ArticlePDF Available

Mitigating the Gender Gap in the Willingness to Compete: Evidence from a Randomized Field Experiment

May 2016
Journal of the European Economic Association 17(4)

DOI:10.1093/jeea/jvy036

Authors:

Sule Alan

European University Institute

Seda Ertac

Koc University

The lower willingness of females to compete is extensively documented, and has a wide range of implications including gender gaps in occupational choice, achievement and labor market outcomes. In this paper, we show that one of the driving forces of competitive behavior is grit, a skill that is highly predictive of achievement, and evaluate the impact on competitiveness of a randomized educational intervention that aims to foster grit. The intervention is implemented in a sample of elementary schools, and we measure its impact using a dynamic competition task with interim performance feedback. We find that when children are exposed to an optimistic worldview that emphasizes the role of effort in achievement and encourages perseverance, the gender gap in the willingness to compete disappears. We propose the effect of this treatment on self-confidence and perseverance as a potential mechanism to explain the results.

Gender Dierence in the Willingness to Compete: Stage 1 and Stage 2

…

Distribution of Treatment Eects on Expected Payo Gains: Stage 1 and Stage 2

…

: Gender Gap in Competition Choice (Control Sample)

…

Gender Dierence in Relative Confidence: Stage 1 and Stage 2

…

: Gender Dierence in Performance (Control Sample)

…

Figures - uploaded by Sule Alan

Content may be subject to copyright.

Content uploaded by Sule Alan

Content may be subject to copyright.

Mitigating the Gender Gap in the Willingness to Compete:

Evidence from a Randomized Field Experimentú

Sule Alan, University of Essex

Seda Ertac, Koc University

May 2016

Abstract

The lower willingness of females to compete is extensively documented, and has a wide range of

implications including gender gaps in occupational choice, achievement and labor market outcomes.

In this paper, we show that one of the driving forces of competitive behavior is grit, a skill that

is highly predictive of achievement, and evaluate the impact on competitiveness of a randomized

educational intervention that aims to foster grit. The intervention is implemented in a sample of

elementary schools, and we measure its impact using a dynamic competition task with interim

performance feedback. We ﬁnd that when children are exposed to an optimistic worldview that

emphasizes the role of eﬀort in achievement and encourages perseverance, the gender gap in the

willingness to compete disappears. We propose the eﬀect of this treatment on self-conﬁdence and

perseverance as a potential mechanism to explain the results.

JEL Categories: C91, C93, D03, I28

Keywords: competition, gender, grit, randomized interventions, experiments

úContact information: Sule Alan: salan@essex.ac.uk, Seda Ertac: sertac@ku.edu.tr. We would like to thank the ING

Bank of Turkey for providing funding. Ertac thanks the Turkish Academy of the Sciences (TUBA-GEBIP program)

and Alan thanks the ESRC Research Center on Micro-Social Change (MISOC) for ﬁnancial support. We thank Thomas

Crossley, Thomas Dohmen, Uri Gneezy, Jonathan Meer and participants at the ASSA 2016 San Francisco meetings,

MISOC Workshop, UCSD Rady School of Management strategy and economics seminar, and the ROA Conference in

Maastricht University for comments and suggestions. We would also like to thank Elif Kubilay, Ipek Mumcu, Nergis

Zaim, Banu Donmez and Enes Duysak, as well as numerous other students who provided excellent research assistance.

All errors are our own.

1 Introduction

It is well-known that fewer women than men occupy top leadership positions in politics and the

corporate world, and fewer women are represented in high-paying occupations that involve competitive

paths. Gender diﬀerences in attitudes and preferences towards competition have been put forward as

an explanation for these ﬁndings; since if fewer women choose to compete, there will be less female

winners in the competition for top positions, or in ambitious careers that usually involve competitive

paths; see, for example, Gneezy et al. (2003), Niederle and Vesterlund (2007, 2011), Buser et al. (2014)

for lab evidence and Flory et al. (2014) for ﬁeld evidence. A number of studies also point out that

these diﬀerences emerge quite early in the life cycle and persist into adulthood1. From an economic

standpoint, an important concern here is eﬃciency: if males and females are equally able in tackling

a task and if females shy away from competition involving such a task, then winners of tournaments

will on average be less able than if males and females had similar entry rates2.

In the economics literature, a number of papers use laboratory experiments to explore the eﬀective-

ness of policies that aim to mitigate gender diﬀerences in competitiveness. For example, one strand

of the literature considers aﬃrmative action and preferential treatment through changes in tourna-

ment rules favoring women (e.g. Balafoutas and Sutter (2012), Niederle et al. (2013), Sutter et al.

(2015)). These studies ﬁnd that women enter tournaments more frequently with such policies, without

sacriﬁcing eﬃciency. Booth and Nolen (2012) show that single-sex schooling might eliminate gender

diﬀerences in competitiveness while Petrie and Segal (2014) ﬁnd that the gender diﬀerence disappears

if tournament prizes are high enough3. Implicit in these studies is that competitiveness may not be a

ﬁxed trait and is likely to be responsive to external and environmental factors. Consistently with this,

Almas et al. (2014) highlight the role of family background and socio-economic status in determining

competitive behavior. Gneezy et al. (2009) and Andersen et al. (2013) show that social structure

may inﬂuence competitiveness, with gender gaps being non-existent in matrilineal societies with more

gender equality.

In this paper, we take a unique approach to the issue of competitiveness. We conjecture that one

of the driving forces of competitive behavior is grit, a non-cognitive skill that has been shown to be

1There is a growing literature studying gender gaps in the willingness to compete among children and adolescents.

While some studies ﬁnd no gender diﬀerences at young ages in some countries (e.g. Cardenas et al. (2012), Khachatrian

et al. (2015)), in many societies a gender gap will have become apparent by the time adolescence is reached and will

persist (Andersen et al. (2013), Dreber et al. (2014), Sutter and Glatzle-Ruetzler (2015), Sutter et al. (2015)).

2Eckel and Fullbrunn (2015) show that one such implication of gender diﬀerences in attitudes relates to asset price

bubbles in ﬁnancial markets where all-male markets generate signiﬁcantly higher speculative bubbles than all-female

markets.

3Lee et al. (2014), on the other hand, ﬁnd that single-sex schooling has no eﬀect on the gender gap.

highly predictive of educational and occupational achievement; see Duckworth et al. (2007), Duckworth

and Quinn (2009), Maddie et al. (2012) and Eskreis-Winkler et al. (2014). Grit involves challenge-

seeking and passion for long-term goals, and it is closely linked to perseverance and tenacity. What

distinguishes a gritty individual is her beliefs regarding the role of eﬀort in the performance process and

her interpretation of performance feedback. A gritty individual tends to set ambitious performance

goals with the belief that those goals are attainable through persistent eﬀort, and attributes success

to hard work4. Choosing to compete essentially means setting an ambitious performance goal, and

sustained competitiveness in many cases requires perseverance in the competitive path by interpreting

failure and success constructively, rather than immediately attributing failures to a lack of ability and

successes to ability or luck. This is why we conjecture that there can be important interactions between

grit and competitiveness, especially in dynamic performance settings.

The aim of the paper is to explore whether the gender gap in the willingness to compete can be

mitigated in childhood by fostering grit. For this, we evaluate a unique educational program targeted

at elementary school children in Turkey. The program is implemented by children’s own (trained)

teachers for about a semester, with the help of a carefully designed curriculum5.Weprovideabrief

review of the content of the curriculum in Section 3. The intervention is designed as a randomized-

controlled trial and implemented twice in this manner using two independent samples. Each study

follows the same randomization procedure and uses the same education materials, providing us with

two large independent samples to estimate the treatment eﬀect on the willingness to compete.

Our outcome measures come from an incentivized mathematical real-eﬀort task, whereby children

choose to compete, receive performance feedback, then make a choice again. The dynamic nature of our

experimental task suits our purposes well, as in real-life performance settings individuals usually do not

make one-time, permanent choices but rather can observe how they fared, interpret the feedback they

receive and revise their decisions. Choosing whether to stick with or quit a diﬃcult degree program or a

competitive career after receiving negative performance feedback, or choosing whether to pursue more

ambitious and competitive paths after doing well in less ambitious ones are important decisions. Using

a competition task with interim feedback given to everyone regardless of the chosen incentive scheme

helps us explore the relationship between competitiveness and grit, as well as the potential mechanisms

4Recent research in psychology shows that believing that skills are not ﬁxed but malleable through eﬀort, i.e. having

a “growth mindset” can increase motivation, perseverance and achievement; see Dweck (2006), Blackwell et al (2007).

5Alan, Boneva and Ertac (2016) show that this intervention is highly eﬀective in inducing ambitious goal-setting,

perseverance and eventual skill accumulation. It also leads to a signiﬁcant increase in standardized test scores, without

any heterogeneity in such eﬀects with respect to gender.

through which an intervention that targets grit might be aﬀecting the willingness to compete6.

We ﬁrst show that, consistently with the literature, the decision to compete in the ﬁrst stage and in

the second stage are highly correlated with self-conﬁdence, risk tolerance and cognitive ability in the

baseline. We add to that a new piece of evidence that grit is also highly predictive of competitiveness:

a one standard deviation increase in the grit score is associated with a 3 percentage point increase in

the probability of competing in the ﬁrst stage, and a 5 percentage point increase in the second stage7.

We then turn to competitive behavior across gender. In our control sample, there is a statistically

signiﬁcant 8 percentage point gender gap in the willingness to compete in the ﬁrst stage of the com-

petition task. Despite the fact that there is no gender diﬀerence in actual performance and therefore

in the probability of receiving negative performance feedback, this gap slightly widens and becomes

10 percentage points in the second stage competition, after feedback. The size of this gap appears to

be independent of the type of feedback received in the ﬁrst stage.

We proceed by documenting that there is no gender gap in the willingness to compete in the

treatment group, either in the ﬁrst or in the second stage. We estimate a statistically signiﬁcant

treatment eﬀect on the willingness to compete for both boys and girls in the ﬁrst stage and this eﬀect

is larger for girls, mitigating the ﬁrst stage gender gap. Moreover, we ﬁnd a statistically signiﬁcant

treatment eﬀect on competitiveness after feedback only for girls, entirely eliminating the second stage

gender gap. The eﬀect on girls is of considerable size: the propensity to compete in the second stage

is about 16 percentage points higher for treated girls relative to untreated girls. These results provide

strong evidence that fostering grit in the classroom environment goes a long way in mitigating the

gender gap in the willingness to compete. Finally, we also show that the higher willingness to compete

induced by the treatment brings signiﬁcantly higher rewards to treated girls relative to the girls in the

control group, implying signiﬁcant eﬃciency gains. Our rich data allow us to contemplate a potential

mechanism that points to improved self-conﬁdence and grit in girls, to explain these results.

The paper contributes to several distinct strands of the literature. First, by being the ﬁrst study

that documents a strong link between grit, an important but understudied non-cognitive skill, and the

willingness to compete, it contributes to the growing literature that strives to understand the role of

non-cognitive skills in achievement outcomes; see Borghans (2008), Heckman et al (2006, 2010, 2013,

2014), Almlund et al (2011), Kautz et al (2014). Second, by using a randomized-controlled design

6Andersen et al. (2014) consider a setting with dynamic competition in a matrilineal and patriarchal society, and

ﬁnd that women in the matrilineal society are more motivated by positive performance feedback.

7The grit score is a standardized factor extracted using a number of item-set questions mostly adapted from the

Duckworth grit scale; see Duckworth (2009). The speciﬁc questions we use to obtain this score are given in the Appendix.

and showing for the ﬁrst time the causal impact of an educational intervention on the gender gap

in competitiveness, it contributes to the large literature on gender and competition; see Croson and

Gneezy (2009) and Niederle and Vesterlund (2011) for reviews. Besides opening new and promising

research avenues, the evidence documented in this paper also provides crucial input into policy actions

targeting gender gaps in achievement, and oﬀers speciﬁc cost-eﬀective recommendations that can be

implemented in the classroom environment. Finally, our two-stage design and related results highlight

the importance of considering competitiveness in a dynamic context, and makes novel contributions to

the recent literature on performance feedback and its eﬀects on choices and performance (Azmat and

Iriberri (2010), Barankay (2011), Eriksson et al. (2009), Ertac and Szentes (2011), Gill and Prowse

(2014), Wozniak et al. (2014)).

The rest of the paper is structured as follows: Section 2 provides background information on

the program we evaluate and presents the evaluation design, Section 3 describes the content of the

intervention, Section 4 describes our experimental outcome measures, Section 5 presents the results,

and Section 6 provides a discussion and concluding remarks.

2 Program Background and Evaluation Design

The Turkish Ministry of Education encourages schools and teachers to participate in socially useful

extra-curricular programs oﬀered by the private sector, NGOs, the government and international orga-

nizations. All elementary school teachers are given a maximum of 5 hours per week to be involved in

these programs. Their participation is voluntary and if they choose not to participate in any program,

there is no restriction on the way in which these hours are used. The program we evaluate in this paper

is implemented in state-run elementary schools in Istanbul as an extra-curricular project under the

oversight of the Education Directorate of Istanbul. The main objective of the program is to improve

key non-cognitive skills in elementary school children in the classroom environment by training their

teachers.

The program is designed as a randomized-controlled trial and implemented twice in this manner,

using two independent samples of elementary schools all across Istanbul. Our ﬁrst sample comes

from an intervention that had two main treatment arms, each of which had a speciﬁc behavioral

target. The ﬁrst arm aimed to improve the ability to make decisions in a forward-looking manner and

encourage patience; see the evaluation of this arm in Alan and Ertac (2014). The second arm, which

was implemented in the same schools after the implementation of the ﬁrst arm, aimed to foster grit.

In the second sample, only the second arm of the ﬁrst study, namely the grit arm, was implemented.

The ﬁrst sample consists of 23 schools (about 1300 students), where a version of a phase-in design

was implemented. Initially, students in a randomly selected 9 schools received the patience treatment

for about 8 weeks, while others were in the control group. The same children in these 9 schools received

the grit treatment in the following semester. While the initial treatment group was receiving the grit

treatment, a randomly selected set of 6 schools, which were previously in the control group, received

the patience treatment. Remaining schools were kept as control8. With this design, we can evaluate

the independent impact of the patience treatment on our outcome measures, and indeed show that the

patience treatment by itself does not have any impact on competitiveness (see Table 16 in Appendix)).

Still, we would not be able to isolate or rule out potential complementarities across grit and patience

training using this study sample. Such an issue does not arise in the second study.

In the second study, we randomly assign only the grit training arm of the ﬁrst study across a new

set of schools in Istanbul. The intervention follows the same procedures, with the same curricular

materials and the same teacher training approach. This sample consists of 16 schools (8 treatment, 8

control) and has a total of about 1,300 students.

In both studies, the randomization was performed in the following way. First, the Istanbul Direc-

torate of Education sent the oﬃcial documentation of the program to all elementary schools in major

districts of Istanbul9. The teachers in these schools were then contacted in random sequence and

invited to participate in the program. Teachers were informed that upon participation they would be

assigned to diﬀerent training phases within the coming two academic years. All teachers who agreed

to participate were promised to eventually participate in training seminars and receive all training

materials, but they were not told when within the next two academic years they would receive the

treatment until the random assignment was completed. The promise of the training oﬀer was made

to the teacher and not to current students, i.e., while children in control groups never received the

training as they moved on to middle school after 4th grade, their teachers did, albeit at a later time.

This was done in order to allow for long-term follow-up.

Once a teacher stated a willingness to participate, we randomly assigned their school into the

8The original program sample of the ﬁrst study includes 36 schools but a random subset of these schools (13 schools)

were exposed to another treatment (a role model exposure intervention) before we collected our follow-up data in May

2014, therefore we remove those schools from our current analysis. A previous version of this paper included evaluation

of that treatment as well as the one we evaluate here but it did not include sample 2 data. Sample 2 was collected later

with the purpose of replication and exploring complementarities that might have arisen in the ﬁrst study. By increasing

the number of clusters, combining Sample 1 and Sample 2 allows us to make powerful inference.

9The program was titled “ﬁnancial literacy, savings and economic decisions” and no further information on the

particulars of the program were disclosed to the teachers prior to the teacher training seminars.

treatment or the control group. Note that in both studies the unit of randomization was the school,

not classroom, in order to prevent potential spillovers across classrooms. In the ﬁrst study, a given

school where there was a teacher willing to participate had a 40% ex-ante chance of being assigned

to the initial treatment group (patience+grit), a 30% chance to be in the second treatment group

(patience only) and 30% chance to be assigned to the control group. In the second study, a given

school where there was a teacher willing to participate had a 50% chance of being in the treatment

group. For each study, we stopped recruiting teachers when we hit the logistical constraint of being

able to physically visit the classrooms10.

The sample generated with this design contains schools in which at least one teacher stated their

willingness to participate in the program. Therefore, the estimated impact of the program is the average

treatment eﬀect on the treated and in principle, is not readily generalizable to the population. However,

in the ﬁrst study, approximately 60% of the contacted teachers accepted our oﬀer and the most common

reason for non-participation was being “busy with other projects, although happy to participate in this

program at a later date” (about 20%). The rest of the non-participation was due to “impending transfer

to a school in another city, with a willingness to participate if the program is implemented there” (about

5%), and “not being in a position to participate due to private circumstances” (about 10%). In study

sample 2, acceptance of the training oﬀer reached 80%. Given these numbers, we conjecture that the

external validity of our results is strong.

In the ﬁrst study, baseline data were collected in Spring 2013, the ﬁrst intervention (patience) was

implemented in Spring 2013, and the intervention on grit was implemented in Fall 2013. The follow-up

data were collected in May 2014. In the second sample, baseline data were collected in Spring 2015,

the intervention (grit only) was implemented in Fall 2015, and the follow-up data were collected in

January 2016. Full details of the evaluation design for each study sample are given in Table A in the

appendix.

After establishing that the two samples represent the same population using the baseline charac-

teristics, we pool them to increase the number of clusters (number of schools) and deﬁne the treated

child as one who received the grit treatment, either as grit only (as in Study 2) or grit plus patience (as

in Study 1). With this, we have a total of 17 schools in treatment and 22 schools in the control, giving

us about 2600 students in total. We provide additional analyses based on disaggregated treatment

10In order to ensure data quality, authors of the paper coordinated the ﬁeld logistics, trained a select group of students

and experienced interviewers to assist with data collection, and physically visited all classrooms to implement the tasks

and collect data. All measurements were conducted with the approval of the local IRB and the permission of the Ministry.

status in the Appendix.

3 The Educational Intervention on Grit

The educational initiative required the production of a rich set of educational materials, which involved

a broad interdisciplinary endeavor. While the target concepts of the materials were determined by the

authors, speciﬁc contents (e.g. scripts) were shaped with input from an interdisciplinary team of

education psychologists, a voluntary group of elementary school teachers, children’s story writers and

media animation artists, according to the age and cognitive capacity of the students.

The program involves providing animated videos, mini case studies and classroom activities that

highlight i) the plasticity of the human brain against the notion of innately ﬁxed ability, ii) the role

of eﬀort in enhancing skills and achieving goals, iii) the importance of a constructive interpretation

of setbacks, failures and success, and iv) the importance of goal setting. The aim of the training is

to expose students to an optimistic worldview in which any one of them can set goals in an area of

their interest and can work towards these goals by exerting eﬀort. The materials highlight the idea

that in order to achieve these goals, it is imperative to avoid interpreting immediate failures as a lack

of innate ability or intelligence. This worldview encompasses any productive area of interest, whether

it be music, art, science or sports11 . Visual materials and stories are supplemented by classroom

activities created and supervised by teachers, based on general suggestions and guidelines put forward

in teacher training. It should be noted that there is no mention of competitiveness or emphasis on

doing better than others in the materials, and success in the examples given to children means reaching

individualistic goals, such as mastering a task that one ﬁnds challenging.

Weekly topics, main materials and supplementary activities are very clearly deﬁned and speciﬁc

guidelines on how to structure each lesson are prepared for the teachers (the teacher kit). However,

the program is not merely a set of materials to be covered in a speciﬁed period of time, like a common

curriculum item. Instead, it aims to change students’ beliefs about the role of eﬀort in performance

processes and the potential return to perseverance partly by changing the mindset of the teachers and

the nature of the classroom environment. To this end, in addition to covering the curricular items

11To give an example, in an animated video, two students who hold opposite views on the malleability of ability engage

in a dialog. The student who believes that ability is innate and therefore there is no scope for enhancing it through

eﬀort, points out that the setbacks she experiences are reminders of the fact that she is not intelligent. Following this

remark, the student who holds the opposite view replies that setbacks are usually inevitable on the way to success; she

interprets them as opportunities to learn, and therefore, they do not discourage her. The curriculum with all covered

topics and classroom activities are available upon request.

as suggested, teachers were strongly encouraged to adopt a teaching philosophy that emphasizes the

role of eﬀort in everyday classroom tasks, e.g. while giving performance feedback and interpreting test

results.

4 The Outcome Measure: A Two-Stage Competition Task

Our outcome measure is designed to estimate the eﬀect of the intervention on initial competitiveness

and competitiveness in response to performance feedback. The task we use is an addition task, which

involves adding two 2-digit numbers and one single-digit number. The task consists of three perfor-

mance periods and children are given 2.5 minutes per each period. One of these periods is selected

randomly at the end, and rewards are determined based on the performance and decisions in the se-

lected period. For all three periods, each student is matched with another student from a diﬀerent

school (whom we will call “opponent” hereafter), who had done the same tasks before and whose

performance was recorded12.

In the ﬁrst period, students perform the addition task under a piece-rate incentive scheme, whereby

they receive 1 token for every addition they are able to do correctly. In the second period (the ﬁrst

competition choice stage), students have a choice between piece-rate and competing with their matched

opponent. If they choose the piece-rate, they are rewarded with one token per correct answer. If

they choose the tournament, they are rewarded with 3 tokens per correct answer, but only if their

performance exceeds that of the opponent. In the case of having a lower performance, a student that

chooses to compete receives zero tokens, whereas in the case of a tie, she gets 1 token per correct

answer.

After choosing the incentive scheme, we elicit beliefs with the purpose of exploring the role of

self-conﬁdence in incentive scheme choices. Speciﬁcally, children are asked to state their beliefs about

(1) the number of correct answers they will have, (2) the number of correct answers of their opponent.

These beliefs are incentivized in the following way: Children are told that three people in their class

will be randomly chosen after the experiment ends, and these three will get an extra small gift for each

correct guess13. After performing the 2nd period task, they receive feedback about (1) the number of

additions they were able to do correctly, (2) whether their performance was better than, worse than,

12We implemented the same tasks in the same order in three classrooms in three pilot schools, and recorded perfor-

mances. Each student in our sample is randomly matched with one student from this “benchmark sample”.

13Children are also reminded that there is always an incentive to do as many additions as they can in the actual task

and it does not make sense to stop just to be consistent with their guess.

or equal to that of the opponent. In the 3rd period (the 2nd competition choice stage), children make

a choice again between piece-rate and competition against the same opponent, to be implemented if

that period is chosen for payment. After making this choice, they state beliefs again about their own

and opponent’s performance, in the same way as in the ﬁrst choice stage. This allows us to understand

how children update their relative self-conﬁdence in response to performance feedback. If the child

chooses to compete in any period, her results are compared with the performance of the opponent in

the corresponding period, and rewards are determined based on this comparison.

Children learn the gender of their (randomly assigned) opponent before making their initial com-

petition choice. That is, they know whether they are matched with a boy or a girl. Before this, there

is also an incentivized question that elicits gender stereotypes: children are told about one girl and one

boy, randomly selected from a diﬀerent school/class, and are asked to guess whether the girl or the

boy has done better in this same task. If they guess who actually did better correctly, they get a small

extra gift. Children were given workbooks, and instructed to turn over pages only at speciﬁed times.

Each child’s workbook had a code that matched it to a set of actual performances (in the second and

third periods of the task) that came from a student in one of the pilot classes. The workbooks were

randomly distributed, rendering a random match in terms of opponent gender and opponent perfor-

mance. After the task is completed in the second period (the ﬁrst choice stage), feedback was given to

children by experimenters by writing the child’s own actual performance and circling the outcome of

the performance comparison (the child having a worse, better or equal performance compared to the

matched child) on the sheet. Instructions are in the Appendix.

5 Results

We have data from approximately 2600 students in 39 schools. About 51% percent of our sample is

male, and the average age is about 9.5 years, as is typically the case for 4th graders in Turkey. We

collected baseline information on students using surveys, cognition tests and experimental elicitation

tasks. While some survey information come directly from children, some are obtained via teacher

assessment surveys. These variables allow us to test the balance of the treatment groups in our data

and provide us with useful covariates that are predictive of our outcome measures.

5.1 Internal Validity

We ﬁrst check whether our pooled data are balanced across treatment status with respect to a number

of student characteristics, collected at the baseline stage. Table 1 shows results from ordinary least

squares regressions of baseline variables on the treatment dummy. While the ﬁrst column gives the

mean of the control for the respective variable, column 2 shows the diﬀerence from the control’s mean.

Panel 1 presents the balance results for demographic variables and baseline attitudes either reported

by the child or the teacher, and Panel 2 presents the balance results pertaining to three variables that

come from our competition task.

The variables in Panel 1 are constructed as follows: Age is reported by the student, while student’s

family income (wealth) and her overall academic standing (academic success) are reported by the

teacher using a 5-point scale. Cognitive ability is measured using Raven’s progressive matrices test

(Raven et al (2004)) and risk tolerance is measured using a task based on Gneezy and Potters(1997)14.

The grit score is a standardized factor extracted using a number of item-set questions mostly adapted

from the Duckworth grit scale; see Duckworth (2009). The speciﬁc questions we use to obtain this

score are given in the Appendix.

As can be seen from the table, diﬀerences are not statistically diﬀerent from zero across treatment

and control for any of the variables. This ensures us that the pooled data are balanced across treatment

status and our results are internally valid. An important ﬁnding to note in this table (Panel 2) is that

there is no diﬀerence in piece-rate performance across our treatment groups. Nor is there any diﬀerence

in the probability of receiving negative feedback after the ﬁrst stage. We will refer to these ﬁndings

when we discuss the gender gaps in competitiveness we observe in the baseline in the next section.

5.2 Willingness to Compete and Gender Gaps in the Control Group

Before moving on to estimating the eﬀect of the intervention, we study the gender gap in the baseline.

Focusing on elementary school children and using a dynamic version of a well-known experimental

task, this initial analysis provides new evidence on the prevalence of the gender gap in the willingness

to compete and gives us the baseline gender gap ﬁgures upon which our treatment operates. Table 2

presents marginal eﬀects from logit regressions where the dependent variables are the binary competi-

14In this task, children have an endowment of 5 tokens, which they can allocate between a riskless and a risky option.

Tokens allocated into the risky option, which is conveyed as putting tokens in a “risky bowl”, are tripled with 50%

chance and lost with 50% chance, based on the color of a ball drawn from an opaque bag. Tokens that are not allocated

into the risky bowl are safe. The number of tokens placed into the risky option is a measure of risk tolerance.

tion choices in stage 1 and stage 2. The ﬁrst point to note in this table is that there is a statistically

signiﬁcant gender gap between girls and boys in the ﬁrst stage, with girls about 8 percentage points

less likely to choose to compete (see column 1). The gender gap remains, and in fact slightly widens

in the second stage, after feedback: We estimate that girls’ willingness to compete is now about 10

percentage-points lower than boys, although the diﬀerence in the gap between the two stages does not

reach statistical signiﬁcance (p-value=0.38).

A major question that the gender-competition literature has focused on is the determinants of com-

petitiveness in girls and boys, which also provides insights about the nature of the gender gap. Columns

2 and 4 of Table 2 include the main potential determinants of competition choice as explanatory vari-

ables: risk tolerance and beliefs (relative self-conﬁdence, deﬁned as the expected number of correct

answers for oneself minus that for the opponent), in addition to opponent gender, teacher gender and

cognitive ability score (Raven score). We add to that the baseline grit score to show the relationship

between competitiveness and grit, which provides the basis for investigating the eﬀect of grit training

on competitiveness, the main motivation of the paper. The variables relative self-conﬁdence, Raven

score and grit score are normalized to facilitate the interpretation of the coeﬃcient estimates.

As expected, higher expected performance relative to the opponent, risk tolerance and cognitive

score increase the propensity to compete in both stages. Note that opponent gender seems to have no

eﬀect on the willingness to compete in either stage. We ﬁnd that the grit score is also highly predictive

of competition choice in both stages. The predictive power of the grit score in the second stage is of

considerable size, rivaling that of risk tolerance and cognitive score: a one standard deviation increase

in the grit score is associated with a 5 percentage-point increase in the willingness to compete after

feedback.

What is the reason for the persistent gender gap in the willingness to compete? One explanation is

that if girls perform generally worse than boys, we could observe girls (rationally) having a higher ten-

dency to shy away from competition. This would be especially true in the second stage, after receiving

performance feedback. However, there is absolutely no gender gap in either piece-rate performance or

the ﬁrst stage performance. As can be seen in Table 3, the performance of boys and girls is not statis-

tically diﬀerent in either the piece-rate stage or the ﬁrst choice stage, and naturally, their propensity

to receive negative/positive feedback after the ﬁrst stage is also statistically the same (41.1% of girls

and 43.6% of boys receive negative feedback, p-value of the diﬀerence=0.38). Interestingly, as shown

in Table 4 the observed gender gap in the second stage seems independent of the type of feedback

received in the ﬁrst stage: girls’ willingness to compete in the second stage is lower than boys among

the children that received positive feedback as well as those that received negative feedback. We now

turn to the question of whether this gap is also prevalent in the treatment group.

5.3 Willingness to Compete and Gender Gaps in the Treatment Group

Table 5 shows that in the treatment group, boys and girls are equally likely to compete, both in the

ﬁrst and the second stage. That is, the gender gap is erased in the treatment group. Figure 1 shows

this in visual clarity: it depicts the coeﬃcient estimates and conﬁdence bands on the male dummy

in logit regressions that are run separately for treatment and control, in each of the two stages of

competition (controlling for all covariates shown in Table 5). It can be seen here that the gender gap,

which exists already in the ﬁrst stage and remains in the second stage in the control group, is closed

in the treatment group. Turning to Table 5, we observe that relative conﬁdence, risk tolerance, Raven

score and grit score are predictive of competitiveness in the treatment group, with grit score emerging

again as a signiﬁcant predictor in both stages. Table 6 shows that there is no gender gap in the second

stage, regardless of the type of feedback received after the ﬁrst stage.

How does the grit treatment work to close the gender gap? Are boys’ competitiveness unaﬀected

by treatment while girls’ is? Do both genders become more competitive, with girls more so? We now

turn to estimate the eﬀect of our treatment on competitive behavior in the ﬁrst and the second stage,

on the whole sample as well as for boys and girls separately.

5.4 Treatment Eﬀects on Competition Choice

In order to test the null hypothesis that the treatment had no impact on the experimental outcome

ys, competition choice in stage s={1,2}, we estimate the average treatment eﬀect by conditioning

on baseline covariates:

ij =–0+–1T reatmentj+Xij “+Áij

where the dependent variable ys

ij is a dummy variable which equals 1 if student iin school jchose to

compete in stage s. The binary variable T reatmentjindicates the treatment status of school j, and

Xij is a vector of observables for student iin school jthat are potentially predictive of the outcome

measures we use. These observables include opponent gender (randomly assigned across all students),

cognitive score and risk tolerance collected at the baseline, and teacher gender. Note however that the

estimated eﬀect sizes are not expected to be aﬀected by the inclusion of these covariates due to random

assignment. The estimated coeﬃcient of the treatment dummy ˆ–1yields the average treatment eﬀect

on the treated. Estimates are obtained via logit regressions since the outcome considered here is binary.

In all empirical analyses where we estimate treatment eﬀects, standard errors are clustered at the level

of the school, which is the unit of randomization.

Table 7 presents the estimated treatment eﬀects on the willingness to compete in the ﬁrst stage. The

ﬁrst ﬁnding to note here is that the ﬁrst-stage competition choices of all children are very responsive to

the treatment. This shows that willingness to compete is malleable in children. The second column in

the same table shows that the intervention is quite eﬀective in inducing competitive behavior among

girls: treated girls are about 15 percentage points more likely to compete than those in the control

group. Boys’ competitiveness also responds positively to the treatment. Treated boys are about 9

percentage points more likely to compete relative to those in the control group. Both eﬀects are

statistically signiﬁcant at the 1% level. Although the point estimate for girls is larger, the diﬀerence

between the treatment eﬀect sizes does not reach statistical signiﬁcance (p-value=0.12). Note, however,

that the diﬀerence in the treatment eﬀects across gender is suﬃcient to mitigate the ﬁrst stage gender

gap observed in the control sample; see Table 5 and Figure 1.

5.5 Treatment Eﬀects on Competition Choice after Feedback

The two-stage nature of our experimental task provides us with an outcome measure that is useful for

assessing the impact of the intervention on behavior after receiving performance feedback. Recall that

before making the second competition decision, students receive feedback. Speciﬁcally, they ﬁnd out

their absolute performance (how many correct answers they had) and whether they did better, worse or

equally well relative to their opponents in the ﬁrst stage. This feedback is given to everyone, regardless

of the choice of incentive scheme in the ﬁrst stage. Performance in the ﬁrst competition stage and the

feedback received based on this performance are balanced across treatment groups (p-value=0.72 and

0.38, respectively).

Table 8 presents the estimated treatment eﬀects on the propensity to compete in the second stage.

Similar to the ﬁrst-stage results, treatment leads to an overall increase in the propensity to compete in

the second-stage, after feedback. However, here, we see a signiﬁcant gender diﬀerence in the response

to treatment: Grit training is very eﬀective in making girls more competitive in response to feedback.

The estimated eﬀect is of considerable size: treated girls are about 16 percentage points more likely

to compete relative to untreated girls. In the case of boys, however, the treatment does not have a

statistically signiﬁcant eﬀect on competitive behavior after feedback. Contrary to the ﬁrst stage results,

the estimated coeﬃcients for girls and boys are signiﬁcantly diﬀerent (p-value=0.002). These results

suggest that while the intervention inﬂuenced the competitiveness of all children, it has diﬀerential

impact across gender in the second stage, after feedback. We will explore the possible factors behind

this result in Section 5.7.

5.6 Treatment Eﬀects on Payoﬀs

Based on performance, for some children it is payoﬀ-maximizing to compete, while for others it is better

to stay out. One can be concerned that interventions such as the one we evaluate in this paper may

lead to unintended inferior outcomes for some children, by inducing decisions that turn out to be bad

for payoﬀs ex-post. Analyzing how children’s decisions fare in terms of expected material payoﬀs can

shed light on these issues. To do this, we ﬁrst calculate the probabilities of winning and tying for any

given performance level, using the empirical distribution of performances. Using these probabilities

along with realized performances, we calculate each child’s expected payoﬀfrom competition, and

analyze whether the child’s actual choice was payoﬀmaximizing ex-post. We do this separately for

each competition stage15. We then estimate the treatment eﬀects on this outcome for stage 1 and 2.

Table 9 presents the estimated treatment eﬀects on the propensity to make payoﬀ-maximizing

choices. We ﬁrst note that overall, a little more than half of all children make a pay-oﬀmaximizing

choice in the control group (52% for Stage 1 and 59% for Stage 2). Next, we observe that treatment

does not cause inferior outcomes on average. On the contrary, treated children appear to be more

likely than untreated children to make the best choice from an expected payoﬀstandpoint. Moreover,

this favorable result seems to be driven by girls: treated girls are about 5 to 6 percentage points more

likely to make a payoﬀ-maximizing choice than untreated girls in both stages. The eﬀect on boys, on

the other hand, is not statistically diﬀerent from zero in either stage.

Next, we look at the estimated treatment eﬀects on expected payoﬀdiﬀerences between the chosen

and unchosen incentive scheme for each stage. The expected payoﬀgain from having chosen a certain

incentive scheme is constructed as the diﬀerence between the expected payoﬀfrom choosing competition

and the (alternative) payoﬀfrom choosing piece-rate for a child who chooses competition, and vice

versa for a child who chooses piece rate. This measure gives us the child’s expected gain from choosing

15It should be noted that this analysis does not make utility comparisons and therefore it is not an optimality analysis

per se, as it disregards eﬀort costs, which are unobservable.

competition over piece-rate (or the other way around), which may be positive or negative ex-post,

depending on the performance of the child16. In Table 10, we clearly see that treated girls fare

signiﬁcantly better than untreated girls in terms of expected gains, while boys’ expected gains are

unaﬀected by the treatment. This analysis provides further evidence that the treatment does not lead

to inferior outcomes on average, either for boys or for girls. On the contrary, it leads to better outcomes

on average for girls.

Although the results are quite positive, we acknowledge that these are average eﬀects, which may

conceal undesired distributional eﬀects. In order to see the distributional impact of the treatment, we

estimate quantile treatment eﬀects on expected payoﬀgains. The four panels in Figure 2 depict the

coeﬃcient estimates and 95% conﬁdence bands for expected payoﬀgains for boys and girls in each

competition stage. At ﬁrst glance, the ﬁgures show that treatment eﬀects are generally non-negative

across quantiles in both stages, for both boys and girls. For girls, the eﬀects are positive across

quantiles in both stages, and statistically signiﬁcant in low and high quantiles. This is because very

low gains (large negatives) result from under-entry into the tournament by girls that would have done

well in the tournament, which is signiﬁcantly lessened by the treatment. Positive treatment eﬀects at

high gains are indicative of optimal entry into competition (from a payoﬀ-maximization perspective),

which is signiﬁcantly more prevalent in the treatment group. Taken together, these results suggest

that the increase in competitiveness brought about by grit training does not harm girls (or boys) in

terms of payoﬀs at any point in the distribution. On the contrary, the intervention results in signiﬁcant

eﬃciency gains, suggesting that many able girls are staying away from competition in the baseline.

Further support for our evidence of eﬃciency gains on the part of girls comes from our analysis

of heterogeneous treatment eﬀects based on cognitive function, as measured by Raven’s progressive

matrices. Speciﬁcally, we re-estimate treatment eﬀects on Stage 1 and Stage 2 choices for students

with below and above mean cognitive scores. Table 11 presents the results of this analysis for males

and females separately. Estimates presented in this table clearly show that the treatment eﬀects on

competition choices are similar for low and high cognitive scores within females and within males. A

notable ﬁnding here is that the treatment seems to be very eﬀective in inducing both groups of girls

to choose to compete, in both stages. This immediately brings to mind that treated girls with low

cognitive scores may fare worse than those in the control group in terms of expected gains. However,

Table 12 shows that this concern is unwarranted. While we observe that treated girls with high

16For calculating expected payoﬀs from competition, the empirical win and tie probabilities corresponding to the

child’s actual performance is used.

cognitive ability are signiﬁcantly more likely to make payoﬀmaximizing choices in both stages (10

percentage points in Stage 1 and about 6 percentage points in Stage 2), there is no negative eﬀect

on girls with lower cognitive ability in either stage: the average treatment eﬀects are statistically

insigniﬁcant for this group (see columns 1 and 5). For boys with high cognitive ability, we estimate

a statistically signiﬁcant treatment eﬀect on payoﬀmaximizing choices in Stage 1 only (see columns

4 and 8). The estimated treatment eﬀects for boys with low cognitive ability are similar to those we

obtain for girls with low cognitive ability, statistically zero.

5.7 Mechanisms of Impact

Which aspects of the treatment are likely to drive girls and boys into competition, and what is the

reason for the diﬀerential impact that virtually eliminates the gender gaps observed in both stages

in the baseline? It may be that multiple factors are at work, and the relative weight of these may

be diﬀerent for girls and boys. While we cannot possibly account for every factor that could play

a role, we can explore several major factors that might be relevant. One such factor that underlies

competition choice is beliefs. Recall that we elicit beliefs by asking children about their absolute

and relative expected performance, in both choice stages. More speciﬁcally, children are asked how

many questions they expect to answer correctly, and how many they expect their opponent to answer

correctly. We construct a relative self-conﬁdence measure by taking (and normalizing) the diﬀerence

between own expected performance minus the expectation about the opponent’s performance, which

can be a positive or a negative number. As a backdrop to what follows, we ﬁrst note that in the

control group, controlling for actual performance, boys are more self-conﬁdent than girls in both the

ﬁrst stage and the second stage (p-values are 0.03 and 0.08 respectively). This is consistent with the

literature on gender gaps in conﬁdence in adults17. We then turn to analyzing the treatment eﬀect on

relative self-conﬁdence, among girls and boys.

Table 13 columns 1 and 4 show that children’s initial conﬁdence levels are positively aﬀected by the

treatment. However, as can be seen in columns 2 and 5, these eﬀects seem to be coming mainly from

girls. Speciﬁcally, treated girls’ average conﬁdence score is about 0.16 standard deviations higher than

that of untreated girls in the ﬁrst stage, and this eﬀect is statistically signiﬁcant at the 5% level. The

eﬀect size is about 0.13 standard deviations in the second stage, and again statistically signiﬁcant at

the 5% level. While the estimated treatment eﬀects are also positive for boys, they are not statistically

17see Croson and Gneezy (2009) for a review.

diﬀerent from zero. The eﬀect of the treatment on relative conﬁdence in both stages can be seen in

Figure 3 as well. Here, we see that the gender gaps in conﬁdence we observe in the control group do

not exist in the treatment group. Given that the piece-rate performance of girls is not diﬀerent across

treatment status (p-value=0.70), this increase in relative beliefs is not driven by improvements in task

performance through treatment but is likely to indicate a true increase in self-conﬁdence for similar

performance. This self-conﬁdence improvement can explain why treated girls opt to compete more.

This ﬁnding is our ﬁrst piece of evidence that the potential mechanism through which the treatment

mitigates the gender gap may be a boost in task-related conﬁdence experienced by girls that received

the treatment18.

To explore the belief channel further and in particular to understand the widely diﬀerential treat-

ment eﬀects across gender in the second stage, we construct another outcome measure related to beliefs.

This measure is a dummy variable that takes the value of 1 if the child expects a better performance

from herself in the second stage relative to her actual ﬁrst stage performance. Recall that we inform

every child about their absolute performance after the ﬁrst stage is completed and ask them to guess

their own performance (and their opponent’s performance) in the second stage before the second stage

starts. Table 14 presents the estimated treatment eﬀects on this measure. Note ﬁrst that about 51%

of the girls and and 54% of the boys believe that they can improve upon their ﬁrst stage performance

in the control group. On this measure, we estimate an about 7 percentage point treatment eﬀect for

girls, which is signiﬁcant at the 1% level. For boys, we estimate a statistically zero treatment eﬀect.

This analysis provides further evidence on the importance of the belief channel in explaining the treat-

ment eﬀects on competitiveness. The treatment seems to induce more optimistic beliefs about future

performance on the part of girls for a given prior performance: treated girls are signiﬁcantly more

likely to believe that they can improve upon their past performance than those in the control group.

This belief may lead them to opt into competition more in the second stage.

The previous analysis brings to mind that grit, which is expected to be positively inﬂuenced by the

treatment, may play a role, especially in the second stage. To explore this potential mechanism we

make use of our surveys, which are administered both at baseline and follow-up. First, we explore the

gender gap in the grit score at baseline and ﬁnd that boys score 0.8 standard deviations higher than

girls on grit and the p-value of this estimate is 0.02. Against this ﬁnding, we estimate a statistically

insigniﬁcant gender gap for the treatment group. Figure 4 summarizes these results. Next, we estimate

18We do have more general conﬁdence measures unrelated to our competition task, collected at baseline and follow-up.

We do not estimate a signiﬁcant treatment eﬀect on those measures.

the eﬀect of the treatment on the grit score, controlling for the baseline grit score. Table 15 presents

the results. Here, we see that treated students report signiﬁcantly higher grit than those in the control

group (the eﬀect size is 0.31 standard deviations). While we estimate signiﬁcant eﬀects for both boys

and girls, the eﬀect size for girls (0.37 standard deviations) is signiﬁcantly larger than that of boys

(p-value of the diﬀerence is 0.08). This is consistent with the ﬁnding that grit is one of the driving

forces of competitive behavior after feedback and with the result that treated girls are signiﬁcantly

more likely to opt into competing than untreated girls after feedback.

While beliefs and improved grit/perseverance are strong candidates for potential mechanisms

through which the intervention may have aﬀected competition choices, there are other possibilities.

One such possibility is that the treatment might have aﬀected risk tolerance levels, inducing girls to

take more risk and choose to compete. While we do not have data on post-treatment risk tolerance for

sample 2, we have data to compare risk tolerance across the groups for sample 1. In unreported regres-

sions, we ﬁnd that the grit treatment does not have a signiﬁcant eﬀect on risk tolerance (p=0.36). This

suggests that the impact of grit training on competitive behavior does not work through increasing

the tolerance for risk.

Overall, improved self-conﬁdence and increased grit on the part of girls are likely to play a large

role in generating the diﬀerential treatment eﬀects we estimate. By boosting their conﬁdence and per-

severance, treatment signiﬁcantly reduces under-entry into the tournament and leads girls to compete

as much as boys do. A possible reason for this is the strong emphasis on ambitious goal setting, the

role of eﬀort and the value of perseverance in the training material. A child who is exposed to this

optimistic view on performance and achievement may now believe that she can tackle a challenging

task, and she can always improve her performance with higher eﬀort, which would lead to increased

selection into the tournament, regardless of the feedback received.

6 Concluding Remarks

Documenting, understanding the reasons for, and exploring policies to mitigate gender gaps in compet-

itiveness has been a very active area of research in economics in recent years. Considering competitive

behavior in a dynamic context where individuals choose an incentive scheme, receive performance

feedback and make a decision again, we document that grit is one of the driving forces of sustained

willingness to compete. We then explore whether competitive behavior can be inﬂuenced by an ed-

ucational intervention that aims to foster grit in the classroom environment, via a carefully designed

curriculum implemented by trained teachers. Using choices in a two-stage competition task as our

outcome variables, we evaluate the impact of this unique educational intervention.

We ﬁrst show that the competition propensity of boys and girls are signiﬁcantly diﬀerent, even

in childhood. We then show that an educational intervention that aims to foster the non-cognitive

skill of grit can have a signiﬁcant impact on both girls’ and boys’ competitiveness. This indicates

that competitiveness as an individual attitude is malleable in childhood through a targeted program

that promotes ambitious goal-setting and perseverance. The eﬀect on competitiveness naturally has

implications for the gender gap as well. We estimate a statistically signiﬁcant treatment eﬀect on the

willingness to compete for both boys and girls in the ﬁrst stage and this eﬀect is bigger for girls, miti-

gating the ﬁrst stage gender gap. Moreover, we ﬁnd a signiﬁcant treatment eﬀect on competitiveness

after feedback, only for girls, entirely eliminating the second stage gender gap. Our data suggest that

improved self-conﬁdence and enhanced grit on the part of girls is likely to be the main mechanism that

generates our results.

Coupled with the fact that the treatment is overall payoﬀ-improving for highly able girls, our results

indicate that fostering grit in the classroom environment can address one of the most debated gender

gaps in the literature without sacriﬁcing eﬃciency. In this sense, our paper provides a complement to

the interventions in the literature that close the gender gap by changing tournament rules and thereby

the material incentives for women. Incorporating the type of intervention proposed in this paper

into the classroom environment (or even the home environment) can be an easy and cost-eﬀective

alternative from a policy perspective.

Finally, our data highlight the importance of studying competitiveness in a dynamic framework

where, as in real life, individuals receive performance feedback and decide whether or not to stick

with the chosen path. Performance feedback is ubiquitous in economic life and in education, and the

way it is interpreted may have profound impact on an individual’s subsequent decisions. Evaluations

of educational and other policies that aim to mitigate ineﬃcient gender gaps in competitiveness in

particular, and achievement in general, will be more comprehensive if their eﬀects on dynamic choices

and their interactions with feedback are taken into account.

References

[1] Alan, S. and Ertac, S. (2014), “Good Things Come to Those Who (Are Taught How to) Wait:

Results from a Randomized Educational Intervention on Time Preference”, HCEO Working Paper.

[2] Alan, S., Boneva T. and Ertac S. (2016), “Ever Failed, Try Again, Succeed Better: Results from

a Randomized Educational Intervention on Grit”, HCEO Working Paper.

[3] Almas, I., Cappelen A., Salvanes, K., Sørensen, E. and Tungodden, B. (2014), “Willingness to

Compete: Family matters”, Management Science, forthcoming.

[4] Almlund, M., Duckworth, A.L., Heckman, J.J. and Kautz, T.D. (2011), “Personality Psychology

and Economics”, Handbook of the Economics of Education, pp. 1-181.

[5] Andersen, S., Ertac, S., Gneezy, U., List, J. A., & Maximiano, S. (2013), “Gender, competitive-

ness, and socialization at a young age: Evidence from a matrilineal and a patriarchal society”,

Review of Economics and Statistics, 95(4), 1438-1443.

[6] Andersen, S., Ertac, S., Gneezy, U., List, J. A., & Maximiano, S. (2014). “Gender, competitiveness,

and the response to performance feedback: Evidence from a matrilineal and a patriarchal society”,

Mimeo.

[7] Azmat, G. and Iriberri, N. (2010), “The Importance of Relative Performance Feedback Infor-

mation: Evidence from a Natural Experiment using High School Students”, Journal of Public

Economics, 94, pp. 435-452.

[8] Balafoutas, L., & Sutter, M. (2012), “Aﬃrmative action policies promote women and do not harm

eﬃciency in the laboratory”, Science, 335(6068), 579-582.

[9] Barankay, I. (2011), “Rankings and social tournaments: Evidence from a crowd-sourcing experi-

ment”, Wharton School of Business, University of Pennsylvania Working Paper.

[10] Blackwell, L., Trzesniewski, K.H. and Dweck, C.S. (2007), “Implicit Theories of Intelligence Pre-

dict Achievement Across an Adolescent Transition: A Longitudinal Study and an Intervention”,

Child Development, 78(1), pp. 246-263.

[11] Booth, A., & Nolen, P. (2012), “Choosing to compete: How diﬀerent are girls and boys?”, Journal

of Economic Behavior & Organization, 81(2), 542-555.

[12] Borghans, L., Duckworth, A.L., Heckman, J.J. and ter Weel, B. (2008), “The Economics and

Psychology of Personality Traits”, Journal of Human Resources, 43(4), pp. 972-1059.

[13] Buser, T., Niederle, M. and Oosterbeek, H. (2014), “Gender, Competitiveness and Career

Choices”, Quarterly Journal of Economics, August 2014, 129 (3): 1409-1447.

[14] Cárdenas, J. C., Dreber, A., Von Essen, E., & Ranehill, E. (2012), “Gender diﬀerences in com-

petitiveness and risk taking: Comparing children in Colombia and Sweden”, Journal of Economic

Behavior & Organization, 83(1), 11-23.

[15] Croson, R., & Gneezy, U. (2009), “Gender diﬀerences in preferences”, Journal of Economic liter-

ature, 448-474.

[16] Dreber, A., Von Essen, E. & Ranehill, E. (2014), "Gender and Competition in Adolescence: Tasks

Matter." Experimental Economics, 17(1): 154-172.

[17] Duckworth, A.L., Peterson, C., Matthews, M.D., and Kelly, D.R. (2007), “Grit: Perseverance and

Passion for Long-Term Goals”, Journal of Personality and Social Psychology, 92(6), pp. 1087-1101.

[18] Duckworth, A. L. and Quinn, P. D. (2009), “Development and validation of the Short Grit Scale

(Grit-S)”, Journal of Personality Assessment, 91, pp. 166-174.

[19] Dweck, C. Mindset: The New Psychology of Success. Random House Digital, Inc., 2006

[20] Eckel, C. C., and Füllbrunn, S. (2015), “Thar ‘She’ Blows? Gender, Competition and Bubbles

in Experimental Asset Markets” The American Economic Review, Volume 105, Number 2, pp.

906-920(15).

[21] Eriksson, T., Poulsen, A. and Villeval, M.C. (2009), “Feedback and Incentives: Experimental

Evidence”, Labour Economics, 16(6), pp. 679-688.

[22] Ertac, S., & Szentes, B. (2011), “The eﬀect of information on gender diﬀerences in competitiveness:

Experimental evidence”, TÜSAD-Koç University Economic Research Forum working paper series.

[23] Flory, J. A., Leibbrandt, A., & List, J. A. (2014), “Do Competitive Workplaces Deter Female

Workers? A Large-Scale Natural Field Experiment on Job-Entry Decisions”, The Review of

Economic Studies, rdu030.

[24] Eskreis-Winkler, L., Shulman, E.P., Beal, S. and Duckworth, A.L. (2014), “Survivor mission:

Why those who survive have a drive to thrive at work”, Journal of Positive Psychology, 9(3), pp.

209-218.

[25] Gneezy, U, Leonard, K.L. and List, J.A. (2009) “Gender Diﬀerences in Competition: Evidence

from a Matrilineal and a Patriarchal Society”, Econometrica, 77(5), 1637-1664.

[26] Gill, D. and Prowse, V. (2014), “Gender diﬀerences and dynamics in competition: The role of

luck”, Quantitative Economics 5 (2014), 351–376.

[27] Gneezy, U., Niederle, M. and Rustichini, A. (2003), “Performance in Competitive Environments:

Gender Diﬀerences”, Quarterly Journal of Economics,118(3), 1049-1074.

[28] Gneezy, U. and Potters, J. (1997), “An experiment on risk taking and evaluation periods”, Quar-

terly Journal of Economics, 112(2), pp. 631-645.

[29] Heckman, J.J., Stixrud, J. and Urzua, S. (2006), “The Eﬀects of Cognitive and Noncognitive

Abilities on Labor Market Outcomes and Social Behaviour”, Journal of Labor economics, 24, pp.

411-482.

[30] Heckman, J.J., Moon, S.H., Pinto, R., Savelyev, P.A. and Yavitz, A.Q. (2010), “The Rate of

Return to the High Scope Perry Preschool Program”, Journal of Public Economics, 94(1-2), pp.

114-128.

[31] Heckman, J.J., Pinto, R. and Savelyev, P.A. (2013), “Understanding the Mechanisms Through

Which an Inﬂuential Early Childhood Program Boosted Adult Outcomes”, American Economic

Review, 103(6), pp. 1-35.

[32] Heckman, J.J., Moon, S.H. and Pinto, R. (2014), “The Eﬀects of Early Intervention on Abilities

and Social Outcomes: Evidence from the Carolina Abecedarian Study”, unpublished manuscript.

[33] Kautz, T., Heckman, J.J., Diris, R., ter Weel, B., Borghans, L. (2014), “Fostering and Measur-

ing Skills: Improving Cognitive and Non-cognitive Skills to Promote Lifetime Success”, NBER

Working Paper 20749.

[34] Khachatryan, K., Dreber, A., Von Essen, E., & Ranehill, E. (2015), “Gender and preferences

at a young age: Evidence from Armenia”. Journal of Economic Behavior & Organization, 118,

318-332.

[35] Lee, S., Niederle, M., Kang, N. (2014), “Do Single-Sex Schools Make Girls More Competitive?”,

Economics Letters, 124(3), pp. 474–477.

[36] Maddie, S.R., Matthews, M.D., Kelly, D.R., Villarreal, B. and White, M. (2012), “The Role

of Hardiness and Grit in Predicting Performance and Retention of USMA Cadets”, Military

Psychology, 24, pp. 19-28.

[37] Niederle, M., Segal, C., & Vesterlund, L. (2013), “How costly is diversity? Aﬃrmative action in

light of gender diﬀerences in competitiveness”, Management Science, 59(1), 1-16.

[38] Niederle, M., & Vesterlund, L. (2007), “Do Women Shy Away from Competition? Do Men Com-

pete Too Much”, Quarterly Journal of Economics, August 2007, vol. 122(3), 1067-1101.

[39] Niederle, M., & Vesterlund, L. (2011), “Gender and competition”, Annual Review of Economics,

3(1), 601-630.

[40] Petrie, R., & Segal, C. (2014), “Gender diﬀerences in competitiveness: The role of prizes”, Avail-

able at SSRN 2520052.

[41] Raven, J., Raven, J.C., & Court, J.H. (2004), “Manual for Raven’s Progressive Matrices and

Vocabulary Scales”, San Antonio, TX: Harcourt Assessment.

[42] Sutter, M., & Glätzle-Rützler, D. (2015), “Gender diﬀerences in the willingness to compete emerge

early in life and persist”, Management Science, forthcoming.

[43] Sutter, M., Glätzle-Rützler, D., Balafoutas, L., & Czermak, S. (2015), “Canceling out early age

gender diﬀerences in competition–an analysis of policy interventions”, Experimental Economics,

forthcoming.

[44] Wozniak, D., Harbaugh, W. T., & Mayr, U. (2014), “The menstrual cycle and performance

feedback alter gender diﬀerences in competitive choices” Journal of Labor Economics, 32(1), 161-

198.

Tables

Table A: Evaluation Design

STUDY 1

STUDY 2

Patience+Grit

(9 schools)

Patience

(6 schools)

Control

(8 schools)

Grit

(8 schools)

Control

(8 schools)

Baseline Data

March 2013

May 2015

Patience Treatment

Spring 2013

Fall 2013

Grit Treatment

Fall 2013

Fall 2015

Follow-up Data

May 2014

Jan 2016

Table 1: Randomization Balance

PANEL 1: Baseline Variables

Control Treatment

Male 0.52 -0.01

(0.02)

Age 9.51 -0.01

(0.03)

Raven Score (normalized) 0.03 -0.07

(0.13)

Math Grades (normalized) 0.02 -0.04

(0.12)

Risk Tolerance 2.28 0.12

(0.17)

Male Teacher 0.22 -0.04

(0.10)

Wealth (teacher reported) 2.72 0.14

(0.18)

Academic Success (teacher reported) 3.43 0.04

(0.08)

Grit Score (normalized) -.00 .00

(.10)

PANEL 2: Within Task Variables

Control Treatment

Correct Answers in Part 1 5.80 -0.30

(0.39)

Negative Feedback in Stage 1 0.42 0.04

(0.05)

Male Opponent 0.50 -0.01

(0.02)

Note: Each row reports coeﬃcients from a regression of the variable shown in the ﬁrst column on

the treatment dummy. The ﬁrst column reports the mean of the control, the second one reports the

diﬀerence between the treatment and control. Panel 1 presents the balance for demographic variables

and baseline attitudes either reported by the child or the teacher. Panel 2 presents the balance for

i) the performance in the ﬁrst period, ii) the proportion of negative feedback in Stage 1 and iii) the

proportion of male opponents. Standard errors, obtained via clustering at the school level, are reported

in parentheses.

Table 2: Gender Gap in Competition Choice (Control Sample)

Stage 1 Stage 2

(1) (2) (3) (4)

W/O Controls With Controls W/O Controls With Controls

Male 0.081*** 0.078*** 0.095*** 0.100***

(0.02) (0.02) (0.03) (0.03)

Male opponent 0.026 0.005

(0.02) (0.02)

Risk tolerance 0.042*** 0.024**

(0.01) (0.01)

Raven Score 0.053*** 0.041***

(0.01) (0.01)

Male Teacher -0.038 -0.031

(0.04) (0.02)

Relative conﬁdence S1 0.070***

(0.01)

Baseline Grit Score 0.028* 0.052***

(0.02) (0.01)

Relative conﬁdence S2 0.133***

(0.01)

Observations 1380 1354 1377 1353

Standard errors clustered at the classroom level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates

are marginal eﬀects from logistic regressions.

Table 3: Gender Diﬀerence in Performance (Control Sample)

Piece-Rate Stage 1

(1) (2) (3) (4)

W/O Controls With Controls W/O Controls With Controls

Male -0.280 -0.215 -0.262 -0.209

(0.19) (0.18) (0.23) (0.21)

Male opponent -0.025 0.025

(0.18) (0.22)

Risk tolerance -0.090 -0.046

(0.06) (0.06)

Raven Score 1.304*** 1.397***

(0.11) (0.13)

Male Teacher -0.621 -0.474

(0.38) (0.42)

Observations 1379 1379 1382 1382

Standard errors clustered at the classroom level. * p<0.10, ** p<0.05, *** p<0.01. Reported

estimates are OLS coeﬃcients.

Table 4: Feedback Type and Gender Gap in Competition Choice (Control Sample)

NonNegative Feedback Negative Feedback

(1) (2) (3) (4)

W/O Controls With Controls W/O Controls With Controls

Male 0.112*** 0.101*** 0.091** 0.090**

(0.04) (0.04) (0.04) (0.04)

Male opponent 0.005 0.010

(0.03) (0.03)

Risk tolerance 0.038*** 0.003

(0.01) (0.01)

Raven Score 0.087*** -0.026

(0.02) (0.02)

Male Teacher 0.038 -0.079

(0.04) (0.06)

Observations 786 786 574 574

Standard errors clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates are

marginal eﬀects from logistic regressions.

Table 5: Gender Gap in Competition Choice (Treatment Sample)

Stage 1 Stage 2

(1) (2) (3) (4)

W/O Controls With Controls W/O Controls With Controls

Male 0.017 0.033 -0.048 -0.028

(0.03) (0.03) (0.03) (0.03)

Male opponent -0.039* -0.003

(0.02) (0.03)

Risk tolerance 0.044*** 0.028**

(0.01) (0.01)

Raven Score 0.023 0.053***

(0.01) (0.02)

Male Teacher 0.025 -0.016

(0.04) (0.03)

Relative conﬁdence S1 0.097***

(0.02)

Baseline Grit Score 0.066*** 0.049***

(0.01) (0.02)

Relative conﬁdence S2 0.122***

(0.02)

Observations 1212 1190 1215 1190

Standard errors clustered at the classroom level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates

are marginal eﬀects from logistic regressions.

Table 6: Feedback Type and Gender Gap in Competition Choice (Treatment Sample)

NonNegative Feedback Negative Feedback

(1) (2) (3) (4)

W/O Controls With Controls W/O Controls With Controls

Male -0.013 -0.013 -0.043 -0.042

(0.04) (0.04) (0.05) (0.05)

Male opponent -0.003 -0.029

(0.03) (0.05)

Risk tolerance 0.031* 0.018

(0.02) (0.01)

Raven Score 0.059*** 0.022

(0.02) (0.02)

Male Teacher -0.013 -0.020

(0.05) (0.05)

Observations 646 646 566 566

Standard errors clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates are

marginal eﬀects from logistic regressions.

Table 7: Treatment Eﬀects on Competition Choice: Stage 1

(1) (2) (3)

All Females Males

Treatment 0.119*** 0.146*** 0.091***

(0.03) (0.03) (0.03)

Male opponent -0.003 -0.004 -0.002

(0.02) (0.02) (0.03)

Risk tolerance 0.041*** 0.044*** 0.037***

(0.01) (0.01) (0.01)

Raven Score 0.049*** 0.064*** 0.037**

(0.01) (0.01) (0.02)

Male Teacher -0.016 -0.016 -0.017

(0.03) (0.03) (0.04)

Control Mean 0.38 0.34 0.42

N 2592 1264 1328

Standard errors are clustered at the school level. * p<0.10

** p<0.05, *** p<0.01. Reported estimates are marginal

eﬀects from logistic regressions.

Table 8: Treatment Eﬀects on Competition Choice: Stage 2

(1) (2) (3)

All Females Males

Treatment 0.089*** 0.156*** 0.022

(0.03) (0.03) (0.04)

Risk tolerance 0.024*** 0.025** 0.023***

(0.01) (0.01) (0.01)

Male opponent 0.003 0.008 -0.001

(0.02) (0.02) (0.03)

Raven Score 0.074*** 0.072*** 0.075***

(0.01) (0.01) (0.01)

Male Teacher -0.040* -0.030 -0.050*

(0.02) (0.03) (0.03)

Control Mean 0.45 0.40 0.49

N 2592 1263 1329

Standard errors are clustered at the school level. * p<0.10

** p<0.05, *** p<0.01. Reported estimates are marginal

eﬀects from logistic regressions.

Table 9: Treatment Eﬀects on PayoﬀMaximizing Choice

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6)

All Females Males All Females Males

Treatment 0.041* 0.059** 0.024 0.029* 0.051** 0.008

(0.02) (0.03) (0.03) (0.02) (0.02) (0.03)

Risk tolerance 0.027*** 0.022* 0.032*** 0.016** 0.012 0.019***

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

Male opponent 0.005 0.003 0.006 0.010 0.012 0.009

(0.02) (0.03) (0.02) (0.02) (0.03) (0.03)

Raven Score 0.007 0.001 0.011 0.023*** 0.017 0.027**

(0.01) (0.02) (0.01) (0.01) (0.01) (0.01)

Male Teacher 0.040 0.063 0.019 -0.004 0.039 -0.043

(0.04) (0.04) (0.04) (0.02) (0.03) (0.03)

Control Mean 0.52 0.48 0.55 0.59 0.56 0.62

N 2592 1264 1328 2592 1263 1329

Standard errors clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates

are marginal eﬀects from logistic regressions.

Table 10: Treatment Eﬀects on Expected PayoﬀGains

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6)

All Females Males All Females Males

Treatment 1.893*** 2.812*** 0.994 0.941 1.718** 0.200

(0.58) (0.68) (0.73) (0.59) (0.74) (0.76)

Risk tolerance 0.765*** 0.825*** 0.706*** 0.286** 0.226 0.338**

(0.14) (0.20) (0.18) (0.13) (0.20) (0.14)

Male opponent 0.223 0.693 -0.210 -0.044 -0.155 0.066

(0.38) (0.61) (0.53) (0.31) (0.49) (0.55)

Raven Score 0.842*** 0.974*** 0.710** 1.513*** 1.376*** 1.594***

(0.19) (0.29) (0.29) (0.17) (0.25) (0.22)

Male Teacher 0.110 0.339 -0.135 -0.535 -0.020 -1.015*

(0.55) (0.74) (0.75) (0.47) (0.63) (0.54)

Control Mean -0.65 -1.74 0.37 1.69 0.81 2.50

N 2580 1259 1321 2584 1260 1324

Standard errors are clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates

are OLS coeﬃcients.

Table 11: Heterogeneous Treatment Eﬀects on Competition Choice (High vs Low Cognitive Score)

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6) (7) (8)

Fem_L Fem_H Male_L Male_H Fem_L Fem_H Male_L Male_H

Treatment 0.153*** 0.145*** 0.111*** 0.079* 0.172*** 0.146*** 0.006 0.031

(0.05) (0.03) (0.04) (0.04) (0.06) (0.03) (0.04) (0.05)

Male opponent -0.036 0.014 0.018 -0.016 -0.032 0.028 -0.071** 0.050

(0.04) (0.03) (0.03) (0.03) (0.05) (0.03) (0.03) (0.04)

Risk tolerance 0.033** 0.051*** 0.029* 0.044*** 0.017 0.031*** 0.015 0.028**

(0.01) (0.01) (0.02) (0.01) (0.02) (0.01) (0.01) (0.01)

Raven Score 0.035 0.097*** 0.028 0.052 0.036 0.170*** 0.038 0.073**

(0.04) (0.04) (0.03) (0.04) (0.03) (0.02) (0.03) (0.03)

Male Teacher -0.051 0.008 -0.036 -0.002 -0.088 0.011 -0.062 -0.040

(0.06) (0.04) (0.06) (0.05) (0.07) (0.04) (0.04) (0.04)

Control Mean 0.27 0.37 0.37 0.45 0.34 0.43 0.42 0.55

N 494 770 559 769 495 768 557 772

Standard errors are clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates are marginal eﬀects

from logit regressions.

Table 12: Heterogeneous Treatment Eﬀects on PayoﬀMaximizing Choice (High vs Low Cognitive

Score)

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6) (7) (8)

Fem_L Fem_H Male_L Male_H Fem_L Fem_H Male_L Male_H

Treatment -0.005 0.100** -0.023 0.062* 0.038 0.058** -0.017 0.028

(0.04) (0.04) (0.05) (0.03) (0.04) (0.03) (0.05) (0.04)

Male opponent -0.082** 0.055 -0.000 0.010 -0.023 0.037 0.004 0.011

(0.03) (0.04) (0.04) (0.03) (0.04) (0.04) (0.05) (0.03)

Risk tolerance 0.013 0.028 0.015 0.046*** 0.008 0.011 0.008 0.028**

(0.02) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01)

Raven Score -0.071** 0.106*** -0.014 0.058* -0.073** 0.096*** 0.007 0.036

(0.03) (0.04) (0.03) (0.03) (0.03) (0.03) (0.02) (0.03)

Male Teacher 0.056 0.082 0.032 0.020 0.132** -0.002 -0.057 -0.027

(0.04) (0.05) (0.06) (0.04) (0.06) (0.03) (0.04) (0.04)

Control Mean 0.52 0.46 0.56 0.54 0.54 0.57 0.60 0.64

N 494 770 559 769 495 768 557 772

Standard errors are clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates are marginal eﬀects

logit regressions.

Table 13: Treatment Eﬀects on Conﬁdence

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6)

All Females Males All Females Males

Treatment 0.142** 0.163** 0.126 0.092* 0.128** 0.061

(0.06) (0.07) (0.08) (0.05) (0.06) (0.06)

Risk tolerance -0.024* -0.034* -0.015 -0.001 -0.021 0.018

(0.01) (0.02) (0.02) (0.01) (0.02) (0.02)

Male opponent 0.013 -0.014 0.037 0.017 0.049 -0.016

(0.05) (0.07) (0.05) (0.05) (0.06) (0.06)

Raven Score 0.004 -0.022 0.029 0.103*** 0.090*** 0.115***

(0.02) (0.02) (0.04) (0.02) (0.03) (0.03)

Male Teacher -0.034 -0.140** 0.071 -0.129** -0.172** -0.088

(0.06) (0.07) (0.08) (0.05) (0.07) (0.06)

Observations 2555 1255 1300 2551 1251 1300

Standard errors clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported

estimates are coeﬃcients from OLS regressions. Dependent variable is the number of

questions the subject expects to solve minus the number of questions she expects the

opponent to solve, normalized.

Table 14: Treatment Eﬀects on Belief to Improve upon First Stage Performance

(1) (2) (3)

All Females Males

Treatment 0.053** 0.073*** 0.032

(0.03) (0.03) (0.04)

Risk tolerance 0.006 0.022** -0.010

(0.01) (0.01) (0.01)

Male opponent 0.000 -0.029 0.029

(0.02) (0.03) (0.02)

Raven Score -0.082*** -0.095*** -0.072***

(0.01) (0.01) (0.02)

Male Teacher -0.052** -0.032 -0.067*

(0.02) (0.02) (0.04)

Control Mean 0.52 0.51 0.54

N 2559 1255 1304

Standard errors clustered at the school level. * p<0.10, ** p<0.05,

*** p<0.01. Reported estimates are marginal eﬀects from logit

regressions. The dependent variable takes the value of 1 if the

subject believes that her performance in the second stage will

exceed her actual performance in the ﬁrst stage, and zero otherwise.

Table 15: Treatment Eﬀects on the Grit Score

(1) (2) (3)

All Females Males

Treatment 0.313*** 0.371*** 0.259***

(0.06) (0.06) (0.07)

Baseline Grit Score 0.312*** 0.308*** 0.314***

(0.03) (0.04) (0.03)

Risk tolerance -0.004 -0.018 0.008

(0.02) (0.02) (0.02)

Male opponent 0.026 0.053 -0.000

(0.04) (0.06) (0.05)

Raven Score 0.165*** 0.156*** 0.173***

(0.02) (0.03) (0.03)

Male Teacher 0.022 -0.048 0.084

(0.07) (0.08) (0.07)

Observations 2216 1091 1125

Standard errors clustered at the school level. * p<0.10, ** p<0.05,

*** p<0.01. Reported estimates are coeﬃcients from OLS regressions.

Figure 1: Gender Diﬀerence in the Willingness to Compete: Stage 1 and Stage 2

The Figure presents the marginal eﬀects (and conﬁdence bands) on the “male” dummy obtained via

logistic regressions. Depicted coeﬃcient estimates are obtained by controlling for opponent’s gender,

risk attitude, Raven score and teacher’s gender. Conﬁdence bands are obtained via clustered standard

errors (classroom level). C: Control, TR: Treatment.

Figure 2: Distribution of Treatment Eﬀects on Expected PayoﬀGains: Stage 1 and Stage 2

Figures plot the quantile treatment eﬀects on expected payoﬀgains and 95% conﬁdence bands. Upper

panels show Stage 1 results and lower panels show Stage 2 results for females and males separately.

The dashed lines indicate the estimated average treatment eﬀects with the dotted lines showing the

conﬁdence bands for the average treatment eﬀect estimates.

Figure 3: Gender Diﬀerence in Relative Conﬁdence: Stage 1 and Stage 2

The Figure presents the OLS coeﬃcients on the “male” dummy. Depicted coeﬃcient estimates are

obtained by controlling for initial performance, opponent’s gender, risk attitude, Raven score and

teacher’s gender. Conﬁdence bands are obtained via clustered standard errors (classroom level). C:

Control, TR: Treatment.

Figure 4: Gender Diﬀerence in Grit Score

The Figure presents the OLS coeﬃcients on the “male” dummy. Depicted coeﬃcient estimates are

obtained by controlling for opponent’s gender, risk attitude, Raven score, teacher’s gender and baseline

grit score. Conﬁdence bands are obtained via clustered standard errors (classroom level).

Appendix

A Screenshot from Education Materials and Sample Class Activities

The first picture is a screenshot from a video that depicts the brain when it is struggling with

difficult math problems. With sustained practice, the brain becomes bigger and stronger. The

first and the second posters (made by children) are about quitting and unfounded fear of

math, with the second one translating as: “I never quit”. The third activity relates to goal-

setting, where each student posts their monthly target and updates it during the month.

Questions Used for Constructing the Grit Score

4-point item scale: completely agree, agree, disagree, completely disagree

1. I don’t like it when people (my teacher, parents, friends) make comments and suggestions about

how to improve my performance in a class, game or task that I am not very good at.

2. When I receive a bad result on a test it is because the test was unfairly hard.

3. I like school work best which makes me think hard, even if I make a lot of mistakes.

4. When I receive a bad result on a test I work harder.

5. Setbacks discourage me.

6. If I think I will lose in a game, I do not want to continue playing.

7. If I set a goal and see that it’s harder than I thought I easily lose interest.

8. When I receive a bad result on a test I spend less time on this subject and focus on other

subjects that I’m actually good at.

9. I work hard in tasks.

10. I prefer easy homework where I can easily answer all questions correctly.

11. If I’m having diﬃculty in a task, it is a waste of time to keep trying. I move on to things which

I am better at doing.

Instructions for the Competition Task

Today we are going to play a game with you. At the end of the game, you will get to choose gifts from

this gift bag [show gift bag]. You will be asked to make some decisions in the game. How many gifts

you earn will depend on your decisions, and your performance in a task. We will now explain the rules

for this. Please listen very carefully. You can raise your hand if you have questions about the rules.

But in all the decisions you make, you should think to yourself and not tell your decision out loud to

others, OK? There is no right or wrong decision in this game. Everyone is free to choose what he/she

wants.

The game will consist of three parts: Part 1, Part 2, and Part 3. Only one of these parts will count

to determine your rewards. At the very end of the game, we will have a draw. Here are pieces of paper

with 1, 2 and 3 written on it. [Show papers] We will put these in a bag and draw one. Whichever part

comes out, only your decisions and performance in that part will count for rewards. This means, the

gifts do not accumulate in every part. You get rewards only based on what happens in the randomly

drawn part. Let’s give an example. Suppose someone would earn 3 gifts from the 1st part, 5 from the

2nd and 4 from the 3rd. In the draw at the end, Part 2 came out. How many gifts does this person

actually get? 5. Not 12, because the gifts do not accumulate. Any questions? OK.

Now, the task in this game will be to add numbers. In each question, there are two 2-digit numbers

and 1 one-digit number to be added. For example, [written on the board] 75+80+4=?, 72+47+8=?.

You will be given many addition questions like this. Now, we will distribute workbooks. There are

many pages in this workbook. It is extremely important that you do not turn over a page before we

tell you to do so, or go back to a previous page. Everyone should be at the same page at any given

time. OK? We are also going to give you a separate sheet of paper with some questions on it. First,

just write your name on everything you get. Then put the single sheet of paper aside with written side

down. That is for later. Once you write your name, please wait and do not open any pages. Now, I

will explain the rules for the 1st part of the game. In this part, you will be given 2.5 minutes. During

this time, you will try to do as many addition questions as you can. There will be many such questions

on your sheet. If this part (Part 1) gets drawn at the end of the game, you will earn gifts as follows:

every question you solve correctly in this part will mean one gift of your choice from the gift bag. The

more questions you solve, the more gifts you get. If you do 1 right, you get 1 gift, if you do 4 you get

4, if you do 7, you get 7, so on and so forth. Is this clear? [If no questions] Are you ready? OK, on

the count of three, everyone will turn the page to start. 3, 2, 1, Start! [At the end of 2.5 minutes] OK,

everyone turn the page! You should now see the waiting page, Page 2. Stay there, OK?

Now, we played this same game, before you, in other classes in other schools. Those children were

also 4th graders like you are. We also timed them on the exact same task with 2.5 minutes, and we

recorded how many sums each student did correctly in all three parts of the game. Now, we will ask

you something about the task you just did. Remember, you will think to yourself and not say anything

out loud, OK? We randomly selected, from those other classes in other schools where we played the

game before, one boy and one girl. One is called Batuhan, another called Merve [note: Batuhan is a

common boys’ name, Merve a common girls’ name in Turkey]. Now, we will ask you to guess: which

of them do you think did more questions right in 2.5 minutes? Batuhan more, Merve more, or did

they do equally well? If you guess correctly, you will get an extra, smaller gift. OK? Now turn the

page. You’ll be on Page 3. Circle your guess there, and turn the page [children see a waiting page]

Now, remember we told you we did this game in other classes in other schools before. We will match

each one of you, randomly, with one student from that group. You will not know who that student

is. You will never see them, you will not get to meet them. For example, let’s say, one of you, got

matched with somebody from another class, called Hakan. We had given Hakan a number when we

visited that class. Say, student number 1024. You will not know that you got matched with Hakan.

The only thing you will know is whether your matched student is a boy or a girl. Let’s take another

one of you. Suppose you were matched with student number 1038. You will not know who this person

is. But you will know whether it’s a girl or a boy. OK? Keep this in mind for now.

Now, we will explain the rules of the 2nd part of the game. In this part, the task is the same.

You will again do addition questions. And again, you will have 2.5 minutes. However, in this part,

you will choose how you earn gifts. You have two options: One option is “per-question”. [Write on

board: Per-question] What does this mean? If you choose this, for every addition question you answer

correctly you earn one gift. That is, 2 questions right=2 gifts, 4 questions right=4 gifts, 7 questions

right=7 gifts, so on and so forth. [Write on board: Every correct answer=1 gift] The other option is

“Competition”. [Write on board: Competition] Remember that every one of you is going to be matched

with another student? If you choose competition, you will be competing with that matched student.

In this case, how many gifts you earn will depend on how many questions you do correctly and how

many the matched student did correctly. We have here, recorded, the number of correct questions

solved by each of the matched students in each part of the game. If you do more questions right than

your matched student in Part 2, that is, if you outperform that student, you get not 1, but 3 gifts for

each question you solve correctly. [Write on board: Better: Each correct answer=3 gifts] But if you

do less questions right than your matched student, that is if he/she does better than you, you get zero

gifts [Write on board] If you do an equal number of correct questions as your matched student, you

get 1 gift for each question you solve correctly. [Write on board: Tied: Each correct answer=1 gift] Is

this clear? Now, let’s give some examples. Suppose that I chose the per-question option. Suppose that

I had 5 questions correct. [Write performances on board] How many gifts do I get? 5x1=5. It does

not matter what my matched student had done. Suppose that I chose to compete. How many gifts

do I get? Suppose my matched student had 4 questions correct. I outperformed the matched student,

so I get 5x3=15 gifts. What if the matched student had 7 questions correct? The matched person

outperformed me, so I get 0 gifts. What if the matched student had 5 questions correct, like I did? It

is a tie, so I get 5x1= 5 gifts. [Give 2 more examples, make sure students fully understand] Now let’s

give another example. Suppose that I chose piece-rate and got 7 questions right. How many gifts do I

get? 7x1=7 gifts. It is 7 regardless of what my matched student’s performance. Now suppose that I

chose competition. Suppose my matched student had less than 7 questions right. How many gifts do I

get? My performance was better, so I get 7x3=21 gifts. Suppose that my matched student had more

questions right than me. What do I get? I get 0 gifts. Suppose that my matched student also had 7

questions right. What do I get? 7x1=7 gifts. Did everyone understand? Do you have any questions?

Now, on your workbooks, there will be an ID number corresponding to the student you were

matched with. And you will see if this person is a boy or a girl. Now everyone turn the page. Do

you see it?. Now please think carefully and silently. If Part 2 is chosen for payment, based on which

option would you like to get gifts? Please circle what you want, and turn the page. OK. Now, we will

ask you two questions. As you know, every one of you is matched with a student from another school

and class. Everyone, even if they did not choose to compete, has a matched student. Now, please

think. Every part of this game has 16 addition questions. How many questions do you think you will

get right in Part 2? And how many do you think your matched student got right? Think carefully

because every correct guess can get you an extra small gift. At the end of the game, we will pick

3 students among you randomly. We will look at their guesses. For each correct guess, we will give

them an extra small gift. Understood? Now please everyone turn the page and make your guesses. Is

everyone done? Now, you will start doing the questions in Part 2 and will have 2.5 minutes again. One

thing to note–you should not stop solving questions just because you’ve already reached your guess

and want to be consistent with it. Why? Because solving more questions brings you more gifts than

being consistent with your guess. So to get more gifts, you should try to do as many as you can in the

time you are given. OK. Is everyone ready? Turn the page, and start on the count of 3. [When time

is up] Everyone turn the page immediately. You’ll see Page 7 (a waiting page).

Now, we will come to each of your desks. We will calculate how many questions you did, and

compare it with your matched student’s performance. We will mark on your sheet how many you did

correctly. We will also mark whether you have done more, the matched student has done more, or you

both did the same number of correct questions. Everyone will get this information. You should be

extremely quiet here and not tell anyone the information you got. Now while you are waiting, please

take that separate sheet we distributed at the start. There are some questions for you there. Please

start ﬁlling that in. [Note: Students work on a short questionnaire to ﬁll the time when they are waiting

for their performance to be calculated and feedback given. When marking, empty answer blanks were

crossed over with a pen so that the child cannot go back and change performance. Experimenters wrote

the number of correctly solved questions and used a code on the workbooks to make the performance

comparison, and circled the outcome of the comparison on the student’s feedback page.] So, did

everyone see how many they got right? Did everyone see whether they did more, less or equal number

of questions with their matched student? You know how many you got right. You know whether it

was more, less or equal. But notice, you do not know how many more or how many less you did than

the matched student.

Now, we will start the 3rd part of the game. In this part, you will again do addition problems in 2.5

minutes. Your matched student is the same student as in Part 2 for this part. Remember that those

matched students also did this same task for 3 periods. We also recorded how many each matched

student did correctly in the 3rd part. Now please think. Again, you have 2 options for how you earn

gifts. You can choose per question, or you can choose to compete with the matched student. If you

choose per-question, for every correct answer, you get 1 gift. If you choose to compete, you get 3 gifts

per correct answer you have, but only if you do more questions than your matched student did in Part

3. If you do less, you get zero gifts. If you do equal, you get 1 gift per correct answer. Now, everybody

turn the next page. There you will make your decision for Part 3. If Part 3 is drawn at the end of the

game for reward, you will get gifts according to what you choose here. Did everyone decide? OK, now

turn the page. You will again make guesses. How many questions do you think you will get right in

Part 3? How many questions do you think your matched student did in Part 3? Don’t forget, again,

every correct guess can get you an extra small gift if you are one of the 3 people chosen. Now make

your guesses and wait, do not turn the page. Everyone ready? OK, now on the count of 3, turn the

page, and start! [When time is up] Everyone stop and turn the page!

Table 16: Treatment Eﬀects by Disaggregated Treatment Status

Stage 1 Stage 2

(1) (2) (3) (4) (5) (6)

All Females Males All Females Males

Grit 0.169*** 0.181*** 0.157*** 0.150*** 0.218*** 0.082**

(0.02) (0.04) (0.03) (0.03) (0.03) (0.04)

Grit+Patience 0.094*** 0.137*** 0.049* 0.019 0.089** -0.051

(0.02) (0.03) (0.03) (0.03) (0.04) (0.04)

Patience 0.061 0.073 0.051 -0.046 -0.040 -0.058

(0.06) (0.07) (0.07) (0.04) (0.04) (0.09)

Male 0.045** 0.030

(0.02) (0.03)

Risk tolerance 0.042*** 0.045*** 0.039*** 0.027*** 0.028*** 0.026***

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

Male opponent -0.002 -0.004 0.001 0.005 0.010 0.002

(0.01) (0.02) (0.03) (0.02) (0.02) (0.03)

Raven Score 0.051*** 0.064*** 0.040** 0.083*** 0.080*** 0.083***

(0.01) (0.01) (0.02) (0.01) (0.01) (0.01)

Male Teacher -0.024 -0.022 -0.029 -0.056** -0.047 -0.065**

(0.03) (0.03) (0.04) (0.02) (0.03) (0.03)

Control Mean 0.36 0.32 0.40 0.45 0.40 0.49

N 2592 1264 1328 2592 1263 1329

Standard errors are clustered at the school level. * p<0.10, ** p<0.05, *** p<0.01. Reported estimates

are marginal eﬀects from logistic regressions.

Відображення в мас-медіа наслідків впливу війни на ґендерну сеґреґацію на українському ринку праці

Article

Full-text available

Jun 2024

The article aims to outline the trends of gender professional segregation occurring in Ukraine during the military conflict. Materials and Methods. The research was carried out with the use of qualitative methods. A content analysis was chosen for this study. The material consists of online textual publications, which include interviews with managers, specialists, and employees of the State Employment Center. It also includes interviews in which women residing on the territory of Ukraine share their experiences of employment and work during the war. Results and Discussion. In Ukraine today, two factors operate in parallel. On the one hand, there is an urgent need to fill vacancies in professions that have traditionally been dominated by men. On the other hand, the financial situation of women is on the decline. Women are forced to rethink their gender identity in the family and in society because of financial difficulties and their inability to guarantee a minimum of financial stability for themselves and their families. This leads them to be willing to accept jobs with working conditions that are particularly harmful and difficult. Employment opportunities for women are increasing due to differentiated demand, which contributes to the reduction of horizontal gender segregation. One factor that impedes this process is the demand for professions traditionally considered women’s domains, which do not require basic skills, and for which demand increased during the war. These professions are mainly related to manufacturing and working specialties. Given that educated and skilled women occupy blue-collar jobs, we can expect vertical segregation to intensify during wartime, when gender differences are manifested in positions and roles with differing status or potential for advancement. In the area of state gender policy, Ukraine has business development initiatives that reduce the gender gap in women’s access to assets and capital to invest in profitable entrepreneurial industries. However, these initiatives only involve a limited number of women and cannot be associated with radical gender changes in society. At the same time, the positive aspects of reducing horizontal gender segregation observed in Ukraine during wartime may be negated by the lack of or insufficient incentives for the development of preschool education. Thus, there are few effective levers to overcome gender segregation in Ukraine in the short and medium term. Achieving gender equality in the labor market is practically declaratory. Conclusions. In the context of the wartime crisis, gender segregation is a significant issue with negative consequences. A market specifically for women’s labor is emerging, and the social status of women’s work is declining compared to pre-war times. This situation may reinforce societal stereotypes that view men as more suitable for leadership positions, potentially decreasing women’s motivation and career aspirations.

Media coverage of the impact of the war on gender segregation in the Ukrainian labor market

Article

Jul 2024

Stimulating creativity and grit of high school students with creative STEM activities: an RCT with noncompliance

Article

Full-text available

Oct 2024
QUAL QUANT

Creativity and grit are widely recognized as vital components for the success and prosperity of individuals, even more so for younger people who will have to deal with the complex challenges connected to the digital era. Yet, whether these skills can be learnt or are innate traits is still subject to debate, which is further complicated by the difficulty of defining and, in turn, assessing these concepts. In this study, we challenge the idea that creativity and grit cannot be learnt and show that creative pedagogy activities, such as those offered by FabLabs, are powerful tools to enhance these non-cognitive skills. We conduct a randomized controlled trial in which 710 students from five Italian high schools are randomly assigned to creative STEM courses. The courses are delivered by FabLabs, small-scale workshops that offer access to tools for digital fabrication and employ a hands-on pedagogical approach expected to impact students’ creativity and grit positively. Assignment to the courses is at the class level: only students in classes randomly selected to join the Fablab activities can participate, and they may decide whether to enrol or not on a voluntary basis. We address noncompliance by adopting an instrumental variable approach. We use two modalities to measure creativity: a self-assessment through the Short Scale of Creative Self and an assessment made by an independent expert using an index of creativity developed by the authors. While the effect on self-assessed creativity is not significant, results show that the external, independent assessment of students’ creativity can capture a significant effect of Fablabs’ courses. It also emerges that FabLab activities have a positive significant effect on students’ grit, measured through the Duckworth scale. The encouraging results obtained for a relatively small sample of students should prompt a replication of the experiment on a broader scale.

Tackling the gender gap in mathematics with active learning methodologies

Article

Jun 2024
ECON EDUC REV

Less Macho, More Mellow: The Malleability of Competitiveness

Article

Jan 2024

The Impact of Role Models on Women’s Self-Selection into Competitive Environments

Article

Apr 2024

In this paper, we show that competitive female role models reduce women’s perceived stereotype threat and increase their willingness to compete. Competitive male role models have the opposite effect: they reduce women’s willingness to compete, and the gender gap in tournament entry increases. Results are strongest for the best performing women who would benefit most from competing. Role models have no impact on low performing women, and on men.

Effects of Workplace Competition on Work Time and Gender Inequality

Article

Jan 2024

High-pay, high-status jobs are competitive and male-dominated and typically demand long work hours. The authors study the role of competition in producing the latter two outcomes using two field experiments. In the first, they find that paying tournament prizes for performance induces both men and women to work longer, but that men respond more than women to the high-prize tournament. In the second, men are more likely than women to choose tournament-based compensation over a wage rate for larger prizes. These results demonstrate that high-stakes workplace competition can fuel gender inequality both directly, because men are more likely to enter and win tournaments, and indirectly, by raising work hours, which hurts women who face greater time demands in household production.

Effect of Social Information on Competition Choice

Article

Jan 2023

Is being competitive always an advantage? Competitiveness, gender, and labour market success

Article

Oct 2023
LABOUR ECON

How Experiments with Children Inform Economics

Article

Jun 2023
J ECON LIT

In the past several decades, the experimental method has lent deep insights into economics. One area that has contributed is the experimental study of children, where advances as varied as the evolution of human behaviors that shape markets and institutions to how early life influences shape later life outcomes, have been explored. We first develop a framework for economic preference measurement that provides a lens into how to interpret data from experiments with children. Next, we survey work that provides general empirical insights within our framework and provide a comprehensive summary of experimental methods used with children. Finally, we provide 10 tips for pulling off experiments with children, including factors such as taking into account child competencies, causal identification, and logistical issues related to recruitment and implementation. We envision the experimental study of children as a high-growth research area in the coming decades as social scientists begin to more fully appreciate that children are active participants in markets who (might) respond predictably to economic incentives. (JEL C90, D11, D83, D91, J13, J16, Z13)

Ever Failed, Try Again, Succeed Better: Results from a Randomized Educational Intervention on Grit

Article

Full-text available

Mar 2016

We show that grit, a non-cognitive skill that has been shown to be highly predictive of achievement, is malleable in the childhood period and can be fostered in the classroom environment. Our evidence comes from the evaluation of a randomized educational intervention implemented in two independent elementary school samples in Istanbul. Outcomes are measured via a novel incentivized real effort task and performance in standardized tests. We find that treated students are 1) more likely to choose to undertake a more challenging and more rewarding task against an easier but less rewarding alternative, less likely to give up after failure, 3) more likely to exert effort to accumulate task-specific ability, and consequently, 4) more likely to succeed and collect higher payoffs. The intervention also has a large positive impact on students' standardized test scores. Treated students score 0.28 standard deviations higher in a standardized math test, and 0.13 standard deviations higher in a standardized Turkish test at follow-up.

Fostering and Measuring Skills: Improving Cognitive and Non-Cognitive Skills to Promote Lifetime Success

Article

Jan 2014

Rankings and Social Tournaments: Evidence From a Crowd-Sourcing Experiment

Article

Sep 2011

Iwan Barankay

There is a growing interest in behavioral incentive schemes exploiting people preference about how they rank compared to others as a non-monetary mechanism to shape effort. In this paper we present evidence from a crowd-sourcing experiment where employees were given feedback about how they rank in terms of performance compared to others doing the same task. The context is such that rank had no implication for current or future compensation. Compared to a control group with no rank feedback, employees who received feedback about their rank were less likely to return to work and also less productive on the job.

The Importance of Relative Performance Feedback Information: Evidence from a Natural Experiment Using High School Students

Article

Jan 2009

We study the effect of providing relative performance feedback information on performance under piece-rate incentives. A natural experiment that took place in a high school offers an unusual opportunity to test this effect in a real-effort setting. For one year only, students received information that allowed them to know whether they were above (below) the class average as well as the distance from this average. We exploit a rich panel data set and find that the provision of this information led to an increase of 5% in students’ grades. Moreover, the effect was significant for the whole distribution. However, once the information was removed the effect disappeared. To rule out the concern that the effect may be driven by teachers within the school, we verify our results using national level exams (externally graded) for the same students, and the effect remains.

Willingness to compete: Family matters

Article

Aug 2016

This paper studies the role of family background in explaining differences in the willingness to compete in a cognitive task. By combining data from a lab experiment conducted with a fairly representative sample of adolescents in Norway and high-quality register data on family background, we show that family background is fundamental in two important ways. First, boys from low socioeconomic status families are less willing to compete than boys from better-off families, even when controlling for confidence, performance, risk preferences, time preferences, social preferences, and psychological traits. Second, family background is crucial for understanding the large gender difference in the willingness to compete. Girls are much less willing to compete than boys among children from better-off families, whereas we do not find any gender difference in willingness to compete among children from low socioeconomic status families. Our data suggest that the main explanation of the role of family background is that the father's socioeconomic status is strongly associated with boys' willingness to compete. We do not find any association between the willingness to compete for boys or girls and the mother's socioeconomic status or other family characteristic that may potentially shape competition preferences, including parental equality and sibling rivalry.

First-Place Loving and Last-Place Loathing: How Rank in the Distribution of Performance Affects Effort Provision

Article

Aug 2015

David Gill

Rank-order relative-performance evaluation, in which pay, promotion and symbolic awards depend on the rank of workers in the distribution of performance, is ubiquitous. Whenever firms use rank-order relative-performance evaluation, workers receive feedback about their rank. Using a real-effort experiment, we aim to discover whether workers respond to the specific rank that they achieve. In particular, we leverage random variation in the allocation of rank among subjects who exerted the same effort to obtain a causal estimate of the rank response function that describes how effort provision responds to the content of rank-order feedback. We find that the rank response function is U-shaped. Subjects exhibit 'first-place loving' and 'last-place loathing', that is subjects increase their effort the most after being ranked first or last. We discuss implications of our findings for the optimal design of firms' performance feedback policies, workplace organizational structures and incentives schemes.

Gender Differences in Reaction to Feedback and Willingness to Compete

Article

Aug 2016
J ECON BEHAV ORGAN

In Western societies, it is generally known that men have a greater taste for competition than women. However, the determinants of the decision to enter competitions are still not fully understood. The aim of this paper is twofold. We first evaluate how participants update their beliefs after receiving feedback informing them of whether their performance is below or above the median performance. Second, we are interested in how men and women react to this information in terms of competitive entry. Our first result is that participants, and women in particular, react more strongly to the feedback they receive than would a Bayesian agent. As far as entry into competition is concerned, below-median participants adjust their entry decision according to the competition they expect to face, while above-median participants do not. However, the behaviour behind these results is quite different for men and women: women mainly react to information on their own performance, while men seem to respond more to their beliefs over the competition they will face. Moreover, most of the effect of feedback and the information regarding the level of the competition on the decision to compete seems to operate via beliefs.

The Effects of Cognitive and Non-Cognitive Abilities on Labor Market Outcomes and Social Behavior

Article

Jul 2006
J LABOR ECON

This article establishes that a low-dimensional vector of cognitive and noncognitive skills explains a variety of labor market and behavioral outcomes. Our analysis addresses the problems of measurement error, imperfect proxies, and reverse causality that plague conventional studies. Noncognitive skills strongly influence schooling decisions and also affect wages, given schooling decisions. Schooling, employment, work experience, and choice of occupation are affected by latent noncognitive and cognitive skills. We show that the same low-dimensional vector of abilities that explains schooling choices, wages, employment, work experience, and choice of occupation explains a wide variety of risky behaviors.

The effects of cognitive and noncognitive abilities on labor market outcomes and socialbehavior

Article

Jan 2006
J LABOR ECON

Mindset: The New Psychology of Success

Book

Jan 2006

Carol S. Dweck

Mitigating the Gender Gap in the Willingness to Compete: Evidence from a Randomized Field Experiment

Abstract and Figures

Recommended publications

Mitigating the Gender Gap in the Willingness to Compete: Evidence from a Randomized Field Experiment

Mitigating the Gender Gap in the Willingness to Compete: Evidence from a Randomized Filed Experiment

Cancelling out early age gender differences in competition: an analysis of policy interventions

Is there a gender gap in preschoolers’ competitiveness? An experiment in the U.S