Publications

Online Images Amplify Gender Bias

Nature (2024)

Best Paper Prize at the 2022 International Conference on Computational Social Science

Semi-finalist at the Wharton People Analytics Competition, 2021

Each year, people spend less time reading and more time viewing images, which are skyrocketing in their prevalence online. Images from platforms like Google and Wikipedia are downloaded by millions every day, and digital advertisers are increasingly harnessing images to capture attention. Here, we show that the proliferation of images online may significantly exacerbate gender bias. Unlike text, which can readily minimize bias via gender-neutral terminology, images of people inherently convey demographic information, amplifying the presence of gender in the depiction of social categories. People also process images more quickly, implicitly, and memorably than text, suggesting that images may be stronger at reinforcing gender bias in people's beliefs. In this study, we develop computational and experimental techniques for comparing gender bias and its psychological impact across images and texts. We examine the gender associations of 3,495 social categories in over one million images from Google, Wikipedia, and IMDb, as well as in billions of words from these platforms. We find that gender bias is more prevalent in images than text for both female- and male- typed categories. We further show that the documented underrepresentation of women online is substantially worse in images compared to not only text, but also public opinion and US census data. A nationally representative, pre-registered experiment demonstrates that googling for images rather than textual descriptions of occupations amplifies gender bias in participants’ explicit and implicit beliefs.

Douglas Guilbeault, Solène Delecourt, Tasker Hull, Ethan Nadler, Bhargav Srinivasa Desikan, and Mark Chu. 

Who believes gender research? How readers' gender shapes the evaluation of gender research 

Social Psychology Quarterly (2024)

Prior research finds that men view scientific evidence of gender bias against women in science, technology, engineering, and mathematics (STEM) less favorably than women, and the researcher’s gender does not influence evaluations of gender research. Do these effects hold for research documenting workplace gender inequalities more generally? In a preregistered survey experiment fielded on Prolific, survey participants were shown tweets from a fictitious researcher – a woman or man – that summarized recent research about workplace gender inequality, and were asked to rate the research. Consistent with prior work, men viewed research findings about workplace gender inequality less positively than women, and researcher gender did not significantly influence evaluations. Men’s higher endorsement of gender system justification beliefs and hostile sexism partially explained their less positive views, suggesting that men view gender research less positively in part because it challenges the idea that men’s relative advantages in the workplace are natural and earned.

Chloe Hart, Charlotte Townsend, Solène Delecourt

Childcare Matters: Female Business Owners and the Baby-Profit Gap

Management Science (2021)

The previous literature documents that female-owned businesses are less profitable than male-owned businesses, including micro-enterprises that make up the majority of firms in developing countries. In this paper, we uncover an overlooked gendered constraint for these businesses: childcare. We collect field data through unannounced visits to a sample of micro-entrepreneurs in select areas of Uganda, combining surveys of business owners and real customers, as well as purchases by confederate buyers (i.e., mystery shoppers). We document that childcare duties in businesses are highly gendered: 37% of female owners bring small children to work, compared with 0% of men. Childcare duties are correlated with a ``baby-profit gap,'' as businesses where children are present earn 48% lower profits than even other female-owned businesses where a child is not present. Using our rich data, we analyze potential reasons why childcare obligations may affect profits. We find that prices, product quality, and other explanations are not robustly correlated with the presence of a baby. However, we find that women with children in the store are more likely to run out of stock than both men and women who do not have children in the store. While we caution that our analysis is not causal, we consistently find that childcare duties are associated with profitability and may relate to the wider gender gap in business performance.

Solène Delecourt and Anne Fitzpatrick

When does advice impact startup performance?

Strategic Management Journal (2019)

Why do some entrepreneurs thrive while others fail? We explore whether the advice entrepreneurs receive about managing their employees influences their startup's performance. We conducted a randomized field experiment in India with 100 high‐growth technology firms whose founders received in‐person advice from other entrepreneurs who varied in their managerial style. We find that entrepreneurs who received advice from peers with a formal approach to managing people—instituting regular meetings, setting goals consistently, and providing frequent feedback to employees—grew 28% larger and were 10 percentage points less likely to fail than those who got advice from peers with an informal approach to managing people, 2 years after our intervention. Entrepreneurs with MBAs or accelerator experience did not respond to this intervention, suggesting that formal training can limit the spread of peer advice.

Aaron Chatterji, Solène Delecourt, Sharique Hasan and Rembrand Koning

Papers in progress

Equalizing business characteristics can close the gender gap in micro-business performance 

Revise & Resubmit at the Strategic Management Journal

Women-owned businesses earn less than men-owned businesses worldwide, particularly among small firms in developing countries. One potential explanation for this gap is that female business owners may generate lower returns for a given business compared to their male counterparts. If there is indeed a gender gap in business returns, then the gender performance gap could be attributed to two potential factors: differences in the owner's human capital, or female-led businesses facing constraints in market demand. To test whether women have lower returns than men, we conducted a pre-registered field experiment where we equalized business characteristics for men and women, including inventory, hours, setup, and location. Specifically, we created our own businesses in real markets. We set up pairs of market stalls in three different markets and supplied them with identical goods. We then recruited experienced sellers from other local markets and randomized them to our stalls. Prior to our experiment, we found that men earned at least 50% more than women. But men and women also differed in various characteristics, such as women having lower levels of education and operating smaller businesses. Our experimental results demonstrate that equalizing business characteristics reduces the gender gap in business performance to zero. This indicates that the gender gap in business performance is not solely rooted in differences in returns. Instead, our findings suggest that men and women have the potential to achieve equal returns on business resources. 

Solène Delecourt and Odyssia Ng (World Bank)  

The Mixed Impacts of Gender Stereotypes in Face-to-Face Market Transactions: A Field Experiment among Microenterprises

Revise & Resubmit at Organization Science

Entrepreneurs often sell goods and services directly to customers in face-to-face transactions, particularly in developing countries. While existing literature suggests that gender stereotypes likely pattern customers’ behavior during in-person sales transactions, research offers divergent expectations about whether gender stereotypes would advantage or disadvantage female sellers relative to men. To reconcile these competing expectations, we introduce a stage-based theory of gender inequality in face-to-face transactions. Our theory suggests that gender stereotypes can disadvantage or benefit female sellers at different stages of customer-seller interactions. To test this theory, we conducted a novel field experiment in India, with research actors posing as sellers in outdoor food markets. The sellers in our experiment ran identical microenterprises and used standardized behavioral protocols. We gathered data about customers’ behavior at three critical stages of the sales process. We found that female sellers’ (dis)advantage ebbed and flowed across these stages: customers were more likely to approach women than men, requested more price discounts from women, and ultimately, purchased more from women. Our theoretical model and field experimental findings reveal how relative (dis)advantages can morph across transaction stages and cumulatively contribute to gendered differences in entrepreneurial experiences and performance.

Solène Delecourt, Laura Doering and Odyssia Ng

The Uneven Impact of Generative AI on Entrepreneurial Performance

Winner of the Wharton People Analytics Competition, 2024

AOM best paper designation, STR Division (AOM 2024)

Nominated for the Carolyn B. Dexter Award (AOM 2024)

Nominated for Research Methods Paper Prize at the Strategic Management Society Annual Conference (2024)

There  is  a  growing  belief  that  scalable  and  low-cost  AI  assistance  can  improve  firm decision-making  and  economic  performance.   However,  running  a  business  involves a  myriad  of  open-ended  problems,  making  it  hard  to  generalize  from  recent  studies showing  that  generative  AI  improves  performance  on  well-defined  writing  tasks. In our five-month field experiment with 640 Kenyan entrepreneurs, we assessed the impact of AI-generated advice on small business revenues and profits.  Participants were randomly assigned to a control group that received a standard business guide or to a treatment group that received a GPT-4 powered AI business mentor via WhatsApp. While  we  find  no  average  treatment  effect,  this  is  because  the  causal  effect  of  generative AI access varied with the baseline business performance of the entrepreneur: high  performers  benefited  by  just  over  20%  from  AI  advice , whereas  low  performers did  roughly  10%  worse  with  AI  assistance.  Exploratory  analysis  of  the  WhatsApp interaction logs shows that both groups sought the AI mentor’s advice, but that low performers  did  worse  because  they  sought  help  on  much  more  challenging  business tasks.  These findings highlight how the tasks selected by firms and entrepreneurs for AI assistance fundamentally shape who will benefit from generative AI. 

Nicholas G. Otis, Rowan Clarke, Solène Delecourt, David Holtz, Rembrand Koning

The effect of relieving time constraints on the business performance of women-owned businesses: A field experiment

Revise & Resubmit at Research Policy

Women business owners typically earn less than their male counterparts. Previous studies have shown that unconditional cash transfers increase business performance for men but not women. Female business owners may face unique household responsibilities and constraints and, as a result, may not spend the unconditional cash transfer on their business. In this paper, I test whether providing time-saving services to women-owned businesses can boost their profits. Unlike cash, this intervention is not transferable to other household members. In a field experiment, female business owners in Kenya (N=557) were randomly assigned to receive cash, time-saving services (free meals and laundry), or to a control group. My results show that those who received time-saving services reported a 70.4% increase in revenue compared to those who received unconditional cash transfers. Cash recipients did not outperform the control group, in line with prior findings. The cash was frequently spent on household expenses like food and school fees, potentially explaining its failure to increase profits. These findings suggest that time-saving interventions may effectively contribute to closing the gender gap in business performance by alleviating burdensome chores that disproportionately affect women. My study highlights the potential impact of targeted interventions to boost the performance of women-owned businesses, emphasizing the need for nuanced approaches to foster inclusive business environments worldwide.

Solène Delecourt 

Work in progress

Which startup ideas attract talent?

 Nominated for Best Conference Paper Award at the Strategic Management Society Annual Conference 2023

Nominated for Research Methods Paper Prize at the Strategic Management Society Annual Conference 2023

Startups employ few women, especially at their earliest stages. An underappreciated implication of this demographic inequality is that startup ideas targeting women may struggle to attract the talent needed for venture success. If talent prefers to work on ideas that resonate with their backgrounds, or if workers feel uncomfortable working on ideas focused on other demographic groups, then a lack of workforce diversity may have an unequal impact on which ideas succeed. We test this thesis using a field experiment and observational data on the universe of high-tech startup workers in the U.S. In our experiment, we randomize the gender-focus of startups ideas, for example changing an e-bike startup to specifically focus on female commuters. Results from our (in-progress) field experiment show that shifting a non-female-focused startup idea to be female-focused reduces the probability a candidate applies to the job by 15 percentage points. This effect size is equivalent to reducing a job’s posted annual salary by $30,000. Concealing the gender focus of an already female-focused startup idea has equivalent but opposite effect. These effects are rooted in the choices of male job seekers; the minority of women who look for startup jobs on our job-search platform are more likely to apply to female-focused ideas. We then turn to our observational analysis covering the near-universe of U.S. startups in PitchBook. We use word-embedding models to measure how much an idea focuses on women. We find that female-focused ideas attract 25% fewer employees, that those employees have on average 30% less work experience, and that these gaps are driven by the fact that 80% of startup employees are men. Finally, building on models of bias in entrepreneurial experimentation, we find that these gaps only hold for early-stage firms, female-focused ventures that raise substantial capital exhibit no talent gap. Our findings illustrate how heterogeneity in beliefs about which early-stage ideas are "worthy of working on" can impact the rate and direction of innovation.

Solène Delecourt, Rem Koning, and Sahiba Chopra

The Invisibility of Older Women Online

Women face disproportionate social pressure to appear young, constituting a pervasive and harmful bias known as gendered ageism, the study of which has been markedly absent in large-scale audits of online stereotypes. Here, we develop a crowdsourcing method for identifying associations between age and gender in over half a million images from Google, Wikipedia, and IMDb, associated with over 3,000 of the most common social categories, including occupations (such as "doctor" and "scientist") as well as generic social roles (such as "friend" and "neighbor"). Across all platforms, women are depicted as substantially younger than men, such that women are most commonly represented as adolescents and men as adults. These findings reveal a crucial pathway for the diffusion of gender bias, given not only that images from these platforms are frequently viewed, downloaded, and circulated, but also that exposure to online images can reinforce stereotypes in people's beliefs, as demonstrated by experimental research. 

Douglas Guilbeault, Solène Delecourt, and Bhargav Srinivasa Desikan 

Sticky constraints: Business location, competition, and the gender profit gap

Gender inequality is ubiquitous, including among small business owners in developing countries, which constitute the majority of businesses worldwide. To understand inequality in business performance, research has primarily focused on improving the performance of existing firms by providing capital, advice, or training. These strategies often increase the performance of male-, but not female-owned, firms. In this paper, we explore whether gender differences in a business' initial conditions can help explain this puzzle. If men and women start different kinds of businesses, and those initial choices affect profitability but are hard to change ex-post (are ``sticky"), this may limit potential profitability. To answer this question, we use rich quantitative data from a representative sample of 3,077 businesses in Kenya. In this context, women make 47% fewer profits than men. We find that women tend to locate their businesses in less profitable locations. While firm profits grow with the distance from home, women are more likely to locate their businesses close to home. Using census data of all firms in the study area, we find that male entrepreneurs locate in places with less competition. Male business owners are over five times more likely to be a monopolist in their sector within a specific radius than women. Though sector and location are important for profits, the majority of owners do not update these initial decisions after founding their business: in the two years preceding our study, only 7% of entrepreneurs changed either sector or location. Our findings suggest that these sticky conditions are more binding for women with greater childcare responsibilities than for men or other women. Our results point to the importance of initial conditions in perpetuating the gender gap among small firms in developing countries.

Solène Delecourt, Anne Fitzpatrick, Layna Lowe, Anya Marchenko