For two generations, U.S. companies, regulators, and activists have grappled with how to increase employment diversity in large firms. Quotas and other explicit hiring targets have tended to fare poorly in the courts. Instead, diversity policies have come to focus on processes rather than outcomes. If a firm can demonstrate that it used fair and objective practices when hiring, evaluating, and rewarding employees, the argument goes, then that firm should not be thought of as discriminatory, even if its resulting workforce does not represent the wider labor market.

This focus on confirming or denying discriminatory intent can obscure the original question. An ultimate goal of the Civil Rights Act and supporting legislation is to remove racial and sexual discrimination in U.S. workplaces. An implication is that, if discrimination were ended, we would see more racially integrated workplaces. Yet there is relatively little research on which employer policies actually increase racial or gender diversity. The importance of policies’ effect on actual workforce composition should be underlined. For if the prevailing approach is to assume employers have fulfilled all of their obligations by adopting policies that do not increase diversity, then the prevailing approach does not live up to the intentions of the act.

In a new working paper, we compare workforce composition across two populations of privately held and publicly traded companies. Public companies are often seen as having the potential to be more diverse, for a variety of reasons. Almost all work on employment diversity stresses that formal rules and objective employment procedures that limit the discretion of potentially biased employers or managers will help women and minority workers. Public firms tend to be larger, older, and more bureaucratic than privately held firms. SEC rules and other reporting requirements force public firms to disclose more and different types of information to the government and the public, meaning they have to be more aware of the implications of their own practices. Additionally, because publicly traded companies must provide information to shareholders, and because problems with a public company’s reputation can affect its bottom line through its share price, many activists think that there are more ways to influence such firms.

At the same time, we know that public companies tend to be more profitable and to offer their employees more secure and varied careers than many privately held ones. For these reasons, jobs with these firms are highly valued, and the historically white, male employees who hold them might fight to keep them more tenaciously. More fundamentally, since we do not know exactly which employment practices produce more diverse workforces, we do know whether greater bureaucracy around HR practices in public firms translates into greater diversity.

The problem with studying this question lies in the comparison group. Which private and public firms should we compare? The decision to offer equity on the public markets is not random, and we have every reason to expect that public firms differ systematically from private ones, in size, profitability, the industries in which they operate, and more. Any of these characteristics might be related to what the firm’s workforce looks like. How might we isolate the effect of being publicly traded?

Our approach is to look at firms that file an intent to make an initial public offering (through filing as S-1 with the SEC) and then to compare the firms that went through with their IPO with the firms that withdrew their initial filing and remained private. These two groups of firms are all but indistinguishable when they make those initial filings, allowing us to control for unobserved differences between public and private firms.

We gathered data on all S-1 filings, withdrawn and successful, between 1985 and 2014. We then tried to match each company with data on their workforce composition gathered from EEO-1 establishment surveys filed with the Equal Employment Opportunity Commission. Any firm with more than 100 employees must complete an EEO-1 survey each year, detailing (for each establishment it operates, across nine occupational categories) the numbers of employees it has, broken out by race and gender. Thus, for more than 2,000 firms, we were able to track their workforce numbers for multiple years before and after their S-1 filing and directly compare diversity in firms with successful IPOs to diversity in those that withdrew.

Of course, a firm’s decision to withdraw an IPO filing is not itself random. Were a firm to suddenly record losses, or to find itself in regulatory trouble, it would be more likely to withdraw, but it would also be more likely to lay off employees or make other changes that could affect the composition of its workforce. To get around this issue, we predict whether or not a company followed through with its IPO based on how the markets did in the first two months after the company’s initial filing, and then compare employment outcomes across the two predicted populations. The procedure works for two reasons. First, swings in the markets are very predictive of whether firms withdraw IPO filings—virtually no one wants make an initial public offering into a bear market. Second, while such short-term swings can lower a firm’s valuation, they tend not to affect the firm’s performance—its products and services are unchanged. Thus, the broader market performance does not affect firm performance, except insofar as it affects whether the firm ultimately goes public. With this research design in place, we can isolate the effect of being publicly traded.

Anyone who is hoping that publicly traded firms have, on average, more diverse workforces will find our results distressing. We find no evidence that going public increases employment diversity. We examined proportions of female and non-white employees, as well as female and non-white managers. In no category did public firms outperform private ones.

It is important to understand that this does not mean that diversity was flat over time in these companies. The percentage of female workers has been fairly constant in public companies over the last 30 years (albeit at more than 40 percent in most firms), but the shares of non-white workers and all female and minority managers have grown steadily. Yet they have grown at the same rates, from the same starting points, as in privately held firms. Public ownership, by itself, seems to have had zero effect.

Given our findings, we think the most important question to ask is why so many people assume that public firms will be better managed or more diverse than their private counterparts. We argue that there is a general problem that affects this type of research on organizations. Firms vary in how willing and able they are to cooperate with researchers. They also vary in how willing they are to diversify their workforces, and how able they are to use internal data to do so. Most existing research on employment diversity uses samples of firms that agreed to share data with researchers. Yet if such cooperating firms are also the firms most committed to changing what their workforces look like, then we will tend to over-estimate the effect of different types of structures and policies. Our work demonstrates that we should not be optimistic about the indirect effects of formalizing employment practices on diversity. Even a comparatively large change in the governance structure, like that associated with going public, seems to have little effect. For those firms, regulators, and activists who want to increase diversity, the answers lie elsewhere.

This post comes to us from professors Rembrand Koning at Harvard Business School and John-Paul Ferguson at McGill University. It is based on their recent article, “Does Public Ownership and Accountability Increase Diversity?: Evidence From IPOs,” available here.