Underrepresentation of Women CEOs

Women are underrepresented in the top management of U.S. companies. According to Catalyst’s latest report, women occupied merely 41 or 8.2% of the CEO positions at S&P 500 companies in 2023.

Why do so few women become CEOs? Three prominent explanations have garnered widespread attention. First, discussions center around potential differences in skills between male and female CEOs. Commonly cited is the notion that women excel in soft skills such as teamwork and communication, while men thrive in hard skills like analysis, strategy, and execution. The second explanation revolves around career paths, asserting that a restricted female labor supply may diminish women’s presence in managerial positions. The intricate balance between work and family responsibilities often places women in challenging situations, impeding their career advancement. This phenomenon results in a constrained pool of qualified female candidates, amplifying the costs associated with CEO recruitment. The third explanation delves into board preferences, proposing a tendency to discriminate against women. This preference may lead to a deliberate avoidance of promoting women to leadership positions, even if it comes at the expense of company profitability.

Despite substantial media coverage and academic discussion, the precise quantitative importance of these explanations remains ambiguous. Our paper, Underrepresentation of Women CEOs, aims to assess the individual contributions of these factors in explaining the persistent existence of a glass ceiling.

We proceed in two steps. First, we gather data on CEOs and boards of major U.S. corporations from 2001 to 2019. Most of our analysis surrounds CEO successions, which are easy to identify in these data. In contrast, it is a challenge to construct the candidate pools for CEO successions.  As background, nearly all the successors to departing CEOs have connections to the firm or the board (see Czirak & Jenter 2020), with S&P 500 firms hiring from a surprisingly small pool of highly familiar candidates. More than 90% of new CEOs are either current executives or executives with whom its directors have worked.  As such, we use two definitions. We start with a narrow definition of a CEO candidate pool as all the firm’s past and current executives and board members. We then use a broader definition that includes the current executives of peer firms in the same industry.

To motivate the second step in our analysis, we note that using regression analysis on these data is insufficient for the main goal of our work, which is disentangling the different explanations. This type of analysis typically requires some sort of variation in the data that cannot be confounded across explanations. Unfortunately, random variation in the possible reasons for hiring women as CEOs is unavailable for U.S. CEOs. As a general rule, one way to deal with a lack of data is to impose some economically motivated structure on the data, and this path is the one we adopt.

Specifically, we develop a dynamic model of the CEO gender decision that contains three sources of gender-based differences: CEO productivity, search frictions that reflect limited female labor supply, and the impact of discrimination against women. In the model, the board strategically selects the gender of the CEO to maximize their own utilities, which consist of CEO-generated profits, search costs, and the additional disutility stemming from gender-based discrimination. The model contains parameters that describe CEO ability, board distaste for female CEOs, and the cost of searching for a female CEO.

Using the model as a lens through which to interpret our data, we estimate the importance of these three reasons behind gender choice by fitting our model to the observed dynamics of CEO gender decisions. This structural approach enables us to uncover board preferences through revealed preferences and to assess the three contributing factors quantitatively, as different model parameters predict different patterns in the data.  For example, the parameters that capture discrimination, ability differences, and search costs all decline with the observed frequency of female CEOs.  However, we can identify ability differences separately, as our model captures the notion that CEOs of large firms have higher marginal products. Similarly, we can identify search costs separately because increase with the paucity of women in the CEO pool.

Our parameter estimation results show that our model closely reproduces the heterogeneous CEO gender decisions observed in the data. Notably, our model successfully matches the rising trend in women CEO appointments since 2001. The model also captures variation across industries, a crucial aspect considering the pronounced gender segregation within the U.S. labor market.

Our parameter estimates reveal insightful findings regarding the gender gap in CEO positions. Surprisingly, genuine board reluctance toward hiring women as CEOs seems to play a small role. Moreover, our findings indicate negligible actual productivity disparities between male and female CEOs. The primary driver behind the stark gender gap lies in the scarcity of women in candidate pools. This paucity of women implies that search costs are high, and these high search costs are the predominant factor contributing to the underrepresentation of women in CEO roles. Intriguingly, when we eliminate this factor in our model, we find that boards actually prefer women, and the fraction of women CEOs rises from the low single digits to over 50%. In the end, our main conclusion is that the challenge of low female representation extends further down the job ladder, rooted in the limited presence of women in the initial applicant pool.

The complete paper is available here.