Breaking glass ceilings in Colombia: Strategies and outcomes in efforts to narrow the gender gap in educational leadership

Elacqua, Gregory; Perez Nunez, Graciela; Cubillos, Pedro; Iglesias, Juliana Working Paper Breaking glass ceilings in Colombia: Strategies and outcomes in efforts to narrow the gender gap in educational leadership IDB Working Paper Series, No. IDB-WP-01678 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Elacqua, Gregory; Perez Nunez, Graciela; Cubillos, Pedro; Iglesias, Juliana (2025) : Breaking glass ceilings in Colombia: Strategies and outcomes in efforts to narrow the gender gap in educational leadership, IDB Working Paper Series, No. IDB-WP-01678, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0013404 This Version is available at: https://hdl.handle.net/10419/315925 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ Breaking Glass Ceilings in Colombia: Strategies and Outcomes in Efforts to Narrow the Gender Gap in Educational Leadership Gregory Elacqua Graciela Perez Nunez Pedro Cubillos Juliana Iglesias WORKING PAPER No IDB-WP-01678 Inter-American Development Bank Education Division January 2025 Breaking Glass Ceilings in Colombia: Strategies and Outcomes in Efforts to Narrow the Gender Gap in Educational Leadership Gregory Elacqua Graciela Perez Nunez Pedro Cubillos Juliana Iglesias Inter-American Development Bank Education Division January 2025 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Breaking glass ceilings in Colombia: strategies and outcomes in efforts to narrow the gender gap in educational leadership / Gregory Elacqua, Graciela Pérez-Núñez, Juliana Iglesias, Pedro Cubillos. p. cm. — (IDB Working Papers Series ; 1678) Includes bibliographic references. 1. Educational leadership-Sex differences-Colombia. 2. Women-RecruitingColombia. 3. Teachers-Selection and appointment. I. Elacqua, Gregory M., 1972II. Pérez-Nuñez, Graciela. III. Iglesias Velasco, Juliana. IV. Cubillos, Pedro. V. Inter-American Development Bank. Education Division. VI. Serie. IDB-WP-1678 Jel Codes: I21, I24, J71, J16, M51, O15 Keywords: Gender Gap, Educational Leadership, School Management, Recruitment Process, Diversity in Leadership, Latin America, Colombia http://www.iadb.org Copyright © 2025 Inter-American Development Bank ("IDB"). This work is subject to a Creative Commons license CC BY 3.0 IGO (https://creativecommons.org/licenses/by/3.0/igo/legalcode). The terms and conditions indicated in the URL link must be met and the respective recognition must be granted to the IDB. Further to section 8 of the above license, any mediation relating to disputes arising under such license shall be conducted in accordance with the WIPO Mediation Rules. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the United Nations Commission on International Trade Law (UNCITRAL) rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this license. Note that the URL link includes terms and conditions that are an integral part of this license. The opinions expressed in this work are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Breaking Glass Ceilings in Colombia: Strategies and Outcomes in Efforts to Narrow the Gender Gap in Educational Leadership* Gregory Elacqua†Graciela Pérez Núñez‡Pedro Cubillos§Juliana Iglesias¶ January 2025 Abstract In Colombia, women represent 65% of the teacher workforce but only 34% of school principals, reflecting a significant gender gap in leadership. This study examines two centralized principal selection processes implemented by Colombia’s National Civil Service Commission: the 2016 nationwide process and the 2018 process targeting disadvantaged PDET regions (Development Programs with a Territorial Focus). Both processes evaluated candidates through standardized tests, minimum requirements, and assessments of education and experience, determining eligibility for leadership vacancies. Our descriptive analysis shows how selection criteria influence gender representation. In 2016, standardized testing dominated, resulting in 45% of applicants being women but only 20% qualifying, with an overall eligible-to-vacancy ratio of just 0.7%. In contrast, the 2018 PDET process prioritized context-specific competencies and practical experience, yielding 35% female eligibility despite women comprising only 38% of applicants (likely due to challenging conditions in PDET regions). Moreover, eligible candidates of both genders outnumbered vacancies by 4.5 times. These findings underscore the critical role of selection design in shaping gender representation in school leadership. However, structural barriers, such as inadequate childcare and rigid work schedules, persist as obstacles to women’s participation. JEL Classification: I21, I24, J71, J16, M51, O15 Keywords: gender gap, educational leadership, school management, recruitment process, diversity in leadership, Latin America, Colombia *We gratefully acknowledge funding from the Inter-American Development Bank. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. The authors have no conflicts of interest or financial or material interests in the results. All errors are our own. †Inter-American Development Bank. E-mail: [email protected]. ‡International Institute for Education Planning, UNESCO. E-mail: [email protected]. §Department of Economics, University College London. E-mail: pedr[email protected] ¶UNICEF-Colombia. E-mail: [email protected] 1 INTRODUCTION 1 Introduction The gender gap in leadership roles remains a global challenge across sectors. In Latin America and the Caribbean (LAC), women hold only 29% of congressional seats, 25% of ministerial positions, 20% of corporate board seats, and 30% of positions on governing boards of multilateral organizations (Gonzalez and Ibanez,March 8, 2024). While these disparities are well-documented, less attention has been paid to the gender gap in school leadership (Bergmann et al.,2022;Wang and Gao,2022). Globally, women comprise 68.3% of the teaching workforce in OECD countries but hold only 47.3% of school leadership positions (OECD,2019). This disparity is even more pronounced in LAC, where in 2013, eight out of 15 countries reported at least a 20 percentage point gap between female teachers and principals (Adelman and Lemos,2021). While countries such as Argentina, Brazil, and Peru have made strides in narrowing this gap, it remains a significant issue in others, including Colombia. In Colombia, where women constitute 65% of the teacher workforce, only 34% of school principals are female (Elacqua et al., 2024), underscoring the pressing need for targeted efforts to address gender inequality in educational leadership. School principals play a pivotal role in shaping student outcomes, second only to teachers in terms of their impact (UNESCO,2018;Grissom, Egalite, Lindsay et al.,2021; Branch et al.,2012;Coelli and Green,2012;Muñoz and Prem,2024;Hallinger and Heck, 1996;Heck et al.,1990;Marks and Printy,2003;Pounder et al.,1995;Waters et al., 2003;Zheng et al.,2017).1Principals’ responsibilities span curriculum implementation, teacher development, student disciplinary policies, financial management, and fostering a positive culture. By setting high expectations and guiding teachers, staff, students, and families, principals create environments conducive to effective learning (DarlingHammond et al.,2022;Leithwood et al.,2004;Ten Bruggencate et al.,2012;Lemos et al., 2021;Bloom et al.,2015). Female leadership provides broader institutional and societal benefits. In schools, greater gender balance fosters inclusive practices in teacher recruitment, retention, and 1Note that empirical evidence on principals’ impact on student performance yields mixed results, with some studies finding no significant effects (Bartanen et al.,2024;Ten Bruggencate et al.,2012). 2 1 INTRODUCTION diversity promotion (Husain et al.,2018;Grissom et al.,2012;Bartanen and Grissom, 2021;Branch et al.,2012;Campos-García and Zúñiga-Vicente,2019). Female leaders often adopt participative leadership styles, encouraging collaboration and inclusion (Eagly and Karau,2002;Shaked et al.,2018;Conto et al.,2023), and they serve as role models, positively influencing the aspirations and achievements of female students (Beaman et al.,2012;Bettinger and Long,2005;Carrell et al.,2010;Dee,2005;Lim and Meer,2017;Paredes,2014). However, challenges such as inadequate childcare, inflexible work policies, and gender biases recruitment practices continue to limit women’s access to leadership positions (Grissom, Timmer, Nelson and Blissett,2021;Bailes and Guthery, 2020;Angelov et al.,2016;Bertrand et al.,2010;Bertrand,2018;Biasi and Sarsons,2022; Blau and Kahn,2017;Buser et al.,2014;Carrell et al.,2010;Howe-Walsh and Turnbull, 2016;Wang and Gao,2022). Recruitment processes play a critical role in shaping school leadership outcomes. Centralized systems, like Colombia’s, promote equity by using competency-based selection tools to enhance transparency and mitigate discrimination (Bertrand and Mullainathan, 2004;Neumark et al.,1996;Muñoz and Prem,2024;Oreopoulos,2011;Riach and Rich, 2002). However, such systems are rare in education globally. In many regions, including LAC, principals are typically selected at the local level, where subjective criteria, such as experience and political connections, often outweigh merit (Ramachandram et al.,2017; Elacqua et al.,2021;Aravena,2020). Colombia’s centralized, competency-based recruitment approach stands out for its transparency and fairness, yet it can unintentionally reinforce disparities when standardized tests are heavily weighted, as women may experience test anxiety or face structural disadvantages in test preparation (Arias et al., 2023;Azmat et al.,2016;Cai et al.,2019;Jurajda and Münich,2011). This paper examines how institutional design impacts gender representation in educational leadership for coordinators, rural directors, and school principals in Colombia.2 It compares two centralized selection processes: the 2016 nationwide process, which heavily weighted standardized testing, and the 2018 process, designed specifically for regions affected by armed conflict, poverty, and limited government presence—known 2In Colombia, rural directors are principals of a single, typically small school in a remote area, while school principals lead either a single large school or a network of schools, often spanning rural and urban sites. A 2001 reform merged some schools into networks, while others, particularly rural schools, remained independent. Research indicates that multi-site schools achieve similar test scores but have lower dropout rates compared to independent schools (Elacqua and Santos,2020). 3 1 INTRODUCTION as Development Programs with a Territorial Focus (Programas de Desarrollo con Enfoque Territorial - PDET). To better address the unique challenges in PDET regions, the 2018 process emphasized practical competencies and contextual experience, while reducing the weight of the high-stakes standardized exam used in 2016. Although a direct causal comparison between the two processes is not feasible, descriptive data suggests that the 2018 process narrowed the gender gap in eligibility and reduced the influence of socioeconomic factors. Our findings reveal that the design of the 2016 selection process created significant barriers for women pursuing school leadership roles. The selection instruments exacerbated disparities: only 0.3% of female applicants were deemed eligible for leadership roles, compared to 1% of male applicants and an overall eligibility rate of 0.7% across all candidates. Women scored lower on the basic test, particularly in the numeric and functional components, reducing their likelihood of advancing beyond the eliminatory phase. Regression analysis further confirms these patterns, showing that only 16% of leadership vacancies were ultimately filled by women. The 2018 selection process achieved notable progress despite challenging conditions in PDET regions. Female participation dropped to 38%, down from 45% in 2016, likely due to due to the remote and resource-scarce nature of PDET areas. Nevertheless, 19% of female applicants qualified for leadership roles, compared to 21% of male applicants and an overall eligibility rate of 20%. Moreover, the total number of eligible candidates in 2018 far exceeded available vacancies by 4.5 times, with 1,787 eligible candidates competing for 405 positions–a stark improvement over 2016, which yielded only 133 eligible candidates for 1,098 vacancies. Further analysis of the 2016 selection process reveals persistent gender disparities even among eligible candidates. Women who qualified for leadership positions tended to select roles other than principal (e.g., coordinator), likely due to the greater work flexibility these positions offered. Conversely, principal positions were predominantly occupied by men. Additionally, women who attained school leadership roles were more frequently assigned to urban schools in less impoverished areas with lower risks of victimization—a pattern consistently observed across all candidates. 4 2 THE COLOMBIAN SCHOOL MANAGER SELECTION SYSTEM The remainder of this paper is organized as follows. Section 2 describes Colombia’s school manager selection system and the features of the 2016 and 2018 PDET processes. Section 3 presents the data and empirical approach. Section 4 discusses the results, and Section 5 concludes with policy implications. 2 The Colombian School Manager Selection System Between 1979 and 2002, both the selection and promotion of school staff in Colombia occurred through direct appointment by local governors and mayors. This changed significantly in 2002 when the National Civil Service Commission (Comisión Nacional de Servicio Civil - CNSC) implemented a centralized public competition system for filling school manager and teacher positions. This shift was meant to promote transparency, meritocracy, and equal opportunity, breaking away from regionally influenced practices. The new system was a reaction to a long-standing issue, where selection processes favored personal and political interests over merit—a common problem when decisions rely on committees, juries, or local authorities (Aravena,2020;Donoso-Díaz et al.,2019; Soto Arango,2013). To manage the selection process, the CNSC established the Equal Merit and Opportunity Support System (Sistema de Apoyo para la Igualdad de Mérito y la Oportunidad - SIMO). This platform streamlines the application process by making information publicly accessible to candidates and providing transparency on available positions through the Public Offering of Career Jobs (Oferta Pública de Empleos de Carrera - OPEC). Additionally, the CNSC established call agreements with each certified territorial entity, requiring detailed information on vacancies. These legally binding and publicly available agreements outline the selection process and specified positions available for rural directors, coordinators, and principals. If positions remained unfilled after the competition, local authorities can make temporary appointments. This centralized approach differs substantially from those used in other Latin American countries. While Colombia uses transparent evaluations and clear merit-based criteria to ensure an equitable selection process for all candidates, other countries in the region employ different models, ranging from democratic to politicized, each with vary5 3 DATA graduate degree (master’s or doctorate), while in the 2018 process there was a modest increase in participants with STEM backgrounds. Despite these two centralized selection processes, data from 2023 shows a persistent underrepresentation of women within school management roles: women make up 65% of the teaching workforce but account for just 43% of managerial positions (Figure 2). This discrepancy highlights the need for greater female participation in centralized selection processes and the importance of addressing the structural barriers that hinder women’s career advancement in education. Figure 2: Share of Teachers and Managers by Gender in 2023 Our 2016 selection process analysis examines three dimensions: contest outcome, employment status in the school labor market, and characteristics of the municipality home to the management position.12 The more limited 2018 PDET selection process data al12The first dimension concerns scores from the basic test and its four components (numerical, verbal, pedagogical, and functional), the psycho-technical test, and the likelihood of candidates passing the two key elimination stages (i.e., the basic test and minimum requirement verification). The labor market dimension captures whether individuals secured a position, using binary variables to indicate if a candidate held a school management role or a specific position such as rural director, coordinator, or principal, distinguishing between interim and permanent appointments. We also analyze salaries using the logarithm of salaries. Finally,the third dimension regards the attributes of municipalities where positions exist, where we use binary variables to indicate whether jobs exist in the poorest or least poor quartile and their level of victimization risk (high, medium-high, low). Additional variables include the ethnic population percentage, whether the school is in a rural area, and the logarithm of GDP 12 4 RESULTS lows to assess contest outcomes, including scores from the basic and psychometric tests and the likelihood of candidates passing the basic test and meeting minimum requirements. We also employ an additional analytical approach to evaluate the 2018 contest outcomes and to conduct a descriptive comparison with the 2016 contest results. 4 Results 4.1 Nationwide 2016 Selection Process for School Managers This section explores which factors shaped outcomes in the 2016 selection process, focusing on candidates’ performance across stages and components and their progression through each phase. We focus on the process’s design and assessment tools, particularly the basic test, which disproportionately hindered female candidates’ advancement. An analysis of test scores and pass rates, differentiated by gender, reveals significant disparities. Women scored lower on the general basic test, especially on the numeric and functional components, but outperformed male candidates on the pedagogical component.13 These performance disparities created gaps in passing rates for the basic test, with only 0.5% of women advancing compared to 1.4% of men (Table 1). Table 1: Descriptive Statistics of the 2016 Selection Process by Gender Male Female Diff. Variables (1) (2) (2)−(1) Basic test score 51.62 49.10 −2.516∗∗∗ Numeric component score 53.96 43.86 −10.10∗∗∗ Verbal component score 58.09 58.18 0.090 Pedagogical component score 46.89 47.19 0.293∗ Functional component score 49.58 48.15 −1.431∗∗∗ Psycho-technical test score 49.28 49.49 0.206∗ Final score 67.71 66.99 −0.720 Probability of Passing Basic test 0.014 0.005 −0.009∗∗∗ Probability of Meeting Minimum Requirements 0.010 0.003 −0.007∗∗∗ Observations 11,061 8,946 − Note: School managerial positions require candidates to achieve a minimum score of 70 out of 100 points on the Basic test. ∗p<0.10,∗∗ p<0.05,∗∗∗ p<0.01. per capita (sourced from Acevedo and Bornacelly Olivella (2014) and compiled by the Center for Economic Development Studies [Centro de Estudios sobre Desarrollo Económico - CEDE]; Link). 13For a graphical distribution of the basic test components by gender, see Figure 4. 13 4 RESULTS We estimate a series of regressions controlling for relevant variables to assess the performance of applicants. Column (1) of Table 2shows that female candidates scored 0.33 standard deviations lower than their male counterparts on the basic test (extended analysis in Table A5). The likelihood of women passing the basic test and minimum requirements was 0.9 and 0.7 percentage points lower respectively (columns (7) and (8)). These reductions translate to discrepancies of 90% and 106% relative to baseline pass rates of 1% and 0.7% for these stages, indicating substantial barriers disproportionately affecting women, especially at the basic test stage. This disparity exists despite data showing no significant gender differences in education levels (Table A6). Table 2: Effects of Gender on Outcomes in the 2016 School Manager Selection Process Basic test components Approval probability Basic test Psycho-technical Ver. of min. score test score Numeric Verbal Pedagogical Functional Basic test requirements (1) (2) (3) (4) (5) (6) (7) (8) Female -0.325*** 0.025* -0.513*** -0.051*** -0.021 -0.185*** -0.009*** -0.007*** (0.013) (0.014) (0.013) (0.014) (0.014) (0.014) (0.001) (0.001) Covariates Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,007 20,007 20,007 20,007 20,007 20,007 20,007 20,007 R-Squared 0.18 0.01 0.22 0.11 0.08 0.04 0.01 0.01 Note: Observations include all individuals who participated in the various tests of the competition. Columns 1 and 2 show standardized scores for the basic test and the psycho-technical test. The components of the basic test—numerical, verbal, pedagogical knowledge, and functional knowledge—are presented in columns 3 to 6 and are also standardized. Columns 7 and 8 display the probability of passing each stage of the competition. Covariates include ethnicity, age, highest level of education achieved, STEM major status, and residency characteristics such as rurality, poverty, and victimization risk indexes. ∗p<0.10,∗∗ p<0.05,∗∗∗ p<0.01. A more detailed analysis of the basic test reveals that women performed less well across multiple components, including numerical and verbal aptitudes, and basic or functional competencies. The gap is widest for the numerical component, even after adjusting for socioeconomic background and academic qualification. This disparity in part reflects the different educational paths chosen by female participants, who are less likely to have pursued STEM-related undergraduate programs compared to their male counterparts (38% of women vs. 49% of men). In addition, historical data shows that women generally score lower than men on aptitude tests across educational levels, from primary through high school (Arredondo et al.,2019;Bernal and Bernal,2016;Gelber et al.,2016). Results from the past seven years of the Saber 11 test, required for higher education admission in Colombia, reveal that women consistently 14 4 RESULTS score lower than men in both mathematics and overall performance. A similar pattern appears on the Saber PRO test, a mandatory college exit exam.14 It has been argued that this performance gap is due to higher levels of test anxiety among women in high-stakes exams, particularly in mathematics—a factor heightened in competitive settings (Arias et al.,2023;Azmat et al.,2016;Cai et al.,2019;Jurajda and Münich,2011). The selection process outcomes show women initially comprising 43.6% of applicants, yet only 24% of this group passed, compared to 76% of male applicants. This suggests that in placing significant emphasis on generic competencies such as numerical and verbal skills, the basic test put female candidates at a disadvantage. Notably, the basic test is modeled after the Saber Pro test, thus indicating broader issues with educational quality and underscoring the learning and performative gaps faced by women in Colombia. 4.1.1 Effects on Labor Market Outcomes Data on the applicants in the 2016 selection process also allows to examine labor market outcomes for school managers in 2023.15 Specifically, we assess gender differences in school management positions and explore how candidates deemed eligible for leadership roles have performed. The variable Eligible serves as an explanatory factor. Table 3investigates the likelihood of securing a managerial position, distinguishing between permanent and temporary appointments (detailed results in Table A7). A permanent position implies having passed the competitive selection process and successfully completed the probationary period, and thus being full integrated into the education managerial profession, including associated career rights.16 14For more details on gender trends in Saber 11 and Saber PRO results, see Figures 5,6, and 7. 15Data constraints limit labor market data to 2023, and prevent matching this data with the 2018 selection process. 16In passing the selection process, choosing a vacancy, and receiving an appointment to that position through a public hearing, the job remains temporary for a probationary period. Successful evaluation of this period secures the position, failure results in termination. "Provisional" and "temporary" appointments designate interim positions that remain open until filled through the formal selection process. These temporary appointments provide continuity in teaching and management roles until permanent staff are selected. 15 4 RESULTS Table 3: Effects of Gender on Labor Market Outcomes in the 2016 School Manager Selection Process School Manager Rural director Coordinator Principal School Manager Permanent Temporary Permanent Temporary Permanent Temporary Permanent Temporary Log(Salary) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Eligible 0.647*** 0.648*** -0.001 -0.010 -0.000 0.243*** -0.000 0.415*** -0.000 0.075*** (0.038) (0.038) (0.003) (0.010) (0.001) (0.033) (0.002) (0.022) (0.001) (0.022) Female -0.067*** -0.066*** -0.001* -0.005*** -0.000 -0.020*** -0.000 -0.041*** -0.000* 0.018** (0.006) (0.006) (0.000) (0.001) (0.000) (0.005) (0.000) (0.003) (0.000) (0.007) Female ×Eligible -0.037 -0.038 0.001 0.005 0.000 0.216*** 0.001 -0.259*** 0.000 -0.049 (0.086) (0.086) (0.006) (0.022) (0.002) (0.074) (0.005) (0.050) (0.003) (0.051) Covariables Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,007 20,007 20,007 20,007 20,007 20,007 20,007 20,007 20,007 3,978 R-Squared 0.07 0.07 0.00 0.00 0.00 0.03 0.00 0.04 0.00 0.28 Notes: Observations include all individuals who participated in the various tests of the competition, except for column 10, which is estimated only for individuals working as principals in 2023. Covariates include ethnicity, age, highest level of education achieved, STEM major status, and residency characteristics such as rurality, poverty, and victimization risk indexes. ∗p<0.10,∗∗ p<0.05,∗∗∗ p<0.01. Eligible individuals (0.7% of all participants) had a substantially higher likelihood (64.8 percentage points) of securing permanent leadership roles in schools. This advantage includes an increase of 24.3 percentage points for coordinator roles and 41.5 percentage points for principal positions, with no impact on the likelihood of becoming a rural director. Eligible candidates earned a 7.5 percentage point higher salary. Among female candidates, women had a 7 percentage point lower chance of attaining permanent leadership positions across all roles. The largest disparity appears for the role of principal, with a 4.1 percentage point gap, followed by that of coordinator at 2 percentage points, and rural directors at 0.5 percentage points. However, women in leadership roles earned 1.8 percentage points more than their male counterparts, suggesting that the women who do secure these positions tend to have higher skill levels. While gender, combined with eligibility, did not significantly impact the overall likelihood of becoming a manager, columns 6 and 8 reveal that eligible women are more likely to secure coordinator roles than principal positions compared to men. This finding indicates potential structural barriers that limit women’s advancement to principal roles, possibly due to preferences for jobs with greater flexibility and fewer working hours. 16 4 RESULTS 4.1.2 Effects on Geographical Outcomes We further explore gender disparities within school management by examining the characteristics of schools where managers are employed. These characteristics are inferred from municipal-level variables, with the exception of school rurality, which comes from the labor market dataset. We analyze how these dynamics shift based on candidates’ eligibility status in the 2016 selection process. Figure 3shows that female school managers work in municipalities with lower levels of poverty and victimization risk compared to their male counterparts. Men are more likely to work in areas with higher poverty and victimization levels. On average, both genders chose positions in municipalities with lower poverty and victimization risks. However, female candidates’ choices showed greater variability, likely due to their lower representation among those who qualified as eligible for leadership roles in the 2016 selection process. The ethnic composition of municipalities where school managers work showed no significant differences, across both the general sample of principals and those who passed the 2016 selection process. However, women work less in rural schools. Eligibility status influenced the preference for urban positions, as eligible men and women assumed leadership roles in urban rather than rural settings. 17 4 RESULTS Figure 3: Characteristics of Principal’s School Location by Gender in the 2016 School Principal Selection Process 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (A): Mid-High Multidim. Poverty 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (B): Mid-Low Multidim. Poverty 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (C): Mid-High Victimization Risk 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (D): Mid-Low Victimization Risk 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (E): % Ethnic 0 .2 .4 .6 .8 1 Share All Eligible from contest Male Female (F): Rural Area 4.2 Selection Process for School Managers in PDET Territories in 2018 We extend the analytical approach from Section 4.1 to analyze the PDET 2018 selection process. As described above, this process introduced substantial modifications designed to recruit applicants to the hard-to-staff PDET territories. It addressed the unique demands of these areas by prioritizing relevant competencies and knowledge and incorporating a test format that included case studies reflective of the PDET context. Unlike the 2016 process, which emphasized numerical abilities, the 2018 process prioritized experience and awarded additional points to candidates affected by conflict or originating from the PDET territories. Table 4reveals no significant gender disparities in basic test scores, indicating equal opportunities in passing this initial stage. However, women scored lower on the psychotechnical test, leading to lower overall scores. A two-percentage-point gap in meeting minimum requirements emerged, largely attributable to differences in experience and female candidates’ qualifications. This shift in selection criteria for the 2018 PDET pro18 4 RESULTS cess reflects an effort to improve candidate suitability for positions in these challenging regions. Table 4: Descriptive Statistics of the PDET 2018 Selection Process by Gender Male Female Diff. Variables (1) (2) (2)-(1) Basic test score 63.75 63.43 -0.314 Psycho-technical test score 64.90 64.32 -0.580∗∗ Final score 59.12 57.79 -1.327∗∗∗ Probability of Passing Basic test 0.282 0.281 -0.001 Probability of Meeting Minimun Requirements 0.211 0.191 -0.020∗∗ Observations 5,477 3,290 - Note: School managerial positions require candidates to achieve a minimum score of 70 out of 100 points on the basic test. ∗p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01. We estimate a series of regressions using standardized scores on the basic and psychotechnical tests, and the probability of passing the elimination stages (basic test and minimum requirements verification) as dependent variables. These estimates control for confounding factors, consistent with the 2016 analyses. Table 5supports the previous results, showing that women scored 0.06 standard deviations lower in both assessed areas. They were also 2.4 percentage points less likely to meet the minimum requirements (see Table A8 for an extended analysis). Unlike the 2016 process, the 2018 PDET contest presented fewer barriers for women, suggesting that adjustments to the selection criteria and process design contributed to narrowing the gender gap in school leadership selection. 19 4 RESULTS Table 5: Effects of Gender on Outcomes in the PDET 2018 School Manager Selection Process Approval probability Basic test Psycho-technical Ver. of min. score test score Basic test requirements (1) (2) (3) (4) Female -0.063*** -0.064*** -0.011 -0.024*** (0.022) (0.022) (0.010) (0.009) Covariables Yes Yes Yes Yes Observations 8,767 8,767 8,767 8,767 R-Squared 0.07 0.03 0.04 0.03 Note: Observations include all individuals who took the various tests of the competition. Columns 1 and 2 show standardized scores for the basic test and the psycho-technical test. Columns 3 and 4 display the probability of passing each stage of the competition. Covariates include ethnicity, age, highest level of education achieved, STEM major status, and residency characteristics such as rurality, poverty, and victimization risk indexes. ∗ p<0.10,∗∗ p<0.05,∗∗∗ p<0.01. Focusing on candidates who qualified as eligible, we compare applicant characteristics between the 2016 selection process and the 2018 PDET competition. In Table A9, we observe minimal gender disparities between the two contests. However, the gender gap in progression through the competition stages was smaller in the PDET 2018 process compared to that of 2016. The 2018 contest also saw a larger representation of participants from economically disadvantaged municipalities with higher rates of victimization, in line with the focus on PDET territories where candidates had higher chances of qualifying as eligible for leadership roles. While individuals holding a postgraduate degree more frequently qualified in both years, this factor was less decisive in 2018 as the proportion of postgraduates remains consistent from registration to eligibility qualification. This descriptive comparison between the 2016 and 2018 contests highlights how selection process design can influence the profile of qualifying candidates, in this case increasing the representation of women and individuals from disadvantaged regions. The 2018 basic test contributed to this outcome by emphasizing competencies more relevant to rural contexts, providing a more equitable evaluation. Notably, the PDET process also filled more vacancies: 16% filled in 2016 compared to 86% in 2018, leading to more women securing permanent positions in 2018. Because the 2018 PDET process targeted specific regions with distinct characteristics, direct causal comparisons with the 2016 nationwide process are not possible. Nonetheless, our analysis demonstrates how selec20 5 DISCUSSION tion criteria can affect gender representation in leadership roles. The findings suggest that carefully designed selection processes may help reduce gender gaps in educational leadership, offering important considerations for future policy design. 5 Discussion Ensuring women’s representation in leadership positions is critical for advancing gender equity and fostering inclusive development across all sectors. Research highlights their significant contributions to societal outcomes, particularly in the provision of public goods such as infrastructure, education, and health. Women leaders are also less likely to engage in corrupt practices (Baskaran and Hessami,2023;Bhalotra and Clots-Figueras,2014;Brollo and Troiano,2016;Chattopadhyay and Duflo,2004;ClotsFigueras,2012). In education, female leaders play a pivotal role in creating collaborative environments, implementing inclusive policies, and serving as crucial role models (Bartanen and Grissom,2021;Branch et al.,2012;Eagly and Karau,2002;Shaked et al.,2018; Xu and Yao,2015). These benefits extend to student outcomes, with evidence showing that female leadership positively influences academic achievement and aspirations, particularly for female students (Bettinger and Long,2005;Carrell et al.,2010;Dee,2005; Lim and Meer,2017;Paredes,2014;Xu and Yao,2015). Despite these documented benefits, gender gaps persist across sectors in Latin America and the Caribbean (LAC), including education. Women occupy less than one-third of leadership roles in government, corporate boards, and multilateral organizations (Gonzalez and Ibanez,March 8, 2024). In Colombia, women make up 65% of teachers but only 34% of school principals, underscoring structural barriers that limit their advancement. These include inadequate childcare, inflexible work schedules, male-dominated leadership networks, and geographic mobility constraints—particularly for roles in remote areas. Additionally, recruitment and selection processes may inadvertently favor male candidates, perpetuating existing inequities. Our analysis of two centralized school manager selection processes in Colombia highlights how institutional design can either reinforce or help mitigate these barriers. The 2016 nationwide process illustrates how seemingly neutral criteria, such as an overre21 REFERENCES Admissions to Czech universities, American Economic Review 101(3): 514–518. Leithwood, K., Seashore, K., Anderson, S. and Wahlstrom, K. (2004). Review of research: How leadership influences student learning, Technical report, University of Minnesota, Center for Applied Research and Educational Improvement. Lemos, R., Muralidharan, K. and Scur, D. (2021). Personnel management and school productivity: Evidence from India, Technical Report Working Paper 28336, National Bureau of Economic Research. Lim, J. and Meer, J. (2017). The impact of teacher–student gender matches, Journal of Human Resources 52(4): 979–997. Marks, H. M. and Printy, S. M. (2003). Principal leadership and school performance: An integration of transformational and instructional leadership, Educational Administration Quarterly 39(3): 370–397. Muñoz, P. and Prem, M. (2024). Managers’ productivity and recruitment in the public sector, American Economic Journal: Economic Policy 16(4): 223–253. Neumark, D., Bank, R. J. and Van Nort, K. D. (1996). Sex discrimination in restaurant hiring: An audit study, The Quarterly Journal of Economics 111: 915–941. OECD (2019). TALIS 2018 results (Volume I): Teachers and school leaders as lifelong learners. URL: https://doi.org/10.1787/1d0bc92a-en Oreopoulos, P. (2011). Why do skilled immigrants struggle in the labor market? a field experiment with thirteen thousand resumes, American Economic Journal: Economic Policy 3: 148–171. Paredes, V. (2014). A teacher like me or a student like me? role model versus teacher bias effect, Economics of Education Review 39: 38–49. Pounder, D. G., Ogawa, R. T. and Adams, E. A. (1995). Leadership as an organizationwide phenomena: Its impact on school performance, Educational Administration Quarterly 31(4): 564–588. Ramachandram, V., Beteille, T., Linden, T. and Dey, S. (2017). Getting the right teachers into the right schools: Managing India’s teacher workforce, Technical report, The World Bank. Riach, P. A. and Rich, J. (2002). Field experiments of discrimination in the market place, The Economic Journal 112. Shaked, H., Glanz, J. and Gross, Z. (2018). Gender differences in instructional leadership: 28 REFERENCES How male and female principals perform their instructional leadership role, School Leadership & Management 38(4): 417–434. Soto Arango, D. E. (2013). Legislación e imaginarios sociales en el escalafón y los salarios de los educadores de primaria en Colombia. 1952-1994, Revista Historia de la Educación Latinoamericana 15(21): 229–262. Ten Bruggencate, G., Luyten, H., Scheerens, J. and Sleegers, P. (2012). Modeling the influence of school leaders on student achievement: How can school leaders make a difference?, Educational Administration Quarterly 48(4): 699–732. UNESCO (2018). Activating policy levers for Education 2030: The untapped potential of governance, school leadership, and monitoring and evaluation policies, UNESCO, Paris. URL: https://unesdoc.unesco.org/ark:/48223/pf0000265951 Wang, T. and Gao, S. (2022). Systematic review of research on women and educational leadership in mainland China, Hong Kong and Taiwan, Educational Management Administration & Leadership 0(0): 17411432221142150. Waters, T., Marzano, R. J. and McNulty, B. (2003). Balanced Leadership: What 30 Years of Research Tells Us about the Effect of Leadership on Student Achievement, Mid-Continent Research for Education and Learning, Aurora, CO. Weinstein, J., Muñoz, G. and Flessa, J. (2019). Liderazgo directivo para la calidad de la educación: aprendizajes desde un campo de investigación emergente, Calidad en la educación (51): 10–14. Xu, Y. and Yao, Y. (2015). Informal institutions, collective action, and public investment in rural China, American Political Science Review 109(2): 371–391. Zheng, Q., Li, L., Chen, H. and Loeb, S. (2017). What aspects of principal leadership are most highly correlated with school outcomes in China?, Educational Administration Quarterly 53(3): 409–447. 29 A Appendix A: Socioeconomic Characterization of Municipalities To capture applicants’ socioeconomic background, we used a set of variables associated with their municipality of birth as a proxy. First, we employed 2018 data from the National Administrative Department of Statistics (DANE) to determine the probability that an individual belongs to an ethnic group, based on whether they were born in a municipality where 70% or more of the population identifies as such (e.g., Indigenous, Afro-Colombian, Romani, or Palenquero).17 We also use the 2018 DANE data to characterize each municipality by its multidimensional poverty level.18 This index, which ranges from 0 to 1, comprises five dimensions and 15 indicators, each contributing to a municipality’s score. We divided the total sample of Colombian municipalities (1,122 administrative entities) into quartiles based on multidimensional poverty levels, ranging from least poor to most poor. To approximate the impact of armed conflict on applicants’ birth municipality, we used 2022 data on victimization risk from the Victims Unit (Unidad para las Víctimas).19 This index provides insight on the threat level, victimization, and vulnerability experienced within each municipality. We incorporated all of this information into the data processing for the school manager selection contest, creating a comprehensive dataset of 1,122 municipalities, each associated with its DANE code and relevant socioeconomic variables. We then linked this dataset to each individual’s municipality and department of birth. 17See the DANE Geoportal, Geovisor de Autorreconocimiento Étnico. 18The multidimensional poverty measure includes 15 indicators spanning five dimensions: education, health, employment, housing, and childhood, capturing factors such as literacy, school attendance, early childhood care, health insurance, and access to improved water sources. See Multidimensional poverty measure - dane.gov.co 19The victimization risk index is based on three dimensions: threat (presence of armed groups and incidents affecting civilians), victimization (restrictions on movement, personal safety, and life security), and vulnerability (corruption, demographics, geography, institutional and socioeconomic factors). See IRV (unidadvictimas.gov.co). 30 A Modifications in the Design of the PDET 2018 Selection Process In the 2016 nationwide selection process, priority was placed on educational training and advancement in the evaluation of candidate’s backgrounds, while work experience was given secondary importance. Candidates’ scores increased with higher levels of education, and additional points were awarded to those with qualifications in education as opposed to other fields.20 When comparing the weight assigned to education versus work experience—which emphasized school management—the former could result in an accumulated 70 points, compared to just 30 points for the latter. In calculating the final score, different percentage weights were also allocated to each test or assessment stage, which in turn determined the list of candidates deemed eligible for leadership roles. These weights were as follows: the basic test, 55%; psycho-technical test, 15%; background assessment, 20%; and interview, 10%. This structure emphasized general abilities and knowledge, with great value placed on education and postgraduate qualifications in the educational field. Several modifications were made to the 2018 PDET selection process that altered the desired candidate profile. Firstly, the basic test focused more on competencies and knowledge directly related to the position, including situational judgment questions for scenarios specific to PDET territories. Secondly, the background assessment placed greater emphasis on work experience, assigning the latter a maximum of 70 points, while education was capped at 30 points. Experience in conflict-affected and community-based settings was particularly valued, indicating a preference for candidates with the skills to effectively manage schools in challenging contexts. Points were also awarded to candidates with a history of being affected by conflict or strong local ties. Thirdly, the interview stage was eliminated, and the weighting of the different stages accordingly reallocated: the basic test was reduced to 45%, the background assessment doubled to 40%, and the psycho-technical test remained at 15%. This reallocation placed greater value on candidates with the necessarily skills to succeed in areas affected by deprivation and armed conflict. Additionally, a policy was introduced to encourage the retention of selected candidates in these regions: applicants appointed to permanent positions could 20For example, a candidate with a teaching degree and a master’s in education could receive up to 20 points, while a candidate with equivalent qualifications in a different field would receive only half those points. 31 A only request transfers to other PDET territories, thereby limiting mobility. Table A1: Assessment Stages in the School Manager Selection Process Across Selected Latin American and Caribbean Countries Tests and assessment moments Argentina Brazil Chile Colombia Ecuador Mexico Peru Dom. Rep. Minimum requirements and/or background assessment X X X X X X X X Technical test X X X X X X X Psycho-technical test X X Interview or colloquium X X X X Case study and/or project solution X X X X Probation period X X Source: Own elaboration. Table A2: Vacancies for School Management Positions in CNSC Selection Processes for the Special Teaching Career System Year Position 2006 2009 2012 2012 bis 2016 2018 2021 2006-2021 Coordinator 1,197 1,695 1,126 173 572 138 1,676 6,577 Rural Director 270 577 55 9 159 94 98 1,262 Principal 324 652 577 49 367 173 931 3,073 Total 1,791 2,924 1,758 231 1,098 405 2,705 10,912 Source: Author’s elaboration based on call agreements for each selection process. Note: The 2012 bis contest refers to a special ethnic-based process conducted that year for AfroColombian, Raizal, and Palenquero ethnic groups. 32 A Table A3: Descriptive Statistics of Covariates in the 2016 School Manager Selection Process Variable Obs. Mean Std. Dev. Min. Max. Female 20,007 0.447 0.497 0 1 Ethnic group 20,007 0.057 0.232 0 1 Age 20,007 42.6 7.846 18 67 Rurality 20,007 0.12 0.325 0 1 Multidimensional poverty index 1st quartile 20,007 0.478 0.500 0 1 2nd quartile 20,007 0.184 0.387 0 1 3rd quartile 20,007 0.200 0.400 0 1 4th quartile 20,007 0.138 0.345 0 1 Risk of victimization index Low 20,007 0.254 0.435 0 1 Medium-Low 20,007 0.439 0.496 0 1 Medium 20,007 0.188 0.391 0 1 Medium-High 20,007 0.089 0.285 0 1 High 20,007 0.029 0.169 0 1 Highest level of education High School/Technician 20,007 0.001 0.032 0 1 Normal School 20,007 0.004 0.067 0 1 Bachelor’s/Undergraduate 20,007 0.250 0.433 0 1 Specialization 20,007 0.409 0.492 0 1 Postgraduate 20,007 0.336 0.472 0 1 STEM Program 20,007 0.445 0.497 0 1 33 A Table A4: Descriptive Statistics of Covariates in the 2018 PDET School Manager Selection Process Variable Obs. Mean Std. Dev. Min. Max. Female 8,767 0.375 0.484 0 1 Ethnic group 8,767 0.078 0.267 0 1 Age 8,767 41.5 8.353 19 68 Rurality 8,767 0.13 0.336 0 1 Multidimensional poverty index 1st quartile 8,767 0.435 0.496 0 1 2nd quartile 8,767 0.178 0.382 0 1 3rd quartile 8,767 0.215 0.411 0 1 4th quartile 8,767 0.172 0.377 0 1 Risk of victimization index Low 8,767 0.145 0.352 0 1 Medium-low 8,767 0.424 0.494 0 1 Medium 8,767 0.228 0.42 0 1 Medium-High 8,767 0.146 0.353 0 1 High 8,767 0.057 0.233 0 1 Highest level of education High School/Technician 8,767 0.005 0.073 0 1 Normal School 8,767 0.019 0.135 0 1 Bachelor’s/Associate’s 8,767 0.408 0.491 0 1 Specialization 8,767 0.293 0.455 0 1 Postgraduate 8,767 0.275 0.447 0 1 STEM Program 8,767 0.493 0.500 0 1 Figure 4: Score Distribution by Basic Test Component and Gender 0 .005 .01 .015 .02 .025 Density 0 20 40 60 80 100 Numeric component score Male Female 0 .01 .02 .03 Density 0 20 40 60 80 100 Verbal component score Male Female 0 .01 .02 .03 .04 Density 0 20 40 60 80 100 Pedagogical component score Male Female 0 .01 .02 .03 .04 .05 Density 0 20 40 60 80 Functional component score Male Female 34 A Figure 5: Trends in Saber 11 Test Results by Gender 240 245 250 255 260 265 Global Score 2016 2017 2018 2019 2020 2021 2022 Year Female Male 49 50 51 52 53 Math Score 2016 2017 2018 2019 2020 2021 2022 Year Female Male Figure 6: Results by Gender on Saber 11 and Saber PRO Tests 46 48 50 52 54 Male Female Math score percentile Saber 11 - 2022 46 48 50 52 54 Male Female Global score percentile Saber 11 - 2022 35 A Figure 7: Results by Gender in Saber 11 and Saber PRO Tests in the Field of Education 40 45 50 55 Male Female Quant. score percentile Area: Education | Saber PRO - 2022 40 45 50 55 Male Female Global score percentile Area: Education | Saber PRO - 2022 36 A Table A5: Test Results and Probability of Passing Stages in the 2016 Selection Process Basic test components Approval probability Basic test Psycho-technical Ver. of min. score test score Numeric Verbal Pedagogical Functional Basic test requirements (1) (2) (3) (4) (5) (6) (7) (8) Female -0.325*** 0.025* -0.513*** -0.051*** -0.021 -0.185*** -0.009*** -0.007*** (0.013) (0.014) (0.013) (0.014) (0.014) (0.014) (0.001) (0.001) Ethnic group -0.160*** -0.002 -0.077** -0.136*** -0.168*** -0.094*** 0.001 0.001 (0.031) (0.034) (0.030) (0.032) (0.033) (0.034) (0.003) (0.003) Age -0.023*** 0.002** -0.024*** -0.021*** -0.014*** -0.004*** -0.001*** -0.000*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.000) (0.000) Rurality 0.100*** -0.097*** 0.118*** 0.086*** 0.065*** 0.008 0.000 -0.000 (0.022) (0.024) (0.021) (0.022) (0.023) (0.023) (0.002) (0.002) Multidimensional Poverty Index 2nd Quartile -0.134*** 0.018 -0.123*** -0.151*** -0.104*** -0.012 -0.002 0.000 (0.018) (0.020) (0.017) (0.019) (0.019) (0.019) (0.002) (0.002) 3rd Quartile -0.296*** -0.018 -0.172*** -0.304*** -0.227*** -0.155*** -0.006*** -0.003* (0.018) (0.020) (0.018) (0.019) (0.020) (0.020) (0.002) (0.002) 4th Quartile -0.343*** 0.016 -0.259*** -0.369*** -0.244*** -0.115*** -0.005* -0.003 (0.024) (0.026) (0.023) (0.024) (0.025) (0.025) (0.003) (0.002) Victimization Risk Index Medium-Low 0.008 0.038** -0.029* 0.037** 0.020 0.011 0.002 0.002 (0.016) (0.018) (0.016) (0.017) (0.017) (0.017) (0.002) (0.001) Medium -0.034* -0.012 -0.000 -0.029 -0.064*** -0.018 0.002 0.002 (0.020) (0.022) (0.019) (0.021) (0.021) (0.021) (0.002) (0.002) Medium-High -0.035 0.022 -0.022 -0.024 -0.050* -0.009 0.001 0.002 (0.028) (0.031) (0.027) (0.029) (0.030) (0.030) (0.003) (0.002) High -0.074* 0.107** -0.110*** -0.040 -0.074* 0.017 -0.001 0.002 (0.041) (0.046) (0.040) (0.043) (0.044) (0.045) (0.005) (0.004) Formal education Normal School 0.685*** 0.120 0.488** 0.364 0.580** 0.486** 0.010 -0.001 (0.224) (0.247) (0.219) (0.233) (0.238) (0.242) (0.025) (0.020) Bachelor’s/Undergraduate 0.633*** 0.208 0.348* 0.497** 0.507** 0.458** 0.004 0.003 (0.203) (0.223) (0.198) (0.211) (0.215) (0.219) (0.022) (0.018) Specialization 0.893*** 0.293 0.522*** 0.657*** 0.664*** 0.682*** 0.005 0.004 (0.203) (0.223) (0.198) (0.211) (0.215) (0.219) (0.022) (0.018) Postgraduate 1.295*** 0.412* 0.823*** 0.977*** 0.980*** 0.885*** 0.018 0.012 (0.203) (0.223) (0.198) (0.211) (0.215) (0.219) (0.022) (0.018) STEM Program 0.153*** -0.009 0.448*** -0.022 -0.047*** -0.067*** 0.007*** 0.005*** (0.013) (0.014) (0.013) (0.014) (0.014) (0.014) (0.001) (0.001) Constant 0.228 -0.426* 0.563*** 0.357* 0.009 -0.333 0.024 0.016 (0.207) (0.227) (0.201) (0.215) (0.219) (0.223) (0.023) (0.018) Observations 20,007 20,007 20,007 20,007 20,007 20,007 20,007 20,007 R-Squared 0.18 0.01 0.22 0.11 0.08 0.04 0.01 0.01 Note: Observations include all individuals who participated in the various tests of the competition. Columns 1 and 2 show standardized scores for the basic test and the psychotechnical test. The components of the basic test—numerical, verbal, pedagogical knowledge, and functional knowledge—are presented in columns 3 to 6 and are also standardized. Columns 7 and 8 display the probability of passing each stage of the competition. ∗p<0.10,∗∗ p<0.05,∗∗∗ p<0.01. 37