation is that the fund will produce equity-like returns, but with about half the volatility of the market. Over the long term, the average use of leverage is expected to provide investors with $3 to $4 in both long and short positions for each dollar invested.
Expected Return
As with the alternative lending, reinsurance and VRP strategies we have discussed, QSPRX has an equity-like expected return (forecasted at about 7 percent net of fees).1 However, its expected volatility (10 percent) is only about half that of equities. In addition, the correlation of its return to traditional stock and bond returns is expected to be low. This makes QSPRX an excellent diversifier, reducing tail risk relative to long-only portfolios.
Location
Due to its relatively high turnover and the tax treatment of futures, the fund generally should be considered for tax-advantaged accounts.
Summary
Niels Pedersen, Sébastien Page and Fei He, authors of the study “Asset Allocation: Risk Models for Alternative Investments,” concluded that risk premiums diversify more efficiently than traditional alternative investments. They also concluded that the returns of an equally dollar-weighted risk premium portfolio are comparable to those of an endowment portfolio (with allocations to venture capital and hedge funds) except with far less risk.
On an interesting note, the authors cited two studies that found a simple, 1/N diversification strategy (equal-weighting the factors it diversified across) was as good as any of the other methods they tested. This type of strategy is referred to as risk parity.
1 Expected return assumptions are based on statistical modeling and are therefore hypothetical in nature and do not reflect actual investment results. They are not a guarantee of future results.
Chapter 9:
Time-Series Momentum
As you may recall, time-series momentum examines the trend of an asset with respect to its own past performance. For that reason, it is often referred to as trend-following. Strategies that attempt to capture the return premium offered by time-series momentum are often called “managed
futures,” as they take long and short positions in assets via futures markets—ideally in a multitude of futures markets around the globe.
Time-series momentum is different from cross-sectional momentum, which compares the performance of an asset relative to another asset, and has just as much, if not greater, support in the academic literature.
The Evidence
AQR Capital Management’s Brian Hurst, Yao Hua Ooi and Lasse Pedersen, authors of the 2017 paper “A Century of Evidence on Trend-Following Investing,” an update of their 2014 study, constructed an equal-weighted combination of one-month, three-month and 12-month time-series momentum strategies for 67 markets across four major asset classes (29 commodities, 11 equity indices, 15 bond markets and 12 currency pairs) for the period from January 1880 to December 2016. Their results include implementation costs based on estimates of trading expenses in these four asset classes. They further assumed management fees of 2 percent of asset value plus 20 percent of profits, the traditional “2/20” fee for hedge funds. The following is a summary of their findings:
Performance was remarkably consistent over an extended time horizon that included the Great Depression, multiple recessions and expansions, multiple wars, stagflation, the global financial crisis of 2008, and periods of rising and falling interest rates.
Annualized gross returns were 18.0 percent over the full period, with net returns (after fees) of 11.0 percent, higher than the return for equities but with approximately half the volatility (an annual standard deviation of 9.7 percent).
Net returns were positive in every decade, with the lowest net return, at 4.1 percent, coming in the period beginning in 1919. Net returns were in the single digits over seven of the 14 decades in the period.
There was virtually no correlation to either stocks or bonds. Thus, the strategy provides a strong diversification benefit. After considering all costs and the 2/20 hedge fund fee, the Sharpe ratio was still 0.76. Even if future returns are not as strong, the diversification benefits would justify an allocation to the strategy.
Hurst, Ooi and Pedersen write that “a large body of research has shown that price trends exist in part due to long-standing behavioral biases exhibited by investors, such as anchoring and herding [and we would add to that list the disposition effect and confirmation bias], as well as the trading activity of non-profit-seeking participants, such as central banks and corporate hedging programs.” They observe, for instance, that “when central banks intervene to reduce currency and interest-rate volatility, they slow down the rate at which information is incorporated into prices, thus creating trends.”
Hurst, Ooi and Pedersen continued: “The fact that trend-following strategies have performed well historically indicates that these behavioral biases and non-profit-seeking market participants have likely existed for a long time.”
They noted that trend-following has done particularly well in extreme up or down years for the stock market, including the most recent global financial crisis of 2008. In fact, the authors found that during the 10 largest drawdowns experienced by the traditional 60 percent stock/40 percent bond portfolio over the 137 years covered in their study, the time-series momentum strategy delivered positive returns in eight of these stress periods, and it delivered significant positive returns in a number of those.
While Hurst, Ooi and Pedersen provided results that included a 2/20 fee structure, today funds with much lower, although still not exactly cheap, expense ratios are available. The authors’ firm, AQR, has found that, in implementing time-series momentum strategies, their actual trading costs have been only about one-sixth of the estimates used in the study for much of the sample period (1880 through 1992) and approximately one-half of the estimates used for the more recent period (1993 through 2002).
Hurst, Ooi and Pedersen demonstrate that the time-series momentum premium has been persistent across time and economic regimes, is pervasive across asset classes, is robust to various definitions, has low correlation to other factors, and is implementable.
Supporting Evidence
The preceding findings are consistent with those from prior research, such as the 2013 study by Akindynos-Nikolaos Baltas and Robert Kosowski, “Momentum Strategies in Futures Markets and Trend-Following Funds.” Their study covered the period from December 1974 through January 2012 and included 71 futures contracts across several asset classes, specifically 26 commodities, 23 equity indices, seven currencies, and 15 intermediate-term and long-term bonds. The following is a summary of their findings:
Time-series momentum exhibited strong effects across monthly, weekly and daily frequencies.
Strategies with different rebalancing frequencies had low cross-correlations, and therefore captured distinct return patterns.
Times-series momentum patterns were pervasive and fairly robust over the entire evaluation period and within sub-periods.
Different strategies achieved annualized Sharpe ratios above 1.2 and performed well in both up and down markets, which rendered them good diversifiers in equity bear markets. The fact that different strategies were successful demonstrates time-series momentum’s robustness.
Commodity futures-based momentum strategies had low correlation with other futures strategies. Thus, despite the fact that they had a relatively (compared to the returns of momentum strategies for stocks, bonds and currencies) lower return, they do provide additional diversification benefits.
Importantly, Baltas and Kosowski found that momentum profitability is not limited to illiquid contracts. Rather, momentum strategies are typically implemented by means of exchange-traded futures contracts and forward contracts, which are considered relatively liquid and have relatively low transaction costs compared to cash, equity or bond markets. In fact, they found that “for most of the assets, the demanded number of contracts for the construction of the strategy does not exceed the contemporaneous open interest reported by the Commodity Futures Trading Commission (CFTC) over the period 1986 to 2011.” The authors also found that the “notional amount invested in futures contracts in this hypothetical scenario is a small fraction of the global [over-the-counter] derivatives markets (2.3% for commodities, 0.2% for currencies, 2.9% for equities and 0.9% for interest rates at end of 2011).” Thus, they concluded: “Our analyses based on the performance-flow regressions and the hypothetical open interest exceedance scenario do not find statistically or economically significant evidence of capacity constraints in time-series momentum strategies.”
We see the findings presented by Baltas and Kosowski again demonstrate that time-series momentum has been persistent, pervasive and robust, as well as that it provides diversification benefits and is implementable. We find further support for the factor in the 2014 study, “Is This Time Different? Trend Following and Financial Crises.”
Using almost a century of data on trend-following, the authors, Mark Hutchinson and John O’Brien, investigated what happened to the performance of the strategy subsequent to the U.S. subprime and Eurozone crises, and whether it was typical of what occurs after a financial crisis. But, as Hutchinson and O’Brien observed, identifying a sample of global and regional financial crises can be problematic. Thus, they chose to use the list from two of the most highly cited studies on financial crises, “Manias, Panics, and Crashes: A History of Financial Crises” (originally published in 1978 by Robert Aliber and Charles Kindleberger) and “This Time Is Different: Eight Centuries of Financial Folly” (originally published in 2009 by Carmen Reinhart and Kenneth Rogoff). The six global crises Hutchinson and O’Brien studied were: the Great Depression of 1929, the 1973 Oil Crisis, the Third World Debt Crisis of 1981, the Crash of October 1987, the bursting of the dot-com bubble in 2000 and the Subprime/Euro Crisis beginning in 2007. The regional crises they studied (with the year of inception in parentheses) were: Spain (1977), Norway (1987), Nordic (1989), Japan (1990), Mexico (1994), Asia (1997), Colombia (1997) and Argentina (2000). The start date for each crisis was the month following the equity market high precedi
ng it. Because neither of the two studies Hutchinson and O’Brien used for their list furnished guidance on the length or end date of each crisis, rather than attempting to define when each individual crisis concluded, the authors instead focused on two fixed time periods: 24 months and 48 months after the prior equity market high.
Hutchinson and O’Brien’s dataset for their global analysis consisted of 21 commodities, 13 government bonds, 21 equity indices and currency crosses derived from nine underlying exchange rates covering a sample period from January 1921 to June 2013. Their results incorporate trading cost estimates as well as the typical 2/20 hedge fund fee. The following is a summary of their findings:
Time-series momentum has been highly successful over the long term. The average net return for the global portfolio from 1925 to 2013 was 12.1 percent, with volatility of 11 percent. The Sharpe ratio was an impressive 1.1 (a finding consistent with other research).
A breakdown in futures market return predictability occurred during crisis periods.
In no-crisis periods, market returns exhibited strong serial correlation at lags of up to 12 months.