Employee Satisfaction and Long-Run Stock Returns, 1984–2020

Abstract

Economic theory predicts that (in the absence of mispricing) the excess return to socially responsible businesses is negative in equilibrium. In contrast, using the state-of-art empirical models and a sample spanning four decades (1984–2020), an equal-weighted portfolio of companies that treat their employees the best earns an excess return of 2% to 2.7% per year. The estimated alphas are positive in most periods within the sample (with no upward or downward trend) and are particularly large during crisis periods. Overall, the results suggest that the stock market (still) undervalues employee satisfaction.

Keywords:

• business social responsibility
• corporate culture
• employee satisfaction
• socially responsible investing
• underreaction

PL Credits: 2.0:

Disclosure statement

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

Notes

1 See http://www.gsi-alliance.org/wp-content/uploads/2021/08/GSIR-20201.pdf.

2 The 1-year, 3-year, 5-year, 10-year, and 15-year returns are 60.1%, 21.2%, 19.5%, 19%, and 14.6%, respectively. The fund has an annual return of 14% since inception and the highest possible rating for all periods from Morningstar. Data retrieved from Morningstar as of August 31, 2021.

3 See https://www.fastcompany.com/3006150/proof-profits-americas-happiest-companies-also-fare-best-financially.

4 See https://fortune.com/franchise-list-page/best-companies-2020-methodology/.

5 We note that some of these characteristics vary significantly across years. For example, comparing the first available BC list to the one in 2020, we observe that in 1984, operating margin was 1.08% (compared to 20.46% in 2020) and the proportion of intangible assets in the balance sheet was 0.85 (compared to 22.29% in 2020).

6 We downloaded factor vintages from the wayBackmachine functionality in www.archive.org. A more complete replication including the industry-adjusted returns and value-weighted BC portfolio is provided in the Internet Appendix Table 1.

7 The fact that the results of industry-adjusted returns (reported in Internet Appendix) are also very close to the original study is consistent with this idea.

8 While we do have factors such as size, value, and momentum included as covariates in the factor regressions, these characteristics could still have explanatory power, as argued in Daniel and Titman (1997).

9 It is important to note that time-series factor models and cross-sectional FM. regressions are methodologically different and could yield different results. Fama and French (2020) provide detailed comparison of the two methods.

10 In the Internet Appendix Table 2, we reproduce Table 6 without any of our sampling filters. The results are robust and all the excess return estimates in the table are statistically significant. For example, the corresponding BC coefficient in Column 1 is 0.39 (similar as the original study) with the t-statistics of 3.71 and 0.37 in Column 4 with the t-statistics of 4.25.

11 In the Internet Appendix Table 5, we provide more granular estimates of excess returns for each boom and crisis period. Those results show that the average monthly excess return in the dotcom boom was in the range of 20–30 bps whereas in the boom preceding the financial crisis was around 0. None of the estimated boom excess returns are statistically significant. On the contrary, the alphas for both crisis periods are sizable, statistically significant, and robust across different factor models.

12 One notable difference between FM. and factor model results is that while still positive, the FM. alpha in the dotcom crash is much smaller than the one estimated from factor regressions.

13 For the sake of brevity, we only report alphas here. We provide detailed estimates in the Internet Appendix Tables 7 and 8. We also provide similar tables but using characteristics-adjusted excess return as the dependent variable for both the straight and capped value-weighted portfolios (Internet Appendix Tables 9 and 10).

14 As we do not have access to these data, we cannot judge their reliability and value-added. The point is to provide an example that asset managers can and do use proprietary information about employee satisfaction and corporate culture to form portfolios including these criteria.

Additional information

Notes on contributors

Hamid Boustanifar

Hamid Boustanifar is an associate professor of finance at EDHEC Business School, France.

Young Dae Kang

Young Dae Kang is a financial economist at the Bank of Korea.