Great Place to Work just released its annual Fortune 100 Best Companies to Work For® 2021. The top five slots went to: Cisco, Salesforce, Hilton, Wegmans Food Markets, Inc., and Rocket Companies in that order.
Is this something that stock pickers, and not just job-seekers, should be paying attention to? We're going to look at that question below -- but before that, there is a potential concern. Historically, the rating has typically been 75% based on confidential employee surveys, and 25% based on an essay question evaluated by Great Place to Work. This most recent ranking, however, has changed the methodology. The weighting of employee survey decreased to 60%, and the essay increased to 40%. In addition, the companies were instructed in advance to focus their essay not on practices in general, but on specific responses to the crisis of 2020, and were also instructed to reveal community support. While community support matters, this raises the problem of metric mission creep. Is this survey about employee satisfaction or general social impact? Comparing the data over time might create problems. If the theory is that employee satisfaction is a highly important proxy for intangible assets, then these changes dilute that. If the theory is that employee satisfaction is a proxy for general social impact, then this year's methodology shift is probably a better test, but that still leaves the problem that the nature of the metric has changed. In an age of increased politicization of our nation and of corporations, and the various agencies which evaluate and nudge them, more subjectivity from the evaluating entity is a genuine cause for added skepticism.
The following analysis is based on past data, not the most recent year and its change in methodology.
In his excellent book, Grow the Pie, London Business School finance professor, Alex Edmans argues that business works best when it grows along with the subsector of society in which it functions, rather than focusing narrowly on maximizing short-term profit. Edmans is for profit, he just thinks that long-term profit, and long-term returns for investors, are generally more of a function of growth and abundance, than they are a function of squeezing down costs or off-loading them onto associates. A good example would be the difference between a company the managers of which are mostly focused on driving a hard bargain with workers and controlling or even cutting wages as the main strategy for profitability as opposed to a company which focuses on getting the most talented workers and training them to high levels of skill and growing the business through the production of excellent and innovative goods and services.
This is not your standard college professor’s “stakeholder capitalism” or “global reset” in which corporate managers’ responsibility is sublimated into a vague haze of general social responsibility. Companies should be profitable, and have to be in order to exist for the long run. But long-term profitability isn’t built at the expense of workers, suppliers, customers or the local community. At least that’s the theory. But what does the evidence show?
Edmans decides to focus on one particular constituency: employees.
“My first decision was how to measure social performance. I chose employee satisfaction – how well a firm treats its colleagues – because there’s a particularly good output measure available. That’s the list of the 100 Best Companies to Work for in America, produced by the Great Place to Work Institute in California and, since 1998, published every year in Fortune magazine. This list is extremely thorough and the ultimate in grass-roots analysis. It surveys 250 employees at all levels, asking them fifty-seven questions on credibility, fairness, respect, pride and camaraderie.”
This makes prima facie sense, employees are a very intimate part of the business; they are in a very real sense the business itself. If human productivity is the key to economic growth, then we need to see how the humans are treated and whether they are afforded the prerequisites to productivity.
There is another reason that Edmans chose employees as the first group of 'partners' in wealth creation.
"There’s a second reason for studying employee satisfaction – there’s clear logic for why it might translate into financial performance. Employees are the most important asset in many modern firms – it’s they who win client relationships and invent new products. Higher employee satisfaction allows an enterprise to recruit and retain top-quality colleagues, and leads to them being more motivated and productive."
This approach seems so much better than standard corporate "green-washing" by giving money, for example, to some tangentially related charity. Focusing on the well-being of your employees is a good deal more likely to affect your bottom line and the welfare of the world than focusing on the well-being of the polar bears, unless your business is a zoo, and probably not even then.
Looking at any particular factor and correlating it with returns is fraught with dangers, especially the danger which comes from confusing correlation with causation. Sometimes things correlate accidentally.
"If a team from the American Football Conference wins the Superbowl, the market tends to fall subsequently; if a team from the National Football Conference wins, the market rises. Some advisors even recommend investing on this effect.7 But there’s no reason for why the Superbowl winner should affect the stock market."
-- Alex Edmans, Grow the Pie
One of the undesirable side effects of the proliferation of cheap computing power is that the more data is available and the easier it is to run thousands of correlations, the more likely we will find random coincidences.
"The ability to mine the data is a particular concern in today’s ‘big-data’ world, where data sources and computing power are becoming limitless. Finance professor Robert Novy-Marx parodied this ability when he was able to predict the performance of trading strategies using the identity of the US President, the weather in Manhattan, global warming, the El Niño phenomenon (the temperature in the Pacific Ocean), sunspots and the alignment of the planets.8 He noted wryly: ‘It seems likely that others could replicate my success, especially given […] the exponential growth in easily obtainable, machine readable data on candidate explanatory variables, and the ease of running these sorts of regressions.’"
-- Alex Edmans, Grow the Pie
One of the ways to mitigate this risk is to restrict oneself to data relationships which have at least a surface plausibility of a causal relationship with one another. If one looks at random relationships long enough, one is bound to find a pattern. Stare at the sky long enough, one might see a cloud that looks like a bunny rabbit. But is it logical that there is a giant vapor-like lapin in the sky? No, so the bunny cloud is not really a bunny.
Another danger is the risk that even if there is an actual correlation and that correlation is plausibly due to a causal relationship, the vector of causation might be running in the opposite direction from the one the analyst is theorizing.
Edmans imagines a company name Super Supermarkets. Super has a great workforce, and has great economic performance. Does that mean that Super's super workforce is the cause of the performance? Not necessarily.
"Now consider a world in which reverse causality holds – employee satisfaction is simply the result of financial performance already being strong. Due to Super’s high profits, its stock price is already 112 today. So the increase to 120 is a return of 7%, no different from the market. Thus, the Best Companies will beat the market only if employee satisfaction improves financial performance, not if financial performance improves employee satisfaction."
-- Alex Edmans, Grow the Pie
If employees are mainly happy only because the company's stockholders (and therefore C-suite managers) are happy, then that is already baked into the stock price. The happiness, in this case, would be the effect of the prosperity of Super's shareholders and the profit-growth which tends to accompany strong performance. But if the over-performance comes after the happiness, then the causal relationship is more likely to go from employee satisfaction to shareholder satisfaction (with good returns) and not the other way around.
First, Edmans had to isolate the employee satisfaction factor:
"To isolate the effect of employee satisfaction, I thus did two things. First, I studied not only Super, but every Best Company traded on the stock market. If Super beat the market, it could be due to its small size or strong recent performance. But if very many Best Companies – with different size and recent performance and in different industries – beat the market, then it’s likely due to their one factor in common, employee satisfaction."
Then, he had to isolate other factors so as not to accidentally include something like size or sector-based returns in the statistical sidecar riding along with employee satisfaction:
"So I compared Super not only to the overall stock market, but also to other firms in the supermarket industry, or other small firms with good recent performance. I did the same for every Best Company. If Automatic Automobiles, a large car firm that performed badly recently, was also in the list, I compared it to other car firms, or other large firms with poor recent performance. Each enterprise thus had its own bespoke comparison group. Importantly, I could also control for risk. There’s no established way to adjust market shares, revenues or profits for risk, but decades of finance research have come up with tools to adjust stock returns. The most well-known is the Capital Asset Pricing Model; I used a more sophisticated version known as the Carhart model."
Having done all of that, he found strong evidence of a causal relationship.
"After all that effort, studying 1,682 firm-years, what was the punchline? I found that the 100 Best Companies to Work for in America delivered stock returns that beat their peers by an average of 2.3 to 3.8% per year over a twenty-eight-year period. That’s 89% to 184% cumulative."
I'm all for the type of bucketing and controlling for variables work that Edmans did, and I found it convincing. But we did our own analysis along much simpler lines. There is a virtue in both approaches. Complex analyses which account for various other causes, controlling for their possible contribution to the effects is a vigorous approach and can deal with certain types of objections. But it can also present a certain "black box" opacity. How so? Often the studies are found in journals which are not readily available to the reading public, including financial analysts. More importantly, the writing style, parlance, jargon, and mathematical rigor is something which even financial professionals (let alone rank and file investors) find difficult to read and digest. Another downside to this type of analysis is that the data is not always made available by journal article authors. In this particular case, however, Edmans has made the data available.
In addition, he has made available an update of the data which can be downloaded along with two papers written about it. See sources at the bottom of this blog for the citations and credits.
Because of some of the complexity-leading-to-opacity issues listed above, we decided to do a much simpler analysis. We simply asked about each company in any given year: Was this company of the Best Places to Work list? If so, we put it in one bucket. If not, we put it in another bucket. Then we asked: What was the average performance in the following calendar year of Bucket One and Bucket Two and what are the differences between those two average returns?
What we found is that Best Places to Work listees did perform slightly better than the rest:
This serves as a kind of cross-check on whether something is hidden in the complex mathematics which would, if not used, reverse the employee satisfaction effect.
In addition, we looked at a different question. Above, we asked whether being on the list was associated with higher returns than not being on the list. Next, we looked only at those companies which were on the list and asked whether the position on that list influenced subsequent returns. We found that there was a relationship.
The scale from left to right is ranking on the list. Over on the furthest left are those companies ranked number 1. On the furthest right are the companies ranked 100, which is the worst. Companies which are not on the list at all are not on this chart. From top to bottom represents highest to lowest return in the subsequent calendar year. The downward slope indicates a negative correlation. But the fit is not very close. In other words, there is a tendency for companies which are ranked better to work for and returns, but not a reliable one. This is consistent with the idea that employee satisfaction is a factor which contributes to outperformance.
For now, it looks like employee satisfaction is additive, notwithstanding current methodology changes and the danger that the list is more subjectively constructed than past methodology.
- Fortune 100 Best Companies to Work For®
- Methodology for 100 Best Companies to Work For (2021), Fortune.com
- alexedmans.com, Data
- Alex Edmans (2011): Does the stock market fully value intangibles? Employee satisfaction and equity prices, (Journal of Financial Economics 101, 621-640) which first used the list from 1984-2009
- Alex Edmans (2012): The Link Between Job Satisfaction and Firm Value, With Implications for Corporate Social Responsibility (Academy of Management Perspectives 26, 1-19) which used the updated list from 1984-2011