Do stocks outperform Treasury bills? (2018)
Journal of Financial Economics 129(3), 440-457, URL
Wealth Creation in the U.S. Public Stock Markets 1926 to 2019 (2021)
The Journal of Investing 30(3), 47-61, URL
I try to be careful with superlatives, but I truly think that this week’s AGNOSTIC Paper(s) are a must-read for everyone seriously interested in stock markets. At the risk of raising expectations further, I even believe that the two papers should be part of any investment-education. So let’s dive into it.
Because of its wide-ranging implications for investors, the original 2018-paper was a big success. Four years later, there are now several follow-ups of similar quality. Full disclosure, I am a fan of all of them. Therefore, I will devote the next seven weeks of AGNOSTIC Papers to this research. I recommend to read this one first, but you can also use the following list to navigate for yourself.
- 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.
Setup and Idea
Aggregate stock markets, measured as passive market-cap-weighted portfolios, produced very decent returns over the last decades and centuries. This is the well-known equity risk premium. Given that the net performance of the aggregate market was positive, one could conclude that the typical stock was quite a good investment. As we will see, this is a bad idea and heavily misleading.
The author of this week’s paper, Hendrik Bessembinder, shows that the majority of US stocks (58%) actually underperformed risk-free treasury bills over their lifetime. In fact, the lifetime return of many companies is negative and for some even close to -100%. So the majority of US stocks were losers. How does this fit with the positive equity risk premium? Well, if you have a lot of losers you need a few really big winners to make the whole distribution net-positive. This is exactly what is happening in stock markets.This pattern is probably very similar for entrepreneurship in general. One Amazon compensates for hundreds of failed internet companies.
Statisticians call this positive skewnessFor the geeks among us: positively skewed distributions are characterized by large positive outliers. The mean is therefore much larger than the median. and Bessembinder shows that this is a pervasive feature of the US stock market. Especially over long time horizons. Just one empirical fact to illustrate this: only 1,172 out of 26,168 US companies created the entire net wealth in the stock market between 1926 and 2019. That’s about 4% of companies creating 100% of the wealth. This is extreme concentration by any standard.
Data and Methodology
The key concept of the paper is Shareholder Wealth Creation (SWC). SWC captures the aggregate dollar-wealth created by companies over their lifetime in excess of risk-free treasury bills. So it’s something like a realized equity risk premium but measured in absolute dollars instead of percentages. Why can’t we just use the total return (incl. reinvested dividends) to measure wealth creation? Two reasons. First, all shareholders together cannot reinvest. If one investor reinvests her dividend, another one must sell shares to her. This is an exchange of shares and cash among existing investors, but not wealth creation. Second, companies not only return capital to shareholders, they sometimes also ask for it via equity offerings. SWC properly accounts for both of these issues and takes the perspective of owners of the entire company.
However, it’s neither this innovative methodology nor sophisticated data what makes the papers special. It’s really just the idea of looking at the distribution of stock returns from another angle. In fact, you don’t need any advanced statistical models to derive the key results. There is not even a regression in the paper.An empirical finance-paper without regression is almost an oxymoron. Simplicity without sacrificing quality, I really like that.
This doesn’t mean the analysis is easy to replicate. Studying more than 90 years of stock market data involves a lot of nasty data work and there are a lot of challenges on the way. The author sources his data from CRSP, a state of the art service for research on US stock returns.CRSP stands for Center for Research in Security Prices and is an affiliate of the University of Chicago. More information here. The original 2018-paper is also published in the Journal of Financial Economics, one of the top finance journals in the world.The follow-ups are not (yet) published in top journals. However, I am quite confident that this is not because of poor quality. An update of an existing paper with three years more data is often a too minor contribution for a top journal. Therefore, I am quite confident that the results are based on solid methodology and clean data.
Important Results and Takeaways
Longer investment-horizons lead to extremer return distributions
Bessembinder starts the original 2018-paper with simple frequency distributions of compounded buy-and-hold returns over different time frames. The sample ranges from 1926 to 2016 and includes all US stocks that existed during this period.Unfortunately, he doesn’t include this statistics in the updated version with data until 2019. The results are very interesting.
Panel A of the right chart starts with annual returns. The distribution is surprisingly symmetric between -100% and 100% and has a pronounced peak around 0%. In addition to that, there are a few but substantial positive outliers (“a fat right tail”). The distribution as a whole is therefore heavily asymmetric and, because of the outliers, positively skewed. This asymmetry is a direct result of limited liability. The downside of stock investments is capped at -100% while the upside is potentially unlimited. So this is not surprising.
Panel B shows the same chart for decade returns. For this longer period, the distribution of returns is now entirely different. The most frequent return is no longer 0% but -100%. So a lot of companies haven’t even survived one decade and investors lost all their money with them. This is a strong reminder of competition. There are again substantial positive outliers and their impact is even stronger. This is largely a result of compounding over longer time horizons.
Finally, the distribution of lifetime returns is even more extreme (Panel C). The most frequent observation remains -100%, indicating that a lot of companies destroyed value over their lifetime and disappear at some point. Most lifetime returns fall within the range of -100% to 200%. However, the range of observed lifetime returns is much broader than for decade returns (-100% to 1,000%). The impact of positive outliers (“big winners”) is therefore once again stronger.
Why are the distributions so different? There are two reasons. First, returns on individual stocks are very volatile. Second, compounding of such volatile returns has a substantial impact over time. A simple example to illustrate this. Suppose a stock has an equal chance to increase by 20% or to decrease by 20% in each year. For one year, the distribution is of course symmetric. But even for just two years, the effect of compounding kicks in. In year two, you have a 25% chance for a return of 44%, a 50% chance for a return of -4%, and again a 25% chance for a return of -36%. The most frequent outcome is -4% but the expected value is 0%. So we just created positive skewness.
Just 4% of all companies created the entire net wealth in the US stock market
The following chart summarizes the most important result of the two papers. It shows the concentration curve of Shareholder Wealth Creation (SWC) for all publicly traded US companies between 1926 and 2019. As mentioned before, the distribution is extremely concentrated and only 4% of all companies created the entire net wealth. However, even more surprising to me was that the majority of companies (58%) actually destroyed wealth over their lifetime (relative to risk-free treasuries). That’s a quite brutal fact.
Of course, we all want to know who those top 4% are. The following tables therefore shows SWC of the top 50 companies. Unsurprisingly, we find a lot of the largest and well-known US corporations. The table also illustrates how SWC is different from annualized total returns. For example, General Motors defaulted in June 2009 but is still one of the top wealth creators. How can this be? Well, it paid a lot of dividends in the 80 years before and SWC properly reflects this.
Given their giant size and tremendous success over the last decades, there are several tech-companies among the top 50 wealth creators. I don’t think this surprises anybody. To further analyze this pattern, Bessembinder (2021) also provides the following statistics on SWC across industries.
Technology firms were indeed the largest wealth creators, but this is not the full story. First, these statistics are again driven by the few big winners. Apple and Microsoft alone account for about 1/3 of the tech-industry’s SWC. Second, there are simply a lot of tech-stock-observations in the sample. To adjust for this, Bessembinder relates the fraction of total SWC to the fraction of data points in the sample (last column). Based on this ratio, the Technology industry is still among the top performers but not as far ahead as in absolute terms.
Just 5 (tech) companies created 22% of total shareholder wealth between 2016 and 2019
In the updated 2021-paper, Bessembinder also analyzes how the concentration of SWC changes over time. For this purpose, he looks at consecutive 3-year periods. There is a clear pattern of increased concentration over time and in particular since 1995. Some argue that this development is consistent with “winner takes it all” competition in the digital economy.
But especially the most recent three years from 2016 to 2019 were particularly striking. During this period, only five companies (who would have thought it: Apple, Microsoft, Amazon, Alphabet, Facebook) created about 22% of the entire net shareholder wealth. There were 4,896 publicly traded companies in the US between 2016 and 2019. So 0.1% of all companies created 22% of the entire wealth. Once again, extreme concentration by any standard.
The biggest wealth creators were not necessarily the best investments
For the last takeaway, I want to again highlight the difference between SWC and total returns. SWC measures the wealth that a company creates for all its owners. In contrast, annualized total returns denote the investment performance of a long-term shareholder who decides to reinvest all dividends into the company. To be great wealth creators, companies must make a lot of money in absolute terms. So they need to earn good returns on a lot of capital. To be great investments, good returns are sufficient.A company with a market cap of $100M that compounds at 30% per year is an outstanding investment. But because of its size, it will probably not become a large wealth creator. The following table gives an impression for this difference among the top-50 wealth creators. Statistics on lifetime returns are unfortunately only available in the original 2018-paper. So the period from 2016 to 2019 is missing in this table and the numbers are therefore slightly different.
Admittedly, almost all of the top wealth creators also generated very good long-term returns. But there are a few interesting cases. For example, Visa generated a better annualized return than Apple (21.1% vs. 16.3%) but is a much smaller wealth creator. There are two reasons for this. First, as mentioned above, Visa is just smaller than Apple. Second, Apple exists longer than Visa (1981 vs. 2008) and simply had more time to create wealth.
Another interesting result in this table is the cumulative lifetime return of the Altria Group a.k.a. the tobacco company Philip Morris. Although the company “just” generated an annualized total return of 17.7%, it did so over a period of almost 91 years. As a consequence, $1 invested in 1926 grew to unbelievable $2,446,638 by 2016. According to Bessembinder, this is by far the highest cumulative return in the sample.
Conclusions and Further Ideas
I always felt that a few very successful firms are driving the stock market. I mean, just look at the concentrated composition of indices like the S&P 500. But really seeing it in the data changed my view on stock markets profoundly. On a higher level, the results are a strong reminder how competitive and dynamic markets actually are. Most companies disappear after a few years and destroy the wealth of their shareholders. At least historically, the most likely lifetime return for a randomly selected stock in the US was -100%. We should definitely keep this in mind when selecting individual stocks.
The extreme concentration within stock markets has two strong implications for investors. Active investors use the results to argue how much value they can create by selecting the few big winners and/or avoiding the underperforming majority. Passive investors use the same results to argue in favor of broad diversification and market-cap weighting to ensure you don’t miss the big winners. The same fact but two very different conclusions. That’s amazing and frustrating at the same time and everybody must decide for himself what he/she believes in.
Properly executed, both strategies are “right” and this brings me to another important implication. You often read about the “concentration risk” of indices like the S&P 500 because the top constituents make up a large fraction of the portfolio.Asset managers frequently use this empirical fact to sell a strategy that attempts to perform better by avoiding this risk. For example, here. But given the extreme concentration in stock markets, you should actually be happy about this. The few big winners outperform and grow to a larger part of the portfolio. As we have seen, those companies drive the entire equity risk premium. So there is nothing wrong with that.
Finally, I want to highlight one important limitation. The whole analysis is built on lifetime returns. Depending on the company, those are very long periods of time and most likely not the horizon of the typical investor. So although most companies destroyed value over their lifetime, you can still make a lot of money by trading them over shorter time intervals. Even the crappiest company is a good investment if you buy it low and sell it high.
Next week, I will continue this series with the same analysis for international stock markets. For the moment, let’s conclude this post with Warren Buffett who – as so often – discovered this pattern much earlier than many others. But knowing that a few big winners are sufficient is of course the easy part. Actually finding them remains very difficult. At least if you are not Warren Buffett.
You don’t need 20 right decisions to get very rich. Four or five will probably do it over time.Warren Buffett – 2013 at Georgetown University
- AgPa #56: The Equity Risk Premium of Small Businesses
- AgPa #55: Backtests in the Age of Machine Learning
- AgPa #54: Transitory Inflation
- AgPa #53: Investing in Interesting Times
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 (*). Please also read the Disclaimer.
|1||This pattern is probably very similar for entrepreneurship in general.|
|2||For the geeks among us: positively skewed distributions are characterized by large positive outliers. The mean is therefore much larger than the median.|
|3||An empirical finance-paper without regression is almost an oxymoron.|
|4||CRSP stands for Center for Research in Security Prices and is an affiliate of the University of Chicago. More information here.|
|5||The follow-ups are not (yet) published in top journals. However, I am quite confident that this is not because of poor quality. An update of an existing paper with three years more data is often a too minor contribution for a top journal.|
|6||Unfortunately, he doesn’t include this statistics in the updated version with data until 2019.|
|7||A company with a market cap of $100M that compounds at 30% per year is an outstanding investment. But because of its size, it will probably not become a large wealth creator.|
|8||Asset managers frequently use this empirical fact to sell a strategy that attempts to perform better by avoiding this risk. For example, here.|