AgPa #13: Concentrated Stock Markets (4/7)

Extreme Stock Market Performers, Part III: What are their Observable Characteristics? (2020)
Hendrik Bessembinder
SSRN Working Paper, URL

This is the fourth of seven AGNOSTIC Papers about the extreme concentration in stock markets. Last week, we have seen that industries are not really helpful to narrow the search for big winners. Therefore, we now go one step further and examine their observable fundamental characteristics. The main insight is promising and quite intuitive: outstanding stock performance usually comes with outstanding fundamental performance of the underlying company.

As always, I recommend to read the posts of this series chronologically. But feel free to use the following list to navigate for yourself if this better suits you.

  • Week 1: Concentration in the US Stock Market between 1926 and 2019
  • Week 2: Concentration in Global Stock Markets between 1990 and 2019
  • Week 3: Dominance of the Tech-Industry?
  • Week 4: Characteristics of Big Winners?
  • Week 5: Identifying Big Winners Upfront?
  • Week 6: Even Big Winners had Bad Drawdowns
  • Week 7: The Same Pattern for US Mutual Funds

Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.

Data and Methodology

Most of the data and methodology is identical to last week. The author uses stock market data from CRSP to construct a sample that includes 26,285 US companies between 1950 and 2019. So it is very similar to the US sample of the original paper. For fundamental data, he uses Compustat. Both services are state of the art for finance-research, so data quality shouldn’t be an issue. The two main concepts remain total-returns in excess of cash1Total return means that all dividends are reinvested. For brevity, I will just refer to “returns” or “decade-returns” in the following. and Shareholder Wealth Creation (SWC).2SWC measures the total wealth that a company creates over its lifetime in excess of risk-free treasuries. Most importantly, it accounts for the fact that dividends cannot be reinvested in aggregate. Therefore, it takes the perspective of a hypothetical investor who owns the entire company.

The author again calculates returns and SWC for the seven calendar-decades between 1950 and 2019. For each of those decades, he identifies the best and worst 200 companies in terms of total returns and SWC (Top 200 and Bottom 200, respectively). Each top- and bottom-list therefore consists of 1,400 companies.3200 firms for seven decades: 200 x 7 = 1,400 The average company-characteristics of those winners and losers are the heart of the following analysis. Those of the more ordinary companies in between (Non-200) serve as benchmark.

With respect to observable characteristics, the author uses various fundamental- and market data. The following table summarizes the variables. Growth measures refer to the annualized growth rate over the respective decade in percent. Means are decade-averages of the respective annual ratios.

Accounting-Based GrowthAccounting-Based OutcomeReturn-Based
Total Asset GrowthMean Dividend-to-Asset RatioMaximum Drawdown
Asset Growth excl. Net Equity IssuanceMean Net Income-to-Assets RatioStandard Deviation of Monthly Returns within Decade
Asset Growth excl. AcquisitionsMean New Equity Issuance-to-Assets RatioSkewness of Monthly Returns within Decade
Cash GrowthMean R&D-to-Assets Ratio
Current Asset GrowthMean Sales-to-Asset Ratio
Fixed Asset GrowthTotal Debt-to-Assets Ratio
Fixed Asset GrowthMarket-to-Book Ratio
Other Asset Growth
Sales Growth
Sales-to-Asset Ratio Growth
Dividend-to-Asset Ratio Growth
Income-to-Asset Ratio Growth
Own illustration of Section 1 of Bessembinder (2020).

The author doesn‘t explicitly comment on his motivation for those particular fundamentals. However, except for a missing cash-flow ratio, I think the variables cover many aspects of the underlying business.4To the defense of the paper, cash-flows are implicitly embedded in other variables like Cash Growth. In fact, it is actually not that easy to come up with a robust set of fundamentals without falling prey to data-mining. Therefore, I think the 22 variables are quite a good starting point.

Important Results and Takeaways

Big winners grow faster, are more profitable, and have smaller drawdowns

The following table summarizes the characteristics of the Top 200, Bottom 200, and Non-200 companies by decade-returns. Unsurprisingly, annualized returns and SWC are insanely high for the Top 200. This is of course by construction of the ranking, but the magnitude is truly remarkable. The average Top 200 company generated annualized returns of almost 368%. That turns $1 into more than $5,000,000 in just one decade.

Own illustration of Table 1 of Bessembinder (2020).

I will not comment on every single number in the table, but I think the best way to summarize the results is the following. Top 200 firms grew insanely fast by virtually any measure. Moreover, they are very profitable, accumulate a lot of cash, and exhibit smaller drawdowns than other companies. All of this seems intuitive, but the magnitude is again striking. Depending on the variable, the Top 200 grew up to 9 times faster than the Non-200. The gap to the Bottom 200 is of course even larger.

Overall, the results suggest that outstanding stock performance is associated with outstanding fundamental performance. In addition to that, the Top 200 grew mostly organically. Total Asset Growth for the Top 200 companies is almost identical to Asset Growth excl. Acquisitions (95.3% and 89.8%, respectively). A surprising observation, however, is the negative value for the Top 200’s Sales-to-Asset Ratio Growth. The decline of this ratio contradicts with the fact that Sales Growth was slightly higher than Total Asset Growth (101.6% vs. 95.3%, respectively). But the table just denotes averages, so this might be driven by outliers.

The other end of the spectrum looks very different. Most growth rates for Bottom 200 companies are dismal. Adjusted for net equity issuance, the average Bottom 200 company lost almost 10% of its assets per year. Over one decade, this accumulates to a decline of about 65%. The Bottom 200 firms also had more volatile returns and much larger drawdowns.

The results are less extreme but generally similar when we rank companies by decade-SWC instead of total returns. The next table summarizes the results. Notably, the author uses the same numbers for the Non-200 companies despite the different ranking. I don‘t know if this is on purpose or an error, but it shouldn’t affect the overall interpretation too much.

Own illustration of Table 2 of Bessembinder (2020).

The results once again highlight the difference between top performers and top wealth-creators. The average Top 200 wealth-creator generated annualized returns of 151%. This is much smaller than the >300% before, but still turned $1 into more than $9,500 over one decade. The differences are mostly driven by company size. To be top wealth-creators, companies must make a lot of money in absolute terms. So they need to earn high returns on a lot of capital. To be top-performers, high returns are sufficient.

Despite some differences, the overall pattern of company characteristics is very similar. The Top 200 wealth-creators grew much faster by virtually every fundamental standard. They are also more profitable and were less volatile with smaller drawdowns. There is also the same inconclusive decline of the Top 200’s Sales-to-Asset Ratio. Similar to before, I suspect that is driven by outliers because sales grew actually slightly faster than assets (44.5% vs. 42.6%, respectively).

Observable fundamentals still explain relatively little

In the second part of the paper, the author regresses SWC on the fundamental characteristics to control for correlations among them. For total-returns, he uses an indicator5Such an indicator- or dummy-variable is 1 if the company is among the Top 200 and 0 else. for Top 200 companies as dependent variable. Because of their relevance, he also adds market capitalization and company-age to the analyses.

The idea is to use the regression coefficients and their statistical significance to identify which fundamental characteristics are most important for the big winners. While I really admire the author for his papers, I think this particular methodology is flawed for two reasons.

First, using a standard linear regression with a dependent indicator-variable comes with several problems and there are better models for this.6For those who are as nerdy as I am (or want to get there): linear regressions with binary dependent variables inevitably suffer from heteroskedasticity. The details are terribly technical but the important message is simple. In such a case, you cannot trust the t-statistics and statistical significance. For binary dependent variables, Probit- or Logit-regressions are therefore much better. Second, the whole point of all these analyses and papers is the presence of extreme outliers in the distribution of companies and stocks. It is well-known that standard linear regression doesn’t work particularly well in such settings as it is very sensitive to outliers.7To his defense, the author winsorizes all accounting-based measures at the 1%- and 99%-quantile to mitigate the effect of outliers. So it is a bit like trying to fly with a ship – it’s just the wrong tool.

As a consequence, I don’t believe the detailed regression results are very useful and I think it‘s very hard to identify the individual effect of one particular fundamental characteristic. Therefore, I haven’t included them here and refer those who are interested to Tables 3 and 4 of the paper.

However, the regressions are of course not entirely useless. We can use the R2 of the models to get some idea about the relevance of fundamentals. As I explained in my post on intangible assets, the R2 tells us how much variation of the dependent variable the model explains.8Specifically, the R2 ranges from 0 to 1. For example, a value of 0.2 means that fundamental characteristics explain 20% of the variation in SWC. In the paper, all R2 are in the range between 0.037 and 0.122. Fundamental characteristics therefore explain only 3.7%-12.2% of the variation in total returns and SWC. This is better than nothing but only a small piece of the puzzle. There are many other relevant factors remaining and looking at fundamentals is (unfortunately) not sufficient to identify the few big winners.

Conclusions and Further Ideas

In the opener to AGNOSTIC Papers, I have written that the selected papers “[…] must meet certain standards of quality.”. Let’s face it, that is not fulfilled for this one. The paper is just a follow-up note that aims to provide an idea about the characteristics of big winners. But it is not a peer-reviewed scientific article like the original studies. So don’t be too hard with every single number and let’s rather focus on the overall patterns.

First, the few big winners are also successful with respect to fundamentals (at least on average). Although this seems intuitive, I think it is still nice to see that the stock market rewards great companies over the long-term. Second, no matter how you measure it, the big winners grew much faster and more sustainable (without many acquisitions) than all other companies. And finally, they are more profitable, accumulate a lot of cash, and had less volatile returns with smaller drawdowns.

The obvious catch of those results is that they are contemporaneous. This is the scientific way of saying that we analyze the big winners after we know that they are the big winners. Don’t get this too negative. Studying the characteristics of big winners in hindsight is absolutely useful and certainly helps to spot them in the future. But the more important (and profitable) part is to find the next big winners today. This is unfortunately also the more difficult part and requires predictive analyses. This will be the topic for next week.



This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). All content that is not my intellectual property is marked as such. If you own the intellectual property displayed on this website and do not agree with my use of it, please send me an e-mail and I will remedy the situation immediately. Please also read the Disclaimer.

Endnotes

Endnotes
1 Total return means that all dividends are reinvested. For brevity, I will just refer to “returns” or “decade-returns” in the following.
2 SWC measures the total wealth that a company creates over its lifetime in excess of risk-free treasuries. Most importantly, it accounts for the fact that dividends cannot be reinvested in aggregate. Therefore, it takes the perspective of a hypothetical investor who owns the entire company.
3 200 firms for seven decades: 200 x 7 = 1,400
4 To the defense of the paper, cash-flows are implicitly embedded in other variables like Cash Growth.
5 Such an indicator- or dummy-variable is 1 if the company is among the Top 200 and 0 else.
6 For those who are as nerdy as I am (or want to get there): linear regressions with binary dependent variables inevitably suffer from heteroskedasticity. The details are terribly technical but the important message is simple. In such a case, you cannot trust the t-statistics and statistical significance. For binary dependent variables, Probit- or Logit-regressions are therefore much better.
7 To his defense, the author winsorizes all accounting-based measures at the 1%- and 99%-quantile to mitigate the effect of outliers.
8 Specifically, the R2 ranges from 0 to 1. For example, a value of 0.2 means that fundamental characteristics explain 20% of the variation in SWC.