Policy Research Working Paper 11152 Gender Discrimination in Entrepreneurial Finance Experimental Evidence from Ethiopia Niklas Buehren Sreelakshmi Papineni Africa Region A verified reproducibility package for this paper is Gender Innovation Lab available at http://reproducibility.worldbank.org, June 2025 click here for direct access. Policy Research Working Paper 11152 Abstract This paper examines implicit gender bias in entrepreneurial than others. Our findings are consistent with discrimina- financing by randomizing screenings of business investment tion against women in traditionally male-dominated sectors ideas pitched in the format of the reality television show and discrimination against men in female-dominated sec- Chigign Tobiya. Keeping business idea and pitch quality tors. Men are perceived as better negotiators and leaders constant, the experiment randomizes whether a female or in sectors that attract higher investment. These are also male entrepreneur delivers the pitch across three different the sectors in which women are typically underrepresented. business sectors. The findings suggest that, on average, Exposure to women in leadership positions and informa- gender does not affect recommended investment; however, tion provided during the screenings can increase investment the sector matters. Some sectors attract greater investment in women’s businesses. This paper is a product of the Gender Innovation Lab, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at nbuehren@worldbank.org and spapineni@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank.org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Gender Discrimination in Entrepreneurial Finance: Experimental Evidence from Ethiopia∗ Niklas Buehren† Sreelakshmi Papineni‡ Gender, Entrepreneurship, Firms, Finance [JEL] C93, D14, D25, J16, L25, L26, O12 ∗ This paper is a product of the World Bank Africa Gender Innovation Lab, within the office of the Africa Region Chief Economist (AFRCE). The project was conducted in partnership with Renew Capital and a microfinance institution in Ethiopia. We thank Maria Emilia Cucagna for excellent research assistance, and Alemayehu Woldu Gedrago and Yemsrach Kinfey Edey for expert field and relationship management. We also thank Rachel Coleman and Toni Weis for invaluable project support and Laura Davis who led the production of the videos. We are grateful to Global Affairs Canada (GAC) under the Innovations in Financing Women Entrepreneurs (IFWE) project, as well as the World Bank Umbrella Fund for Gender Equality (UFGE) and other World Bank sources for funding. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. The study is registered on the AEA RCT registry at https://doi.org/10.1257/rct.9064-1.1 † World Bank, nbuehren@worldbank.org ‡ World Bank, spapineni@worldbank.org (Corresponding author) 1 Introduction Gender-based discrimination is often emphasized to help explain a persistent gender gap in access to finance and firm performance (Hardy and Kagy, 2018; World Bank, 2019; Buehren et al., 2019; Delecourt and Ng, 2020; Fang et al., 2022). Both debt and equity financing disproportionately flow to male-led firms which limits the potential of female entrepreneurs to grow their firms (World Bank, 2021).1 For instance, in credit markets, despite an expansion of microfinance loans to support female-led enterprises (Aterido, Beck and Iacovone, 2011), women entrepreneurs are often disadvantaged with loan terms deemed insufficient to promote firm growth (Meager, 2019). Similarly, in equity markets, a gender gap in venture capital (VC) and private equity (PE) investments is observed with only 2% of all VC in the United States and Europe allocated to female-led businesses (PitchBook, 2023); and an estimated 6% of VC going to female-led firms in Sub-Saharan Africa (IFC, 2019). Women’s lower participation in the funding pool is often attributed to inaccurate perceptions and over- evaluation of risk due to gender discrimination (e.g., Bartos et al., 2023; Miller et al., 2023; Alibhai et al., 2019). In this paper we measure gender discrimination in business investment allocation decisions using a novel lab-in-the-field experiment in Ethiopia. In the experiment we isolate a gender and sector effect in business funding allocations using the format of a reality television show. The television show called “Chigign Tobiya” roughly translated as “Ethiopia Emerges” is Ethiopia’s version of the highly successful tele- vision show “Shark Tank” shown in the US and many other countries.2 While we use an existing television show as the foundation for our study on gender discrimination, we do not aim to measure the effect of the educational entertainment or “edutainment” program on 1 Debt financing refers to loans through a traditional lender, e.g., a bank or microfinance institute (MFI). Equity financing involves securing capital in exchange for part business ownership. Private equity represents shares in a business not traded on a public exchange, and public equity is publicly traded stocks or shares. 2 The format of the show is used globally and is called “Tigers of Money” in Japan and “Dragon’s Den” in the UK, for example. The premise of the show is that a panel of investors called “sharks” decide whether to invest as entrepreneurs make investment pitches about their company or product. 2 attitudes or behavior per se.3 Instead, in our experiment we record videos of actors and ac- tresses pitching scripted investment ideas to the four business tycoons on the show. We keep the business idea and pitch quality constant, but randomize whether the investment pitch is given by a male or female actor across three business sectors.4 The study participants are a sample of 1,600 men and women who are students in vocational and technical training colleges, and 224 credit officers from a microfinance institution in urban Ethiopia.5 The experiment involves randomizing the study participants to watch different video content and comparing investment outcomes across groups. Our study examines whether there is an im- plicit gender bias by comparing differences in recommended investment for pitches delivered by either a male or female actor (gender bias). We also test whether effects differ by the gen- der of the viewer (gender homophily) and examine whether the sector of the business pitch influences investment decisions (heterogeneity by sector).6 The three experimental sectors include construction, textiles, and food services.7 We explore whether men and women are rewarded or penalized in investment decisions when they violate gender-sector stereotypes; and measure the effect of a role model and information add-on intervention. The experimental findings suggest limited evidence of gender discrimination in recom- mended investment, on average. Both the sample of students and credit officers are equally likely to recommend investment for a female pitch as for a male pitch. While female viewers are more likely to recommend investment than male viewers (83% vs 77%, respectively), we 3 Existing literature on entrepreneurship edutainment focuses on using television to get non-entrepreneurs interested in starting a business and giving them information on where and how to start. Bjorvatn et al. (2020) find that exposure to an edutainment program in Tanzania, aimed at secondary school students, had a positive impact on entrepreneurial activities and a negative effect on educational performance. Smith and Viceisza (2018) analyze the impact on firms that pitch on Shark Tank in the US to show that female contestants are less likely to receive offers, and receive lower valuations and less capital from the investors than men, partly because they tend to ask for less. 4 We also experiment with good quality versus bad quality pitches and test several intervention add-ons. 5 The research was carried out in partnership with RENEW Strategies, an investment firm in Ethiopia, a microfinance institution and three TVET colleges in Addis Ababa, Ethiopia. 6 Gender homophily analysis assesses the tendency of viewers to prefer an entrepreneur of the same gender. 7 Sectoral sex segregation is the phenomena that women tend to sort into different sectors than men, and earnings in female-dominated sectors tend to be less than in male-dominated sectors (World Bank, 2022). The three sectors were intentionally chosen to represent a male-dominated (construction), gender-neutral (textiles), and female-dominated (caf´ e) sector based on previous research in Ethiopia (Alibhai et al., 2017). 3 find no evidence of gender homophily, i.e., male viewers are not more likely to recommend investing in a male pitch and female viewers are not more likely in a female pitch. However, there are differences by sector. The construction and textile sectors attract higher recom- mended investment than the food services sector. We find a gender gap in favor of men in the recommended amount of investment within the construction sector, and a gender gap in favor of women in the likelihood of investment in the food services sector. Overall, these findings are consistent with discrimination against women in traditionally male-dominated sectors and discrimination against men in female-dominated sectors. Hebert (2023) finds a similar pattern of gender discrimination using French administrative data of venture capital deals. Kanze et al. (2020) provides evidence from the US of a “lack of fit” effect whereby female-led ventures catering to male-dominated industries receive significantly less funding at significantly lower valuations than female-led ventures catering to female-dominated indus- tries. We show that viewers appear to perceive men to be better negotiators and leaders than women within the construction sector. We find that the patterns of bias are not observed among more gender progressive viewers measured by pre-experiment beliefs. In our paper we also examine the impact of role model and information add-on interven- tions. For the role model intervention, a woman chief executive officer (CEO) talks about women’s entrepreneurial abilities prior to the business pitch video. In our experiment we find a positive impact of the role model intervention on the likelihood of investment in a female pitch. In addition, we show that providing information about women’s earnings in tradi- tionally male-dominated sectors increases the recommended amount of investment towards the female pitch within the construction sector. Recent studies that randomize exposure to videos containing information on role models focus on changing aspirations. For exam- ple, Bernard et al. (2023) provide experimental evidence from Ethiopia linking exposure to successful role models with higher aspirations and increased future-oriented investments. Similarly, Lubega et al. (2021) find a positive impact of a role model intervention in Uganda on women’s business start-ups. 4 Earlier economic studies of discrimination often relied on “decomposition” measurement approaches but due to their limitations gave rise to audit and correspondence studies (see Bertrand and Duflo (2016) and Neumark (2018) for a summary). In this paper we use an audit study design to examine gender discrimination in financing. A seminal audit study by Bertrand and Mullainathan (2004) examined race discrimination in the US labor market by esum´ sending fictitious r´ es to job ads in newspapers. The authors manipulate perceived race esum´ by randomly assigning white-sounding names to half the r´ es and African-American- sounding names to the other half and find that white names receive 50% more call-backs for job interviews. Similarly, an audit study by Neumark, Bank and Van Nort (1996) investigat- ing the role of sex discrimination in vertical segregation within the restaurant sector in the US using actors and actresses finds discrimination against women in higher-priced restau- rants and discrimination in women’s favor in low-price restaurants. A key contribution of our paper is to provide evidence of gender discrimination in entrepreneurial financing in a low- and middle-income country context. Existing research has largely focused on discrimi- nation in labor and rental housing markets within high-income countries, with fewer studies exploring the extent of discrimination in access to equity finance (some recent exceptions in credit finance include Ayalew, Manian and Sheth, 2021; Miller et al., 2023; Alibhai, Bell and Conner, 2017). Economic theory traditionally identifies two sources of discrimination: taste-based (Becker, 1957) or statistical (Arrow, 1972, 1973; Phelps, 1972) discrimination.8 More recently, Bertrand, Chugh and Mullainathan (2005) introduced a third interpretation of implicit discrimination described as unconscious and therefore outside a discriminator’s awareness. In terms of ac- cess to finance, women entrepreneurs might experience discrimination due to stereotypes or norms that identify women as less skilled or productive (e.g., taste-based discrimination); or information-specific frictions that lead to establishing financial decisions on the average characteristics of the group (e.g., statistical discrimination); or due to unconscious biases 8 Taste-based discrimination may stem from preferences that favor allocating capital to a specific gender, and statistical discrimination from differing beliefs about the profit-potential of female- versus male-led firms. 5 (e.g., implicit discrimination). Our paper provides evidence of gender discrimination in the context of sectoral sex segregation and equity financing. While field experiments have been overall successful at documenting that discrimination exists, they have (with a few excep- tions) struggled with linking the patterns of discrimination to a specific theory as described in Bertrand and Duflo (2016). Our paper suggests that discrimination can be attributed to both implicit gender biases and taste-based discrimination. We also demonstrate successful strategies to help overcome gender stereotypes in entrepreneurship through the use of role models and information. The remainder of this paper is organized as follows. Section 2 describes the experimental design and sample. Section 3 introduces the data, section 4 summarizes the estimation strategy, while section 5 discusses the results. We provide concluding thoughts in section 6. 2 Experimental Design and Study Sample The study was done in collaboration with Renew Capital, an investment firm headquartered in Addis Ababa, Ethiopia whose co-founder is a producer of the reality television show. Chigign Tobiya is a nationally-televised show where entrepreneurs pitch their business idea for a chance to receive an equity investment from a panel of angel investors that has the same format as “Shark Tank” in the US. In the study, we use actors and actresses to create mock business pitches that are based on the television show’s format. The actors and actresses pitch a business investment idea to the panel of judges — the business idea stays constant but we vary the sex of the person delivering the pitch, pitch quality, and sector of the business in the experiment.9 We randomly screen videos of different pitches to examine the gender biases among viewers. The different video content will help to examine the following research questions: is there a gender bias in assessing business pitches; does the gender bias vary by pitch quality or business sector; and can we shift the gender bias by resetting the norm of what it means to be a successful entrepreneur through the use of role models and 9 The Chigign Tobiya mock episodes were filmed in December 2020. 6 information. For the experiment, mock episodes were recorded where either a male or a female actor presents the same business idea by either delivering a good or bad quality pitch;10 another female or male actor provides general advice on entrepreneurship and thereby plays the part of a role model. The video clips were also recorded for three different business sectors: e/restaurant, and textiles which are chosen to represent business sectors construction, caf´ that are typically male-, female-dominated or gender-neutral, respectively. The different combinations of pitch features that were shown are summarized in Appendix Table A1. We examine whether the investment bias differs across the gender of the person pitching, the business sector, and pitch quality. The scripted business pitch was delivered by either an actor or actress in front of the panel of tycoons who also appear in the actually aired TV show. The business tycoons were instructed to ask the same question in all pitches: “what is your marketing strategy?” In Appendix A.2 we include the business scripts used in the experiment. Our study sample involves two types of participants: students from technical and voca- tional training colleges (TVETs) and credit officers from a microfinance institution based in Addis Ababa. The experiment measures the opinions of these “viewers”. While students are unlikely to have to make investment decisions of this scale in their daily lives, credit officers are often involved in making debt-financed investment decisions for entrepreneurs.11 The lab-in-the-field experiments included a pre-experiment survey during which basic socio- demographic data were collected from the viewers, and then videos were shown to 1,600 TVET students (800 young men and 800 young women). The student sample was randomly assigned to one of eight groups that was shown different combinations of the experimentally varied aspects of the pitches. The student study sample was drawn from college adminis- trative lists provided to the research team for three training centers.12 The rationale for 10 To vary pitch quality, the same pitch is of “bad quality” when the actors were instructed to act with less confidence and, when reading the script, use filler words without changing the content of the pitch. 11 Future research with fund managers or angel investors making equity financing decisions is encouraged. 12 The three training centers include Nefas Silk, Tegrabeid, and Wingate. 7 sampling students from training centers was that the students who are taking vocational or technical skills are likely to be aspiring entrepreneurs themselves. The sampling lists included a total of 6,225 male students and 2,925 female students across different departments. For the sampling we excluded any departments that had less than 15 observations per gender (4 departments were excluded). The survey firm were instructed to recruit individuals into the study at the training colleges where they had to stratify the sample across departments and gender of the viewer based on sample size guidance. Eligible participants are those that consented to the study.13 The largest student population in the training centers are in In- formation and Communication Technology (ICT) courses and this meant approximately one third of our sample was taking courses in the ICT sector. In the student experiment we ran- domize individuals to watch one of eight different combinations of video clips (see Appendix Table A1). The randomization for the student sample was conducted at the individual level and stratified by gender and department of study.14 For the microfinance institution (MFI) sample, the 224 credit officers were assigned to two different combinations of the pitches.15 The credit officer sample were mostly men (212 men and 12 women) drawn from the universe of credit officers working in the Addis Ababa branches of the microfinance institution. Tables A3 and A4 show the characteristics of the study participants. The average age of the students in our sample is 21 and credit officers are 29 years old, on average. 13 While students in regular classes were prioritized, students in extension classes could also be included. 14 The randomization process was automated on a tablet as individuals were recruited into the study. 15 Power calculations were conducted to determine the minimum sample size required to detect a treatment effect of 20% over the control mean at 80% power. We assume the variation in outcome measure as the ratio of the standard deviation to the mean for both treatment and control group is calculated using a ratio of 0.5 and an outcome of level of investment in a business idea. Power calculations recommended a sample size of 99 observations per treatment arm. The number of observations in each group was 100. 8 3 Data The data collection and video intervention all took place in lab-in-the-field experimental sessions conducted in November and December 2021. Enumerators conducted a 30 minute face-to-face interview with participants that included watching the video content and a short data collection before and after. The short baseline survey was conducted before watching the video clip and included demographic characteristics, work, explicit gender attitudes, perceived norms, and family history. In order to avoid priming before watching the video on the gender-related topic of the experiment, we included attitudinal questions related to age and education biases in addition to gender. The baseline survey was conducted face-to-face before the videos were screened. After the baseline survey then the participants are randomized into groups to watch the video clips on a tablet. The videos are approximately 10 minutes total length and are shown to each individual. A follow-up survey was conducted after watching the first video and then again after watching a second video. The main outcomes we measure include whether the respondent would recommend the four business tycoons invest in the business pitch and how much would they recommend they invest (max investment of USD 50,000). A number of additional questions were asked to assess the perceived views of the entrepreneur and the business idea. For example, questions that asked whether the study participant perceived the entrepreneur to be a good negotiator, leader, or manager in their business; and whether the business idea is considered to be risky. Before the respondent exited the lab session they are asked to complete an implicit association test (IAT) based on the Gender-Career topic (Greenwald, McGhee and Schwartz, 1998). The IAT measures associations between concepts (e.g., Family and Career) and evaluations (e.g., Female and Male). The gender-career IAT used in this study assesses the implicit association in people’s minds between two pairs of concepts: male-and-career and female-and-family (stereotype pairings) versus male-and-family and female-and-career 9 (non-stereotype pairings). Participants were instructed to sort words that appeared on the computer screen, also called stimuli, as quickly as possible into different categories. Stimuli were male and female nouns, career-related words and family-related words. The IAT is based on the logical premise that easier pairings (faster response and fewer errors) are more strongly implicitly associated in the participant’s mind than more difficult pairings (slower response and more errors).16 For each study participant, the scoring algorithm as described by (Greenwald, Nosek and Banaji, 2003) was used to calculate a d-score. The IAT d-score is an effect size measure where higher, positive values indicate a stronger association (bias) of the stereotype pairings; and lower, negative values indicate a stronger association of the non-stereotype pairings. To use the IAT in the Ethiopian context the tool was first translated into Amharic and names that are more common in Ethiopia were programmed in the IAT. 4 Empirical Strategy 4.1 Gender Bias Our main estimation uses the following ordinary least squares (OLS) specification: Yi = β0 + β1 M aleP itchi + β2 Xi′ + ϵi , (1) where Yi is the outcome variable measured immediately after the video is screened. M aleP itchi is a binary variable taking the value of one if the individual was assigned to the group to watch a male entrepreneur pitching the business idea; and 0 if assigned to watch the female entrepreneur. β1 gives the pure gender bias effect i.e., whether the viewers are more likely to recommend investment in a business idea pitched by a male entrepreneur compared to a female entrepreneur. Xi is a vector of respondent demographic controls in- 16 The IAT assumes that people are quicker to respond when items that are more closely related in their mind share the same button. For example, an implicit preference for Family relative to Career means that you are faster to sort words when ’Family’ and ’Female’ labels share a button relative to when ’Career’ and ’Female’ labels share a button. 10 cluding gender, age, years of schooling and marital status of the viewer. We also control for the experimental design (e.g., good/bad quality, sector of business pitch, and first/second video) in a pooled regression. Fixed effects for randomization strata in the student sample include location of interview (TVET college) and TVET course subject and level. We report error terms ϵi robust to individual heteroskedasticity. 4.2 Gender Homophily Yi = β0 + β1 M aleP itchi + β2 M aleV ieweri (2) + β3 M aleP itchi × M aleV ieweri + β4 Xi′ + ϵi where Yi and Xi are defined as above in equation 1. We interact the M aleP itchi binary variable with M aleV ieweri (equal to one if the viewer is male; and 0 if female) to explore the hypothesis that male (female) viewers are more likely to invest in a male (female) en- trepreneur. 4.3 Gender Bias Heterogeneity by Business Sector Yi = β0 + β1 M aleP itchi + β2 T extilesi + β3 Constructioni + β4 M aleP itchi × T extilesi + β5 M aleP itchi × Constructioni (3) + β6 M aleV ieweri + β7 Xi′ + ϵi where Yib,t=1 and Xi are again defined as above in equation 1. For this analysis we interact the M aleP itchi dummy variable with the three sector dummy variables. We examine whether there is a gender bias by sector of business operation. There are three sectors: F oodServicesi , T extilesi and Constructioni . The omitted variable in the regressions is F oodServicesi . We present the results from the student study sample in Tables 1 to 4 and the results of the MFI credit officer sample in Table 5. Additional heterogeneity analysis by pre-experiment implicit and explicit gender bias is also shown in Tables 3 and 5. 11 5 Results 5.1 Investment Outcomes The main investment outcomes we use to measure a potential gender bias in Table 1 include a binary variable equal to 1 if the viewer thinks the tycoons should invest in the business idea, and equal to 0 if not; and the proposed investment amount that the viewer thinks should be invested (unconditional and conditional on investing). Table 1 column 1 suggests that there is no evidence of discrimination based on the gender of the actor delivering the pitch, on average. Results show no significant differences in the likelihood of investment and no significant effect on the amount of investment. Column 2 suggests differences in recommended investment based on the gender of the viewer, whereby female students are more likely to recommend investment than male students. Overall, female viewers are more likely to think that the tycoons should invest in the business than male viewers (83% vs 77%, respectively). In Table 1 column 3 we present evidence of differences in the investment decisions by business sector. The negative coefficient on M aleP itch suggests that women are favored over men when pitching a business idea in the food services sector. Viewers are 4 percentage points more likely to recommend investing in women than men within the food services sector. The positive and significant coefficients on ConstructionSector and T extilesSector suggest these sectors, on average, attract more investment than the F oodServices sector (both on the extensive and intensive margins).17 The p-values at the bottom of Table 1 for columns 3, 6, and 9 shows that the recommended amount of investment is approximately 8% higher for a male pitch than a female pitch in the construction sector. Viewers favor investment in men’s businesses over women’s within the male-dominated sector only. 17 The construction sector was intentionally chosen as it is considered to be a male-dominated sector, textiles is gender-neutral, and food services is female-dominated in Ethiopia (Alibhai et al., 2017). 12 5.2 Entrepreneur and Business Perception Outcomes Outcomes presented in Table 2 include index measures of the viewers’ perceptions of the entrepreneur and the business. The entrepreneur is rated overall on a scale of 0 to 10 as well as on their perceived knowledge and skills, and interpersonal skills (based on being a good negotiator, leader, or manager). The final outcome in Table 2 captures how risky the viewer perceives the business idea to be. Table 2 columns 1, 4, 7, and 10 presents the pure gender bias. Overall, there is no evidence of a gender difference in the perceived view of the entrepreneur and riskiness of the business idea. Columns 2, 5, 8, and 11 suggest that male viewers, on average, rate the entrepreneur and their interpersonal skills lower than female viewers, irrespective of the gender of the pitch. Male viewers also perceive the business ideas to be riskier, on average, than female viewers. In terms of sectoral differences, men are perceived less favorably than women pitching in the food services sector. Both in the overall rating (column 3) and interpersonal skills of the entrepreneur (column 9). The average rating of the entrepreneur and their skills in the textiles sector is higher overall but we find no evidence of a gender bias within the textiles sector. Women in the construction sector are rated lower in terms of their perceived knowledge and interpersonal skills than men in the construction sector (see p-value [A]+[G] at the bottom of Table 2). These findings align with the results on gender differences in investment within the construction sector shown in Table 1. Further decomposition of the results on the interpersonal skills outcomes suggests that men are perceived as better negotiators and leaders compared to women pitching the same idea in the construction sector (breakdown not shown). It is interesting that the perceived riskiness of the business idea does not vary by gender or sector (column 12). 13 5.3 Heterogeneity by Gender Attitudes and Norms In Table 3 we present the same outcomes as explored in Tables 1 and 2. However, we conduct heterogeneity analysis using baseline measures of gender attitudes and norms. In Table 3 Panel A we explore whether there is a gender bias in investment based on the viewer’s gender-career IAT score. In Figure A.1 in the Appendix we show that male students and male MFI credit officers have very similar implicit association test score distributions. The female students (yellow line) have a d-score that is slightly higher on average, i.e., they have a stronger implicit bias for Male with Career and Female with Family than the male students in their college.18 In Panel A we code the d-score as a binary variable equal to 1 if the viewer had a moderate or strong stereotypical association between Male with Career and Female with Family in their IAT (equivalent to 26% of the sample). Results suggest that the viewers with more gender-stereotypical implicit biases at baseline are more likely to recommend investing in a male pitch versus a female pitch (see p-value [A]+[C] at the bottom of Table 3 column 1). Those with more gender implicit biases also rate the male entrepreneur more highly than the female entrepreneur (column 3). In Table 3 Panel B we use a more explicit gender bias measure that is a binary variable equal to 1 if the viewer personally agrees that men make better executives than women do (31% of the sample). Using this measure of gender bias we find no association with the viewers’ explicit beliefs and investment outcomes. Finally, in Panel C, we consider a perceived community norms measure where a higher value is associated with greater gender bias in their community with respect to entrepreneurship.19 The gender norms measure is somewhat predictive of investment decisions of a male versus female pitch. The more the viewer perceives their community favors men as business executives over women, the more likely they are to recommend a higher investment amount to a male pitch. 18 Note that the sample size for female credit officers is very small because of the limited number of women credit officers actually working for the MFI relative to men. Therefore we are not able to conduct gender-disaggregated analysis for the MFI sample. 19 The question asks the viewer at baseline “out of 10 of your neighbors, how many do you think would think men make better business executives than women do?” 14 5.4 Role Model Add-on In Table 4 Panel A we examine the results by the role model add-on intervention. The role-model add-on video was used to frame the perception of success before the viewer made their investment decisions. Those who watched a video with a male pitch were shown a male role model and those who watched a female pitch were shown a female role model prior to watching the pitch. Column 1 suggests that the role model add-on increases the likelihood of investment in a female pitch by 7 percentage points. The female role model also increases the perceived rating of the female entrepreneur and her skills, and considers her business idea as less risky. 5.5 Information/Education Add-on In Table 4 Panel B we explore the impact of providing information about women’s average earnings in male-dominated sectors prior to watching the business pitch video (experiment is within the construction sector only). We show that providing information about women’s ability to succeed in a male-dominated sector increases the recommended amount of invest- ment towards women entrepreneurs seeking investment within the construction sector. 5.6 Good Quality versus Bad Quality Pitch In Table 4 Panel C we examine the results by the quality of the investment pitch. The explanatory variable Bad Quality Pitch is a binary variable equal to 1 if the viewer was randomly assigned to view the bad quality video in the experiment. Overall, viewing a bad quality pitch reduces the likelihood of investment at the extensive margin, lowers the perceived rating of the entrepreneur, and increases the perceived risk of the business idea. However, we find no gender differences in investment outcomes for a bad quality pitch. 15 5.7 MFI Credit Officers In Table 5 we present the gender and sector differences in recommended investment among the sample of credit officers. Among the 224 credit officers in the sample, only 12 of them were women. Table 5 suggests no evidence of a gender differential in the likelihood of investment or the amount of recommend investment. The credit officers are more likely to rate a female entrepreneur more highly than a male entrepreneur on average. In Table 5 Panel B we show that similar to the sample of TVET students, credit officers also recommend more investment for the construction sector relative to the food services sector both at the extensive and intensive margin.20 However, credit officers show no bias between male and female pitches within the food services sector. However, they are more likely to recommend investment in women compared to men in the construction sector. There is no difference in the amount of investment recommend towards a female and male pitch. MFI credit officers rate women entrepreneurs in the construction sector more highly than women in the food services sector, and men in the construction sector. Women operating in nontraditional sectors may be seen more favorably by credit officers who are already expe- rienced in credit decisions. Perhaps since the expectation is women who have to go against societal norms to operate a business are more likely to be growth-oriented entrepreneurs or receive support from the men in their lives. In panels C and D we conduct heterogeneity analysis by the baseline implicit and ex- plicit gender attitudes. MFI credit officers present a more perverse pattern with respect to investment decisions. Higher implicit gender bias based on IAT is associated with lower recommend investment among women overall. However, credit officers are more likely to favor investment in the female pitch compared to the male pitch when they are not assessed as implicitly biased based on their IAT score. Similarly, panel D, suggests that those who are more explicitly biased are less likely to invest in a male pitch relative to female pitch. 20 Note that only the food services and construction sectors were used in the experiment for credit officers given the total sample size of 224. 16 6 Conclusion Gender discrimination is not a prevalent driver of entrepreneurial investment outcomes in our experimental study setup. However, our analysis suggests that women in the construc- tion sector receive less financing than men when seeking equity finance for their business. Conversely, men seeking equity finance in the food services sector are less likely to receive financing than women for their business. Since sectors that typically attract higher invest- ment are often male-dominated, our results can be reconciled with real-world findings that women often receive less investment finance than men. Our findings suggest that sector stereotypes are internalized. Given that earnings are often higher in sectors dominated by men, then the gender bias in investment observed in our experiment has implications for gender-based differences in earnings among entrepreneurs. Women study participants generally are more likely to recommend investment than men. However, we find no evidence that women invest more in women, or men invest more in men. When women entrepreneurs are presented as strong leaders in CEO positions, participants are more likely to recommend investment in women’s businesses. Our results suggest that gender of the person pitching a business idea affects investment decisions based on sectoral sex segregation, and signals of ability may be an important tool for supporting investment. Taken at face value, our results have important policy implications. They suggest that if women recognize that gender discrimination in financing is not the modus operandi, they may rationally be more willing to apply for business financing and go after more funding opportunities. Our results also highlight the importance of occupational sex segregation — that there is potentially still stigma within certain industries. Gender-sensitization training programs to investors may help to illuminate gender-sector stereotypes and alleviate the gender gap in equity market outcomes. Exposure to successful role models through career counseling or job fairs may benefit some women who are willing to break traditional sector stereotypes and enter sectors that attract more financing. 17 Table 1: Pure Gender Bias, Gender Homophily, and Heterogeneity By Sector Investment Outcomes Tycoons should choose Investment Amount Investment Amount to invest Unconditional Conditional (Yes 1 No 0) (ETB in thousands) (ETB in thousands) (1) (2) (3) (4) (5) (6) (7) (8) (9) ∗ ∗ Male Pitch [A] -0.01 0.01 -0.04 10.14 47.19 -18.88 5.95 38.74 -8.89 (0.01) (0.02) (0.02) (18.00) (25.84) (29.12) (17.47) (25.07) (31.15) Male Viewer [B] -0.06∗∗∗ -0.06∗∗∗ -10.76 -20.95 7.47 -5.15 (0.02) (0.01) (25.76) (18.10) (24.78) (17.70) Male Pitch × Male Viewer [C] -0.01 -18.98 -22.03 (0.03) (35.84) (34.63) Textile Sector [D] 0.12∗∗∗ 205.25∗∗∗ 171.84∗∗∗ (0.03) (37.51) (36.88) Construction Sector [E] 0.06∗∗ 199.90∗∗∗ 167.42∗∗∗ (0.03) (32.66) (32.05) Male Pitch × Textile Sector [F] 0.07∗∗ 36.12 -2.64 (0.03) (44.34) (44.80) Male Pitch × Construction Sector [G] 0.08∗∗ 103.33∗∗ 74.27∗ (0.04) (44.00) (43.47) Obs 3,256 3,256 3,256 3,256 3,256 3,256 2,615 2,615 2,615 Base Mean (Female Pitch) 0.81 0.83 0.75 1259.14 1261.40 1115.81 1358.61 1350.35 1230.38 p-value [A]+[F] .182 .607 .72 p-value [A]+[G] .16 .011 .033 Controls ✓ ✓ ✓ ✓ ✓ ✓ Strata FE ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Note: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 [1] Outcome variables: Tycoons should choose to invest is a dummy variable equal to 1 if the viewer thinks the tycoons should invest in the idea, and equal to 0 if not. Amount tycoons should invest is the proposed investment amount that the viewer thinks should be invested, and Amount tycoons should invest (conditional) is the amount conditional on investing. [2] Male Pitch is a dummy variable equal to 1 if a male entrepreneur pitches the business idea and equal to 0 if a female entrepreneur pitches the idea. Male Viewer is a dummy variable equal to 1 if the viewer is male and equal to 0 if the viewer is a female. Business sector dummy variables include Textile sector if the business is in the gender-neutral textile sector, Construction sector if the business is in the male- dominated construction sector, and the omitted business sector is the female-dominated food services (caf´ e and restaurant ) sector. [3] Ordinary Least Squares (OLS) regression specification with robust standard errors. [4] Controls include experimental design controls (pitch quality, second video, and experimental add-ons), and respondent characteristics including age and years of schooling. [5] Fixed effects for randomization strata include location (TVET college), TVET course subject and level. 18 Table 2: Pure Gender Bias, Gender Homophily, and Heterogeneity By Sector Entrepreneur and Business Perception Outcomes Entrepreneur Entrepreneur Entrepreneur Business idea overall rating knowledge and skills Interpersonal Skills risk index (Score 0-10) (Score 0-10) (Index 0-3) (Score 0-4) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Male Pitch [A] -0.08 -0.05 -0.24∗∗ 0.01 0.02 -0.04 -0.03 -0.04 -0.13∗∗∗ 0.01 -0.03 0.06 (0.07) (0.10) (0.12) (0.03) (0.05) (0.05) (0.03) (0.04) (0.05) (0.04) (0.06) (0.06) Male Viewer [B] -0.36∗∗∗ -0.32∗∗∗ -0.08 -0.07∗∗ -0.11∗∗∗ -0.09∗∗∗ 0.18∗∗∗ 0.19∗∗∗ (0.10) (0.07) (0.05) (0.03) (0.04) (0.03) (0.06) (0.04) Male Pitch × Male Viewer [C] 0.07 0.00 0.04 0.03 (0.14) (0.06) (0.06) (0.08) Textile Sector [D] 0.54∗∗∗ 0.26∗∗∗ 0.10∗∗ 0.02 (0.15) (0.07) (0.05) (0.08) Construction Sector [E] 0.09 -0.03 -0.11∗ 0.08 (0.13) (0.06) (0.06) (0.07) Male Pitch × Textile Sector [F] 0.30∗ 0.03 0.13∗∗ -0.15 (0.17) (0.08) (0.06) (0.10) Male Pitch × Construction Sector [G] 0.41∗∗ 0.16∗∗ 0.25∗∗∗ -0.14 (0.18) (0.08) (0.08) (0.09) Obs 3,256 3,256 3,256 3,256 3,256 3,256 3,256 3,256 3,256 3,256 3,256 3,256 Base Mean (Female Pitch) 7.47 7.62 7.21 4.49 4.51 4.40 2.56 2.61 2.55 2.23 2.15 2.28 p-value [A]+[F] .64 .886 .961 .272 p-value [A]+[G] .232 .053 .049 .301 Controls ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Strata FE ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Note: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 [1] Outcome variables: Entrepreneur overall rating asks viewers to rate the entrepreneur on a scale from 0 to 10. Entrepreneur knowledge and skills asks viewers to rate the entrepreneur’s skills on a scale of 0-10, and Entrepreneur interpersonal skills is an aggregated index that ranges from 0 to 3 calculated as the sum of three dummy variables for if the viewer perceives the entrepreneur as a good negotiator, good leader and good manager in business. Business idea risk level measures how risky the respondent perceives the business idea to be on a scale of 0 (entirely risk-free) to 4 (very risky). [2] Male Pitch is a dummy variable equal to 1 if a male entrepreneur pitches the business idea and equal to 0 if a female entrepreneur pitches the idea. Male Viewer is a dummy variable equal to 1 if the viewer is male and equal to 0 if the viewer is a female. Business sector dummy variables include Textile sector if the business is in the gender-neutral textile sector, Construction sector if the business is in the male- dominated construction sector, and the omitted business sector is the female-dominated food services (caf´ e and restaurant ) sector. [3] Ordinary Least Squares (OLS) regression specification with robust standard errors. [4] Controls include experimental design controls (pitch quality, second video, and experimental add-ons), and respondent characteristics including age and years of schooling. [5] Fixed effects for randomization strata include location (TVET college), TVET course subject and level. 19 Table 3: Heterogeneity Analysis by Implicit and Explicit Gender Bias, and Gender Norms Heterogeneity Analysis by measures of gender bias and perceived norms Tycoons should Investment Amount Entrepreneur Entrepreneur Entrepreneur Business idea choose to invest Conditional overall rating knowledge and skills Interpersonal Skills risk index (Yes 1 No 0) (ETB in thousands) (Score 0-10) (Score 0-10) (Index 0-3) (Score 0-4) (1) (2) (3) (4) (5) (6) Panel A (Implicit Gender Attitudes) Male Pitch [A] -0.01 27.34 -0.15∗ -0.02 -0.05 0.00 (0.02) (21.53) (0.09) (0.04) (0.04) (0.05) Implicit Bias (IAT) [B] -0.02 55.13∗ -0.22∗ -0.10∗ -0.08 -0.14∗∗ (0.02) (29.08) (0.12) (0.06) (0.05) (0.07) Male Pitch × Implicit Bias (IAT) [C] 0.06∗ -4.61 0.44∗∗ 0.12 0.08 0.02 (0.03) (40.99) (0.17) (0.08) (0.07) (0.09) Obs 3,090 2,477 3,090 3,090 3,090 3,090 Base Mean (Female Pitch No Bias) 0.81 1349.02 7.54 4.53 2.59 2.28 p-value [A]+[C] .07 .529 .065 .143 .621 .785 Panel B (Explicit Gender Attitudes) Male Pitch [A] 0.02 34.24 0.02 0.02 -0.04 -0.04 (0.02) (21.71) (0.09) (0.04) (0.04) (0.05) Explicit Bias (Attitude) [B] 0.02 15.84 0.16 0.01 0.02 0.06 (0.02) (25.92) (0.11) (0.05) (0.04) (0.06) Male Pitch × Explicit Bias (Attitude) [C] -0.03 -18.79 -0.12 0.03 0.07 0.06 (0.03) (37.65) (0.15) (0.07) (0.06) (0.08) Obs 3,256 2,615 3,256 3,256 3,256 3,256 Base Mean (Female Pitch No Bias) 0.80 1352.02 7.43 4.49 2.56 2.22 p-value [A]+[C] .61 .631 .45 .462 .495 .749 Panel C (Perceived Gender Norms) Male Pitch [A] 0.06 -48.95 0.09 0.04 0.17∗∗ 0.05 (0.04) (44.92) (0.19) (0.09) (0.08) (0.10) Gender Norm [B] -0.03 -68.02 -0.12 -0.04 0.12 0.17 (0.04) (53.69) (0.24) (0.11) (0.09) (0.13) Male Pitch × Gender Norm [C] -0.10 150.48∗ -0.21 -0.04 -0.35∗∗ -0.14 (0.06) (78.96) (0.34) (0.16) (0.14) (0.18) Obs 3,256 2,615 3,256 3,256 3,256 3,256 Base Mean (Female Pitch No Perceived Bias) 0.77 1526.96 7.47 4.70 2.44 2.00 p-value [A]+[C] .255 .016 .521 .93 .013 .353 Controls ✓ ✓ ✓ ✓ ✓ ✓ Strata FE ✓ ✓ ✓ ✓ ✓ ✓ Note: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 [1] Gender Bias specification heterogeneity analysis by measures of gender bias. Panel A heterogeneity analysis by a dummy variable equal to 1 if the viewer had a moderate or strong stereotypical association between Male with Career and Female with Family in their Implicit Association Test (26% of sample). Panel B by a dummy variable equal to 1 if the viewer personally strongly agrees or agrees that men make better executives than women do (31% of sample). Panel C by a gender norm that is a continuous variable indicating the viewer’s perception of the proportion of the community that think men make better business executives than women do. [2] OLS regression specification with robust standard errors. [3] Controls include experimental design controls and respondent characteristics including gender, age and years of schooling. Fixed effects for randomization strata include location (TVET college), TVET course subject and level. 20 Table 4: Impact of Experimental Role Model, Education, and Bad Quality Pitch Impact of Role Model, Education, and Pitch Quality Experiments Tycoons should Investment Amount Entrepreneur Entrepreneur Entrepreneur Business idea choose to invest Conditional overall rating knowledge and skills Interpersonal Skills risk index (Yes 1 No 0) (ETB in thousands) (Score 0-10) (Score 0-10) (Index 0-3) (Score 0-4) (1) (2) (3) (4) (5) (6) Panel A (Role Model Add-on) Male Pitch [A] -0.03∗ 3.31 -0.22∗∗ -0.02 -0.07∗∗ 0.03 (0.02) (22.02) (0.09) (0.04) (0.03) (0.05) Role Model Add-on [B] 0.07∗∗ 57.61 0.57∗∗∗ 0.35∗∗∗ 0.09 -0.35∗∗∗ (0.03) (36.19) (0.14) (0.08) (0.06) (0.09) Male Pitch × Role Model Add-on [C] 0.07∗ -43.24 0.13 -0.12 0.07 -0.03 (0.04) (50.34) (0.20) (0.11) (0.07) (0.13) Obs 2,418 2,003 2,418 2,418 2,418 2,418 Base Mean (Female Pitch No Add-on) 0.83 1358.42 7.71 4.57 2.68 7.71 p-value [A]+[C] .22 .378 .621 .159 .962 .998 Panel B (Education Add-on Construction) Education Add-on -0.00 161.60∗∗∗ -0.08 0.01 -0.04 -0.13 (0.04) (43.31) (0.20) (0.09) (0.08) (0.12) Obs 405 349 405 405 405 405 Base Mean (Female Pitch No Add-on) 0.86 1462.51 7.91 4.59 2.67 2.16 Panel C (Quality of the Pitch Experiment) Male Pitch [A] -0.02 18.61 -0.23∗∗∗ -0.07∗ -0.07∗∗ 0.04 (0.02) (21.47) (0.08) (0.04) (0.03) (0.05) Bad Quality Pitch [B] -0.07∗∗∗ 20.05 -0.90∗∗∗ -0.38∗∗∗ -0.39∗∗∗ 0.14∗∗ (0.03) (31.10) (0.12) (0.05) (0.05) (0.06) Male Pitch × Bad Quality Pitch [C] 0.04 14.01 0.34∗ 0.17∗∗ 0.10 -0.07 (0.03) (42.37) (0.18) (0.08) (0.08) (0.09) Obs 3,256 2,615 3,256 3,256 3,256 3,256 Base Mean (Female Good Quality Pitch) 0.84 1367.96 7.79 4.62 2.69 2.17 p-value [A]+[C] .582 .371 .479 .134 .751 .708 Controls ✓ ✓ ✓ ✓ ✓ ✓ Strata FE ✓ ✓ ✓ ✓ ✓ ✓ Note: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 [1] Panel A role model add-on experiment sample restricted to the good pitch quality videos only. Panel B education add-on experiment sample restricted to the good pitch quality in construction sector videos only. In Panel C bad quality pitch experiment includes the full study sample. [2] OLS regression specification with robust standard errors. [3] Controls include experimental design controls and respondent characteristics including gender, age and years of schooling. Fixed effects for randomization strata. 21 Table 5: MFI Sample: Pure Gender Bias and Heterogeneity by Sector and Gender Atttiudes Investment, Entrepreneur and Business Outcomes Tycoons should Investment Amount Entrepreneur Entrepreneur Entrepreneur Business idea choose to invest Conditional overall rating knowledge and skills Interpersonal Skills risk index (Yes 1 No 0) (ETB in thousands) (Score 0-10) (Score 0-10) (Index 0-3) (Score 0-4) (1) (2) (3) (4) (5) (6) Panel A (Pure Gender Bias) Male Pitch -0.07 66.81 -0.46∗∗ -0.04 -0.05 -0.07 (0.04) (41.24) (0.20) (0.10) (0.09) (0.12) Obs 448 350 448 448 448 448 Base Mean (Female Pitch) 0.82 1398.46 7.65 4.58 2.63 2.24 Panel B (Heterogeneity by Sector) Male Pitch [A] 0.02 104.87 -0.33 0.11 0.13 -0.12 (0.06) (64.67) (0.25) (0.13) (0.12) (0.15) Construction Sector [B] 0.18∗∗∗ 349.88∗∗∗ 0.47∗∗ 0.29∗∗ 0.16 -0.14 (0.05) (56.74) (0.21) (0.12) (0.11) (0.14) Male Pitch × Construction Sector [C] -0.18∗∗ -70.42 -0.27 -0.31∗ -0.35∗∗ 0.10 (0.08) (78.44) (0.32) (0.17) (0.16) (0.20) Obs 448 350 448 448 448 448 Base Mean (Female Pitch) 0.73 1196.23 7.41 4.44 2.55 2.31 p-value [A]+[C] .004 .485 .024 .166 .082 .897 Panel C (Implicit Gender Attitudes) Male Pitch [A] -0.11∗∗ 55.22 -0.55∗∗ 0.01 -0.09 -0.15 (0.05) (48.93) (0.22) (0.12) (0.11) (0.14) Implicit Bias (IAT) [B] -0.16∗∗ 38.87 0.16 0.05 -0.13 -0.02 (0.07) (69.81) (0.28) (0.17) (0.14) (0.19) Male Pitch × Implicit Bias (IAT) [C] 0.15 24.27 0.34 -0.14 0.14 0.37 (0.11) (97.21) (0.47) (0.24) (0.22) (0.27) Obs 436 340 436 436 436 436 Base Mean (Female Pitch) 0.85 1396.79 7.61 4.55 2.68 2.24 p-value [A]+[C] .694 .336 .622 .55 .825 .343 Panel D (Explicit Gender Attitudes) Male Pitch [A] -0.02 117.04∗∗ -0.07 0.07 0.19∗ -0.23 (0.06) (52.85) (0.23) (0.12) (0.11) (0.14) Explicit Bias (Attitude) [B] 0.03 70.95 0.19 -0.02 0.13 -0.08 (0.06) (60.54) (0.23) (0.14) (0.12) (0.16) Male Pitch × Explicit Bias (Attitude) [C] -0.14 -128.68 -1.02∗∗∗ -0.30 -0.62∗∗∗ 0.40∗ (0.09) (90.11) (0.38) (0.20) (0.20) (0.23) Obs 448 350 448 448 448 448 Base Mean (Female Pitch No Bias) 0.80 1379.80 7.57 4.57 2.58 2.26 p-value [A]+[C] .033 .868 .001 .185 .012 .348 Controls ✓ ✓ ✓ ✓ ✓ ✓ Strata FE ✓ ✓ ✓ ✓ ✓ ✓ Note: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01 [1] MFI credit officer sample gender bias specification and split to explore heterogeneity by sector and measures of gender bias. Panel A is the pure gender bias. Panel B by a binary variable equal to 1 for the construction sector and 0 for the food services sector. Panel C by a binary variable equal to 1 if the viewer had a moderate or strong stereotypical association between Male with Career and Female with Family in their Implicit Association Test (23% of sample). Panel D by a binary variable equal to 1 if the viewer personally strongly agrees or agrees that men make better executives than women do (38% of sample). [2] OLS regression specification with robust standard errors. Controls include respondent characteristics including gender, age, and years of schooling. Fixed effects for randomization strata include MFI branch and position of the MFI officer. 22 References Alibhai, Salman, Aletheia Amalia Donald, Markus Goldstein, Alper Ahmet Oguz, Alexander Pankov and Francesco Strobbe. 2019. “Gender Bias in SME Lending : Experimental Evidence from Turkey.” World Bank Policy Research Working Papers 9100. Alibhai, Salman, Niklas Buehren, Sreelakshmi Papineni and Rachael Pierotti. 2017. “Crossovers: Female Entrepreneurs Who Enter Male Sectors - Evidence from Ethiopia.” World Bank Policy Research Working Paper 8065. Alibhai, Salman, Simon Bell and Gillette Conner. 2017. “What’s Happening in the Missing Middle?: Lessons from Financing SMEs.” World Bank Working Paper Series . Arrow, Kenneth. 1972. Some Mathematical Models of Race in the Labor Market. A. H. Pascal ed. Racial Discrimination in Economic Life. Lexington Books. Arrow, Kenneth. 1973. The Theory of Discrimination. in O. A. Ashenfelter and A. Rees, eds. Discrimination in Labor Markets. Princeton University Press. Aterido, Reyes, Thorsten Beck and Leonardo Iacovone. 2011. Gender and finance in Sub- Saharan Africa: are women disadvantaged ? Policy Research Working Paper Series 5571 The World Bank. Ayalew, Shibiru, Shanthi Manian and Ketki Sheth. 2021. “Discrimination from below: Ex- perimental evidence from Ethiopia.” Journal of Development Economics 151. Bartos, Vojtech, Silvia Castro, Kristina Czura and Timm Opitz. 2023. “Gendered Access to Finance: The Role of Team Formation, Idea Quality, and Implementation Constraints in Business Evaluations.” CESifo Working Paper No. 10719 . Becker, G.S. 1957. The Economics of Discrimination. Chicago. University. Economics re- search center. Economics research studies University of Chicago Press. Bernard, T, S Dercon, K Orkin, G Schinaia and A Seyoum Taffesse. 2023. “The Future in Mind: Aspirations and Long-term Outcomes in Rural Ethiopia.” CEPR Discussion Paper No. 18492 . Bertrand, Marianne, Dolly Chugh and Sendhil Mullainathan. 2005. “Implicit Discrimina- tion.” American Economic Review 95(2):94–98. Bertrand, Marianne and Esther Duflo. 2016. Field Experiments on Discrimination. NBER Working Papers 22014 National Bureau of Economic Research, Inc. Bertrand, Marianne and Sendhil Mullainathan. 2004. “Are Emily and Greg More Employ- able Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” American Economic Review 94(4):991–1013. Bjorvatn, Kjetil, Alexander Cappelen, Linda Sekei, Erik Sørensen and Bertil Tungodden. 2020. “Teaching Through Television: Experimental Evidence on Entrepreneurship Edu- cation in Tanzania.” Management Science 66:2308–2325. 23 Buehren, Niklas, Markus Goldstein, Paula Gonzalez Martinez, Adiam Hagos Hailemicheal, Daniel Kirkwood, Patricia Paskov, Michelle Poulin, Chandni Raja and Andrew Tarter. 2019. “Ethiopia Gender Diagnostic Report : Priorities for Promoting Equity.” World Bank Policy Research Working Paper Series . ene and Odyssia Ng. 2020. Do buyers discriminate against female-owned Delecourt, Sol` businesses? Two field experiments. Technical report Working Paper, Stanford University. Fang, Sheng, Chorching Goh, Mark Roberts, Lixin Colin Xu and Albert Zeufack. 2022. “Female entrepreneurs and productivity around the world: Rule of law, network, culture, and gender equality.” World Development 154. Greenwald, Anthony, Brian Nosek and Mahzarin Banaji. 2003. “Understanding and Using the Implicit Association Test: I. An Improved Scoring Algorithm.” Journal of Personality and Social Psychology 85:197–216. Greenwald, Anthony, Debbie McGhee and Jordan Schwartz. 1998. “Measuring Individual Differences in Implicit Cognition: The Implicit Association Test.” Journal of personality and social psychology 74:1464–80. Hardy, Morgan and Gisella Kagy. 2018. Mind The (Profit) Gap: Why Are Female Enterprise Owners Earning Less than Men? In AEA Papers and Proceedings. Vol. 108 pp. 252–55. Hebert, Camille. 2023. “Gender Stereotypes and Entrepreneur Financing.” 10th Miami behavioral finance conference . IFC. 2019. “Moving Toward Gender Balance in Private Equity and Venture Capital.”. Kanze, Dana, Mark Conley, Tyler Okimoto, Damon Phillips and Jennifer Merluzzi. 2020. “Evidence that investors penalize female founders for lack of industry fit.” Science Ad- vances 6. Lubega, Patrick, Frances Nakakawa, Gaia Narciso, Carol Newman, Archileo N. Kaaya, Cissy Kityo and Gaston A. Tumuhimbise. 2021. “Body and mind: Experimental evidence from women living with HIV.” Journal of Development Economics 150:102613. Meager, Rachael. 2019. “Understanding the Average Impact of Microcredit Expansions: A Bayesian Hierarchical Analysis of Seven Randomized Experiments.” American Economic Journal: Applied Economics 11(1):57–91. Miller, Amisha, Saurabh A. Lall, Markus Goldstein and Joao Montalvao. 2023. “Asking Bet- ter Questions: The Effect of Changing Investment Organizations’ Evaluation Practices on Gender Disparities in Funding Innovation.” World Bank Policy Research Working Papers 19625. Neumark, David. 2018. “Experimental Research on Labor Market Discrimination.” Journal of Economic Literature 56(3):799–866. 24 Neumark, David, Roy J. Bank and Kyle D. Van Nort. 1996. “Sex Discrimination in Restau- rant Hiring: An Audit Study.” The Quarterly Journal of Economics 111(3):915–941. Phelps, E. S. 1972. “The Statistical Theory of Racism and Sexism.” The American Economic Review 62(4):659–661. PitchBook. 2023. “US All In: Female Founders in the VC Ecosystem.”. Smith, Baylee and Angelino Viceisza. 2018. “Bite me! ABC’s Shark Tank as a path to entrepreneurship.” Small Business Economics 50(3):463–479. World Bank. 2019. “Profiting from Parity : Unlocking the Potential of Women’s Business in Africa.” World Bank Report 88(1):19–31. World Bank. 2021. “In Search of Equity : Exploring Africa’s Gender Gap In Start-Up Financing.” Africa Gender Innovation Lab and Briter Bridges . World Bank. 2022. “Breaking Barriers : Female Entrepreneurs Who Cross Over to Male- Dominated Sectors.” World Bank Report . 25 A Online Appendix A.1 Implicit Association Test (IAT) Distribution Figure A1: Implicit Gender-Career Bias using the Implicit Association Test (IAT) Notes: IAT measures the strength of association between Male/Female and Career/Family. 26 Table A1: Experimental Design: Randomized Groups for TVET Students Sample 15 second break with Video 2 Groups Video 1 Video 3 instruction (Add-on Experiments) 1 Business Script 1 - Female - Good Pitch Please PAUSE this video Educational – Female Business Script 3 - Female - Good Pitch 2 Business Script 2 - Female - Good Pitch Please PAUSE this video Emotional - Female 1 Business Script 1 - Female - Good Pitch 3 Business Script 3 - Female - Good Pitch Please PAUSE this video Role Model - Female Business Script 2 - Female - Good Pitch 4 Business Script 1 - Male - Good Pitch Please PAUSE this video None Business Script 3 - Male - Good Pitch 5 Business Script 2 - Male - Good Pitch Please PAUSE this video Emotional - Male 1 Business Script 1 - Male - Good Pitch 6 Business Script 3 - Male - Good Pitch Please PAUSE this video Role Model - Male Business Script 2 - Male - Good Pitch 7 Business Script 1 - Male - Bad Pitch Please PAUSE this video Emotional - Male 3 Business Script 3 - Male - Bad Pitch 8 Business Script 1 - Female - Bad Pitch Please PAUSE this video Emotional - Female 3 Business Script 3 - Female - Bad Pitch Business 1 = Café Key: Business 2 = Textile 27 Business 3 = Construction Good pitch = quality of the pitch was "good". Bad pitch = quality of the pitch was intentionally "bad". Table A2: Sample Size of Strata for the TVET Students Randomized Sampling Total Number of Students in 3 TVETS in Proposed Sampling all levels Regular + for the Study Extension Female Male Female Male Automotive 29 1780 16 16 Biomedical Equipment Servicing Management 80 216 40 40 Business 582 210 120 120 Construction 83 1135 40 40 Electricity and Electronics 22 343 16 16 Furniture Making 15 233 8 8 Health Information Technician 65 24 16 16 Hotel and Tourism 277 59 32 32 ICT 1163 1038 312 312 28 Manufacturing 21 469 16 16 Surveying and Drafting 129 516 48 48 Textile, Garment and Leather 459 202 136 136 Agro-processing 17 5 0 0 Building Construction 13 268 0 0 Mineral Exploration, Mining and Processing 14 12 0 0 Urban Land Development 1 6 0 0 2925 6225 800 800 Table A3: Mean Differences in Characteristics of Youth Sample by Gender (1) (2) (3) (2)-(3) Total Male Female Pairwise t-test Variable N Mean/(SE) N Mean/(SE) N Mean/(SE) N Mean difference Age of respondent (Years) 1628 21.426 813 22.022 815 20.832 1628 1.190*** (0.084) (0.132) (0.100) Education Years of Schooling (Years) 1628 12.643 813 12.736 815 12.551 1628 0.185*** (0.033) (0.053) (0.040) Married (Yes=1; No=0) 1628 0.037 813 0.036 815 0.038 1628 -0.002 (0.005) (0.007) (0.007) Generalized self efficacy (standardized 0-1) 1628 0.839 813 0.839 815 0.838 1628 0.001 (0.002) (0.003) (0.003) Gender Att: % Agree men make better business execs than women 1628 0.313 813 0.403 815 0.223 1628 0.180*** (0.011) (0.017) (0.015) Age Att: % Disagree older men make better business execs than younger men 1628 0.689 813 0.701 815 0.676 1628 0.025 (0.011) (0.016) (0.016) Educ Att: % Agree uni grads make better business execs than highschool grads 1628 0.519 813 0.524 815 0.514 1628 0.010 (0.012) (0.018) (0.018) Norm: % of community believe men make better business executives than women 1628 0.519 813 0.538 815 0.500 1628 0.038*** (0.005) (0.007) (0.008) Gender Att: % Agree women should not enter occupations dominated by men 1628 0.060 813 0.066 815 0.053 1628 0.014 (0.006) (0.009) (0.008) Norm: % of community believe women should not enter occupations dominated by men 1628 0.453 813 0.453 815 0.453 1628 -0.000 (0.006) (0.009) (0.009) IAT: No Association between Female & Male with Career & Family (Yes=1; No=0) 1443 0.282 731 0.309 712 0.254 1443 0.055** (0.012) (0.017) (0.016) Currently Engaged in Paid Work (Yes=1; No=0) 1628 0.370 813 0.518 815 0.223 1628 0.295*** (0.012) (0.018) (0.015) Aspires to be a business owner when they graduate (Yes=1; No=0) 1628 0.447 813 0.475 815 0.420 1628 0.055** (0.012) (0.018) (0.017) Aspires to work in wage employement when they graduate (Yes=1; No=0) 1628 0.275 813 0.252 815 0.297 1628 -0.045** (0.011) (0.015) (0.016) Aspires to work in other employment when they graduate (Yes=1; No=0) 1628 0.278 813 0.273 815 0.283 1628 -0.010 (0.011) (0.016) (0.016) Had a male role model while growing up (Yes=1; No=0) 1628 0.467 813 0.585 815 0.350 1628 0.236*** (0.012) (0.017) (0.017) Had a female role model while growing up (Yes=1; No=0) 1628 0.406 813 0.278 815 0.534 1628 -0.256*** (0.012) (0.016) (0.017) Breadwinner mother: main income earner while growing up (Yes=1; No=0) 1628 0.329 813 0.317 815 0.341 1628 -0.024 (0.012) (0.016) (0.017) Breadwinner father: main income earner while growing up (Yes=1; No=0) 1628 0.685 813 0.708 815 0.661 1628 0.047** (0.012) (0.016) (0.017) * 0.1 ** 0.05 *** 0.01. Value for t-tests are the differences in means across genders 29 Table A4: Mean Differences in Characteristics of MFI Credit Officers Sample by Gender (1) (2) (3) (2)-(3) Total Male Female Pairwise t-test Variable N Mean/(SE) N Mean/(SE) N Mean/(SE) N Mean difference Age of respondent (Years) 224 28.960 213 28.437 11 39.091 224 -10.654*** (0.364) (0.342) (1.132) Education Years of Schooling (Years) 224 15.763 213 15.671 11 17.545 224 -1.874*** (0.142) (0.140) (0.857) Married (Yes=1; No=0) 224 0.259 213 0.239 11 0.636 224 -0.397*** (0.029) (0.029) (0.152) Generalized self efficacy (standardized 0-1) 224 0.867 213 0.868 11 0.844 224 0.025 (0.005) (0.005) (0.029) Gender Att: % Agree men make better business execs than women 224 0.375 213 0.390 11 0.091 224 0.299** (0.032) (0.033) (0.091) Age Att: % Disagree older men make better business execs than younger men 224 0.701 213 0.695 11 0.818 224 -0.123 (0.031) (0.032) (0.122) Educ Att: % Agree uni grads make better business execs than highschool grads 224 0.625 213 0.629 11 0.545 224 0.084 (0.032) (0.033) (0.157) Norm: % of community believe men make better business executives than women 224 0.542 213 0.548 11 0.409 224 0.139** (0.013) (0.013) (0.061) Gender Att: % Agree women should not enter occupations dominated by men 224 0.031 213 0.033 11 0.000 224 0.033 (0.012) (0.012) (0.000) Norm: % of community believe women should not enter occupations dominated by men 224 0.413 213 0.413 11 0.409 224 0.004 (0.016) (0.016) (0.092) IAT: No Association between Female & Male with Career & Family (Yes=1; No=0) 213 0.286 204 0.294 9 0.111 213 0.183 (0.031) (0.032) (0.111) Had a male role model while growing up (Yes=1; No=0) 224 0.598 213 0.615 11 0.273 224 0.342** (0.033) (0.033) (0.141) Had a female role model while growing up (Yes=1; No=0) 224 0.335 213 0.319 11 0.636 224 -0.317** (0.032) (0.032) (0.152) Breadwinner mother: main income earner while growing up (Yes=1; No=0) 224 0.335 213 0.333 11 0.364 224 -0.030 (0.032) (0.032) (0.152) Breadwinner father: main income earner while growing up (Yes=1; No=0) 224 0.759 213 0.761 11 0.727 224 0.033 (0.029) (0.029) (0.141) * 0.1 ** 0.05 *** 0.01. Value for t-tests are the differences in means across genders 30 A.2 Business Pitch Scripts The following includes a description of the scripts given to the actors and actresses to mem- orize before the experiment. BUSINESS SCRIPT 1: CAFE/RESTAURANT (Female-Dominated Sector) Opening • Thank you investors for giving me the opportunity to present my business to you today • My name is Dagmawit XXX and I am the founder of “The Ethiopia Pizza Company” • As a little (boy/girl) I had a dream – of creating the best pizza in the world and now I am making that a reality. • We have a longstanding history of pizza in Ethiopia, and I am perfecting it in partner- ship with my husband/wife. The Problem • Good pizza starts with good ingredients • Supply chains are traditionally hard to build and maintain in Ethiopia. • Good cheese, good flour are the building blocks and here in Ethiopia, the wheat has low levels of gluten and the cows are not producing consistent and safe milk to make cheese. • Also, much of the society, fasts from animal products for 155 days a year, so our product offering must accommodate these needs. The Opportunity • There is a longstanding history of eating pizza and pasta, so the palate is already in my culture. • People are beginning to understand the importance of high quality, consistent products and we have a consistent clientele. • The inputs for our pizza come directly from Dawit’s farm and we are among the first farm to table restaurants in Ethiopia. • We also see an opportunity to utilize the delivery services and are exploring a take and bake model similar to what you might find in Europe or America. The Business Model • Our model focuses on high quality ingredients and consistent products that are locally sourced. We have one restaurant in CMC where we have perfected our products. We have eat-in and take out services and last year our revenue was X. 31 The Ask • To expand The Ethiopia Pizza Company we’re seeking $50K USD, or 1.75 M ETB. • The investment will help us expand our farm and open a second location in Bole. • Our goal is to eventually have 5 locations in Addis Ababa, and then consider expanding to other cities throughout the country. Closing • Investors – it’s time Ethiopia experienced the high quality homegrown pizza it deserves. • The Ethiopia Pizza Company will deliver on that promise. • We welcome you to join us on this pizza adventure. BUSINESS SCRIPT 2: TEXTILES (Gender Neutral Sector) Opening • Thank you investors for giving me the opportunity to pitch my businesses to you today • My name is Dagmawit XXX and I am the founder of The Textile Trainer – “your one-stop for trained and skilled textile factory workers in Ethiopia.” • As a little (boy/girl) I had a dream – that my country would be known around the world for beauty and excellence • When the textile industry started to shift from Southeast Asia to Africa – I knew my calling! • Help make Ethiopia the focal point for the global textile industry for the next 50 years! The Problem • As many know, textiles move to where labor is cheap. • Cheap labor drives the industry. • But with cheap comes problems – problems that take a decade to fix and by that time the industry has developed a tarnished brand for how they treat their low skilled workers • As a result, the industry is dominated by large international firms, that work through a shadow network of large companies that mistreat local employees and, when the world finally discovers, packs up shop and moves to another country and do it all over again • Local textile firms also feel the pain from this – when the larger firms move out they lose their customers. Or for the smaller cut and sew shops they can never compete with the larger companies so they are forced to take small orders, pay high input costs and are always on the sidelines. Local firms simply cannot compete! 32 The Opportunity • Here in Ethiopia the textile infrastructure – transportation, supply chain, and man- power – is still in the early stages of development relative to the eastern countries; • And I see problems on the horizon if someone does not enter the space and begin building the soft infrastructure for the industry to flourish. • Therefore, we see an opportunity. • One of the major problems local, small textile outfit face is the high cost of quality raw materials that we need to import from overseas to meet the specifications of inter- national buyers; even small-batch buyers. So they normally lose those bids, or if they win small bids their margins are horrible. • Local firms have to wait in long queues for LCs and then when we do get access to forex, we don’t have the negotiating power with supplier to get a good price – not like the major international. • For larger multinational companies that are coming to Ethiopia they struggle with high turnover – they have many sewing lines dedicated to training – and as soon as a worker is trained, they leave because they either get pregnant or are needed at home • Is there a way to bridge these two problems? • We say, YES WE CAN! Our Solution • May I introduce you to The Textile Trainer – “your one-stop for trained and skilled textile factory workers.” • “Triple T” as we call it, partners with international multinational brands that are looking to set up operations in Ethiopia; and we’re their soft-landing manager! • We hire and begin training every type of worker to have a turn-key workforce ready for them when they come – “from the guard to the line manager.” • Our trainers will travel to international factories before they come, to learn and un- derstand their needs. • We will also play a role in convincing them to come to Ethiopia – as we will be proactive in reaching out to these companies. • Then, we will ask that they train our trainers on every role they need to make a factory work – the low skilled roles. • We then assume the burden of recruiting, training and then placing the new workers – the janitors, guards, managers, cooks, sewers, purchasers, etc. in the company once they arrive. 33 • “Everyone but the management level technical staff, Triple T handles!” • We also do cross cultural training and teach incoming managers about Ethiopians culture, and vice versa. Our trained Ethiopian cooks will know how to prepare Turk- ish food for Turkish companies; Chinese food for Chinese companies. The Vietnam manager will have a secretary that can greet him in Vietnamese! The Business Model • Here’s how we make money – our business model consists of three diversified revenue streams: 1. Outsourced training fees charged to multinationals – they pay us to get their workforce ready. 2. Revenue generated from small-batch orders – likewise, in the agreements we sign with multinationals we also require that they provide a certain number of small batch orders to local manufacturers so we help develop the cottage industry into a larger and skilled local industry; but the MNCs need to order our raw mate- rials along with their orders so we leverage the full power of their strength with suppliers and banks. 3. Placement fees - 1 month of pay after 6 months of service - for every worker we train and place at a company. • As you can see, we combine the best models from a training, textile and recruitment firm, all in one! The Ask • To expand Triple T we’re seeking $50K USD, or 1.75M ETB • With this • We’ll rent a training center in Hawassa. • Purchase machines and a dedicated training room with skilled local trainers • We will also develop marketing materials and work with the textile association to run an event with the large multinationals to showcase our services to them. • We are modeling sales of 200M ETB by year 5 and gross margins of 60% and a net margin of 35%. Closing • Everyone needs a soft landing when entering a new country. • We want “Triple T” to be that soft-landing provider for the huge textile industry that we see is coming to Ethiopia, and help spark a vibrant, ethical and wonderful industry here – to not only create jobs but to bring the beauty of Ethiopia around the world • Thank you for your time. 34 BUSINESS SCRIPT 3: CONSTRUCTION (Male-Dominated Sector) Opening • Thank you investors for giving me the opportunity to pitch my businesses to you today. • My name is Dagmawit XXX and I am the founder of The Ethiopia Construction Company – “the fastest road construction company in Africa.” • Ethiopia is one of the fastest industrializing countries in the world. • And we need roads to get our people, goods and services from the center heartland of our country to the coasts and cities to drive growth and create prosperity • Roads are our future. • And I want The Ethiopia Construction Company to pave the way to prosperity for Ethiopia. The Problem • As you know, Ethiopia has one of the most underdeveloped road systems in the world. • As the 12th largest population, and one of the fastest growing countries in the world, there’s a lot of potential for road-driven-growth (no pun intended). • But the road industry is fraught with corruption. • Many people win tenders, do the bare minimum and purchase inferior raw materials that are hard to identify to the untrained eye, but fall apart in months, long after the contractor has fled with the profits. The Opportunity • That’s why we believe there is an opportunity to design a new type of road construction company. • One that puts safety at the center. • Our model is what makes us unique. • First, we will have a fully transparent billing process – only do cost plus – and we will post our contracts and have a live video and pictures on our website for the general population and the government to see. • Then we will get a profit payment based on the number of road accidents we prevent on our roads – as measured by the national average, using that as the benchmark – • And also get a bonus on the total cost of ownership of the road – meaning we will get paid a fee to maintain the road at a fixed price per year. 35 • If we build a good road from the start, that price will be much lower than the national average per kilometer, • Thus, safety and quality excellence are how we will make our money. The Business Model • The business model is pretty straightforward. We will do cost plus bids and be known for our transparency. • We will provide updates on our report The Ask • To expand The Ethiopia Construction Company, we’re seeking $50K USD, or 1.75M ETB. • The investment will be used to buy machines and raw material, develop marketing material and for training our team. • We are estimating that our initial 18 months will be break even. Closing • Investors – it’s time we put safety and quality at the center stage of our road system. • We need a new model to do this – one that incentivizes the right outcomes. • We believe The Ethiopia Construction Company will build Ethiopia and make it a model for countries around the world. • We welcome you to join us on this innovative new business that is going to transform corrupt and old practice and make the construction industry great again. • Thank you for considering our company for an investment. ROLE MODEL ADD-ON SCRIPT • My name is XXX (gender-neutral name) and I am the CEO of COMPANY. • You may know our company; it is one of the largest textile companies in Ethiopia. • We started our business more than 15 years ago with only 10,000 ETB of capital. • It was challenging at first but because of hard work, dedication and the grace of God, we made it. • I wanted to share a few things I have learned about entrepreneurship over the years. • It takes perseverance, focus and passion. 36 • Perseverance to not give up – you need to put in the time and learn the ropes to be successful. I had this business idea from when I was a teenager and against the advice of my parents who wanted me to work in a more stable job, I pursued it because I believed in the vision. I received help from mentors along the way to get to this position. • Focus – many things will take your attention. Without a plan, written down that you are committed to and you will lose focus. But with focus your energy can be channeled and you will get results. • Passion – blended with faith, passion can become your engine to get you up early before everyone else and working hard through the day to come out successful. • This is why I believe we have been so successful, and I hope others can learn from my experience. • My COMPANY now employs more than 200 employees and the future looks bright despite the challenges we have faced in the past year. Figure A2: Women Entrepreneur Pitching a Business Idea during a Chigign Tobiya Filming Notes: In the main format of the television show, the four tycoons make an investment decision after the pitch. In the gender discrimination experiment we ask viewers to report on whether and how much they think the tycoons should invest. To measure gender bias we compare across groups of viewers who are randomized to view videos of men and women pitching the same business idea; and explore effects by sector. 37