‘Value investing’ seems like a fairly straightforward concept, but many investors have realized that there are nuances to value investing—especially in light of large losses in some value-oriented strategies in 2008-2009. The idea seems simple: buy stocks that are relatively low-priced, as measured by dividend yield, price-to-earnings ratio, or price-to-book. There are studies going back decades that have noted that buying stocks at low P/E ratios tends to yield have average total returns than buying stocks at higher P/E ratios (see Malkiel’s A Random Walk Down Wall Street, for example). Fama and French identify ‘value’ as a statistically robust predictor of higher returns. Rob Arnott has further popularized this idea with ‘fundamental indexing.’
One challenge for value investors is that any given statistical metric of value can mean different things. Dividend yield, in particular, is a problematic measure. The dividend yield can be high because a company is solid and throws off robust earnings that are returned to shareholders, but dividend yield can also be high for a company in distress that has seen its stock price plummet but that has not yet seen its dividends reduced to reflect its current status. There are other challenges, too. Some companies employ ‘leveraged’ dividend strategies in which they will pay out more in dividends than their actual earnings. This surely cannot be compared to companies that only pay shareholders a fraction of real earnings.
An additional pitfall to investing entirely on the basis of a metric like dividend yield is that firm managers are well aware that dividends are considered an important ‘signal’ to investors. Mutual funds and retail investors use the signal of confirmed or increased dividends as the basis for a purchase---particularly funds that have a stated ‘value’ tilt. Companies may have an incentive to increase leverage to boost dividends, thereby increasing the risks associated with future earnings. While this may not be in the best interests of shareholders, it is a rational response on the part of managers.
Further, it has been noted that ‘value’ investing goes through periods of out-performance, followed by periods of under-performance relative to ‘growth.’ This is something of a tactical investing challenge. After a period in which value stocks have out-performed, it is quite likely that value will go out of style and growth will be popular. Investors need to be cognizant of this effect.
What all of this means for ‘value’ investors is that there are three key components to the investing decision that require consideration:
1) dividend yield (or other fundamental factors)
2) risk (to determine if a company is in distress)
3) trailing returns relative to ‘expected’ returns
The first of these factors is the basis for the concept of value investing: judging the price of a stock in terms of the value of the company as determined from earnings, dividends, or book value. The second factor, risk, is very important as well. If a value stock has very high risk, as reflected in trailing volatility, implied volatility, or model-based estimates of expected volatility, this is a warning that the very high yield might be due to the company being in distress—the classic value trap.
In March of 2008, I examined a series of individual stocks in attempting to determine their expected future volatility and default risk. At that time, Citigroup had a trailing 12-month yield of 8.1% and closed out February at a price of $23.71. My analysis at the time (using Quantext Portfolio Planner with all default settings) projected 34% annualized volatility for C. Consolidated Edison (ED) was trading at $40.89 at the end of February 2008, and had a trailing 12-month yield of 5.7%. Using data through February 2007, Quantext Portfolio Planner (QPP) projects a 16% annualized volatility—less than half that for C. Today, C is trading at $4.68 and ED is trading at $39.8.
What I am suggesting is that value investors should look at implied volatility and/or use a model to generate a meaningful expected volatility estimate as a key decision point is looking at value-driven investment decisions. It is very simple to get implied volatilities using an online tool such as that provided by iVolatility.com. As of this writing, Citi has an implied volatility of around 70%, while ED has implied volatility of around 20% for long-dated options.
The key question for people who are not familiar with implied volatility is whether it is, indeed, a good estimate of future actual volatility (and, by extension, downside risk). This issue has been explored in a range of studies. The results tend to show that implied volatility is a remarkably good estimator of future volatility.
The study linked above finds that there appear to be some systematic biases in option pricing having to do with Beta and volatility of the underlying stock. This effect has also been found Monte Carlo simulations (using Quantext Portfolio Planner, QPP). The study linked above finds that options on low-Beta / low-volatility stocks (which are often value stocks like utilities) tend to be over-priced (and vice versa). QPP finds the same result. In my analysis, I suggested a new value-oriented strategy:
1) Buy stocks with high yield, low Beta, and low volatility
2) Sell covered calls against these stocks
3) Buy call options on high Beta, high volatility, non-dividend-paying stocks
Developing a value investing strategy using options is clearly more complex than simply using implied volatility to inform the process of vetting value stocks. Regardless of which path is taken, however, there is no question that options prices (and the implied volatilities that are derived from options prices) provide valuable information.
An additional consideration for value-oriented investors is to use some measure of expected returns for a value-driven portfolio strategy. High-quality value stocks (such as the Dividend Aristocrats) tend to allow the creation of highly diversified portfolios which have high risk-adjusted returns. This occurs largely because these stocks tend to have low Beta and low correlations to one another. Two and a half years ago, Dividend Aristocrats looked like a great deal—and they subsequently out-performed the broader market. Even four months ago, with higher correlations among all asset classes, Aristocrats still had a significant projected strategic (long-term) expected benefit for a portfolio, albeit diminished from previous years.
The final issue that I believe needs additional consideration is the tactical side of value investing—the examination of shorter-term price / value anomalies. There will be periods when value-based investing under-performs, even if a portfolio with a value tilt tends to out-perform over the long haul. Stocks have some rational expected return and they cannot exceed this forever. When value stocks have out-performed their reasonable expected return levels, it is likely that they will under-perform for some period to get their valuations consistent with a reasonable expectation. Evaluating stocks on this basis is a form of tactical asset allocation. Using Monte Carlo simulations, I often compare trailing returns to projected ‘fair’ returns to determine if a potential investment is over-priced or under-priced. This approach works when evaluating value investing, too.
To demonstrate this point, I obtained the holdings of DVY in June 2008 and examined the top forty holdings. Using Quantext Portfolio Planner (QPP) with all baseline settings and three-years of trailing data through June 2008, I calculated the difference between the model-generated expected return and the trailing three-year return. Expected return is a measure of long-term ‘fair’ returns in order for there to be a balance between risk and return. When returns have been unsustainably high, the expected return minus the trailing return is negative—and this indicates that a stock may be over-valued. The opposite is also true.
The table below shows the top forty holdings of DVY as of June 2008, sorted by the difference between modeled expected return and trailing three-year return through June 2008. The table also shows the average annual returns in the next 14 months for the half of the list projected to be most over-valued vs. the half of the list projected to be most under-valued. The average annual return for the group of stocks that were most over-valued is -6.1% for the subsequent 14 months, while the average annual return for the group of stocks that were most under-valued is +9.1% for the same period.
These results suggest that we have experienced a substantial mean reversion—and the prices of the most over-valued stocks have come down while those of the most under-valued stocks have gone up. This is, of course, only one example of this effect but my prior research supports this approach. Further, I hasten to note that the mean reversion did not occur for every stock in the list—this is a statistical tendency, but surely not an infallible indicator.
The average trailing three-year dividend yield for the most over-priced half of the list and the most under-priced half of the list are essentially identical through June 08: 3.98% and 4.02%, respectively.
A notable difference between the over- and under-priced groups is trailing historical volatility. The under-priced list averages annualized volatility 1.4 times the trailing volatility of the over-priced list for the three years through June 2008. The under-priced list has a trailing historical Beta (with respect to the S&P500) of 114% vs. 80% for over-priced list. Interestingly, the under-priced list actually averaged 1.6 times the volatility of the over-priced list in the following fourteen months, remarkable close to the historical ratio. The portfolio of stocks comprised of the most under-valued stocks was also clearly much riskier than the portfolio of over-valued stocks.
We can look at the criterion of relative valuation in conjunction with trailing volatility as a measure of risk. The median trailing three-year volatility (through June 2008) for this list of stocks is 20.9%. If I screen for stocks in the total list that have trailing volatility less than 20.9% and have projected return greater than trailing three year return, I end up with a list of stocks that are relatively low volatility and under-valued. This list of stocks averaged 5% in return over the next 14 months with lower average volatility for the stocks in the list than for the entire list.
There are a range of reasons why investors choose to invest on the basis of value. There are numerous studies that demonstrate that value investing has historically increased return—the so-called value premium. There are many firms that clearly have a policy of delivering high-quality (stable) earnings streams. These firms may tend to be under-valued because they are not ‘exciting’ to investors. By my hypothesis, these firms should tend to have relatively low Beta and low volatility. As such, when choosing between possible value-based investments, it is very important to look at both Beta and volatility. I believe that this is likely the reason why Fama and French (2004) find that low-Beta portfolios generate alpha.
Value investors may have interest in companies that are in distress and are highly volatile, but they may also want companies with stable high-quality earnings. Beta and volatility provide valuable indicators about which stocks are appropriate for which strategy.
The idea that that low-Beta / low-volatility / high-dividend stocks can provide value for investors over long periods does not mean that these stocks will always out-perform. In fact, as I showed in the example for the components of DVY, we have just seen a period in which higher-Beta and higher-volatility value-oriented stocks have out-performed.
I believe that value investors need to have clarity as to their additional goals as value investors—notably where they wish to be on the continuum of Beta and total volatility. The traditional simple distinction between ‘value’ and ‘growth’ ignores the fact that there are other dimensions to the specification of a value-based strategy, as well as how to pursue it. One area that has not received sufficient attention, for example, is as to how options can be used to execute a value-oriented strategy. Options allow investors to pursue value-oriented strategies at varying levels of risk.
The key theme here is that effectively choosing between value-oriented investments is informed by the use of metrics such as Beta, implied and projected volatility, and estimation of relative over- or under-performance. A value investing strategy based entirely upon metrics such as dividend yield or P/E which ignores these other critical dimensions of the value investing process may expose investors to risk levels they are not expecting. This problem was experienced first-hand by investors in the RAFI fundamental large-cap strategy (PRF), which tended towards the higher Beta / higher volatility side of the value universe and is overweight in Financials. Over the three years through August 2009, PRF has a Beta of 121% and a trailing volatility 25% higher than the S&P500. Because RAFI’s fundamental index model does not filter investments on the basis of risk measures, it has ended up substantially more risky than the S&P500. While this may be what some value investors are looking for, it is not what most income-oriented investors are seeking.