Discrimination from below: Experimental evidence from Ethiopia

Highlights

• Lab-in-the-field experiment in Ethiopia testing whether subjects are less likely to follow advice provided by women leaders.

• Gender discrimination affects adherence to leadership.

• Information about a leader's ability reduces gender discrimination against female leaders.

Abstract

Globally, women are underrepresented in leadership positions. A potential explanation is that gender discrimination by subordinates reduces the effectiveness of female leadership. Using a lab-in-the-field experiment in Ethiopia, we test whether leader gender affects the way subjects respond to leadership. We find subjects are ten percent less likely to follow the same advice from a female leader than an otherwise identical male leader. Subjects also give lower evaluations to hypothetical female managerial candidates. However, we find that ability information reverses discrimination. When leaders are presented as highly trained and competent, subjects are more likely to follow advice from women than men. This pattern suggests that beliefs about men and women's ability (i.e., statistical discrimination) play an important role in driving this discriminatory behavior. Our results show that gender discrimination affects adherence to leadership, and signals of ability may be an important tool for gender equity policies aimed at increasing female representation.

Introduction

Improving gender equity in leadership positions is a key priority of the global development agenda.¹ Globally, women remain underrepresented in leadership roles, with the largest gender gaps concentrated in low-income countries. For example, women hold just 17 percent of board directorships in the world's 200 largest companies (African Development Bank, 2015). In this paper, we explore a potential explanation: that gender discrimination makes subordinates less likely to adhere to female leadership. Successful performance in leadership often depends on how well others adhere to one's advice and direction. If women face such discrimination from below by subordinates, then female leaders may appear less effective than male leaders despite taking identical actions.

Using a novel lab-in-the-field experiment in Ethiopia, we study whether individuals follow advice differently when they are randomly assigned to a male versus female leader. Strikingly, although the female and male leaders are otherwise identical, subjects respond differently to the same guidance when provided by a woman rather than a man. When given no information on their leader's ability, subjects are 10 percent less likely to follow a female leader's guidance. As a result, these female-led subjects earn fewer total points. That is, female leadership appears less effective despite being no different from male leadership, and subjects are made worse off by discounting female leadership.²

We then estimate whether information about the leader's ability can mitigate discrimination by subordinates. We inform a random subset of subjects that their leader is of high ability. This information has significantly higher returns for female leaders, and this differential response is large enough to reverse the gender gap. Among those who were informed that their leader was of high ability, subjects were more likely to follow the guidance of a female leader, making male leaders appear less effective despite identical behavior. Together, the results of the experiment show that subordinates can be less likely to follow female leadership due to gender discrimination. However, a signal of high ability can be sufficient to reduce or even reverse such gender discrimination.

This pattern of results also provides suggestive evidence of the mechanisms underlying this discrimination. The fact that information about the leader's ability reverses the gender gap—rather than simply reducing it—implies that differing beliefs about the underlying ability of male and female leaders are contributing to gender discrimination.³ That is, our results suggest that a simple dislike of following female leaders cannot fully account for this discrimination, and that statistical discrimination plays an important role in the results.⁴ The reversal suggests that the same information about leader ability differentially changes beliefs about male versus female ability.

There are two key strengths to the study design. First, we conduct a framed lab-in-the-field experiment with a sample of full-time employees. In general, discrimination from below is difficult to identify using correspondence or audit studies because it requires varying the behaviors and characteristics of those in a position of relative seniority. By conducting a lab experiment we are able to overcome this constraint and provide clean identification of discrimination towards others in senior positions. Furthermore, we document these results in a unique sample of highly educated, high-skilled employees at a large Ethiopian university. The individuals in our sample work in a hierarchical management setting and face daily decisions about following directions from those in more senior positions. Second, we supplement our lab-in-the-field experiment by asking subjects to evaluate a resume for a hypothetical senior management position in which the candidate gender is randomly assigned. Subjects gave lower evaluations to female candidates for the position, providing additional evidence that subjects discriminate based on gender when evaluating leadership positions.

This paper has three main contributions. First, to the best of our knowledge, we are among the first to show that gender discrimination affects adherence to leadership. Leaders will be less effective if subordinates are unwilling to listen to them, and we show that leader gender directly affects subjects’ decision to follow leadership even when that decision is costly. Thus, our results show that gender discrimination from subordinates may make one leader appear less effective than another, despite both having taken identical actions. This highlights a potential mechanism underlying gender gaps in leadership positions.

This finding builds on several important studies documenting differential responsiveness to female versus male leaders, advisers, and experts, particularly in low-income countries (Gangadharan et al., 2016; Grossman et al., 2019). This recent evidence includes female manager trainees in Bangladeshi garment factories being seen as less effective, female-owned businesses in Ghana receiving fewer customers, and farmers perceiving female agricultural trainers in Malawi as less knowledgeable despite the female trainers being equally effective at diffusing new technologies (Macchiavello et al., 2015; Hardy and Kagy, 2018; BenYishay et al., 2020).⁵ This literature documents a consistent differential response to women, but leaves open the question of whether it is driven by gender discrimination. In these natural settings, men and women often differ on a number of characteristics. And, even when men and women are observably similar, subtle differences in communication style, confidence, and risk preferences can drive gender gaps in adherence to leadership.⁶

In our study, we build on this literature by offering explicit identification of discrimination based on gender. We observe all information presented to subjects and experimentally assign leader gender. We limit the interaction between subjects and leaders to written communications, and pre-scripted messages are used to ensure that leader gender is the only difference between the two groups. Thus, we capture how individuals respond to gender itself, as opposed to correlates of gender.⁷ Our lab-in-the-field results support the notion that gender discrimination contributes to the gaps documented in field experiments; likewise, the field experiments highlight the external validity and real-world consequences of our findings.

Our second contribution is to show that providing information about female leaders' ability can alleviate gender discrimination. Since the reduced adherence to female leaders documented in the literature can be driven by multiple factors, it remains an open question whether information on ability can help close such gender gaps. We find that it can. Information about ability raised subjects’ adherence to female leaders by 12 percentage points, while it had no detectable effect for male leaders. The pattern we observe, in which ability information reverses the gender gap in adherence, implies that subjects make inferences about leader ability based on leader gender (i.e., it is consistent with statistical discrimination as a dominant driver of the results).

By showing that information can reduce discrimination, we contribute to the literature on anti-discrimination interventions at the individual level, which has previously focused primarily on encouraging contact between groups rather than information interventions.⁸ Our finding that ability information reduces discrimination at the individual level reinforces evidence from the United States that interventions that signal ability can reduce gender and racial gaps in the labor market. For example, education and occupational licensing have been shown to have higher returns for black men and women (Arcidiacono et al., 2010; Blair and Chung, 2020). Our results suggest that signals of ability, such as credentials, may be an important tool for gender equity policies directed at increasing female representation.

Our third contribution is the finding that a signal of ability can reverse gender discrimination outside a dynamic context. Bohren et al. (2019) show that an ability signal can reverse a gender gap in evaluations because evaluators account for discrimination faced in obtaining the ability signal. A key difference between this paper and Bohren et al. (2019) is that our experiment has no dynamic component: subjects have no reason to believe that it would be more difficult for women to obtain the ability signal in our experiment. This suggests a broader phenomenon in which subjects respond particularly favorably to women of high ability, perhaps due to a general environment in which women commonly face barriers to attaining skills or accolades. Importantly, such reversals indicate that positive discrimination in favor of high-ability women does not preclude the existence of discrimination against women in other contexts.

The rest of the paper proceeds as follows. Section 2 provides details on the experimental design of our study. In Section 3 Leadership Game results, 4 Supporting evidence: resume evaluation experiment, we present our findings and supporting evidence. Section 5 discusses potential mechanisms and policy implications of the results, and Section 6 concludes.