Tag Archives: undefined

The Low Volatility Anomaly And The Delegated Agency Model

Summary This series offers an expansive look at the Low Volatility Anomaly, or why lower risk stocks have historically produced stronger risk-adjusted returns than higher risk stocks or the broader market. This article hypothesizes that the combination of a cognitive bias and an issue around market structure could contribute to the Low Volatility Anomaly. This article covers a deviation between model and market that may contribute to the outperformance of low volatility strategies. In the last article in this series , I demonstrated that the aversion of certain classes of investors to employing leverage flattens the expected risk-return relationship as leverage-constrained investors bid up the price of risky assets. In addition to the inability to access leverage for long-only investors, the typical model of benchmarking an institutional investor to a fixed benchmark (i.e. the S&P 500 represented through SPY ) could also potentially produce a friction to exploiting the mispricing of low volatility assets (represented through SPLV ). If a security with a beta of 0.75 produces the same tracking error as a security with a beta of 1.25, investors may be more willing to invest in the higher beta security with the belief that it is more likely to generate higher expected returns per unit of tracking error. In this framework, if the investor believes that the higher beta security is going to deliver 2% of alpha and that the higher and lower beta assets are going to have the same tracking error relative to the index, then the investor would not purchase the lower beta asset unless it was expected to earn alpha of more than 2%. An undervalued low beta stock with a positive expected alpha, but an alpha below the expected alpha of a higher beta stock with an equivalent expected tracking error, would be a candidate to be underweight in this framework despite offering both higher expected return and lower expected risk than the broad market. This investor preference results in upward price pressure on higher beta securities and downward price pressure on lower-beta securities that could be a factor in the lower realized risk-adjusted returns of higher beta cohorts depicted in the introductory article in this series . In a foreshadowing of the next article on the potential influence that cognitive biases have on shaping the relationship between risk and return, the difference between absolute wealth and relative wealth could be an important distinction that influences the behavior of delegated investment managers. Richard Easterlin (1974) found that self-reported happiness of individuals varied with income at a point in time, but that average well-being tended to be very stable over long time intervals despite per capita income growth. The author argued that these patterns were consistent with well-being depending more closely on relative income than absolute income. This preference for relative outperformance rather than absolute outperformance may signal why some managers think of risk in terms of tracking error rather than absolute volatility. In perhaps a more salient example, Robert Frank (2011) illustrated the relative utility effect through an experiment that showed that the majority of people would rather earn $100,000 when peers were earning $90,000 than earn $110,000 when peers were earning $200,000. Among the assumptions underpinning CAPM is that investors maximize their personal expected utility, but these studies suggest that investors in effect seek to maximize relative and not absolute wealth. Similar to leverage aversion detailed in the last article, the preference for relative utility could be another CAPM violation that contributes to the Low Volatility Anomaly. Gauging performance versus a benchmark is a form of maximizing relative utility, and has become an institutionalized part of the investment management industry perhaps to the detriment of the desire to capture the available alpha in our low beta asset example. I am not trying to minimize tracking error in my personal account, I am trying to generate risk-adjusted returns to grow wealth over time. As I have demonstrated in this series, academic research has shown that low volatility stocks have outperformed on a risk-adjusted basis since the 1930s. Disclaimer My articles may contain statements and projections that are forward-looking in nature, and therefore inherently subject to numerous risks, uncertainties and assumptions. While my articles focus on generating long-term risk-adjusted returns, investment decisions necessarily involve the risk of loss of principal. Individual investor circumstances vary significantly, and information gleaned from my articles should be applied to your own unique investment situation, objectives, risk tolerance, and investment horizon. Disclosure: I am/we are long SPLV, SPY. (More…) I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Global X Serves Up A New Alternative ETF

By DailyAlts Staff Global X Funds has a growing line of “SuperDividend” ETFs, and the latest is sure to be of interest to liquid alts investors: the Global X SuperDividend Alternatives ETF (NASDAQ: ALTY ) , which began trading on the Nasdaq on July 14. The ETF is the sixth in Global X’s SuperDividend ETF series and is designed to closely track the INDXX SuperDividend Alternatives Index. The underlying index tracks the performance of the highest dividend-yielding securities in each category of alternative income investments, as defined by index sponsor INDXX. This includes MLPs (master limited partnerships), REITs (real estate investment trusts), BDCs (business development companies), and other nontraditional income-producing investments. The Global X SuperDividend Alternatives ETF’s aim is to provide income from alternative sources with low correlation to dividend stocks and other traditional income investments. The strategy also seeks to limit volatility by screening for lower-volatility investments and overweighting categories that have been less volatile, historically. “The alternatives space encompasses a broad range of investments with risks, returns and correlations that differ from traditional equity and fixed income securities,” said Jay Jacobs, research analyst at Global X Funds, in a recent statement. “Investors are increasingly looking for alternative solutions that can potentially generate high income while diversifying their portfolios. Applying the SuperDividend approach to the alternative income space is a natural extension of our suite.” In practice, the Global X SuperDividend Alternatives ETF carries exposure to MLPs and other infrastructure companies, REITs and other real estate investments, alternative managed portfolios, and fixed-income and derivative strategies. As of July 13, REITs accounted for the largest share of the fund’s holdings at over 26%, while private equity and BDCs accounted for the next-largest share at over 19%. MLPs constituted less than 9% of the fund’s holdings, according to the fund’s fact sheet . On the downside, the Global X SuperDividend Alternatives ETF’s expense ratio is rather high at 3.03%. This is largely a function of the fact that the fund invests in other funds, according to ETF.com , and, as required, includes the underlying expense ratios of those fund in its expense ratio.

Eureka! A Valuation-Based Asset Allocation Strategy That Might Work

By Wesley R. Gray, Ph.D. We’ve had a few posts showing that asset allocation systems relying on market valuation indicators (e.g., Shiller CAPE ratios) as a timing signal may end up in disappointment… Nonetheless, we’ve continued on the quest to improve tactical asset allocation using market valuation data. The data speaks clearly when it comes to the association between valuations and long-term realized returns – high valuations are associated with low long-term realized returns. However, as Michael Kitces highlights, tactically allocating using valuation information is challenging . Moreover, there are arguments that the association between CAPE and LT returns may be more complex than was previously thought. In short, valuation-based asset allocation strategies haven’t been that exciting, but… The folks at Gestaltu inspired us with a unique twist on basic valuation-based timing methodologies: … we chose the cyclically adjusted earnings yield as the valuation metric, which is just the reciprocal of the Shiller PE. We then adjusted the yield value for the realized year-over-year inflation rate to find the real earnings yield. Finally, we used an ‘expanding window’ approach to find the percentile rank of the real earnings yield to eliminate as much lookahead bias as possible. Note that because we are using real earnings yield rather than nominal earnings yield, markets can get cheap or expensive in three ways: changes in inflation changes in earnings changes in price Gestaltu’s post used 1/CAPE as the valuation metric, or the “earnings yield,” as a baseline indicator; however, they “adjusted the yield value for the realized year-over-year (yoy) inflation rate” by subtracting the year-over-year inflation rate from the rate of 1/CAPE. To summarize, the metric looks as follows if the CAPE ratio is 20 and realized inflation (Inf) is 3%: Real Yield Spread Metric = (1/20)-3% = 2% Fairly simple. Strategy Background: We performed our own replication of the first two strategies from the post: Average Valuation-based asset allocation: Own S&P 500 when valuation < long-term average, otherwise hold cash. In other words, if last month's CAPE valuation is in the 50 percentile or higher, buy U.S. Treasury bills (Rf); otherwise stay in the market. 80th Percentile Valuation-based asset allocation: Own S&P 500 when valuation < 80th percentile, otherwise hold cash. In other words, if last month's CAPE valuation is in the 80 percentile or higher, buy U.S. Treasury bills (Rf); otherwise stay in the market. Some adjustments are applied in the replication: The Bureau of Labor Statistics (BLS) publishes the CPI on a monthly basis since 1913; however, the data is one-month lagged (possibly longer). For example, the CPI for January won't be released until February. So, when we subtract the year-over-year inflation rate from the rate of 1/CAPE, we do a 1-month lag to avoid look-ahead bias. We use the S&P 500 Total Return index as a buy-and-hold benchmark. Our back test period is from 1/1/1934 to 12/31/2014, while the article looks over the period from 1/1/1934 to 12/31/2012. The results are gross of any fees. All returns are total returns and include the reinvestment of distributions (e.g., dividends). Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Our Replication Results The first table shows the results from the Gestaltu post: The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Our backtest results show similar CAGRs, but higher volatilities than the results from Gestaltu. This could be due to changes in experiment design. Overall, the "Abs Return 80%" strategy outperforms buy-and-hold, while the "Abs Return 50%" strategy underperforms buy-and-hold. We include a long-term moving average rule for reference (S&P 500 if above the 12-month MA, risk-free if below the 12-month MA). Summary statistics are below: (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Robustness Tests: Adjusting the Starting Point for the Look-Back Window Gestaltu set 1/1/1924 as the starting date, and then uses an expanding window as the look-back period. We investigate how changing the start date affects the results. The results shown are from 1/1/1934 to 12/31/2014. The table below shows the results of the "Abs Return 80%" strategy using different starting dates for the expanding window: 1924, 1900, and 1881. The starting date for the expanding window calculation can create marginal differences in the results. For example, the Sharpe ratios vary from 0.57 to 0.63. Overall, the results appear robust to the expanding look-back window start date. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Robustness Tests: Rolling Look-Back Window In this section, we try a 10-year rolling look-back period calculation. For example, we measure the percent rank of CAPE on 12/31/2014 relative to the past 10 years (12/31/2004 to 12/31/2014); while an expanding window (results already shown above) would measure the percent rank of CAPE on 12/31/2014 relative to the whole time period (from the start date to 12/31/2014). The results below highlight that a rolling-window technique yields similar results to the expanding-window technique. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Robustness Tests: Inflation-Adjusted P/E Ratio In this section, we use the old-fashioned price-to-earnings ratio in place of the CAPE ratio. We use a rolling 10-year window look-back method and adjust inflation with a 1-month lag. Full Sample Results: 1/1/1934 to 12/31/2014 Inflation-adjusted P/E strategies work better than simple Moving Average rules and buy-and-hold. They also work better than CAPE-based strategies. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. First-Half Results: 1/1/1934 to 12/31/1974 Inflation-adjusted P/E strategies work well in the first half. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Second-Half Results: 1/1/1975 to 12/31/2014 Inflation-adjusted P/E strategies work well in the second half. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Robustness Tests: Different Thresholds In this section, we look at different percentile thresholds to determine the timing signal. For example, the results are strong when the timing signal is based on the average or the 80th percentile, but what happens if we use different signals? We use a rolling 10-year window look-back method and adjust inflation with a 1-month lag. Full Sample: 1/1/1934 to 12/31/2014 Higher thresholds increase maximum drawdowns (relative to lower thresholds, such as the 50th and 80th percentiles). The results are better than pure buy-and-hold, but this does highlight a potential robustness issue. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system may have robustness issues. Robustness Tests: Staggered Allocations In this robustness test, we vary holding percentages based on the percentile rank of earnings yield - realized inflation. For example, if last month's E/P - CPI is in the 12th percentile based on the past, then we allocate 12% to stock and 88% to T-bills. We use a rolling 10-year window look-back method and adjust inflation with a 1-month lag. Full Sample: 1/1/1934 to 12/31/2014 Staggered allocations strategies are better than buy-and-hold. (click to enlarge) The results are hypothetical, are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system looks promising. Robustness Tests: Changing the Real Inflation Component of the Signal In the post, " Market Valuations based on CAPE - A Deeper Dive ", we take the 1/CAPE and subtract the inflation adjusted 10-year U.S. Treasury yield, so that we can examine how expensive the market is relative to real returns available via a bond alternative (a stock investor would prefer a higher spread, all else being equal). To summarize, the metric looks as follows if the CAPE ratio is 20, realized inflation (Inf) is 3%, and the 10-Year Treasury is 5%: Real 10-Year Spread Metric = (1/20)-(5-3)% ~ 3% Full Sample: 1/1/1934 to 12/31/2014 This new measure doesn't work - at all. Understanding why a seemingly small change in technique destroys the results is puzzling and worthy of more investigation... (click to enlarge) The results are hypothetical, and are NOT an indicator of future results, and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request. Bottom line: The system may have robustness issues. Conclusion After enduring years of frustration trying to identify a valuation-based asset allocation technique - that actually worked - I think the team at Gestaltu is on to an interesting concept. By simply looking at real spreads between equity valuations and realized inflation (high spreads are good for equity; low spreads are bad for equity), one can devise a timing rule that captures most of the upside, but protects on the downside. Of course, this is all historical data and could very well be an exercise in data mining. That said, the concept of buying equity assets when they have much higher yields than current inflation is intuitively appealing. We'll continue our investigations into the subject, but we wanted to give a quick view into some of our high-level research on the subject. Original Post