Almost every major company proclaims a commitment to diversity. Some do this because it’s the right thing, others because it works well for public relations. Still others do this with the conviction that diversity — differences in where we come from and how we think and what we believe — can provide a wellspring of creativity and innovation.
This final conviction has its share of skeptics. Isn’t some baseline similarity, some level of conformity, necessary to coordinate activities across a firm? Differences, these skeptics argue, are not an asset to be cultivated, but an impediment to be avoided.
For decades, this schism has divided both academic literature and the way in which practitioners shape company culture. A new paper in Administrative Science Quarterly provides a potential way to reconcile the two perspectives: As it turns out, both sides may be correct.
Assessing the Bag of Words
Stanford Graduate School of Business organizational behavior professor Amir Goldberg, with colleagues from McGill University and the University of California, Berkeley, probed this issue using data from Glassdoor, a job recruitment site where people review their employers.
They collected roughly 500,000 reviews from 492 publicly traded companies and analyzed what people wrote:
- How similarly do different employees talk about the culture of their workplace?
- Is it perceived as monolithic or expansive in the ideas it contains?
- And how do these differences relate to the firm’s profitability or market valuation?
To conduct this analysis, they used a machine-learning algorithm known colloquially as topic modeling, which works by scanning documents and mapping patterns of word co-occurrences in them. By doing this, the algorithm is able to infer discrete topics without actually understanding what the words or these topics mean; a topic that features expressions such as “moving fast” and “making mistakes,” for example, probably refers to an experimental culture. Ultimately, the algorithm uses those word occurrences to describe each document by the statistical probability of its discussing each of the topics identified.
The researchers trained the algorithm on a subset of sentences that contained either the word “culture” or a close synonym. Once the algorithm had been trained on this dataset, they used it to analyze every sentence of every review, looking for ways in which employees of a single company converged or diverged in their description of the company’s culture.
“This way, we were able to ascertain the cultural topics of a sentence or review even if the writer did not explicitly refer to culture or one of its synonyms,” Goldberg says. “Importantly, we didn’t care about what the specific cultures under discussion were, but only the level of similarity in what people perceived.”
Good and Bad Differences
The researchers found a vital distinction for thinking about diversity. On one hand, companies in which employees held widely differing views about the firm’s culture tended to have lower profits, as measured by return on assets. Companies where people agreed tended to perform better. (The researchers define this type of diversity, which is an indication of how different groups of people vary in their assessment of corporate culture, as a firm’s “interpersonal heterogeneity.”)
On the other hand, the researchers note that individuals can also embrace an internalized form of diversity, one that enables them to perceive a vibrant range of cultural ideals embedded in a company. This form of diversity, which is a sign of an individual’s ability to acknowledge multiple and sometimes polarized beliefs, they dub “intrapersonal heterogeneity.” Firms that “commonly encourage their employees to adopt a broad and potentially inconsistent set of values” were found to have greater market valuation and more patents of higher quality — measures of a firm’s innovativeness and growth potential, the researchers write.
In short, the findings suggest companies should maximize intrapersonal diversity, as this may spur innovation, while minimizing interpersonal diversity, which can damage coordination.
“We still don’t know precisely what caused what,” Goldberg cautioned. Despite the linguistic “pyrotechnics” used to measure culture, their findings ultimately boiled down to a standard correlational model. It’s not clear whether intrapersonal heterogeneity leads to innovation, or if more innovative firms simply draw people who are more inclined to be culturally open-minded.
To tighten this relationship, Goldberg and his colleagues did side-by-side comparisons of companies that “otherwise looked very, very, very similar, except one was high on one or the other measure of diversity,” Goldberg says. These comparisons made more plausible the possibility that innovation and profitability resulted from their measures of diversity. “It doesn’t meet the full standard of causality, but it enabled us to more accurately compare apples to apples.”
Getting Diversity Right
For most of their existence, efforts to diversify corporations have centered on demographics, like gender and race. And behind this effort has been a limited and only partially defensible assumption that demographic diversity translates into a diversity of beliefs and ways of thinking about the world. The tool developed by Goldberg and his colleagues provides a method for distinguishing between the two — for understanding where demographic diversity overlaps with cultural diversity and where it does not.
In recent weeks particularly, as protests related to George Floyd’s killing have flared across the U.S. and overseas, the issue of diversity cannot be ignored, and the researchers contend that their conclusions extend beyond the corporate world. “Our findings,” they write, “suggest that communities — whether at the local or national level — whose members draw on a wide range of cultural elements will exhibit, all else equal, a greater capacity for innovation.”
“The broader implications relate to the political sentiment, which I share, that the underrepresentation of disadvantaged groups is politically and morally indefensible,” Goldberg says. “And many people who promote this argument have said cultural diversity is good for firms.”
But what if an unexamined push for demographic diversity ends up increasing interpersonal heterogeneity, Goldberg wonders? What if it damages, or is perceived as damaging to, firm performance? What if companies see this and, in self-interested response, retreat from the moral high ground?
“The reason I say this is because the problem of diversity won’t solve itself, and it definitely won’t be solved merely by creating diversity along whatever dimension people care about,” Goldberg says. “It won’t solve itself if, by increasing diversity, companies undermine their performance.”
He made clear that this would not necessarily be the case: The diversity of ideas embodied in a demographically varied workforce can catalyze growth if they are exchanged between people. But the fact that the opposite could occur — that these divergent ideas could remain siloed in different demographic groups, potentially driving a wedge between employees and long-term organizational success — should be enough to give pause.
“With few exceptions, people don’t change their gender; people don’t change their race; but what they do change is how they think about the world,” Goldberg says. “If you want to create a more diverse workplace, and want that diversity to be congruent with positive outcomes, then organizations must find ways for this heterogeneity to inhere within people as opposed to between people.”
Goldberg says it is “essential” to develop a true understanding of how cultural diversity operates “so that we achieve these overarching moral and political goals more effectively — so that we can create and work in a company, and in a world, that is more diverse but that is not also falling apart.”