Introduction

The dynamics of academic impact and the determinants of scholarly success are central topics in the study of science and innovation systems. The measurement of academic performance through bibliometric indicators (citation counts, number of papers, and the H-index) has become a cornerstone of how research productivity and scholarly influence are evaluated. These metrics are widely used in decisions on promotion, funding and prestige, and can shape academic careers in significant ways (Bornmann and Daniel, 2007[4] Bornmann, L., & Daniel, H. D. (2007). What do we know about the h-index? Journal of the American Society for Information Science and Technology, 58(9), 1381-1385. https://doi.org/10.1002/asi.20609
; Ellison, 2013[12] Ellison, G. (2013). How does the market use citation data? The Hirsch index in Economics. American Economic Journal: Applied Economics, 5(3), 63-90. https://doi.org/10.1257/app.5.3.63
; Huang et al., 2020[21] Huang, J., Gates, A. J., Sinatra, R., & Barabási, A.-L. (2020). Historical comparison of gender inequality in scientific careers across countries and disciplines. Proceedings of the National Academy of Sciences, 117(9), 4609-4616. https://doi.org/10.1073/pnas.1914221117
; Krauss et al., 2023[24] Krauss, A., Danús, L., & Sales-Pardo, M. (2023). Early-career factors largely determine the future impact of prominent researchers: Evidence across eight scientific fields. Scientific Reports, 13, 18794.https://doi.org/10.1038/s41598-023-46050-x
). Despite their pervasiveness, the factors behind the markedly divergent career trajectories of Mexican researchers remain understudied, particularly those related to coauthorship networks and implicit gender differences.

This paper studies the gender citation gap among researchers affiliated with Mexico's National System of Researchers (SNI), the centralized governmental program that since 1984 evaluates and rewards scientific productivity through a four-rung ladder of recognition (Candidate, Levels I-III). The SNI compensation can represent up to half of a researcher's wage and conditions access to evaluation committees, grant funding and academic visibility (Gonzalez-Brambila & Veloso, 2007[16] Gonzalez-Brambila, C., & Veloso, F. M. (2007). The determinants of research output and impact: A study of Mexican researchers. Research Policy, 36(7), 1035-1051. https://doi.org/10.1016/j.respol.2007.03.005
; Sandoval-Romero & Larivière, 2020[35] Sandoval-Romero, V., & Larivière, V. (2020). The national system of researchers in Mexico: Implications of publication incentives for researchers in social sciences. Scientometrics, 122(1), 99-126. https://doi.org/10.1007/s11192-019-03285-8
; Galaz Fontes et al., 2020[13] Galaz Fontes, J. F., Martínez Stack, J., & Gil Antón, M. (2020). The emergence of the new Mexican academic meritocracy. Higher Education Governance and Policy, 1(2), 138-151. https://dergipark.org.tr/en/pub/hegp/article/827043
). SNI offers a particular empirical setting in which to study how observable productivity translates into recognition, for three reasons. First, it is a centralized, national system of periodic evaluation: certified committees organized by area of knowledge apply publicly defined criteria to every applicant in their area, irrespective of the applicant's home institution, in contrast with the decentralized, institution-by-institution promotion systems that dominate the international literature on gender gaps in academia (CONACYT, 2020, Arts. 11-12[6] CONACYT. (2020, 7 de agosto). Reglamento del Sistema Nacional de Investigadores (Acuerdo 03-07/2020 de la Junta de Gobierno del CONACYT, 3.ª sesión ordinaria 2020). Diario Oficial de la Federación. https://sidof.segob.gob.mx/notas/docFuente/5600871
). Second, SNI membership comes with a periodic monetary transfer calculated by formula and tied to the level assigned in the evaluation; previous work documents that this transfer can represent up to half of a member's salary, which means that any gap in bibliometric visibility translates into a relevant income gap (Sandoval-Romero & Larivière, 2020[35] Sandoval-Romero, V., & Larivière, V. (2020). The national system of researchers in Mexico: Implications of publication incentives for researchers in social sciences. Scientometrics, 122(1), 99-126. https://doi.org/10.1007/s11192-019-03285-8
; Dorantes-Gilardi et al., 2022[10] Dorantes-Gilardi, R., Ramírez-Álvarez, A. A., & Terrazas-Santamaría, D. (2022). The role of highly intercited papers on scientific impact: The Mexican case. Applied Network Science, 7(1), Artículo 58. https://doi.org/10.1007/s41109-022-00497-5
). Third, in practice the evaluation weighs indexed bibliometric output heavily, directly linking the metrics we study (Scopus citations) to effective institutional recognition. The SNI thus makes the connection from published output to symbolic and monetary recognition which is very transparent and measurable. Moreover, the magnitude of the phenomenon is not marginal in the Mexican case: women remain underrepresented in senior academic ranks, particularly in STEM fields and in higher SNI levels (Dorantes-Gilardi et al., 2022[10] Dorantes-Gilardi, R., Ramírez-Álvarez, A. A., & Terrazas-Santamaría, D. (2022). The role of highly intercited papers on scientific impact: The Mexican case. Applied Network Science, 7(1), Artículo 58. https://doi.org/10.1007/s41109-022-00497-5
; Rivera León et al., 2017[33] Rivera León, L., Mairesse, J., & Cowan, R. (2017). Gender gaps and scientific productivity in middle-income countries: Evidence from Mexico (IDB Working Paper Series No. IDB-WP-800). Inter-American Development Bank. https://doi.org/10.18235/0011791
), and the magnitude and sources of the citation gap have not been systematically documented within the SNI.

We focus on three disciplines, Economics, Engineering and Health Sciences, which differ substantially along three relevant margins. First, in female representation: Engineering has the lowest female share in the SNI (about 22%), Economics is intermediate (around 34%), and Health Sciences is female-majoritarian (about 55%). Second, in publication conventions: the three disciplines disseminate knowledge mainly through peer-reviewed journal articles (NSB, 2023[30] National Science Board. (2023). Publications output: U.S. trends and international comparisons (NSB-2023-33). National Science Foundation. https://ncses.nsf.gov/pubs/nsb202333
; Savage and Olejniczak, 2022[36] Savage, W. E., & Olejniczak, A. J. (2022). More journal articles and fewer books: Publication practices in the social sciences in the 2010s. PLOS ONE, 17(2), e0263410. https://doi.org/10.1371/journal.pone.0263410
), which are likely to be indexed in Scopus, providing comparable bibliometric records. Third, in collaboration patterns, which are highly relevant for citation accumulation: research produced by larger and more diverse teams tends to receive more citations (Larivière et al., 2015[25] Larivière, V., Sugimoto, C. R., Tsou, A., & Gingras, Y. (2015). Team size matters: Collaboration and scientific impact since 1900. Journal of the Association for Information Science and Technology, 66(7), 1323-1332. https://doi.org/10.1002/asi.23266
; van der Wal et al., 2021[40] Van der Wal, J. E. M., Thorogood, R., & Horrocks, N. P. C. (2021). Collaboration enhances career progression in academic science, especially for female researchers. Proceedings of the Royal Society B: Biological Sciences, 288(1958), Artículo 20210219. https://doi.org/10.1098/rspb.2021.0219
), and collaboration with prominent coauthors yields lasting visibility benefits (Li et al., 2019[27] Li, W., Aste, T., Caccioli, F., & Livan, G. (2019). Early coauthorship with top scientists predicts success in academic careers. Nature Communications, 10(1), artículo 5170. https://doi.org/10.1038/s41467-019-13130-4
; Dorantes-Gilardi et al., 2023[11] Dorantes-Gilardi, R., Ramírez-Álvarez, A. A., & Terrazas-Santamaría, D. (2023). Is there a differentiated gender effect of collaboration with super-cited authors? Evidence from junior researchers in Economics. Scientometrics, 128(4), 2317-2336. https://doi.org/10.1007/s11192-023-04656-y
; Krauss et al., 2023[24] Krauss, A., Danús, L., & Sales-Pardo, M. (2023). Early-career factors largely determine the future impact of prominent researchers: Evidence across eight scientific fields. Scientific Reports, 13, 18794.https://doi.org/10.1038/s41598-023-46050-x
). These differences make discipline-specific decomposition particularly informative.

Methodologically, we apply an Oaxaca-Blinder (OB) decomposition (Oaxaca, 1973[31] Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review, 14(3), 693-709. https://doi.org/10.2307/2525981
; Oaxaca and Ransom, 1994[32] Oaxaca, R. L., & Ransom, M. R. (1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61(1), 5-21. https://doi.org/10.1016/0304-4076(94)90074-8
; Jann, 2008[22] Jann, B. (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal, 8(4), 453-479. https://doi.org/10.1177/1536867X0800800401
) to log citations, splitting the raw gap into an endowments component (differences in observable characteristics) and a coefficients component (differences in returns to those characteristics). We pair Scopus-based citation data with information on coauthor prominence, team size, publication output and academic age. The outcome and right-tailed covariates are winsorized at the 99th percentile within the discipline to limit the influence of extreme outliers, given the well-documented skewness of bibliometric distributions (Seglen, 1992[37] Seglen, P. O. (1992). The skewness of science. Journal of the American Society for Information Science, 43(9), 628-638. https://doi.org/10.1002/(SICI)1097-4571(199210)43:9<628::AID-ASI5>3.0.CO;2-0
; Aksnes et al., 2019[1] Aksnes, D. W., Langfeldt, L., & Wouters, P. (2019). Citations, citation indicators, and research quality: An overview of basic concepts and theories. SAGE Open, 9(1), 1-17. https://doi.org/10.1177/2158244019829575
). We report robustness to the no-winsorization alternative in Appendix C (row 2). Following Criscuolo et al. (2019)[7] Criscuolo, C., Martin, R., Overman, H. G. y Van Reenen, J. (2019). Some causal effects of an industrial policy. American Economic Review, 109(1), 48-85. https://doi.org/10.1257/aer.20160034
and MaCurdy and Pencavel (1986)[28] MaCurdy, T. E., & Pencavel, J. H. (1986). Testing between competing models of wage and employment determination in unionized markets. Journal of Political Economy, 94(3, Part 2), S3-S39. https://doi.org/10.1086/261398
, variables are transformed as log(x+1) to retain zero-valued observations (17%-37% of the outcome variable observations).

We find substantial, discipline-specific gender gaps. In Economics and Health Sciences, women receive approximately 23% and 21% fewer citations than men, respectively, and in both cases the gap is more than fully explained by endowments, primarily lower publication output and shorter academic careers. In Engineering, where the raw gap is statistically indistinguishable from zero, women have weaker endowments but higher returns in each observable characteristic, which offsets the difference. Because the unexplained component is small (and statistically non-significant in Economics), our preferred reading of the results is that the documented citation gap reflects gaps in the underlying academic trajectories of men and women, rather than differential citation dynamics per se. We confirm the robustness of these findings to alternative transformations of the dependent variable (asinh), to dropping the (potentially endogenous) maximum coauthor H-index, to including academic age in levels with a quadratic term, to excluding the top 1% of citations and publications, to using the researcher's H-index as the dependent variable, to a Heckman correction for selection into Scopus (Appendix B), and to stratification by SNI level (Appendix C).

Our contribution is threefold. Substantively, we provide the first discipline-specific Oaxaca-Blinder decomposition of the citation gap among SNI researchers, and we document that this gap operates mainly through endowments. Conceptually, we clarify that what is often labelled as a “citation gap” is, in our setting, primarily a gap in academic trajectories; this distinction has direct implications for policy design. Methodologically, we follow current bibliometric standards in dealing with the heavy-tailed nature of citation data and document the implications of common practices through extensively checking robustness.

The paper proceeds as follows. Section 1 reviews relevant literature. Section 2 describes the data, the construction of the sample, and the empirical strategy. Section 3 presents the results. Finally, the conclusions.

1. Related literature

Coauthorship in academia has become standard practice, facilitated by advances in travel and telecommunications. Thelwall and Maflahi (2022)[39] Thelwall, M., & Maflahi, N. (2022). Research coauthorship 1900-2020: Continuous, universal, and ongoing expansion. Quantitative Science Studies, 3(2), 331-344. https://doi.org/10.1162/qss_a_00188
find that the average number of authors per article has been in ongoing expansion across all 27 broad Scopus fields. The composition of coauthorship in a team plays a substantive role in researchers' trajectories: complementary expertise can be pooled, and association with established collaborators raises long-term visibility (Li et al., 2019[27] Li, W., Aste, T., Caccioli, F., & Livan, G. (2019). Early coauthorship with top scientists predicts success in academic careers. Nature Communications, 10(1), artículo 5170. https://doi.org/10.1038/s41467-019-13130-4
; Dorantes-Gilardi et al., 2023[11] Dorantes-Gilardi, R., Ramírez-Álvarez, A. A., & Terrazas-Santamaría, D. (2023). Is there a differentiated gender effect of collaboration with super-cited authors? Evidence from junior researchers in Economics. Scientometrics, 128(4), 2317-2336. https://doi.org/10.1007/s11192-023-04656-y
; Krauss et al., 2023[24] Krauss, A., Danús, L., & Sales-Pardo, M. (2023). Early-career factors largely determine the future impact of prominent researchers: Evidence across eight scientific fields. Scientific Reports, 13, 18794.https://doi.org/10.1038/s41598-023-46050-x
). In coauthored work, major findings tend to be attributed to better-known researchers, while less established collaborators may be overlooked in the short run, although they may benefit from associations in the long run, a dynamic consistent with Merton's (1968)[29] Merton, R. K. (1968). The Matthew effect in science. Science, 159(3810), 56-62. https://doi.org/10.1126/science.159.3810.56
Matthew effect of cumulative advantage in science. DiPrete and Eirich (2006)[8] DiPrete, T. A., & Eirich, G. M. (2006). Cumulative advantage as a mechanism for inequality: A review of theoretical and empirical developments. Annual Review of Sociology, 32, 271-297. https://doi.org/10.1146/annurev.soc.32.061604.123127
review the cumulative advantage tradition in sociology and show how repeated small inequalities compound into large differences over time.

A standard metric to assess a researcher's prestige is the total citation count, together with citation-based composite indices such as the H-index (Bornmann and Daniel, 2007[4] Bornmann, L., & Daniel, H. D. (2007). What do we know about the h-index? Journal of the American Society for Information Science and Technology, 58(9), 1381-1385. https://doi.org/10.1002/asi.20609
). Total citations capture the extent to which a researcher's work has been used in the scientific community, while the H-index balances productivity and impact by counting how many publications have reached a minimum threshold of citations. The H-index was proposed by Hirsch (2005)[20] Hirsch, J. E. (2005). An index to quantify an individual's scientific research output. Proceedings of the National Academy of Sciences, 102(46), 16569-16572. https://doi.org/10.1073/pnas.0507656102
; Bornmann and Daniel (2007)[4] Bornmann, L., & Daniel, H. D. (2007). What do we know about the h-index? Journal of the American Society for Information Science and Technology, 58(9), 1381-1385. https://doi.org/10.1002/asi.20609
show that it rewards sustained above-average impact rather than a few highly cited papers, providing a perspective distinct from total citation counts. Because the H-index summarizes the breadth and depth of an academic career in a single number, we use the maximum H-index of a researcher's coauthors as a proxy for the prominence of her collaboration network.

A persistent finding in the literature is that women in academia receive fewer citations than men. This disparity is shaped not only by citation practices but also by differences in research productivity and career trajectories. Huang et al. (2020)[21] Huang, J., Gates, A. J., Sinatra, R., & Barabási, A.-L. (2020). Historical comparison of gender inequality in scientific careers across countries and disciplines. Proceedings of the National Academy of Sciences, 117(9), 4609-4616. https://doi.org/10.1073/pnas.1914221117
show that men typically experience longer and more continuous research careers, while women are more likely to exit the discipline at multiple stages, contributing to differences in citation accumulation across countries and disciplines. Spoon et al. (2023)[38] Spoon, K., LaBerge, N., Wapman, K. H., Zhang, S., Morgan, A. C., Galesic, M., Fosdick, B. K., Larremore, D. B., & Clauset, A. (2023). Gender and retention patterns among U.S. faculty. Science Advances, 9(42), Artículo eadi2205. https://doi.org/10.1126/sciadv.adi2205
document that, in U.S. faculty, women are 6-19% more likely to leave academia than men at each career stage. King et al. (2017)[23] King, M. M., Bergstrom, C. T., Correll, S. J., Jacquet, J., & West, J. D. (2017). Men set their own cites high: Gender and self-citation across fields and over time. Socius: Sociological Research for a Dynamic World, 3, 1-22. https://doi.org/10.1177/2378023117738903
show that men self-cite more frequently than women, inflating their visible citation metrics, and in male-dominated fields the gap may be amplified by within-group citation behaviour (Dion et al., 2018[9] Dion, M. L., Sumner, J. L., & Mitchell, S. M. (2018). Gendered citation patterns across political science and social science methodology fields. Political Analysis, 26(3), 312-327. https://doi.org/10.1017/pan.2018.12
). Hengel (2022)[18] Hengel, E. (2022). Publishing while female: Are women held to higher standards? Evidence from peer review. The Economic Journal, 132(648), 2951-2991. https://doi.org/10.1093/ej/ueac032
finds that women economists are held to higher standards in peer review and spend more time revising their work, mechanically reducing their publication output. Ghiasi et al. (2015)[14] Ghiasi, G., Larivière, V., & Sugimoto, C. R. (2015). On the compliance of women engineers with a gendered scientific system. PLOS ONE, 10(12), Artículo e0145931. https://doi.org/10.1371/journal.pone.0145931
document that women engineers tend to publish in higher-impact journals than their male counterparts, which may compensate for lower publication volume.

A separate strand of literature studies what happens to gender disparities when discrimination is made salient. Hensvik and Nilsson (2025)[19] Hensvik, L., & Nilsson, P. (2025). Equality for granted: What happens when discrimination in academia becomes salient? (CEPR Discussion Paper No. DP20210). Centre for Economic Policy Research. https://cepr.org/publications/dp20210
exploit the public disclosure of substantial male bias in the competence assessments of newly minted PhDs applying for grants from the Swedish Medical Research Council and document that, post-disclosure, male-only review committees were rapidly phased out, reviewers' decision processes changed, the average male bias was eliminated, and the long-run research output of grantees assigned to previously biased committees increased by 27 percent of a standard deviation. The Mexican SNI, with its centralized evaluation system, offers a comparable institutional setting where the visibility of gender gaps in citations could plausibly affect future evaluation practices.

A growing body of work has applied the Oaxaca-Blinder framework to outcomes beyond wages, including educational attainment, health, and academic productivity. The interpretation of the coefficients component as “discrimination” is contested, since it absorbs both differential returns to observed characteristics and the contribution of any unobserved characteristic that is correlated with the group indicator (Oaxaca and Ransom, 1994[32] Oaxaca, R. L., & Ransom, M. R. (1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61(1), 5-21. https://doi.org/10.1016/0304-4076(94)90074-8
; Jann, 2008[22] Jann, B. (2008). The Blinder-Oaxaca decomposition for linear regression models. The Stata Journal, 8(4), 453-479. https://doi.org/10.1177/1536867X0800800401
). In our setting, potential unobservables include institutional prestige, journal quality, international collaboration intensity, and access to research funding, none of which we observe directly. We are therefore careful to interpret the unexplained component as the joint contribution of differential returns and omitted variables, rather than as a direct measure of discrimination.

2. Data and methodology

3. Results

Before turning to the three discipline-specific results, a brief note on how the role of coauthors enters the analysis. The number of coauthors appears in this paper in three distinct ways, each answering a different question: (i) in Table 2 we report the unconditional male-female difference in cumulative coauthors, which is significant in Engineering (p=0.009) and Health Sciences (p=0.001) but not in Economics; (ii) in Tables 3-5 we report the contribution of the endowment in Ln(Coauthors+1) to the male-female gap in log citations, conditional on publications, max coauthor H-index and academic age - this contribution is statistically significant only in Health Sciences (0.061**, SE 0.030); and (iii) in Table 6 we report the OLS coefficient of Ln(Coauthors+1) in the citations equation by gender and discipline, which is statistically significant only in Health Sciences (0.18* for males, 0.24** for females). The three results are mutually consistent: a sizeable raw gap in coauthors translates into a significant explainable component of the citation gap only when, conditional on the other covariates, coauthors actually predict citations - which in our sample occurs only in Health Sciences. Differences in coauthor counts in Engineering, although significant in raw terms, do not propagate into the citation gap because it is conditional on publications, the elasticity of citations with respect to coauthors is statistically indistinguishable from zero in that discipline (Table 6)

Conclusions

This paper documents the gender citation gap among researchers affiliated with Mexico's National System of Researchers in three disciplines: Economics, Engineering and Health Sciences. Using a Oaxaca-Blinder decomposition on Scopus-based citation records for 2,089 SNI researchers active in 2021, we find substantial discipline-specific gaps in Economics and Health Sciences (about 23% and 21%, respectively) that are more than fully explained by career opportunities, chiefly fewer publications and shorter careers for women. In Engineering, the raw gap is statistically indistinguishable from zero, but this null result reflects offsetting forces: women have weaker endowments and higher returns to those endowments.

These results are robust to alternative transformations of the dependent variable (inverse hyperbolic sine), to removing the winsorization, to including academic age in levels with a quadratic term, to dropping the maximum coauthor H-index, to using the researcher's own H-index as the outcome, to excluding the top 1% of citations and publications, to stratifying by SNI level, and to a Heckman correction for selection into Scopus.

We frame our findings in terms of gaps in academic trajectories rather than discrimination in the assignment of citations per se. The unexplained component of the decomposition, small and mostly non-significant, is the part of the gap that could in principle reflect differential returns or unobserved factors. In our setting, plausible omitted variables include institutional prestige, journal quality, international collaboration intensity, and access to research funding, none of which we observe directly. We are therefore careful not to interpret the unexplained component as a measure of citation discrimination.

The reading of the gap as a gap in trajectories has direct policy implications. In Economics and Health Sciences, where the citation disadvantage for women is primarily attributable to endowments, policies should focus on expanding research opportunities and supporting longer career trajectories: targeted grant schemes, structured mentoring programs, and family-friendly tenure and promotion policies. In Engineering, where women already obtain higher returns to each observable characteristic, the structural priority is recruitment and retention at early career stages, given that women in the field do not face a penalty in how their characteristics translate into impact.

Several limitations of our design should be acknowledged. First, the sample is restricted to SNI researchers identifiable in Scopus, which excludes researchers whose work circulates primarily in non-Scopus-indexed outlets. The Heckman correction reported in Appendix B addresses this concern econometrically but cannot fully eliminate the possibility of selection on unobservables. Importantly, this correction addresses selection into observation in Scopus, not publication bias at the article level: we observe citations only to published papers, and a fuller treatment would require data on unpublished or rejected work, which is not available for all SNI researchers. Second, our analysis is cross-sectional and based on cumulative variables, which limits our ability to distinguish cohort effects from career-length effects. Future work using longitudinal data, tracking publications and citations year by year over the course of researchers' careers, is the natural next step, and is the focus of our ongoing research. Third, the analysis is silent on the structural origin of the differences in publication output and career length between men and women, differences that the international literature attributes to a combination of higher peer-review standards (Hengel, 2022[18] Hengel, E. (2022). Publishing while female: Are women held to higher standards? Evidence from peer review. The Economic Journal, 132(648), 2951-2991. https://doi.org/10.1093/ej/ueac032
), differential exit rates (Spoon et al., 2023[38] Spoon, K., LaBerge, N., Wapman, K. H., Zhang, S., Morgan, A. C., Galesic, M., Fosdick, B. K., Larremore, D. B., & Clauset, A. (2023). Gender and retention patterns among U.S. faculty. Science Advances, 9(42), Artículo eadi2205. https://doi.org/10.1126/sciadv.adi2205
), and structural inequalities in access to resources (Rivera-Romano et al., 2020[34] Rivera-Romano, L. S., Fresno, C., Hernández-Lemus, E., Martínez-García, M., & Vallejo, M. (2020). Gender imbalance in executive management positions at the Mexican National Institutes of Health. Human Resources for Health, 18(1), Artículo 21. https://doi.org/10.1186/s12960-020-0463-4
; Hensvik and Nilsson, 2025[19] Hensvik, L., & Nilsson, P. (2025). Equality for granted: What happens when discrimination in academia becomes salient? (CEPR Discussion Paper No. DP20210). Centre for Economic Policy Research. https://cepr.org/publications/dp20210
). Identifying which of these mechanisms is most relevant in the Mexican case remains an important agenda for future research.

References

Appendix A. Sample construction

This appendix details the construction of the analysis sample. Starting from the 2021 SNI registry (“Padrón de Beneficiarios”), we (i) classify researchers by gender using the academic-title field NOBILIS or using manual classification as described in section 3.1, (ii) match them by CVU to Scopus profiles built from the researcher's full bibliographic record up to 2021, and (iii) construct researcher-level cumulative measures of citations, publications, coauthors, maximum coauthor H-index and academic age.

Appendix B. Selection correction (Heckman)

To assess whether selection into Scopus affects our estimates, we follow Heckman (1979)[17] Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47(1), 153-161. https://doi.org/10.2307/1912352
. On the full 2021 SNI registry of researchers with identifiable gender, we jointly estimate a probit for the probability of being observed in Scopus and the citations equation, using the Heckman (1979)[17] Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47(1), 153-161. https://doi.org/10.2307/1912352
two-step estimator at the researcher level. The selection equation includes region (five categories) and grouped-institution indicators as exclusion restrictions; the outcome equation uses the same log-transformed cumulative covariates as the main specification, winsorized at the 99th percentile within discipline, with SNI-level dummies. Table A2 reports the results.

The inverse Mills ratio is not statistically significant in any of the three disciplines (Table A2), indicating no detectable selection bias on observables once the model is estimated. Selection into Scopus is least restrictive in Health Sciences (74.8% of SNI Health researchers with identifiable gender in Scopus), more selective in Engineering (53.5%), and most selective in Economics (44.6%), consistent with the differential coverage of Scopus across disciplines noted in Section 2.1. Within the Heckman estimation, discipline cells with no within-cell variation in Scopus status are dropped to avoid perfect separation, so the selection equation is fitted on 608, 580 and 901 selected observations (i.e. estimation-selection rates of 44.6%, 53.5% and 74.8%; Table A2). The coefficient on the female indicator remains positive in all three disciplines after the selection correction and is statistically significant in Engineering and Health Sciences. We conclude that selection into Scopus does not overturn the central finding of a gender gap in citations driven by endowments.

Appendix C. Robustness

This appendix collects the robustness checks summarized in Section 3.4. Each row reports the aggregate Oaxaca-Blinder components for an alternative specification. Robust standard errors (HC1) in parentheses.