- When the market is in a phase of rich valuation, it is tempting to broadly identify the market for these stocks as “overvalued”.
- Making the distinction between stocks deserving high valuations and those that are overvalued is key, particularly in the current market.
- In part 1 of this series, we looked at fundamental factors based on financial statements to help us make this distinction.
- In this piece, I share a new strategy I’ve developed, the “A_Score”, that exploits market “anomalies” to find outperforming stocks with high valuations.
- We look at the strategy and its performance in detail, and I provide some high and low scoring names by the A_Score.
In Part 1 of this series, we looked at how to potentially find winners in the universe of stocks with high valuations. We discussed how it is important to separate valuation from "valuation sentiment" - i.e., stocks with high valuations are often deemed overvalued, however, they can be under-, fairly- or overvalued. The same goes for stocks with low valuations. This two-dimensional approach is summarized below.
Source: Author Table
Within those stocks with high valuations (as measured by the stocks with high price to book (P/B) ratios, top 20%), we looked at the G_Score. This quantitative scoring system, first developed by Mohanram in 2003, compares various metrics from financial statements of these stocks to the industry median. When a metric for the stocks is higher (or lower) than the industry median, a score of 1 is assigned for the category, 0 if less (or higher). There are 8 factors total, so stocks could score a maximum of 8, and as low as 0.
The factors in the G_Score are as follows, grouped in 3 main categories:
- G1 - Profitability - Return on Assets
- G2 - Profitability - Cash flow Return on Assets
- G3 - Quality of Earnings - Low Accruals
- G4 - Stability - Earnings Growth
- G5 - Stability - Sales Growth
- G6 - Intangible Asset Investment - R&D
- G7 - Intangible Asset Investment - CAPEX
- G8 - Intangible Asset Investment - Advertising
In Part 1, we looked at each factor in detail and their individual performance. We also looked at performance of high-scoring stocks (G_score of 6 or greater), and low-scoring (G_score of 2 or less), summarized below.
Source: Portfolio123 data, Author Table
For our study, the G_Score was able to find winning stocks within the high P/B universe. These stocks, as a group, also outperformed the Russell 3000 index, except from 1999 thru 2005 (tech bubble and recovery).
In terms of low scoring stocks, the G_score was able to find underperforming stocks in the high P/B universe (high negative spread) in all periods:
Source: Portfolio123 data, Author table
These stocks consistently underperformed the Russell 3000 on a rolling 1-year basis as well, except for over the last 4 years.
Introducing the "A_Score"
The original G_Score looks at metrics contained in financial statements. In this piece, we will look at some other factors that are based more on market "anomalies". In efficient market academic circles, the CAPM (capital asset pricing model) should be able to explain market returns. In reality, the market is not entirely efficient, and there are many other drivers of excess returns. These factors are considered "anomalies" as they are not explained by the governing theory. Practically speaking, there are so many anomalies (or exceptions) to the CAPM, this in itself is a testament to just how inefficient the efficient market theory is.
These anomalies could be driven by investor and human behavior, such as overreactions, underreactions, etc. For an investor aware of these anomalies, there can be an opportunity to exploit them to help find alpha and excess returns.
Anomalies can persist over time, or they can be arbitraged away after they are discovered and publicized. Some are relevant for certain types of stocks and less for others.
Based on my research on anomalies, particularly those applicable to stocks with high valuations, I have developed a new scoring system called the "A_Score". The factors included are not directly related to fundamentals as with the G_Score, but rather on broader market factors.
The individual factors comprising the A_Score are:
- Lottery effect
- Post earnings announcement drift (beating investor's expectations), and
- Change in short interest
Similar to the G_Score, stocks are scored on each factor, being assigned a score of 1 if they score higher (or lower) than the industry median for highly valued stocks, and 0 if otherwise. The higher the spread of performance between those stocks scoring a 1 (higher than the industry median) compared to those scoring a 0, the higher the chance that the factor has predictive power in future returns. We take this a step further, and look at the "spread of the spread" between our highly valued universe compared to the All Stocks universe, with higher values suggesting that the factor is particularly effective for our highly valued stocks.
Factor scores are tallied, and a stock will receive a total A_Score, with the highest score of 3, and low score of 0.
Let's now look at each of these factors in more detail.
We will look at each of our factors in the same way as we did with the G_Score. As with all of my strategies, I designed the A_Score in Portfolio123. Our main testing universe, referred to as High P/B universe, is constructed as follows:
- All US stocks from 1999, in the top 20% ranked Price to Book
- Utilities, REITs, and financials excluded
- Minimum median daily volume of $100k
- Minimum price $1
- Include ADRs
- This is roughly 600-700 stocks at any given time
We will also compare to an "All Stocks" universe, which is essentially the same as the above but includes stocks of all valuations.
Our main performance metric will be 1-year performance, on a median rolling basis (to account for seasonal effects during rebalancing). All stocks in the universe for the given factor (pass/fail) are held for one year and rebalanced annually. We will look at the period 1999 thru 2020, but also look at smaller periods within to account for different market regimes:
- 1999 thru 2005 (tech bubble and collapse)
- 2006-2010 (run-up to financial crisis, starting recovery)
- 2011-2015 (continued recovery)
- 2016-2019 (most recent period, without COVID)
- 2016-2020 (above, including COVID impact)
Note the last periods, 2016 thru 2019 and 2016 thru 2020. COVID initially rattled markets in early 2020, but over the year the market experienced exceptional performance (and volatility) in some areas, for a multitude of reasons. This will likely end up being a temporary effect over time. However, we will look at the most period with and without the effects of COVID.
Let's now look at the individual factors making up the A_Score.
A_Score Factor #1 - The Lottery Effect
The stock market is a wonderful innovation and has allowed countless individuals to accumulate wealth over time (key being "over time"). There can be rare opportunities or a perception of opportunities, that can turn an ordinary investor into a millionaire almost overnight.
Those stocks that attract "get rich quick" investors are referred to as "lottery stocks", and generally have underperformed.
There has been much research done to understand these types of stocks, and what drives some investors to invest, or gamble, with them. Kumar, in his 2006 paper "Who Gambles in the Stock Market?", defines lottery stocks as those with properties not unlike the characteristics of say a casino game or a national lottery:
- very low prices relative to the highest potential payoff (i.e., the size of the jackpot),
- negative expected returns,
- have risky payoffs (i.e., the prize distribution has extremely high variance), and
- extremely small probability of a very high reward - i.e., they have positively skewed returns.
Putting this in the context of stocks, there are different ways to measure the "lottery-ness" of stocks.
Kumar defines lottery-type stocks as those that are lower-priced
with very small but positive potential for high returns as lotteries. I further assume that stocks with higher variance (or higher idiosyncratic volatility or large extreme returns) and positively skewed returns are likely to be perceived as high payoff potential stocks."
For our stocks, we already screen out very low-priced issues - i.e., stocks must be at least $1 per share. We can, however, measure the skew of daily returns. Skew, for our purposes, is simply how asymmetric the return distribution is relative to the mean. Kumar refers to positive skew, which actually corresponds to a greater number of lower returns relative to the mean ("positive" refers to the right tail being longer than the left), illustrated below.
While this is interesting, in my own tests for skew, I have only found only a minor improvement in returns when the skew in returns is greater than the industry median. Again, we do not allow low-priced stocks in our universe (as we are only interested in stocks that are liquid and investable); perhaps the effect is more pronounced with these lower-priced stocks.
Another factor that has been studied to assess the "lottery-ness" of lottery stocks is MAX, written about by Bali et al in their 2011 paper "Maxing out: Stocks as lotteries and the cross-section of expected returns". MAX is simply the average of the five highest daily returns of the given stock in the last month; the higher the score, the higher the chances that the stock the returns are "positively skewed", and thus more likely to be a lottery-type stock. Bali et al returned in 2017 with a follow-up, "A Lottery Demand-Based Explanation of the Beta Anomaly", where the well documented "beta anomaly" (stocks with lower variability/correlation in returns compared to the market tend to outperform) was tested against those stocks with lottery-type characteristics. They found that lottery-type stocks are also high beta.
In my own research, I have found that the strongest predictor of returns in terms of lottery and low-beta stocks (in isolation, when compared to industry median) is simply the variance of daily returns in the past 3 months. This factor captures the essence of avoiding lottery-type stocks (naturally have low skew), and their beta value is generally lower as well. For our A_Score, we can theorize that there can be stocks in the high valuation group that have some lottery characteristics and have been bid up by investors perceiving a quick payoff.
For our A_Score, I refer to this factor as A1_Lottery Effect. The table below summarizes 1-year rolling performance of stocks with lower variance in returns, compared to the industry median, for both our all stocks universe and our highly valued universe. Source: Portfolio123 data, Author Table
Our lottery factor, when lower than the industry median, has been very effective for both our highly valued universe and our All Stocks universe on their own (even more so in the highly valued universe). The spread has reduced in the last few years, but still exists.
As an investor who invests predominately in small and microcap stocks, I am curious to see how the size factor works with other factors. In the table below, we control for size by looking at performance for the largest and smallest 50% of stocks, when ranked by market cap. The table below summarizes Lottery Effect performance based on size:
Our lottery effect factor has clearly had more of an edge in the smaller stocks (other than in the tech bubble & recovery period). This should actually not be surprising; Kumar found that it was low-priced stocks that exhibited lottery properties. While our small stocks are not necessarily as low in share price as Kumar suggested (penny stocks, pink sheets, etc.), they are often perceived as being more risky, but with higher reward, than their large-cap counterparts. Let's move on to our next A_Score factor.
A_Score Factor #2 - Beating Expectations
At any given time, market participants set prices for stocks based on their perceived value. When a stock's perceived value changes, the price changes accordingly. Earnings reports have a very significant effect on perceived value; when earnings exceed analyst estimates, perceived value typically increases, and the stock price follows. On the flip side, missing expectations can wreak havoc on a stock price. If the market were truly efficient, then these earnings surprises (positive or negative) would be instantly reflected in the stock price. In reality, what typically happens is that the market may have a knee-jerk reaction to the surprise, but it usually takes time for the information to become fully reflected in the stock price. This is known as Post Earnings Announcement Drift or PEAD.
The PEAD anomaly was first discovered by researchers Brown and Ball over 50 years ago, in their 1968 paper "An empirical evaluation of accounting income numbers". From my own research, this is a powerful factor, and I included a variation of it in my SaaS Scorecard strategy. So how do we exploit the PEAD factor for our highly valued stocks?
One method is to look at the history of a stock in beating expectations. A stock that manages to consistently beat expectations through earnings surprises either
- provides conservative guidance to the market (or doesn't provide any for that matter),
- analysts do not quite understand the business model and therefore provide low estimates, or
- there may be few analysts covering the stock, if any, to provide estimates to beat, or
- a combination of the above.
For our Beating Expectations factor, we will take the average of the earnings surprise over the last 4 quarters. The surprise is simply the difference between the actual earnings per share compared to the median analyst estimate; a surprise can be positive or negative. If this average is higher compared to the industry median, the stock receives a score of 1, and 0 otherwise.
The table below summarizes results for those stocks with the average earnings surprise of the last 4 quarters beating the industry median. Note the change in start date for the results, earnings surprise data availability from Portfolio123 started on 01 Jan 2003:
For both high-valued and All Stocks universes, there is generally a decent spread between stocks with a "beating expectations" score greater than their industry median. On average, the spread is more pronounced for the highly valued universe, particularly since 2011.
On average, we see a slightly higher spread of returns between the smaller stocks than the larger caps, where the spread was larger (and factor assumed to be more prominent) from 2003 thru 2010 for larger caps. Let's move on to our next factor in the A_Score.
A_Score Factor #3 - Change in Short Interest
It's tough to be a short seller. In having to use margin to "borrow" stocks to short, there are more frictional costs to account for. There is of course the risk of the dreaded "short squeeze" (which is great for long investors, particularly recently for long GameStop (GME) investors). To be a successful short seller, you really need to have a strong conviction on your positions. For this reason, short-sellers have earned the reputation of being the "smart money".
Prime targets of short-sellers are those stocks they believe are overvalued and ripe for a correction. The further the drop they expect, the greater the payoff for them. To see this in play, compare the median % short interest of our Highly Valued stock universe compared to the All Stocks universe, over the last few years:
Source: Portfolio123 data, Author Table
Clearly, there has been a consistently higher proportion of short interest in our highly valued universe than our All Stocks universe. This may then present an opportunity in taking long positions in our highly valued universe: avoid high short interest. We can take this a step further. While some factors are used at an instant in time, I find in many cases factors are far more helpful when they are measured over time.
Readers may be familiar with my interest in these types of factors in my Rule of 40 and SaaS Scorecards. For example, a high-profit margin is helpful, but is profitability improving or deteriorating? High overhead costs may not be an issue if they are improving over time.
The same goes for short interest. We believe high short interest stocks should be avoided, but what is the short interest sentiment? Are short sellers piling into the stock, or are they leaving? This tells us much more than just the amount of short interest at a specific point in time.
For our next A_Score factor, A3_Short Interest Sentiment, we will look at the change in short interest, relative to the industry median. If short interest over the last 3 months is reducing faster than the industry median, the stock receives a score of 1, and 0 otherwise. The table below summarizes the performance of this factor: Source: Portfolio123 data, Author Table
This factor has performed consistently well in each universe, with positive spread between 1 and 0 scores. The factor performed considerably better in our highly valued universe for the first 15 years of our study, but has lagged the All Stocks universe since 2016 (perhaps the GME phenomenon has been at play sooner than we thought).
Let's now look at how the factor performs when size is taken into account:
Clearly, decreasing short interest compared to the industry median is a much more powerful factor for our smaller high-valued stocks. Other than the period 2011 thru 2015, there is negligible difference between the larger cap stocks when it comes to short interest sentiment change. We could infer from this that short sellers are indeed the "smart money", particularly with the smaller cap stocks.
We've looked at the rationale and performance of the 3 main components of the A_Score, namely:
- A1_Lottery Effect
- A2_Beating Expectations
- A3_Short Interest Sentiment
Let's now combine the factors and test performance on total score of the A_Score.
As of today, the distribution of stocks in our high P/B universe with the scores from 0 to 3 is as follows: Source: Portfolio123 data, Author Table
This data shows us that the highest score of 3 is a high bar, with only 12% of our universe passing. This is typical across different periods as well. On the opposite end of the spectrum, there is a proportion of stocks that do not register on the score (NA); these would be stocks with no analyst coverage (so the A2 factor would not apply), or perhaps ADR stocks where short interest (A3 factor) is not available.
The table below summarizes performance with various A_Scores over our periods. Note that the A2 factor data is not available until 2003, so I have adjusted performance dates for the overall A_Score to suit: Source: Portfolio123 data, Author Table
Generally, the higher the A_Score, the higher the performance. Note the consistent and widespread between the high and low-scoring stocks over all periods. Note the outperformance of the A_Score of 2 or greater compared to 3 in the last 4 years.
Over the full period, our highest scoring stocks outperformed all other universes and benchmarks R3000 (IWV) and the S&P500 (SPY). The outperformance is particularly notable compared to the parent P/B>80 universe, which is what we would like to see, as we are trying to find those truly winning stocks in the high valuation universe.
When breaking down into smaller periods, high scoring A_Score stocks have outperformed the best in the last 10 years, but with a narrower spread, or slight underperformance, from 2003 thru 2010. Still, this is quite promising.
The A_Score vs the G_Score
As we covered in Part 1, the G_Score is built on fundamental factors from financial statements. The A_Score, in contrast, is built on market factors, or "anomalies". The table below compares performance between the two strategies, from 2003 thru 2020.
Source: Portfolio123, Author Table
Performance between the two strategies is similar, with the A_Score having a slight edge overall, and in the earlier periods in our study. While the G_Score relies on fundamentals and the A_Score market anomalies, there is another key difference: the number of factors. Recall that the G_Score looked at 8 different factors, and our A_Score uses only 3. It is reasonable to assume that there is some overlap, or correlation, in these strategies. For our A3 factor, for example, short-sellers will rely on some fundamentals, or at least interpret them differently than the broader market. Similarly, it is reasonable to assume that those stocks beating expectations also have some strong fundamentals (Factor A2).
To illustrate, take stocks today. Of the 83 stocks with a G_Score of 6 or greater, the distribution of A_Scores is as follows:
- A_Score of 0 = 5 stocks (6%)
- A_Score of 1 = 25 stocks (30%)
- A_Score of 2 = 33 stocks (40%)
- A_Score of 3 = 20 stocks (24%)
While there is a low proportion of A_Scores of 0, the remaining scores are fairly well distributed. This would suggest that the two scoring systems are more "apples and oranges" than "apples and apples", so there may even be some opportunity in combining them. More on this in the next installment…
Putting it All Together
In this series, we've been attempting to find stocks deserving of their high valuations, on a price-to-book basis. In Part 1, we relied on a stock's fundamental factors (or accounting metrics available on financial statements as per the G_Score) that were stronger than their industry's median. We found this method to be effective. In this recent installment, I shared a new method that I've developed that I refer to as the A_Score, based on market factors that exploit market "anomalies". We've seen that the A_Score is also effective at finding winning stocks within the high valuation universe.
In the next installment, we will refine this methodology and look at a strategy that combines both fundamental and market factors to find truly winning stocks.
Until then, happy investing!
But before we go…
As of time of writing, the largest stocks with A_Scores of 2 or 3 are shown below:
On the flip side, the largest stocks with an A_Score of 1 or less are shown below:
This article was written by
Analyst’s Disclosure: I am/we are long CRWD, ZM, SNOW. 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.
This article is for informational purposes only. I am an individual investor and writer, not an investment advisor. Readers should always engage in his or her own research and consider (as appropriate) consulting a fee-only certified financial planner, licensed discount broker/dealer, flat fee registered investment adviser, certified public accountant, or specialized attorney before making any investment, income tax, or estate planning decisions.
Seeking Alpha's Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.