Content uploaded by Florencia Alejandra Fiorentin
Author content
All content in this area was uploaded by Florencia Alejandra Fiorentin on Mar 27, 2023
Content may be subject to copyright.
Observable and unobservable causes of the gender gap in
S&T funding for young researchers
Diana Suarez1,2,*, Florencia Fiorentin
1,3 and Mariano Pereira1,2
1Interdisciplinary Center for Studies in Science, Technology and Innovation (CIECTI), 2390 Godoy Cruz St., Ground Floor (1425), Buenos Aires
Federal District, Argentina, 2National University of General Sarmiento, Industry Institute (IDEI/UNGS), 1150, J.M.Gutierrez St., Los Polvorines
(1663), Buenos Aires, Argentina and 3National Council for Scientic and Technical Research (CONICET), 2290 Godoy Cruz St., Ground Floor
(1425), Buenos Aires Federal District, Argentina
*Corresponding author. E-mail: dsuarez@campus.ungs.edu.ar
Abstract
The objective of this paper is to study the relationship between academic performance, gender, and science and technology grants. The empirical
analysis is based on the Argentinean Fund for Scientic and Technological Research ‘Scientic and Technological Research Projects’ (FON-
CYT-PICT). The methodology is based on a multivariate decomposition for non-linear response models, an extension of the Oaxaca–Blinder
decomposition. Results conrm the presence of a gender gap. Women have lower probabilities of being awarded with funds than men, even
when academic trajectories are alike. Results show that even if the productivity gap is closed, men’s and women’s trajectories are differ-
ently valued against women, and this negatively impacts their probability of being awarded. Therefore, even if women manage to publish
more and—somehow—be more cited, the gap will persist. Explicit public policy measures are required to close the current gap and stop
reproducing it.
Key words: gender gap; productivity puzzle; S&T policy.
© The Author(s) 2023. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
1. Introduction
The objective of this paper is to study the gender gap in being
awarded with public funds for research among young schol-
ars in Argentina. The focus is on the relationship between
academic productivity, gender, and grants. The literature
has largely proved that female researchers are disadvantaged
within the process of scientic evaluation, for both developed
and developing countries (Bukstein and Gandelman 2019;
Ranga et al. 2012; Waisbren et al. 2008; Wullum Nielsen
and B
orjeson 2019). To a great extent, this is related to the
well-known Matilda effect (Rossiter 1993), which refers to
the marginalization, under-estimation, and lower recognition
of women in the eld of science and technology (S&T) just
because they are women. Our main hypothesis is that aca-
demic productivity impacts differently the probability of being
granted even when women’s and men’s trajectories are alike.
The novelty of this paper lies in the fact that it provides
evidence on the observable and unobservable causes of the
gender gaps in S&T funding. We aim at making a contribu-
tion to public policy on S&T by characterizing a problem that
is out of sight since it is part of unconscious practices of the
academic process of evaluation.
The empirical analysis is based on the funding line for
young researchers (‘inicial’ type of call, in Spanish) from
the ‘Scientic and Technological Research Projects’ (PICT,
in Spanish), which is the most important public program
to foster scientic and technological activities in Argentina
(Suarez and Fiorentin 2018). The database includes structural
information from PICT’s administrative data combined with
bibliometric information retrieved from Scopus. The empiri-
cal strategy is based on analyzing the academic productivity
as a proxy for the academic quality of researchers’ proposal.
Then, we analyze how academic quality impacts the probabil-
ity of being awarded and test the presence of a gender gap in
accessing public funds for science.
Results conrm the presence of a gender gap. Women
have 3.3 percentage points (p.p.) lower probabilities of being
awarded with a PICT grant than men. Looking at the gap
over time, results lead to rejecting the presence of the Matilda
effect at the beginning of the period under analysis (2003–11).
However, the gap is conrmed for the rest of the sub-periods.
It climbs up to 3.7 p.p. during 2012–5 and then climbs again
up to 4.0 p.p. during 2016–9. When looking at observable and
unobservable characteristics (differences in actual academic
productivity between men and women and how it is val-
ued), results show that even if the productivity gap is closed,
men’s and women’s work are differently valued, and this
negatively impacts women’s probability of being awarded.
During the period 2012–9, 14.5 per cent of difference in the
rate of awarding between men and women is explained by
actual differences in their trajectories in the matter of aca-
demic productivity (0.48 p.p. of the gap), while 85.5 per cent
is explained by differences in the coefcients, meaning unob-
servable characteristics and if women and men had the same
productivity trajectories (2.82 p.p.). Therefore, even if women
manage to publish more and—somehow—be more cited, the
gap will persist, given the different valuation assigned to
female researchers’ work.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
2Science and Public Policy
The remainder of this paper is organized as follows:
after this introduction, the theoretical framework is pre-
sented in Section 2, together with the presentation of the
PICT instrument and the discussion of the research ques-
tions; in Section 3, the methodological strategy is explained;
Section 4 presents descriptive statistics and the results of
the econometrical model. Some reections related to the
results and the theoretical framework are also included in this
section; and nally, Section 5 provides the conclusions of the
paper.
2. Theoretical framework
2.1 Gender bias in S&T and the productivity gap
The presence of a productivity gap between men and women is
a historical phenomenon, which Cole and Zuckerman (1984)
called the ‘productivity puzzle’. Measured in terms of publi-
cations in specialized journals, women have lower levels of
productivity than men. There are many studies that verify the
productivity gap, for both developed and developing countries
(Beaudry and Larivière 2016; Gonzalez-Brambila and Veloso
2007; Graddy-Reed et al. 2019; Jiménez-Rodrigo et al. 2008;
Nygaard and Bahgat 2018; Prpi
c 2002; Ranga et al. 2012).
Recent evidence on the subject shows that differences in the
publication rate between men and women are decreasing,
although signicant differences remain, even when women are
increasing their participation in academic publishing (Huang
et al. 2020). By 2016, women explained around one-third of
total publishing indexed by the Web of Science (Huang et al.
2020), which is about 8 p.p. lower than women’s participation
in science (UNESCO 2016).
Ever since Cole and Zuckerman’s work (Cole and Zuck-
erman 1984), the literature has searched for clues to solve
that puzzle. Several concepts have emerged to refer to the
presence and persistence of discrimination against women
in S&T (Carrillo et al. 2014; Le
on et al. 2017; Lerchen-
mueller and Sorenson 2018). They can be grouped in terms
of the difference and the decit approaches. The differ-
ence approach refers to observable characteristics of women’s
trajectory that impact their performance in the labor mar-
ket. Some of them are specic factors of academic activ-
ity that inuence women’s publication performance, and
some others are general factors that affect women’s perfor-
mance at any job. Regarding specic factors, evidence shows
that women are more frequently assigned ‘university house-
work’, which negatively impacts their availability of time
for researching and publishing (Ranga et al. 2012). In addi-
tion, there is the impact of the glass ceiling, which refers
to the lower probability of women to move upward in the
scientic career, which negatively impacts funding oppor-
tunities and, again, the possibilities of doing research and
publishing.
The decit explanation of gender-based discrimination
refers to unobservable—thus hardly measurable—aspects of
the academic career that negatively affects women’s perfor-
mance (Mairesse and Pezzoni 2015). Applied to the produc-
tivity gap, the literature alerts about the lower level of recogni-
tion women receive from peer-review evaluations (Cislak et al.
2018). Evidence shows that women tend to be less cited than
men (Desai et al. 2018; Zigerell 2015), and even those papers
co-authored between men and women receive lower recogni-
tion than male-only articles (Beaudry and Larivière 2016). In
addition, the peer review–based system is male-dominated—
partially due to the higher positions they reach—and evalu-
ations tend to be even more biased (Jiménez-Rodrigo et al.
2008; Ranga et al. 2012), which prolongs publishing tim-
ing for women. Moreover, papers about the gender gap tend
to be under a stricter scrutiny, generating a gender bias in
publications trying to shed light on the gender bias (Cislak
et al. 2018). There is evidence that also shows that papers
and research projects written by women tend to be less rec-
ognized (and more rejected) as they are written in a ‘feminine
way’ (Melin 2007; Steinþ
orsd
ottir et al. 2020; Witteman et al.
2019).
2.2 Productivity gap and the award system of
science
The process of funds’ allocation is mostly based on peer-
review evaluations (Bornmann et al. 2007), which consider
the submitted research project and the academic perfor-
mance of the researcher—or research team—that applies for
funding. Performance is mostly estimated through publi-
cations on specialized journals. When evaluations are not
blind, reputation also affects the selection process, which
has been dened by Merton (1968) as the ‘Matthew effect’.
In that line, Rossiter (1993) highlighted that the positive
accumulation of reputation is particularly veried on men.
Then, she named the ‘Matilda effect’ as the systematic lower
recognition of women’s work by the scientic community.
In terms of S&T policy, this is reected in the misalloca-
tion of funds against women based on gender-related—and
prejudice-based—issues instead of strictly scientic elements
of the submitted research project. Literature has largely
proved the Matilda effect, without prejudice to the level of
development of the country (Bukstein and Gandelman 2019;
Fiorentin et al. 2022; Ledin et al. 2007; Waisbren et al. 2008;
Witteman et al. 2019; Wullum Nielsen and B
orjeson 2019). In
fact, in the cases where the Matilda effect is not veried, this is
because probabilities are estimated as if men and women were
equally treated in the S&T system (Aboal and Vairo 2017;
Cañibano et al. 2009), when in fact they are not.
In the case of developed countries, the literature conrms
that productivity and citations positively impact the probabil-
ity of being awarded with different intensities between men
and women (Head et al. 2013; Svider et al. 2014; Waisbren
et al. 2008). In the case of developing countries, there is no
such consensus, and evidence is scarce. In the case of Uruguay,
Bukstein and Gandelman (2019) conrm the expected rela-
tionship between the Matilda effect and productivity. In the
case of Paraguay, Aboal and Vairo (2018) also found evidence
on the productivity gap, although their results do not cor-
roborate its impact on being awarded with public grants. In
Mexico, evidence does not verify different levels of productiv-
ity between men and women; in fact, in some cases, women
outperform men (Le
on et al. 2017). Previous studies on the
Argentinean case point to the existence of the Matilda effect
and a gap in productivity between men and women (Fiorentin
et al. 2022).
2.3 The Argentinean Fund for Scientic and
Technological Research
Argentina is a relatively egalitarian system in terms of S&T
and gender; at the aggregate level, −53 per cent of researchers
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
Science and Public Policy 3
are women, but it is not when trajectories are considered: only
11 per cent of them reaches the highest positions, while that
value among men climbs up to 17 per cent (SiCyTAR 2020).
The publishing gap is around 24.4 per cent against women
for the period 1900–2016, which results from the lower
annual productivity rate (−8.8%) and career length (−17.7%)
of women (Huang et al. 2020). The highest difference between
men and women is veried on the citations, where the gap is
32 per cent. In this paper, we wonder how this gap impacts the
probability of being awarded with public funds. We assume
the better the academic productivity of researchers, the higher
their research capabilities and then the higher their abilities to
formulate and submit a research project to PICT. In addition,
since we work with young researchers, their publishing expe-
rience is a good proxy for their capabilities to lead a research
project, even when he or she lacks research-leading experience
at all.
The empirical study is based on the line of funding called
PICT, which is the acronym for ‘Scientic and Technological
Research Projects’ and it is part of the Argentinean Fund for
Scientic and Technological Research (FONCYT), the main
public source of funding for S&T projects with national scope
(Suarez and Fiorentin 2018). PICT’s calls are organized of
three types and four categories. Based on the eld of science
of the submitted project, researchers can apply to the fol-
lowing categories: (1) ‘open themes’ (all elds of science), (2)
‘Argentina 2030’ (strategic elds dened in the national strate-
gic plan), (3) ‘international cooperation’, and (4) specic calls
depending on the S&T and national agendas (e.g. COVID-
19-related calls). Researchers can apply for three types of
funding, depending on the responsible researcher and the
team’s age: consolidated teams (at least one researcher over
48 years of age), new teams (all researchers below 48 years
of age), and young researchers (without a team, researchers
below 36–38 years of age,1 with a PhD). In this paper, we will
look at those researchers who apply to this last type of call
(‘inicial’ in Spanish).
To apply for these funds, researchers must submit a pro-
posal, and three dimensions are valued based on a single-
blind, peer-review process: (1) relevance of the theme (35 per
cent), (2) quality of the project (35 per cent), and (3) academic
background of the research team (30 per cent). Since infor-
mation about the evaluation of the proposals is not available,
we decided to explore the relationship between academic pro-
ductivity and the probability of being granted. This is based
on two assumptions. On the one hand, academic productivity
has been considered as a proxy for the quality of the pro-
posal since researchers with more publishing experience on
specialized journals might be better in formulating a research
project based on a relevant subject [dimensions (1) and (2) of
the evaluation]. In this regard, ‘quality’ is dened in similar
terms to the way it is dened by academic practices on pub-
lishing on specialized journals. Regarding papers, this means
adjustment to the scientic method, the novelty of the subject,
knowledge about the state-of-the-art, and clearly elaborat-
ing research questions and hypotheses, among others; and
in the case of citations: relevance of results and contribu-
tions to the literature. The rationality behind this selection lies
on the assumption that the higher the academic productivity,
the higher the capabilities to do research, including the for-
mulation of a research project. In addition, publications and
citations also account for the relevance of the research subject.
On the other hand, since one-third of the evaluation is
based on the academic background of the research team
and in the case of the ‘inicial’ call only the responsible
researcher is valued, academic productivity is a good proxy
for the academic trajectory under evaluation (third dimen-
sion). Of course, usual controls are also included, such as
the eld of science, institutional afliation, and location.
Similar assumptions are made for this type of policy evalu-
ations in literature (Section 2.2), in some cases due to con-
dentiality issues connected to the research project and the
double-blind review process.
Descriptive statistics of PICT for the period 2003–19
show a reduced gap in terms of the distribution of awarded
researchers (Table 1).2 Women’s rate of success is 49 per cent
(the ratio of awards to submissions), while men’s rate is 53
per cent, which represents a gap of 4 p.p. (the difference in
the awards-to-submissions ratio between men and women).
Looking at the gap over time, it shows a decreasing trend, with
a peak during the period 2012–5 when it reached 6 per cent
and then dropped up to 5 per cent. The higher number of sub-
missions made by women is coherent with the higher number
of female researchers in the total population of researchers.
The lower rate of success is also coherent with what was men-
tioned before in terms of lower participation of women at the
top of the pyramid of the academic career.
Regarding academic productivity, and consistent with the
evidence reviewed in, Table 2 shows the existence and mag-
nitude of the productivity gap: male researchers publish more
and are more cited than their female counterparts, even when
they both have the same experience in terms of years since
the rst time they published. Among awarded researchers,
the gender gap is still veried. While awarded men publish
1.98 papers per year, this value drops up to 1.61 among
awarded women. In addition, women receive fewer citations
than men: 1.72 vs. 1.64 for male and female researchers,
respectively. Publishing differences remain in the cases of non-
awarded researchers, although women received, on average, a
higher number of citations. Regarding experience (years since
the researcher’s rst publication), the gender gap is reduced,
most probably explained by the age limit of this type of call.
Among awarded researchers, men and women are alike (13.29
and 13.27years of experience, respectively), while among
non-awarded researchers, men have 0.64 additional years of
experience.
Results show a reduction in the differences among awarded
researchers over time (Table 3), consistent with an intensi-
cation of competition within this line and the intensication
of publishing activity in Argentina.3 It is interesting noticing
the signicant increase in papers during the last sub-period,
most probably connected to the increase in sensibilization and
explicit policy measures to close the gender gap in science.4
This reinforces the importance of analyzing the gender gap
over time, to the extent that general conditions for women
changed during the last years of the period under analysis,
but structural gaps related to the male bias in S&T might
persist.
2.4 Research question and hypothesis
Given in the aforementioned analysis, to the extent that
men and women are not in the same initial condition in
terms of academic productivity, the award system—and any
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
4Science and Public Policy
Table 1. Number of submissions and awards (researchers) (2003–19).
Male researchers Female researchers
Period
Awards
(a)
Submissions
(b)
Rate
(a)/(b)
Awards
(a)
Submissions
(b)
Rate
(a)/(b)
Gender gap
(rate female − rate male)
2003–11 324 588 0.55 453 807 0.56 0.01
2012–5 532 1,036 0.51 671 1,484 0.45 −0.06
2016–9 486 902 0.54 634 1,286 0.49 −0.05
2003–19 1,342 2,526 0.53 1,758 3,577 0.49 −0.04
Source: own elaboration based on the PICT database.
Table 2. Academic productivity (2003–19).
Male researchers Female researchers
Awarded Non-awarded Total Awarded Non-awarded Total Total sample
Experience Mean 13.29 13.40 13.35 13.37 12.76 13.06 13.18
(years since the researcher’s rst publication) SD (4.87) (6.17) (5.51) (4.90) (5.82) (5.39) (5.45)
Papers Mean 1.98 1.65 1.83 1.61 1.43 1.52 1.65
(average accumulated number of papers per
year)
SD (1.90) (1.52) (1.74) (1.49) (1.24) (1.37) (1.54)
Citations Mean 1.72 1.37 1.56 1.64 1.46 1.54 1.55
(average accumulated number of citations per
year to accumulated papers)
SD (1.96) (1.44) (1.74) (1.61) (1.50) (1.56) (1.64)
Observations 1,342 1,184 2,526 1,758 1,819 3,577 6,103
Source: own elaboration based on the PICT database and Scopus repository.
Table 3. Academic productivity (awarded researchers).
Male researchers Female researchers
2003–11 2012–5 2016–9 2003–11 2012–5 2016–9
Experience
(years since the researcher’s rst publication)
Mean 17.80 13.48 10.08 17.83 13.37 10.18
SD (4.71) (3.39) (3.77) (4.61) (3.49) (3.78)
Papers
(average number per year)
Mean 1.76 2.04 2.07 1.26 1.43 2.05
SD (2.14) (2.14) (1.37) (1.53) (1.45) (1.39)
Citations
(average number per year)
Mean 1.67 2.44 0.96 1.57 2.29 0.99
SD (1.94) (2.15) (1.36) (1.58) (1.79) (1.08)
Observations 324 532 486 453 671 634
Source: own elaboration based on the PICT database and Scopus repository.
practice based on this dimension—will necessarily reproduce
existent gaps with cumulative impacts over time. If discrim-
ination against women were inexistent and differences were
based only on academic trajectories, then once differences
derived from the starting point are properly controlled, and
there are not reasons to expect productivity to impact dif-
ferently the probability of being awarded. However, given
the already proven Matilda effect, women’s production is less
recognized—as it happens in other academic practices—and
then it differently impacts the probability of being awarded,
even when publishing and citation rates between men and
women are alike. Therefore, our research question is about
the observable and unobservable impacts of productivity on
women’s and men’s probability of being awarded with pub-
lic grants for S&T. Observable impacts are derived from
differences in publishing rates between men and women.
Unobservable impacts appear when a male researcher and a
female researcher with the same productivity have different
probabilities of being awarded. Of course, part of the unob-
servable impacts in our model might be derived from actual
differences in the characteristics of the submitted proposal.
However, there are not a priori reasons to expect this omitted
variable to impact differently between men and women—in
other words, assuming characteristics (e.g. relevance and fea-
sibility), women’s proposals are of a lower quality than men’s,
which is in fact the bias we claim that does exist in the
evaluation process of science.
Derived from the research question and based on previous
evidence about the PICT, we will test three hypotheses:
H1: The gender gap in accessing public funds for
science is partially explained by women’s lower
productivity than men’s one.
H2: Even when academic productivity between men
and women is alike, the gap in accessing public funds
persists, due to unobservable effects of the
productivity gap.
H3: The evolution of the gap over time is explained by
changes in both the productivity gap and how
productivity is valued.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
Science and Public Policy 5
H1 and H2 aim at testing the impact of productivity on
the probabilities of being awarded. H1 refers to the difference
in explanations and the Matthew effect of science. Women
publish less, and hence they are less recognized and then less
granted. H2 is derived from the decit explanation. The patri-
archal structure of science impacts the evaluation system in
the form of under-recognition of women’s production. For
instance, the scrutiny of papers written by women is harder
than men’s, which is not necessarily because of open preju-
dices of reviewers—being them men or women—but because
of unconscious biases that lead to differently look at women’s
works and expectations. Hence, projects submitted by female
researchers are less punctuated than those submitted by male
ones, and therefore, projects of females face less probabilities
of crossing the funding line.
Finally, H3 recognizes the increase in the awareness regard-
ing the gender bias and the implementation of soft and hard
policy measures to reduce the bias. On the one hand, sensibi-
lization about the gender gap and the women’s participation
increase in science is supposed to impact the gap. On the
other, explicit measures such as the ‘stop-the-clock’ policies
have been implemented during the last years in the S&T
Argentinean system. It is worth mentioning that this paper
does not aim at testing the impact of these measures on
the gap, just to acknowledge them and include the proper
controls.
3. Data and methodology
To explore our research question and test the hypotheses,
we will analyze the impact of three of the main indicators
of academic productivity: publications, citations, and pub-
lishing experience (from Scopus) on the probability of being
granted (PICT-Inicial). Publications relate to the rate of com-
munication of results, which is affected not only by the actual
production of the researcher but also by the process of peer
review and editorial policies. We selected accumulated pub-
lications since evaluations look at the researcher’s publishing
trajectory instead of the latest articles. Citations are measured
as the accumulated number of citations per year to accu-
mulated publications per year, to account for the scope of
publishing. This estimation allows us to include a measure
of recognition of researchers’ academic productivity as well
as the relative relevance of the published topics. Publication
experience is the number of years since the researcher’s rst
publication indexed in Scopus, and it is used as a proxy for
researchers’ trajectory.
We will perform the analysis only on young researchers,
who according to the type of call are scholars holding a
PhD, working in S&T institutions from the Argentinean sys-
tem, and are less than 36–38 years of age (three mandatory
requirements). The call is named ‘Inicial’, and only isolated
researchers can apply. In addition, even when multiple sub-
missions can be done, this award can only be granted once
and only if the researcher was not granted before with any
other PICT award (from the other two types). This selection
allows us to avoid the impact of past awards (the Matthew
effect) and also the impact of the research team on the eval-
uation of the proposal. Therefore, we will look at the impact
of productivity on the probability of a female versus a male
researcher being granted for the rst (and only) time.
The database was constructed from all the applications
to PICT-Inicial during 2003–19. Information retrieved from
the PICT administrative records provides information about
demographic and institutional characteristics. It also includes
all historical information about submissions and awards, in
this case, narrowed to the type of call under analysis. Biblio-
metric information was added from Scopus, which includes
publications on scientic journals and citations from 1996
up to 2019. Information from Scopus was retrieved using
Scopus’s application programming interface. Full name, insti-
tutional afliation, eld of science, and geographic location
of each researcher were used to match published papers and
citations for each year under analysis. Multiple alternatives
were allowed to match PICT’s researchers with Scopus IDs:
full name, middle name, and surname; common abbrevia-
tions of name and middle name; changes in the institutional
afliation; and eld of science. Bibliometric information was
added with an error rate of 5 per cent (the probability of a
Scopus ID incorrectly matched with a researcher from the
PICT database). The resulting database is an unbalanced
dynamic panel at the level of researcher and years—4,753
researchers and 6,103 observations (henceforth, the PICT
database).
Performing causal inference between academic productiv-
ity and access to public research funds is different from testing
observable and unobservable factors (Angrist and Pischke
2009; Wooldridge 2010). On the one hand, factors such as
age, gender, academic degree, and research experience (among
others) affect both productivity and the probability of being
awarded a research grant, and if we do not control for these
dimensions, we may bias the effect we attribute to productiv-
ity. Since these factors can be observed and measured, we can
include them as covariates in our baseline specication and
thus address the potential bias they generate in the estima-
tion of the causal effect of productivity. On the other hand,
there are other factors such as innate ability, gender stereo-
types, and social conditioning (among others) that cannot be
observed and measured and therefore cannot be incorporated
as a control variable. Designing an identication strategy to
deal with these unobservable factors relies on the existence of
instrumental variables or longitudinal data and the applica-
tion of xed effects (Abadie and Cattaneo 2018). As we will
see below, The characteristics of the database do not allow us
using this strategy.
The methodology is based on a multivariate decomposi-
tion for non-linear response models, which is an extension
of the Oaxaca–Blinder decomposition (Blinder 1973; Oaxaca
1973). This model allows us to control for path dependency,
add dummy variables to the estimation, and identify the
impact of each variable on the gap (Powers et al. 2011). The
decomposition essentially assesses how much of a gap is due
to the differences in characteristics (explained variation) and
how much is due to the same feature giving different returns
(unexplained variation). The objective is to decompose the
probability of being granted into one component caused by
differences in characteristics of the two groups (female and
male researchers) (H1) and another component caused by dif-
ferences in returns to the same character across groups (the
so-called ‘unexplained component’) (H2). The general model
is as follows:
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
6Science and Public Policy
Table 4. Summary of the variables.
Variable Description Values
PICTit Awarded PICT at t1 if awarded, 0 otherwise
GenderiA binary variable that indicates the gender of the researcher 1 if woman, 0 otherwise
Papersit Accumulated number of papers published on specialized journals and indexed in Scopus 0–∞
Citationsit Accumulated number of citations per year to accumulated number of papers per year and
indexed in Scopus
0–∞
Experienceit Years since the researcher’s rst publication indexed in Scopus (t0=OPU) 0–20
Institutionit Set of binary variables that indicates the institution of researchers (national universities,
National Scientic Research Council [CONICET], other S&T institutions)
1, 0
Regionit Set of binary variables that indicates the geographical location of researchers (north-west,
north-east, south, west, center)
1, 0
Presentationsit The total number of projects submitted to PICT 0–∞
Disciplineit Set of binary variables that account for the main eld of science of the researchers, based on
Scopus classications
1, 0
YeartSet of twelve binary variables that account for time xed effects 2003–19
Source: own elaboration based on the PICT database.
Here, 𝑌𝑡𝑖 denotes the probability of researcher 𝑖 at time 𝑡
being awarded with a PICT grant and depends on a matrix of
independent variables 𝑋𝑖𝑡 and a matrix of 𝛽 coefcients. The
𝑋𝑖𝑡 matrix is compounded by a set of structural characteris-
tics of the researchers (geographical location and institutional
belonging), academic background (publications, citations,
and experience on publishing), and PICT-related variables
(past submissions and category of calls). All variables are
summarized in Table 4.
In order to test the sensitivity and robustness of our model,
we have checked three alternative estimations. Firstly, we have
estimated a pooled probit model clustered by researchers.
Additionally, since researchers can only be awarded once and
75 per cent of the dataset is compounded by researchers who
only applied once, estimations constrained only to this group
were also included. Finally, we have estimated a random
effect probit model based on a yearly periodization. As we
shall see in the next section, since most of the observations
appeared only once and since a time control dummy variable
was included in all cases, we used the pooled probit model as
our selected baseline model for the decomposition, which best
suits the conformation of the dataset.
The gender gap in being awarded with public funds for
science can be estimated as a mean difference between two
groups (male and female researchers) and decomposed as
follows:
The rst two terms in Equation (3) refer to differences
in endowments (E term)—in this case, in the academic
backgrounds of female researchers (observable characteris-
tics)—and the third and fourth terms account for differences
in the coefcients (C term), meaning that similar backgrounds
in both groups are valued differently (unobservable charac-
teristics). Then, the presence of a gender gap is veried if one
or both sets of terms show negative results. In this formula-
tion, we selected male researchers as the comparison group
and female researchers as the reference one. Therefore, the
rst set of terms accounts for the counterfactual situation,
where men have the same distribution of backgrounds as
women, and the second set accounts for the counterfactual sit-
uation, where women receive the same behavioral responses
as men. Finally, since ours is a probabilistic model, the estima-
tion will decompose average differences in predicted outcomes
(marginal effects).
The use of this multivariate decomposition for non-linear
response models allows us to estimate the impact of each
covariate on the gap. This means, to account for the impact
of 𝐸𝑘 and 𝐶𝑘 portions, with (k=1, …, K), represents the
unique contribution of each variable included in our base-
line model (structural, academic, and PICT-related variables).
Estimations of these weights (W) to the gap are based on Yun’s
(2004) solution. Formally,
Regarding the time-related dimension, we have grouped
the period into three sub-periods, in order to analyze the evo-
lution of the gap. The segmentation of the period is based on
the total number of submissions, so each sub-period groups
similar numbers of applications. Looking at the evolution
of the gap in terms of the number of submissions allows us
to better analyze the data, run better estimators (otherwise,
rst years will account for a reduced number of observa-
tions), and follow the evolution of PICT in terms of the
evolution of the gap observed in Section 2.3. Then, when
information from Scopus is added, some observations are lost,
especially for the rst and last periods. It is worth mention-
ing that different periodizations of the data showed similar
results in terms of the gap and its composition. Table 4 pro-
vides information about the variables used in the model.
Besides PICT-related variables (have been awarded or not
and past submissions) and academic productivity variables
(papers, citations, and experience), the model includes institu-
tional afliation, geographic region, and discipline as control
variables.
The database is described in Table 5. Differences in the
gender gap are observed across classications, which are
higher in the case of other S&T institutions different from
the national system (private labs), the north-east region, and
agricultural sciences and engineering and materials sciences.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
Science and Public Policy 7
Table 5. Description of the database—ratio of granted projects to total
submissions.
Male Female Total Gender gap
Institution
CONICET 0.56 0.52 0.54 −0.04
National universities 0.51 0.46 0.48 −0.04
National S&T institutions 0.47 0.54 0.51 0.07
Other S&T institutions 0.65 0.47 0.53 −0.18
Region
Center 0.54 0.49 0.51 −0.05
North-west 0.49 0.47 0.48 −0.02
North-east 0.54 0.44 0.48 −0.10
West 0.46 0.48 0.47 0.02
South 0.56 0.54 0.55 −0.01
Discipline
Social and human sciences 0.53 0.50 0.51 −0.04
Natural sciences 0.54 0.51 0.52 −0.03
Life and health care sciences 0.48 0.47 0.47 −0.01
Agricultural sciences 0.55 0.50 0.52 −0.05
Engineering and materials
sciences
0.53 0.45 0.48 −0.08
Total 0.53 0.49 0.51 −0.04
Source: own elaboration based on the PICT database.
As expected, those groups with the majority of observa-
tions have a gap similar to the one observed for the total
panel (CONICET and national universities, the center region,
and social and natural sciences). In this last case, discipline-
related gap is like the one observed in the literature regarding
STEM (science, technology, engineering, and mathematics)
elds.
It is worth mentioning that changes in the type of call
(PICT-Inicial) over the period under analysis were based only
on the age limit of the researcher who applied, some minor
changes in the strategic elds (connected to the evolution of
technology). No policy measures aimed at closing the gen-
der gap were implemented during the period and instrument
under analysis.
Unfortunately, information about other aspects of the gen-
der gap such as motherhood is not available in the S&T
Argentinean databases. This is something that affects not
only the academic sector but also the labor market, generally
speaking. Based on existing evidence for the academic work
regarding this issue, we can expect results to overestimate
unobserved effects since parental and familial responsibilities
affect women’s productivity (Mairesse et al. 2019). If sen-
sibilization measures about parenthood responsibilities are
effective, this overestimation should decrease over time. If
not, one can expect its impact to remain the same. However,
evidence about the labor market in Argentina shows that con-
trolling for motherhood and family situation does not affect
the intensity of the gender gap, in terms of both participation
and wages (Goren 2017). Therefore, our results might not be
affected by the lack of these types of controls. In any case, we
hope that the increase in the availability of information would
allow better estimations in the near future.
In this regard, another contribution of this paper is that it
highlights the need for databases that contain more informa-
tion, especially, since the gender gap is a systemic phenomenon
(Sato et al. 2020). Available information allows us to focus on
a very relevant topic for S&T literature and policy, related to
academic productivity and granting. Nevertheless, given the
characteristics of our database, we cannot know whether the
gender of the evaluators affects the allocation process, as it
was evidenced by Wenneras and Wold (1997) and Witteman
et al. (2019), if the bias is also manifested in the amounts
granted (Steinþ
orsd
ottir et al. 2020; Zhou et al. 2018), in the
score that projects receive in evaluations (Bol et al. 2022; Tam-
blyn et al. 2018), and even if the selection process of PICT is
also affected by the research topics selected by women and
men researchers (Burns et al. 2019; Philipps et al. 2022) or
language styles (Cheng et al. 2011; Urquhart-Cronish et al.
2019), which could be approached by a semantic analysis on
submitted projects or abstracts. As it was stated in the afore-
mentioned paragraph, we are making efforts to complement
our database with more information, while being aware that
there are issues that can never be included. In the meantime,
as also mentioned, we decided to investigate a topic that the
available information allows us to address and that it is, of
course, relevant.
4. Results
Table 6 presents the estimation of the baseline models. Since
the original estimation is a probit model, marginal effects
are reported. Results for the three models are similar, which
accounts for the robustness of the estimation, and consistent
with other studies on the Argentinean case (Fiorentin et al.
2022). There is a gender gap in being awarded with pub-
lic funds for S&T activity once researchers’ academic career
determinants are controlled. The estimation for the whole
sample shows that being equally productive and recognized
and having the same experience in publishing, young female
scholars have 3.4 p.p. lower probabilities of being awarded
with a PICT grant than men (columns 1 and 3). Publishing
experience, citations, and publications also impact the prob-
ability of being awarded, although with different signs. One
additional paper increases the probability of being awarded
in 0.3 p.p., while one additional citation does in 1.3 p.p.
Conversely, one additional year of accumulated experience
does the opposite in 0.7 p.p. The explanation of this nega-
tive impact lies on the type of call we are analyzing. Since it
is oriented toward young researchers, and given the expected
correlation between publishing experience and age, younger
researchers have higher probabilities of being awarded. In
addition, since experience is net of productivity, it also refers
to researchers who present a low publishing rate.
When only the rst submission is considered (column 2),
the Matilda effect drops up to 2.5 p.p., which is consistent
with evidence about the Matthew and Matilda effects of sci-
ence. Past presentations positively impact the probability of
being granted in the present (the Matthew effect) but only (or
at least higher) among male researchers (the Matilda effect).
Therefore, there are cumulative effects in the causes of the gap.
The impact of citations, publications, and academic produc-
tivity is similar in signicance and signs, with an increase in
the impact of publications, which rises up to 1.8 p.p.
Results for the baseline model lead to reject the presence of
the Matilda effect at the beginning of the period (2003–2011),
becomes signicant in 3.8 p.p. during 2012–2015 and then
climbs up to 4.1 p.p. in 2016–2019 (Table 7).
Regarding the impact of academic productivity, the evolu-
tion over time shows that yearly citations are only signicant
during the rst and second sub-periods, with an intensity of
3.4 p.p. and 1.3 p.p. higher probabilities of being awarded,
respectively. The impact of publishing experience remains
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
8Science and Public Policy
Table 6. Baseline model total panel—Dep. Variable: PICT—Marginal
effects.
Clustered
probit
Pooled probit
(only one
submission)
Dynamic
RE probit
Gender −0.034***
(0.013)
−0.025*
(0.015)
−0.034***
(0.013)
Paper 0.003***
(0.001)
0.003**
(0.001)
0.003***
(0.001)
Citations 0.013**
(0.006)
0.018**
(0.008)
0.013**
(0.006)
Experience −0.007***
(0.002)
−0.008***
(0.002)
−0.007***
(0.002)
Presentations 0.013
(0.010)
0.013
(0.027)
Observations 6,103 4,486 6,103
Year xed effects Yes Yes Yes
Discipline FE Yes Yes Yes
Regional FE Yes Yes Yes
Institutional FE Yes Yes Yes
No. of researchers 4753
Source: own elaboration based on the PICT database.
(1) Standard errors are in parentheses. (2) ***P<0.01, **P<0.05, and
*P< 0.1. (3) Dynamic random effect (RE) probit model. Sigma-u = 0.00232;
Rho = 5.40e−06.
Table 7. Baseline model per sub-periods panel—Dep. Variable: the proba-
bility of being granted—Marginal effects.
2003–11 2012–5 2016–9
Gender 0.012
(0.026)
−0.038*
(0.020)
−0.041*
(0.021)
Papers −0.002
(0.002)
0.011***
(0.003)
0.012***
(0.003)
Citations 0.034*
(0.020)
0.013*
(0.007)
−0.002
(0.006)
Experience −0.016***
(0.003)
−0.001
(0.004)
−0.003
(0.004)
Presentations 0.118*
(0.050)
0.012
(0.016)
−0.006
(0.014)
Observations 1,395 2,520 2,188
Discipline FE Yes Yes Yes
Regional FE Yes Yes Yes
Year FE Yes Yes Yes
Institutional FE Yes Yes Yes
Source: own elaboration based on the PICT database.
(1) Clustered probit model. (2) Standard errors are in parentheses. (3)
***P< 0.01, **P< 0.05, and *P< 0.1.
negative, but only during the rst sub-period (−1.6 p.p.), and
after that, this variable loses signicance. The opposite hap-
pens with the impact of one additional paper, which is not
signicant for the rst sub-period, and then, it turns positive
and becomes signicant for the second and third sub-periods
(1.1 p.p. and 1.2 p.p., respectively). One possible explanation
for this is that to the extent that the competition increases—
the number of submissions rises—it becomes a positive and
more determinant attribute.
These results are consistent with evidence for both devel-
oped and developing countries (Melin 2007; Sato et al. 2020;
Uhlmann and Cohen 2007; van der Lee and Ellemers 2018;
Witteman et al. 2019). According to the literature about the
‘productivity puzzle’ (Cole and Zuckerman 1984), evalua-
tion processes based on academic productivity are a source
of discrimination since women tend to publish less than men.
Table 8. Decomposition results—Dep. Variable: PICT.
Total panel Sub-periods
2012–9 2012–5 2016–9
Coeff. % Coeff. % Coeff. %
Endowment 0.0085**
(0.0036)
14.3 0.0173***
(0.0066)
28.4 0.0012
(0.0043)
2.6
Coeff. 0.0509***
(0.0159)
85.7 0.0438**
(0.0209)
71.6 0.0453**
(0.0218)
97.4
Observations 4,708 2,520 2,188
No. of researchers 3,577 2,314 1,813
Source: own elaboration based on the PICT database.
(1) Standard errors are in parentheses. (2) ***P<0.01, **P<0.05, and
*P< 0.1. (3) FE: discipline, region, and institution.
Theoretically speaking, if women manage to publish more,
then the productivity puzzle might be turned into a driving
force to close the gap. And this explains why in some stud-
ies, when trajectories are exactly alike, the gap disappears
(Aboal and Vairo 2018; Arensbergen et al. 2012). In the case
of our results and considering only 2016–9, female researchers
should accumulate three additional papers per year in order to
close the gap. However, and according to the decit approach,
as women’s work tends to be less recognized than men, then
even when the productivity gap is closed, the gap might per-
sist. Once again, this explains the studies that prove the
Matilda effect even between men and women with the same
backgrounds and productivity (L
opez-Aguirre 2019; Magua
et al. 2017; Uhlmann and Cohen 2007; van der Lee and Elle-
mers 2018). The decomposition model allows us to explore
this heterogeneity of results, in this case for the Argentinean
PICT.
Table 8 depicts the results of the decomposition model.
Since the gap is not signicant for the period 2003–11, we
have estimated the model for the total panel excluding the
rst sub-period and the sub-periods 2012–5 and 2016–9.
For the whole period, the gender gap in the probability of
being awarded is explained mostly due to differences in how
women’s work is valued. The lower 3.4 p.p. probabilities
of being awarded of female researchers is explained 92 per
cent for differences in coefcients (how women’s trajectory
is valued), being the rest of the explanation based on actual
(observable/endowment) differences between academic pro-
ductivity of male and female researchers. When only 2012–9
is considered, these two sides of the explanations are even
clearer, being both percentages positive and signicant: 14.3
per cent of difference in the rate of awarding between men
and women is explained by observable differences in their tra-
jectories, while 85.7 per cent is explained by differences in
the coefcients (meaning unobservable characteristics, and if
women and men had the same productivity rate). Therefore, if
women would publish more and–somehow—were more cited,
the gap would be closed by 0.49 p.p. (14.3 per cent of 3.4
p.p.); the remainder 2.91 p.p. of the gap would persist.
Results for the two sub-periods show an increase in the
impact of unobservable over observable characteristics. Dur-
ing 2012–5, 28.4 per cent of the gap was explained by
differences in academic productivity between men and women
(women tended to publish less and be less cited). However,
as women increase their productivity, during 2016–9, the
gap changes in nature, and only 2.6 per cent is explained
by observable differences. Therefore, the unobservable aspect
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
Science and Public Policy 9
Table 9. Decomposition results—components—Dep. Variable: PICT.
Total panel Sub-periods
2012–9 2012–5 2016–9
Coeff. % Coeff. % Coeff. %
Endowment
Papers 0.0113**
(0.0021)
18.9 0.0180***
(0.0040)
29.4 0.0051
***
(0.0021)
11.0
Citations 0.0002
(0.0001)
0.3 0.0001***
(0.0000)
−0.14 −0.0004
(0.0005)
−0.9
Experience −0.0012
(0.0007)
−1.9 −0.0009
(0.0011)
−1.4 −0.0018
(0.0013)
−3.9
Characteristics
Papers −0.0602**
(0.0252)
−101.39 −0.0607**
(0.0309)
−99.25 −0.06182
(0.0415)
−132.83
Citations 0.0013
(0.0115)
2.12 0.0662***
(0.0218)
108.22 −0.0199
(0.0159)
−42.7
Experience −0.0710
(0.0613)
−119.58 −0.0520
(0.0890)
−85.09 −0.08995
(0.0807)
−132.83
Observations 4,708 2,520 2,188
No. of researchers 2,239 2,314 1,813
Source: own elaboration based on the PICT database.
(1) Standard errors in parentheses. (2) *** P< 0.01, ** P< 0.05, and *P< 0.1. (3) FE: discipline, region, institution, and year.
of the gap connected to academic productivity increases
from 71.6 per cent during 2012–5 to 97.4 per cent during
2016–9.
Table 9 depicts the results regarding the impact of each
covariate on the gap. Given the specication of the model pre-
sented in Section 3, a negative coefcient in the endowment
term indicates the expected reduction in the women–men
probability of being awarded if women were equal to men
on the distribution of their academic productivity (gap reduc-
tion). If the sign is positive, then the covariates contribute to
widening the gap. Results show that equalizing the publishing
rate between men and women would contribute to close the
gap by 18.9 per cent. The evolution over time shows that this
percentage tends to decrease by the end of the period: from
29.4 per cent during 2012–5 to 11 per cent during 2016–9.
Citations are only signicant for the sub-period 2012–5, but
with a negative sign, meaning that it contributes to widen the
gap. Finally, and consistent with the results observed for the
whole panel, equalizing the publishing experience would not
contribute to close the gap.
Regarding the impact of coefcients, the specication of the
model implies that a positive coefcient in covariates indicates
the expected increase in the women–men gap if women’s work
was valued as same as men’s. Conversely, if the coefcient is
negative, it contributes to close the gap in the allocation pro-
cess. In this case, results are more heterogeneous, although still
consistent with the baseline model and the literature review.
The impact of publishing experience is not signicant, mean-
ing that if the experience gap is closed then men’s and women’s
years of experience are valued alike. Regarding publishing,
during 2012–5 if women’s and men’s publications were val-
ued alike, it would contribute to close the gap by 99.25 per
cent. This impact increases over time: during 2016–9, the con-
tribution to closing the gap climbs up to 132.83 per cent. A
value over 100 implies that it would have impacted higher
on women’s probability of being awarded. Citations have the
opposite effect, with a negative and positive impact of 108.22
per cent.
Results imply that at the beginning of the period, women
not only had lower publishing rates but also their papers
were less valued than men’s. The opposite happens with cita-
tions. During 2012–5, if women’s and men’s received citations
were valued alike, this would have increased the gap in the
probability of being awarded, meaning that women’s cita-
tions were actually valued higher than men’s. However, as
women’s citations reach similar levels to men’s, the impact
disappears. Again, meaning that even when academic produc-
tivity of men and women were alike, women still face lower
probabilities of being awarded than men. In addition, closing
the productivity gap does not guarantee an equal impact of
this dimension of academic activity on the probability of being
awarded.
Summing up, the results conrm the hypotheses. The lower
productivity rate among young female researchers negatively
impacts their probabilities of being awarded with public
grants for science (H1), and differences tend to get stronger
as women increase their participation on the call. In addition,
the impact of young female researchers’ academic produc-
tion on the probability of being awarded is lower than the
impact veried among male researchers, even when they show
equal productivity in terms of publications and citations (H2).
When looking at the gap from a dynamic perspective (H3), the
results show that the nature of the gender gap tends to change
in terms of the causes, as women increase their publishing
rate. Recognition of women’s academic productivity used to
be an attribute of discrimination at the beginning of the period
(citations positively impact the gap—reduce it), but as women
increase the number of published papers, the impact disap-
pears. Conversely, papers’ impact remains positive on the gap
(reduce it), meaning that women’s higher productivity might
contribute to close the gender gap.
All in all, results conrm—once again—the gender gap in
science (see Section 2.1), particularly in public subsidy allo-
cation (Section 2.2.). They are also in line with the previous
studies on the Argentinean case (Section 2.3). Even when all
parameters of the academic career are controlled—and even
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
10 Science and Public Policy
after all the increase in the public awareness of the gap during
the last years—young female researchers still face discrimina-
tion based on some unobservable characteristics linked more
to their gender than to their scientic performance, which in
fact reinforces gender discrimination.
In terms of public policy, our results allow us to claim that
the presence of bias at the beginning of the career might lead
to a wider gap as women move upward in their career, given
the widely veried Matthew effect of science. They also show
that one of the dimensions of almost every aspect of the evalu-
ation of academic activity—that is publishing—is biased since
even if women increase their participation, lower recognition
levels of their work might persist. Hence, a process based on
biased inputs will denitely lead to biased results, reinforc-
ing the bias in productivity (if women’s research projects tend
to be less funded, then women will show lower levels of aca-
demic production). In short, evaluations based on academic
productivity—a cornerstone of evaluation of science—lead to
amplifying the gap.
Regarding PICT, and public S&T policy, generally speak-
ing, the verication of the gap provides evidence about
the negative impact of a horizontal public policy, where
the lack of consideration of pre-existent bias leads the pol-
icy to deepen it. Since no gender-related policy measures
were implemented in PICT during the period under analysis,
nor changes in Scopus indexation (except, of course, better
access to information than in previous years), results lead to
reect on the role of public policy on the gender gap and
how the lack of gender agenda leads to biased instruments
that generate, increase, and perpetuate the gender bias of
science.
5. Conclusions
This paper aimed at studying the impact of the productivity
gap on the Matilda effect in the allocation process of grants
in the Argentinean case. We studied the impact of publica-
tions, citations, and publishing experience on the probability
of being awarded with public funds among young female and
male researchers. The literature largely corroborates both the
productivity gap and the Matilda effect, but it is inconclu-
sive in the matter of how the productivity gap impacts the
probability of being awarded. Therefore, we explored to what
extent the gender gap in being awarded is explained by differ-
ences in academic productivity and to what extent it is the
result of different valuation of similar backgrounds between
female and male researchers.
The empirical analysis was based on the Argentinean case
for the period 2003–19, and the line of funding analyzed was
the Scientic and Technological Research Projects aimed at
young researchers (PICT-Inicial). Results conrmed the pres-
ence of a gender gap in the Argentinean case: women are
3.4 p.p. less prone to being awarded than men. This differ-
ence is explained both by observable differences in academic
productivity and by differences in how women’s productivity
is valued. Results show that even if the productivity gap is
closed, men’s and women’s work are differently valued, and
this negatively impacts their probability of being awarded.
The results also show that 0.49 p.p. of the gap is explained
by observable differences in their trajectories, while 2.91 p.p.
by differences in the coefcients (meaning unobservable char-
acteristics and if women and men had the same productivity
rate).
Limitations of our results are linked to the lack of atten-
tion to the relevance and scope of publications, to the extent
that they are considered regardless the impact factor of the
journal where they were published or the number of citations
the paper has received. We hope that this can be checked in
the near future, but evidence for other countries shows the
same results—and even a deeper bias—when these elements
are taken into account. Another limitation of our study is
the lack of information about the submitted research project.
This is a general limitation of studies about the gender gap in
science, which can hardly be approached with a quantitative
large-scale study. Finally, there is the limitation of not being
able to control for marital and family status. This is a limita-
tion of the available information in Argentina and a call to pay
attention to the scarce available information to fully account
for the gender bias in the labor market, generally speaking.
Despite all these limitations, the results lead to some con-
clusions in terms of a policy gender agenda and the evaluation
system of science. Although the results show that increasing
productivity and visibility could be a way of dealing with the
gender gap in the allocation of funds, our research provides
robust evidence on the presence of a structural gender gap
in Argentina derived from differences in how women’s work
is recognized in the evaluation system. This gap is linked to
the lower productivity of women, the lower recognition they
receive from their colleagues, and then the lower probabili-
ties of being awarded. Therefore, the results point out that
explicit policy measures are required to close the gap, espe-
cially within the evaluation system. If the evaluation system
does not change, the productivity gap within the Matilda
effect will be impossible to close. Productivity-based evalu-
ations reinforce the gender bias and perpetuate the higher
obstacles women must face to move upward in their career.
Data Availability
Data available upon request.
Conict of interest statement. None declared.
Notes
1. The age limit changed during the period under analysis; in some
years, it was set up to 35 and other 38 years old. In 2019, the limit
was 35 years. In addition, researchers might ask for an exception
on the age limit in specic cases, for instance sons or daughters or
studies abroad.
2. Details about the database are presented in Section 3.
3. The National Council for Science and Technology (CONICET)
standardized measures to access scholarships and the national
research career, based on publishing. This has impacted the whole
S&T system in Argentina, which led to an increase in publishing
activity.
4. For instance, during the last years, national S&T agencies and
national universities in Argentina have implemented ‘stop-the-
clock’ measures in order to consider the impact of maternity on
academic productivity.
References
Abadie, A. and Cattaneo, M. D. (2018) ‘Econometric Methods for
Program Evaluation’, Annual Review of Economics, 10: 465–503.
Aboal, D. and Vairo, M. (2017) ‘Impact of Subsidies for Researchers on
the Gender Scientic Productivity Gap. Evidence from Paraguay’,
IDB Technical Note No. IDB-TN-1303, IDB, Washington DC.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
Science and Public Policy 11
—— (2018) ‘The Impact of Subsidies for Researchers on the Gender
Scientic Productivity Gap’, Science & Public Policy, 45: 515–32.
Angrist, J. D. and Pischke, J.-S. (2009) Mostly Harmless Econometrics:
An Empiricist’s Companion. Princeton: Princeton University Press.
Arensbergen, P. V., Weijden, I. V. D., and Besselaar, P. V. D. (2012)
‘Gender Differences in Scientic Productivity: A Persisting Phe-
nomenon?’, Scientometrics, 93: 857–68.
Beaudry, C. and Larivière, V. (2016) ‘Which Gender Gap? Factors
Affecting Researchers’ Scientic Impact in Science and Medicine’,
Research Policy, 45: 1790–817.
Bol, T., de Vaan, M., and van de Rijt, A. (2022) ‘Gender-Equal Funding
Rates Conceal Unequal Evaluations’, Research Policy, 51: 104399.
Blinder, A. S. (1973) ‘Wage Discrimination: Reduced Form and Struc-
tural Estimates’, Journal of Human Resources, 8: 436–55.
Bornmann, L., Mutz, R., and Daniel, H. D. (2007) ‘Gender Differences
in Grant Peer Review: A Meta-analysis’, Journal of Informetrics, 1:
226–38.
Bukstein, D. and Gandelman, N. (2019) ‘Glass Ceilings in Research:
Evidence from a National Program in Uruguay’, Research Policy,
48: 1550–63.
Burns, K., Straus, S., Liu, K., et al. (2019) ‘Gender Differences in Grant
and Personnel Award Funding Rates at the Canadian Institutes of
Health Research Based on Research Content Area: A Retrospective
Analysis’, PLoS Medicine, 16: e1002935.
Cañibano, C., Otamendi, J., and Andújar, I. (2009) ‘An Assessment of
Selection Processes among Candidates for Public Research Grants:
The Case of the Ram
on Y Cajal Programme in Spain’, Research
Evaluation, 18: 153–61.
Carrillo, P., Gandelman, N., and Robano, V. (2014) ‘Sticky Floors
and Glass Ceilings in Latin America’, The Journal of Economic
Inequality, 3: 339–61.
Cheng, N., Chandramouli, R., and Subbalakshmi, K. P. (2011) ‘Author
Gender Identication from Text’, Digital Investigation, 8: 78–88.
Cislak, A., Formanowicz, M., and Saguy, T. (2018) ‘Bias against
Research on Gender Bias’, Scientometrics, 115: 189–200.
Cole, J. R. and Zuckerman, H. (1984) ‘The Productivity Puzzle: Persis-
tence and Change in Patterns of Publication among Men and Women
Scientists’, in M. W. Steinkamp and M. Maehr (eds) Advances in
Motivation and Achievement, Vol. 2, pp. 217–58. London: JAI Press
Inc.
Desai, N., Veras, L. V., and Gosain, A. (2018) ‘Using Bibliometrics to
Analyze the State of Academic Productivity in US Pediatric Surgery
Training Programs’, Journal of Pediatric Surgery, 53: 1098–104.
Fiorentin, F., Pereira, M., Suarez, D., et al. (2022) ‘When Matthew Met
Matilda: The Argentinean Gender Gap in the Allocation of Science
and Technology Public Grants’, Economics of Innovation and New
Technology, 23: 1–13.
Gonzalez-Brambila, C. and Veloso, F. M. (2007) ‘The Determinants of
Research Output and Impact: A Study of Mexican Researchers’,
Research Policy, 36: 1035–51.
Goren, N. (2017) ‘Desigualdades sociolaborales. Una aproximaci
on a
sus marcos interpretativos desde la perspectiva feminista’, Revista
Latinoamericana de Antropoligía del Trabajo, 2: 1–21.
Graddy-Reed, A., Lanahan, L., and Eyer, J. (2019) ‘Gender Discrep-
ancies in Publication Productivity of High-Performing Life Science
Graduate Students’, Research Policy, 48: 103838.
Head, M. G., Fitchett, J. R., Cooke, M. K., et al. (2013) ‘Differences
in Research Funding for Women Scientists: A Systematic Compari-
son of UK Investments in Global Infectious Disease Research during
1997-2010’, BMJ Open, 3: e003362.
Huang, J., Gates, A. J., Sinatra, R., et al. (2020) ‘Historical Compari-
son of Gender Inequality in Scientic Careers across Countries and
Disciplines’, Proceedings of the National Academy of Sciences of the
United States of America, 117: 4609–16.
Jiménez-Rodrigo, M. L., Martínez-Morante, E., García-Calvente, M.
D. M., et al. (2008) ‘Through Gender Parity in Scientic Pub-
lications’, Journal of Epidemiology and Community Health, 62:
474–5.
Ledin, A., Bornmann, L., Gannon, F., et al. (2007) ‘A Persistent Prob-
lem. Traditional Gender Roles Hold Back Female Scientists’, EMBO
Reports, 8: 982–7.
Le
on, L. R., Mairesse, J., and Cowan, R. (2017) ‘Gender Gaps and
Scientic Productivity in Middle-Income Countries. Evidence from
Mexico’, IDB Working Paper No. IDB-WP-800, IDB, Washington
DC.
Lerchenmueller, M. J. and Sorenson, O. (2018) ‘The Gender Gap in
Early Career Transitions in the Life Sciences’, Research Policy, 47:
1007–17.
L
opez-Aguirre, C. (2019) ‘Women in Latin American Science: Gen-
der Parity in the Twenty-First Century and Prospects for a Post-
War Colombia’, Tapuya: Latin American Science, Technology and
Society, 2: 356–77.
Magua, W., Zhu, X., Bhattacharya, A., et al. (2017) ‘Are Female Appli-
cants Disadvantaged in National Institutes of Health Peer Review?
Combining Algorithmic Text Mining and Qualitative Methods to
Detect Evaluative Differences in R01 Reviewers’ Critiques’, Journal
of Women’s Health, 26: 560–70.
Mairesse, J. and Pezzoni, M. (2015) ‘Does Gender Affect Scientic
Productivity?’, Revue économique, 66: 65–113.
Mairesse, J., Pezzoni, M., and Visentin, F. (2019) ‘Impact of Fam-
ily Characteristics on the Gender Publication Gap: Evidence for
Physicists in France’, Interdisciplinary Science Reviews, 44: 204–20.
Melin, G. (2007) ‘Reviewing Applications by Women-Critical Use of
Additive and Reasoning Evaluation Methods’, Royal Institute of
Technology, CESIS-Centre of Excellence for Science and Innovation
Studies.
Merton, R. K. (1968) ‘The Matthew Effect in Science: The Reward and
Communication Systems of Science are Considered’, Science, 159:
56–63.
Nygaard, L. P. and Bahgat, K. (2018) ‘What’s in a Number? How (And
Why) Measuring Research Productivity in Different Ways Changes
the Gender Gap’, Journal of English for Academic Purposes, 32:
67–79.
Oaxaca, R. (1973) ‘Male-Female Wage Differentials in Urban Labor
Markets’, International Economic Review, 14: 693–709.
Philipps, A., Weymann, J., and Kahmann, C. (2022) ‘Gender Differences
in Grant Submissions and Research Topics: An Approach to Study
Proposals of the Experiment! Program of the Volkswagen Founda-
tion across Science and Engineering Fields’, LCSS Working Paper
No. 11, Leibniz Center for Science and Society.
Powers, D. A., Yoshioka, H., and Yun, M.-S. (2011) ‘Mvdcmp: Multi-
variate Decomposition for Nonlinear Response Models’, The Stata
Journal, 11: 556–76.
Prpi
c, K. (2002) ‘Gender and Productivity Differentials in Science’,
Scientometrics, 55: 27–58.
Ranga, M., Gupta, N., and Etzkowitz, H. (2012) ‘Gender Effects in
Research Funding. A Review of the Scientic Discussion on the
Gender-Specic Aspects of the Evaluation of Funding Proposals and
the Awarding of Funding’, Gutachten für die DFG.
Rossiter, M. W. (1993) ‘The Matthew Matilda Effect in Science’, Social
Studies of Science, 23: 325–41.
Sato, S., Gygax, P. M., Randall, J., et al. (2020) ‘The Leaky Pipeline
in Research Grant Peer Review and Funding Decisions: Challenges
and Future Directions’, Higher Education, 82: 145–62.
SiCyTAR. (2020) The Argentine Science and Technology Information
Portal. Ministerio de Ciencia, Tecnología e Innovaci
on Productiva.
Steinþ
orsd
ottir, F. S., Einarsd
ottir, Þ., Pétursd
ottir, G. M., et al. (2020)
‘Gendered Inequalities in Competitive Grant Funding: An Over-
looked Dimension of Gendered Power Relations in Academia’,
Higher Education Research and Development, 39: 362–75.
Suarez, D. and Fiorentin, F. (2018) Federalizaci
on y efecto Mateo en la
política cientíca. El caso del PICT en la Argentina (2012-2015)
(Documentos de trabajo CIECTI. Nro 12. Ciudad Aut
onoma de
Buenos Aires, Argentina, Recuperado de) <http://www.ciecti.org.ar/
dt12-federalizacion-y-efecto-mateo-en-la-politica-cientica-el-caso-
del-pict-en-la-argentina-2012-2015/> accessed 21 Sep 2022.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
12 Science and Public Policy
Svider, P. F., D’Aguillo, C. M., White, P. E., et al. (2014) ‘Gender
Differences in Successful National Institutes of Health Funding in
Ophthalmology’, Journal of Surgical Education, 71: 680–8.
Tamblyn, R., Girard, N., Qian, C. J., et al. (2018) ‘Assessment of Poten-
tial Bias in Research Grant Peer Review in Canada’, Cmaj, 190:
E489–99.
Uhlmann, E. L. and Cohen, G. L. (2007) “‘I Think It, Therefore It’s
True”: Effects of Self-perceived Objectivity on Hiring Discrimina-
tion’, Organizational Behavior and Human Decision Processes, 104:
207–23.
UNESCO. (2016) ‘Género y Ciencia’, Proyecto SAGA, UNESCO.
Urquhart-Cronish, M., Otto, S. P., and Coe, I. (2019) ‘Gender and
Language Use in Scientic Grant Writing’, Facets, 4: 442–58.
van der Lee, R. and Ellemers, N. (2018) ‘Perceptions of gender inequal-
ity in academia’. In: Bastiaan, R. Mark, B. (eds.) Belief Systems and
the Perception of Reality in Belief Systems and the Perception of
Reality. London: Routledge.
Waisbren, S. E., Bowles, H., Hasan, T., et al. (2008) ‘Gender Differences
in Research Grant Applications and Funding Outcomes for Medical
School Faculty’, Journal of Women’s Health, 17: 207–14.
Wenneras, C. and Wold, A. (1997) ‘Nepotism and Sexism in Peer-
Review’, Nature, 387: 341–3.
Witteman, H. O., Hendricks, M., Straus, S., et al. (2019) ‘Are Gender
Gaps Due to Evaluations of the Applicant or the Science? A Nat-
ural Experiment at a National Funding Agency’, The Lancet, 393:
531–40.
Wooldridge, J. (2010) Econometric Analysis of Cross Section and Panel
Data, 2nd edn. London: MIT Press.
Wullum Nielsen, M. and B
orjeson, L. (2019) ‘Gender Diversity in
the Management Field: Does It Matter for Research Outcomes?’,
Research Policy, 48: 1617–32.
Yun, M.-S. (2004) ‘Decomposing Differences in the First Moment’,
Economics Letters, 82: 275–80.
Zhou, C. D., Head, M. G., Marshall, D. C., et al. (2018)
‘A Systematic Analysis of UK Cancer Research Fund-
ing by Gender of Primary Investigator’, BMJ Open, 8:
e018625.
Zigerell, L. J. (2015) ‘Is the Gender Citation Gap in International
Relations Driven by Elite Papers?’, Research and Politics, 2:
2053168015585192.
Downloaded from https://academic.oup.com/spp/advance-article/doi/10.1093/scipol/scad008/7084754 by guest on 27 March 2023
