Long-Term Shareholder Returns: Evidence from 64,000 Global Stocks

Abstract

We study long-run shareholder outcomes for more than 64,000 global common stocks during the January 1990 to December 2020 period. The majority, 55.2% of U.S. stocks and 57.4% of non-U.S. stocks, underperform one-month U.S. Treasury bills in terms of compound returns over the full sample. Focusing on aggregate shareholder outcomes, we find that the top-performing 2.4% of firms account for all of the $US 75.7 trillion in net global stock market wealth creation from 1990 to December 2020. Outside the United States, 1.41% of firms account for the $US 30.7 trillion in net wealth creation.

Keywords:

• compound returns
• long-term returns
• return skewness
• stock investing and shareholder wealth

PL Credits: 2.0:

Acknowledgment

The authors thank William Goetzmann (executive editor), Daniel Giamouridis (co-editor), and two anonymous referees for constructive and insightful comments and suggestions that have significantly improved the article. Bessembinder acknowledges financial support from Baillie Gifford & Company. Wei acknowledges financial support from the Research Grants Council of the Hong Kong SAR, China (GRF15505518).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Many studies report unconditional arithmetic mean returns to characteristic-sorted portfolios, while many others estimate conditional arithmetic mean returns by implementing regression analyses with short-horizon returns as the dependent variables.

2 We focus on returns and wealth measured in U.S. dollars to provide a common yardstick that can be compared across stocks traded in currencies with differing inflation rates. The comparison to the short-term U.S. Treasury rate reflects that this rate is often viewed as the best available proxy for the “risk-free” interest rate envisioned by theory.

3 The positive skewness arises in part because the distribution of monthly individual stock returns is positively skewed, but mainly due to the effects of compounding. See, for example, Simkowitz and Beedles (1978), Albuquerque (2012), Heaton, Poulson, and Witte (2017), Bessembinder (2018), Fama and French (2018), and Farago and Hjalmarsson (2023), all of which focus on the U.S. markets.

4 Fang et al. (2021) document that the majority of monthly local-currency stock returns in a global sample are less than local-currency short-term interest rates in the same months. However, they do not study compound returns, nor do they study wealth creation outcomes. Further, their “Treasury bill” proxies are local currency interest rates as diverse as the Luxembourg 10-year Government Bond Yield, the Peru Time Deposit Rate, and the Zimbabwe 3-month Time Deposit Rate. We convert all returns to U.S. dollars to allow comparisons of returns across stocks from different markets and to the common U.S. Treasury bill interest rate, which arguably comprises the best global proxy for the risk-free interest rate envisioned in theory.

5 Indeed, Samuelson (1969) acknowledges (his footnote #1) that his results hold only under the assumption of power utility. For other utility functions outcomes will depend on the preference for skewness relative to other moments of the return distribution (e.g., kurtosis), which also depend on horizon, as Farago and Hjalmarsson (2023) show.

6 In addition, while the issue is not skewness per se, Bessembinder, Cooper, and Zhang (2022) show that alpha and beta parameters (and estimates thereof) differ when returns are measured over long vs. short horizons.

7 The Compustat data upon which we rely for non-U.S. stock returns does not include information regarding post-delisting share values or post-delisting payments to shareholders. Following Shumway (1997), we set the final return on non-U.S. stocks with an incomplete return series, as well as stocks indicated to be delisted for reasons of bankruptcy or liquidation, to −30%. For U.S. stocks, we incorporate CRSP delisting returns where available, while setting the final return to −30% in the few cases where the delisting return is missing and the CRSP delisting code is 500, 520, 551–573, 580, 574, or 584.

8 Examples of “homeless ADRs” include Baidu, Inc. and BioNTech, SE. Firms that were formerly listed only as ADRs but also listed on a local market before the end of the sample (e.g., Alibaba Group) are included with the relevant local market.

9 See, for example, Chui, Titman, and Wei (2010), Hou, Karolyi, and Kho (2011), and Fama and French (2017).

10 Hong Kong SAR and Singapore are exceptional in terms of market capitalization relative to GDP, with many large firms listed on their exchanges. A number of large Chinese firms in particular are listed in Hong Kong SAR, and five members of the Jardine Group, which is headquartered in Hong Kong SAR, shifted from Hong Kong SAR to Singapore in 1994 (Chan, Hameed, and Lau, 2003). Prior to the change in listing, Jardine composed about 10% of the total market capitalization in Hong Kong SAR.

11 Percentages can sum to less than 100% because minor exchanges are excluded from the study.

12 As three examples among many, Fama and French (2017), Jacobs and Muller (2020), and Bartram and Grinblatt (2021) study arithmetic mean portfolio returns and estimates regressions with returns as dependent variables in their international stock market studies.

13 A notable feature of the distribution of monthly returns to U.S. stocks is the peak at zero, which is presumably attributable to non-trading and price rounding. For non-U.S. stocks the peak at zero is less notable, which reflects that a zero return in local currency may not equate, even with rounding, to a zero return in U.S. dollars.

14 We define decades as January 1990 to December 1999, January 2000 to December 2009, and January 2010 to December 2020.

15 The data on Figures 3 and 4 indicate that returns very close to −100% are more frequently observed for U.S. as compared to non-U.S. stocks. However, this observation is likely an artifact of the fact that CRSP provides actual delisting returns for the United States, while in the absence of accurate delisting returns we follow the prior literature in imputing a −30% return when non-U.S. firms exit the database.

16 The match of simulated and actual monthly log returns is almost perfect in terms of these parameters. In particular, the monthly mean log return across all stocks is −0.3% and the standard deviation of monthly log returns is 16.2%, in both the simulated and the actual data. However, skewness is not as well matched. The average skewness in simulated monthly log returns is zero by construction, while the average skewness of monthly log returns in the sample is −0.78. This discrepancy reflects that the sample monthly returns do not conform to the log-normal assumption.

17 The dollar-weighted return corresponds to the calculated wealth creation figure more cleanly than the buy-and-hold return, as it also allows for net equity issuances and the fact that dividends are not, in aggregate, reinvested in stock. See Dichev (2007) and Dichev and Zheng (2020) for discussion the computation of dollar-weighted returns.

18 The Japanese stock market performed very well in the years preceding 1990 (the Nikkei Index reached its all-time high on December 29, 1989), so the result that the worst-performing firms were predominantly Japanese would differ over a longer sample period.

19 The issue, described for example by Ellenberg (2014), is that a few observations can explain far more than 100% of a figure that is obtained by summing across positive and negative observational outcomes, particularly when the sum is itself modest in magnitude. Ellenberg goes so far as to suggest that one should not report percentages when studying the sum of positive and negative outcomes. It is, however, unclear how far this reasoning should be pushed when studying stock market outcomes, where the natural object of interest is the gain to the investor as defined by the net of many individual up and down price movements. For example, focusing on accumulated outcomes from only those days with positive price changes would be of little or no practical interest.

20 Note that the double counting issue arises only due to the ownership of equity in sample stocks by other companies also included in the sample. The fact that a given non-corporate shareholder may hold positions in multiple companies does not lead to double counting in our setting.

21 More specifically, we focus on the holdings of firms with owner type codes equal to Bank and Trust (101), Finance company (103), Investment advisor (107), Insurance company (108), Brokerage firms (200), Research firm (201), Independent research firm (202), Corporation (301), and Holding company (302).

22 The Refinitiv ownership data begins in the first quarter of 1997. We backfill the initial data to earlier quarters.

23 Perhaps the most striking observation in the cross-holding data concerns Naspers’ holdings of Tencent, which averaged over 30% of outstanding shares. Tencent’s full sample wealth creation for non-sample shareholders was $463 billion, as compared to $692 billion for all shareholders (including Naspers).

24 Despite that Berkshire Hathaway obtained a substantive position in Apple just before the end of the sample, outcomes for Apple are also little affected. Apple’s wealth creation outcome was reduced to $US 2.57 trillion (from $US 2.67 trillion) and its share of global net wealth creation was reduced to 3.40% (from 3.53%).

25 We note that the individualism variable is marginally significant when explaining the concentration of wealth creation (t statistic = −2.04) and that the dummy variable indicating the individualism variable to be missing is marginally significant when explaining the percentage of stocks that outperform Treasury bills, but it is unlikely that either of these results would survive an adjustment for multiple testing of the set of Hofstede variables.

Additional information

Notes on contributors

Hendrik Bessembinder

Hendrik Bessembinder is at the W. P. Carey School of Business, Arizona State University, Tempe, Arizona.

Te-Feng Chen

Te-Feng Chen is at the School of Accounting and Finance, Hong Kong Polytechnic University, Hong Kong Special Administrative Region (SAR), China.

Goeun Choi

Goeun Choi is at the A. B. Freeman School of Business, Tulane University, New Orleans, Louisiana.

K. C. John Wei

K. C. John Wei is a distinguished research professor in the School of Accounting and Finance at the Hong Kong Polytechnic University, Hong Kong Special Administrative Region (SAR), China.