Gender and Preferences at a Young Age: Evidence from Armenia

Gender and Preferences at a Young Age: Evidence from Armenia

Karen Khachatryan, Anna Dreber, Emma von Essen, Eva Ranehill

Forthcoming in Journal of Economic Behavior and Organiza6on (2015)

Why are these preferences relevant?

o  Individuals that display more compeJJveness and less altruism earn more

o  Some of the gender gap in earnings can be explained with a set of psychological factors, including aNtudes toward risk and compeJJon

o  Individuals who self-‐report that they are less willing to take risks work in occupaJons with more stable earnings, that tend to pay less on average due to compensaJng wage differenJals

Bonin et al. 2007, Manning and Swaffield 2008, Flory et al. 2010, Manning and Saidi 2010, Dohmen et al. 2011

CompeJJveness

Measures •  Performance in tournaments vs piece-‐rate schemes •  Self-‐selecJon into tournaments vs piece-‐rate schemes

What do we know about adults?

Women are (if anything)…

•  Less compeJJve – Math related tasks

•  Less risk taking •  More altruisJc

Croson and Gneezy 2009, Engel 2011

CompeJJveness in children/adolescents

o  If gender gap, boys more compeJJve –  Israel, Sweden, Austria, Colombia: Running, skipping rope, dancing, children and adolescents

–  Austria and UK: Math and mazes, children and adolescents o  ExcepJon

–  Sweden: Skipping rope, math •  Performance change in both

Gneezy & RusJchini 2004, Booth & Nolen 2009, Dreber et al. 2011, Suber & Rutzler 2010, Cárdenas et al. 2011

Impacts of culture and context

•  Massai vs Khasi –  Patriarchal vs matrilineal society

•  Running –  Difference between GR04 and DvER11

•  Single-‐sex vs coeducaJonal schools •  SelecJon issues?

•  Colombia vs Sweden –  Unexpected findings: No gender gaps in Colombia in all tasks, but some gender gaps in Sweden

Gneezy & RusJchini 2004, Booth & Nolen 2011, Dreber et al. 2011, Cárdenas et al. 2011, Dreber et al. 2011, Grosse & Riener 2010, Gunther et al. 2009, Wozniak et al. 2010

Risk taking in children and adolescents

o  If gender gap, boys more risk taking –  US: 9-‐13 and 14-‐20 yr olds

•  Only study with no gap, smallest sample size –  Netherlands: 15-‐16 yr olds –  UK: 15 yr olds –  Sweden and Colombia: 9-‐12 yr olds

Harbaugh et al. 2002; Booth & Nolen 2011, Borghans et al. 2009, Cárdenas et al. 2011,

Impact of tasks

•  Some influence –  ShooJng baskets vs solving anagrams –  Running vs skipping rope and dancing –  Solving mazes/math vs word search

Gneezy & RusJchini 2004b, Dreber et al. 2011, Grosse & Riener 2010, Gunther et al. 2009, Wozniak et al. 2010, Cárdenas et al. 2011

QuesJons and ContribuJon

o  What causes the gender gap in these preferences?

– When does it emerge? • What is the influence of the country in which the study is performed (culture and context)? • What is the influence of tasks? •  Extrinsic vs intrinsic moJvaJon?

o  How do these preferences develop? –  InteresJng to compare different age groups

What we do: Our Experiments

o  824 children aged 7-‐16 in Yerevan, Armenia

-  in school grades 2 to 10

-  48% girls

o  CompeJJon tasks in the gym class

o  CompeJJon tasks in the classroom

o  Risk task in the classroom

o  Dictator game in the classroom

CompeJJon in the Gym Class Years 2 to 9

o  Running vs Skipping rope (“gendered tasks”?) –  Stage 1: run individually 4 x 13 meters / Skip rope individually

–  Stage 2: repeat stage 1 but in matched pairs

o  Measure of compeJJveness

–  speed in stage 2 — speed in stage 1

–  number of jumps in stage 2 — number of jumps in stage 1

–  could also use a relaJve measure

CompeJJon in the gym class Boys run faster than girls in any age group, but Girls improve more, significantly more compeJJve in running task Girls are beber in skipping rope, but no significant difference in performance change reacJon to compeJJon

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Table 1 – Individual and competitive performance in running (seconds)and skipping rope (number of jumps)

Running task Skipping rope task

Age Group Gender Round 1 Round 2 SR testa Round 1 Round 2 SR test Nb

(Ind.) (Comp.) (p≠value) (Ind.) (Comp.) (p≠value)

Grades 2 to 6 Boys 17.39 16.98 < 0.001 5.92 6.39 0.340 154/153Girls 18.95 18.32 < 0.001 8.38 7.12 0.692 147/145

Grades 7 to 9 Boys 15.13 14.92 < 0.001 67.10 72.27 < 0.001 100/96Girls 16.94 16.27 < 0.001 70.92 79.45 < 0.001 91/88

All Grades Boys 16.50 16.17 < 0.001 29.66 31.79 0.002 254/249Girls 18.19 17.54 < 0.001 32.50 34.44 0.016 238/233

aWilcoxon matched pairs signed-rank test, testing whether (Ind.)=(Comp.), henceforth SR test in the tablesb5 boys and 5 girls took part in the running task, but did not complete the skipping rope task.

-0.41

-0.63

-0.21

-0.66

p=0.023 p=0.001

0

-.2

-.4

-.6

-.8

Sec

onds

Grades 2 to 6 Grades 7 to 9

Running

0.47 -1.26

5.51

8.93

p=0.800

p=0.527

02468

10

Jum

ps


Skipping Rope

Boys Girls

By Task, Age Group and GenderAverage Performance Change: PE Tasks

Figure 1. Gender and competitiveness in running and skipping rope

9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Table 1 – Individual and competitive performance in running (seconds)and skipping rope (number of jumps)

Running task Skipping rope task

Age Group Gender Round 1 Round 2 SR testa Round 1 Round 2 SR test Nb


Grades 2 to 6 Boys 17.39 16.98 < 0.001 5.92 6.39 0.340 154/153Girls 18.95 18.32 < 0.001 8.38 7.12 0.692 147/145

Grades 7 to 9 Boys 15.13 14.92 < 0.001 67.10 72.27 < 0.001 100/96Girls 16.94 16.27 < 0.001 70.92 79.45 < 0.001 91/88

All Grades Boys 16.50 16.17 < 0.001 29.66 31.79 0.002 254/249Girls 18.19 17.54 < 0.001 32.50 34.44 0.016 238/233

aWilcoxon matched pairs signed-rank test, testing whether (Ind.)=(Comp.), henceforth SR test in the tablesb5 boys and 5 girls took part in the running task, but did not complete the skipping rope task.

-0.41

-0.63

-0.21

-0.66

p=0.023 p=0.001

0

-.2

-.4

-.6

-.8

Seco

nds


Running

0.47 -1.26

5.51

8.93

p=0.800

p=0.527

02468

10

Jum

ps


Skipping Rope

Boys Girls

By Task, Age Group and GenderAverage Performance Change: PE Tasks

Figure 1. Gender and competitiveness in running and skipping rope

9

Running: Gender of the opponent effect?

Change in performance in the compeJJve environment according to the gender composiJon of the pairs

Both boys and girls compete more with girls!

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Table 2 – Impact of opponent’s gender on competitiveness

Running againsta Skipping rope againstb

Sample of Boys Girls p≠value N Boys Girls p≠value N

Boys -0.25 -0.49 0.008 162/91 2.12 2.74 0.558 129/118Girls -0.46 -0.76 0.006 94/144 2.57 2.60 0.996 118/118

aAverage performance change (in seconds)bAverage performance change (number of jumps)

However, in the skipping rope task we do not find any gender di�erences in competitiveness (forthe whole sample: p = 0.917).13

Comparing the size of the gender gaps in performance change across age groups, we do notfind any significant di�erences in neither running (p = 0.341) nor skipping rope (p = 0.107).

In both running and skipping rope tasks, the students could observe each other while per-forming in the second round. In Table 2 we therefore explore whether the response to com-petition varies with the gender of the opponent in our sample. In running, both boys andgirls improve their performance significantly more when competing against a girl versus a boy(boys: p = 0.002; girls: p = 0.006), and this di�erence is mainly driven by the older students.14

However, we find no significant opponent gender e�ects in the skipping rope task.Summarizing, we have the following result:

Result 1. Girls are significantly more competitive than boys in the running task, while thereis no significant gender di�erence in the skipping rope task.

3.1.2. Competitiveness as a reaction to competition in math and word search

Table 3 shows average performance in the two classroom tasks under individual (piece rate)and competitive (tournament) incentives across ages, for boys and girls separately. We observe,again, that the gender gap in the first stage piece rate performance largely corresponds to thegender stereotype of the task. Girls perform significantly better in the verbal task (p < 0.05

13We have also conducted the same analysis using two measures of relative performance. We calculate theperformance change for each individual relative to the class mean, since the classmates can see each other whenthey perform in this part of the experiment. Further, older and younger children conducted di�erent types of theskipping rope tasks. This measure of relative performance change takes this into account. The second measureis the relative change in individual performance over the two rounds, i.e. (performance in round 2 - performancein round 1)/(performance in round 1). For both relative measures used, the main results are qualitatively similar(for younger children in the running task while the first relative measure of performance change is significantat conventional levels, the second relative measure is significant at 10% in a two sided test, and at 5% in a onesided test). Furthermore we have also checked if there are any di�erences in the variance of performance changebetween boys and girls. While for the running task there are no significant di�erences for any age group (p < 0.05for all pairwise comparisons, Levene’s test), in the skipping rope task girls have a significantly higher variance:at the median p = 0.033 for younger, p = 0.086 for older and p = 0.016 overall, Levene’s test. However, thisgender di�erence in variance also vanishes when we look at the relative performance change (all p > 0.5).

14In the sample of younger children, while this di�erence is not significant for boys (p = 0.378), it is marginallysignificant for girls (p = 0.073).

10

Math task

Appendix B

In this appendix I present examples of the experimental tasks used in the classroom for

the competitiveness part of the study.

Table 14 shows a few examples of the math exercises that the children solved for

various years. Children in Years 2 to 5 had only to add two 2-digit numbers, while

children in Years 6 to 7 had to both add and subtract two 2-digit numbers. Children in

Years 8 to 10 had to add and subtract in a random sequence of three 2 digit numbers.

All the numbers and mathematical operations were randomly generated to insure that

the level of di✏culty of the math task was the same throughout all the stages of the

experiment for each of the year categories.

Table 14: Examples of Math Tasks for Various Years

Years 2 to 3 Years 4 to 5 Years 6 to 7 Years 8 to 101 + 12 = . . . 82 + 18 = . . . 93 + 67 = . . . 96 + 93 + 3 = . . .

3 + 5 = . . . 48 + 10 = . . . 63 � 38 = . . . 33 � 9 � 85 = . . .

11 + 4 = . . . 47 + 14 = . . . 2 � 38 = . . . 83 + 97 + 14 = . . .

17 + 18 = . . . 39 + 6 = . . . 71 + 52 = . . . 31 � 39 + 28 = . . .

10 + 23 = . . . 68 + 16 = . . . 89 � 47 = . . . 9 � 41 � 75 = . . .

Figure 20 shows an example of the word search task that was used in the experiment.

The children had to find and circle words in any direction on a straight line. Since these

word search puzzles were not generated by a computer there might have been slight

di�erences in the di✏culty of the word search task in di�erent stages of the experiment.

We used the same puzzle within each stage of the experiment for all the children.

37

Word search task

!"# 1

$ % & ' ( ) ( * ( + , & -

. , / & 0 - 1 2 3 2 4 25 6

7 8 25 6 9 : $ ; < = 25 > -

9 2 ? 9 ( ? 3 ! @ A B ( :

. * . 7 . ) ( * & ? ( B 0

? ( = . ? 6 & C D ; ? 2 7

. 6 . B . ? . 7 & 0 ( @ .

7 % A ) . - 2 ( 8 25 7 & C

. 7 $ = C @ : 6 % < 25 ! 25

C 9 & 7 ( ? 2 + 7 0 4 ( E

. > 25 C = $ . 25 % . . , .

< . ? F 8 - ; * / G ) & 1

25 , & D 7 . 1 . B E . * A

CompeJJveness in Math and Verbal tasks 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Table 3 – Individual and competitive performance in the mathematicaland verbal tasks

Math taska Verbal taskb

Age Group Gender Stage 1 Stage 2 SR test Stage 1 Stage 2 SR test N


Gr. 2 to 6 Boys 17.44 20.83 < 0.001 5.31 5.56 0.071 224c

Girls 16.36 20.12 < 0.001 5.80 6.09 0.137 204

Gr. 7 to 10 Boys 8.12 10.32 < 0.001 9.20 9.74 0.112 164Girls 7.96 8.29 < 0.001 10.61 11.61 0.003 173

All Grades Boys 13.50 16.18 < 0.001 6.95 7.33 0.018 388Girls 12.51 15.62 < 0.001 8.01 8.62 0.001 377

aAverage number of correctly solved math exercisesbAverage number of correct words found in the word search puzzlecOne boy did not participate in the second stage of the verbal task

for all pair-wise comparisons). However, while boys solve slightly more exercises in the mathtask, this di�erence is not significant (p > 0.10 for all pairwise tests). In the second stage, whenstudents compete against each other, both boys and girls of all ages are competitive and reactto competition by increasing their performance significantly in both tasks.15

In Figure 2 we compare the change in performance between the first and the second stagesin the two classroom tasks. We find that boys and girls of all ages react similarly to competitionand improve equally in both tasks (p≠value for the whole sample, comparing boys vs girls is0.195 for math and 0.328 for the verbal task).16 Moreover, there is no di�erence in the size ofthe gender gap across age in both tasks (math: p = 0.771; word: p = 0.428).

Summarizing, we have the following result:

Result 2: Boys and girls do not di�er significantly in terms of performance change whenreacting to competition in both a math and a verbal task.

3.1.3. Competitiveness as willingness to compete in math and word search

Contrary to the two physical tasks, both the math task and the verbal task included a thirdstage, where participants could choose to perform the task under either piece rate incentivesor competitive incentives. In the math task, 54% of the boys and 52% of the girls choose tocompete, and in the verbal task the share is 57% for boys and 56% for girls. Neither gender

15In the second stage, again, boys perform better in the math task and girls perform better in the verbal task,but only in the verbal task the gender di�erence is significant (math: p = 0.384, verbal: p = 0.001).

16This finding is also robust to using relative measures of performance change. Furthermore, we have alsochecked whether there is a gender di�erences in the variance of performance change for both tasks. While girlshave a higher variance in performance change in the verbal task (all p < 0.01, Levene’s test), this di�erencevanishes when we look at relative performance change (all p > 0.19, Levene’s test).

11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

3.39 3.76

1.712.36

p=0.322

p=0.327

0

1

2

3

4

5

Num

ber o

f Exe

rcis

es


Math Task

0.25 0.290.54

0.99p=0.920

p=0.195

0

.5

1

Num

ber o

f Wor

ds


Verbal Task

Boys Girls

By Task, Age Group and GenderAverage Performance Change: Classroom Tasks

Figure 2. Gender and competitiveness in math and word search

di�erence is significant (math task: p = 0.708, verbal task: p = 0.697; ‰2 test for equality ofproportions).

In Figure 3 we explore competitive choices of boys and girls in each task for the two agegroups separately. While we do not find any gender di�erence within any age group, 61% ofyounger and 43% of older students choose to compete in math, and this di�erence is significant.This is also true for each gender over age (p = 0.001 for all pairwise comparisons, ‰2 test).In the verbal task, 55% of younger and 59% of older students choose to compete. Howeverwe do not find any significant gender or age di�erences in this task (p > 0.19 for all pairwisecomparisons, ‰2 test).

Summarizing, we have the following result:

Result 3: Boys and girls do not di�er significantly in their willingness to compete.

Comparing this result to those of Cardenas et al. (2012) who use similar tasks, it appearsthat Armenian children are more competitive than both Colombian and Swedish children, inparticular if we compare Armenian girls with Swedish girls. Among Colombian children, 35% ofthe boys and 32% of the girls choose to compete in the math task and 26% of the boys and 29%choose to compete in the verbal task. Among Swedish children, the corresponding numbers are44% of the boys and 19% of the girls in the math task and 39% of the boys and 27% of the girlsin the verbal task.

12

No significant gender gap in selecJon into compeJJon 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

0.61 0.60

0.43 0.43

p=0.855

p=0.924

0

.2

.4

.6

.8Co

mpe

te


Math Task

0.56 0.530.59 0.60

p=0.519p=0.942

0

.2

.4

.6

.8

Com

pete


Verbal Task

Boys Girls

By Task, Age Group and GenderShare of Individuals Willing to Compete

Figure 3. Gender and the choice to compete

3.1.4. Relative performance beliefs in the math and word search

As a measure of confidence related to performance in the classroom tasks we asked the childrenand adolescents about relative performance beliefs, both in the first stage piece rate setting andin the second stage tournament setting. In contrast to some previous research (e.g., Niederleand Vesterlund, 2007) — and interesting in relation to our null results on gender di�erences incompetitiveness in the classroom tasks — we find no gender di�erences in relative performancebeliefs for either setting (piece rate math: p = 0.286; piece rate verbal: p = 0.163; tournamentmath: p = 0.594; tournament verbal: p = 0.633). Moreover, our main results pertainingto gender and competitiveness (both performance change and willingness to compete) do notchange if we control for relative performance beliefs in a parametric regression framework.17

This result is in line with that of Cardenas et al. (2012), who also did not find that performancebeliefs mattered for willingness to compete in Colombia.

17For both tasks, when regressing willingness to compete on female, age, actual performance in the tournamentand confidence as measured via performance beliefs from the tournament, performance is significant and positivewhile confidence is significant and negative. There are also no significant gender and confidence interactions, seeSupplementary Online Appendix.

13

SelecJon into compeJJon

•  No gender effect in the regression analysis (OLS, logit, probit)

•  The only significant variables are either performance under

compeJJon or under piece rate and risk aversion

Risk Preferences o  Choose between a sure amount and a gamble, 10 points if

heads, 0 points if tails o  Sure amount increasing from 2 points to 6 in increments of 1,

and then 7.5 points o  Our main measure of risk aversion relies on the unique

switching point (certainty equivalent taken as the midpoint of the switching interval)

o  18% of subjects are inconsistent, leaving us with a sample of 624 subjects

o  AlternaJve measure of risk taking: number of risky choices –  Results are qualitaJvely the same

Risk Preferences Boys are more risk taking than girls .4

0.1

.2.3

Frac

tion

0 1 2 3 4 5 6Number of risky choices

Grades 2 to 6

0.1

.2.3

.4Fr

actio

n


Grades 7 to 10

0.1

.2.3

Frac

tion


All Grades

Number of risky options choses

0.1

.2.3

.4Fr

actio

n

2 4 6 107.5Sure amount in the switching row

Boys Girls

Grades 2 to 6

0.1

.2.3

.4Fr

actio

n


Grades 7 to 10

0.1

.2.3

Frac

tion


All Grades

Sure option in the switching row

By Risk Measure, Age Group, and GenderDistribution of Risk Preferences

Figure 1. Distribution of risk preferences

3

Risk Preferences Boys are more risk taking than girls

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

2.862.52

3.24

2.48

p<0.001p=0.116

0

1

2

3

4

Aver

age

num

ber o

f ris

ky c

hoic

es


By Age Group and GenderRisk Preferences

Boys Girls

Figure 4. Gender di�erences in risk preferences

3.2. Risk preferences

In this section we test whether there are gender di�erences in risk preferences in our sample.We measure risk preferences from six incentivized choices in the classroom. In our sample of762 students that took part in the risk task, about 19% are inconsistent in their choices, i.e.have multiple switching points. (There is no gender di�erence in being inconsistent in the wholesample, or any of the age groups considered: all p > 0.25, ‰2 equality of proportions test, seeSupplementary Online Appendix.) Inconsistent participants are excluded if we look only atthe unique switching point. We therefore use the number of risky options chosen as our mainmeasure of risk preferences. We find that, on average, boys choose the risky option significantlymore often than girls (boys: 3.02, girls: 2.50, p < 0.001). On average boys make 21% morerisky choices compared to girls.

In Figure 4 we display gender di�erences in risk preferences in the two age groups separately.While boys make significantly more risky choices than girls do, this gender di�erence doesnot reach statistical significance for younger children. It appears that in our sample genderdi�erences in risk taking arise around the age of puberty, and this is mainly due to increasedrisk taking of boys with age, see also Table 7 in Appendix C. Using the other risk measurebased on the unique switching point, and thus excluding inconsistent participants, our resultsare qualitatively the same. Testing whether the gender gap in risk taking is larger among olderchildren than younger ones, we find that while this di�erence is marginally significant with ourmain measure of risk (p = 0.06), it is significant for the other measure that excludes inconsistentparticipants (p = 0.04). However, we should also note that the distribution of risk preferencesis remarkably di�erent across our age groups, see Supplementary Online Appendix.

Summarizing, we have the following result:Result 4: Boys are more risk taking than girls, and this gender gap gets bigger in adoles-

cence.

14

Altruism from DG Girls donate 9% more on average then boys do, but doesn’t vary with age

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

50.58

58.88

52.70

61.66

p=0.007 p=0.014

30

40

50

60

70Av

erag

e po

ints

giv

en to

cha

rity


By Age Group and GenderAltruistic Preferences

Boys Girls

Figure 5. Gender di�erences in altruism

These results on risk preferences are in line with those of Cardenas et al. (2012), who findthat Colombian boys on average take 40% more risk than girls, with the corresponding numberin Sweden being 15%.

3.3. Altruism

In this section we look at gender di�erences in altruism as measured via donations in a dictatorgame. We find that girls give significantly more than boys (p < 0.001), and this is the case forboth age groups, see Figure 5. Girls donate on average 60.16 points whereas boys donate 51.48points out of 100 to the charity organization (an orphanage). The modal allocation is the 50–50split for the sample as a whole, as well as for both genders and age groups.

Testing whether the gender gap in altruism is bigger among the older students, we do notfind a significant di�erence (p = 0.89). Hence, we have the following result:

Result 5: Girls are significantly more altruistic than boys.

3.4. The relationship between competitiveness, risk preferences, and altruism

We also explore the relationship between competitiveness, risk taking, and altruism, since thethree areas often are related, yet are three separate concepts, and gender di�erences are oftenfound in all three preferences. A correlation analysis between all the behaviors we examineshows that while competitiveness and risk taking are related, our measure of altruism is neitherrelated to competitiveness nor to risk taking, see Supplementary Online Appendix.18

Like many others, e.g. Niederle and Vesterlund (2007); Niederle and Yestrumskas (2008),we find a positive relationship between risk taking and willingness to compete in math and

18We have also tested whether the distributions of donations di�er between those who choose to compete andthose who do not, across tasks, age and gender. For both tasks none of these (pairwise) comparisons is significant,all p > 0.20.

15

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

50.58

58.88

52.70

61.66

p=0.007 p=0.014

30

40

50

60

70

Aver

age

poin

ts g

iven

to c

harit

y


By Age Group and GenderAltruistic Preferences

Boys Girls

Figure 5. Gender di�erences in altruism

These results on risk preferences are in line with those of Cardenas et al. (2012), who findthat Colombian boys on average take 40% more risk than girls, with the corresponding numberin Sweden being 15%.

3.3. Altruism

In this section we look at gender di�erences in altruism as measured via donations in a dictatorgame. We find that girls give significantly more than boys (p < 0.001), and this is the case forboth age groups, see Figure 5. Girls donate on average 60.16 points whereas boys donate 51.48points out of 100 to the charity organization (an orphanage). The modal allocation is the 50–50split for the sample as a whole, as well as for both genders and age groups.

Testing whether the gender gap in altruism is bigger among the older students, we do notfind a significant di�erence (p = 0.89). Hence, we have the following result:

Result 5: Girls are significantly more altruistic than boys.

3.4. The relationship between competitiveness, risk preferences, and altruism

We also explore the relationship between competitiveness, risk taking, and altruism, since thethree areas often are related, yet are three separate concepts, and gender di�erences are oftenfound in all three preferences. A correlation analysis between all the behaviors we examineshows that while competitiveness and risk taking are related, our measure of altruism is neitherrelated to competitiveness nor to risk taking, see Supplementary Online Appendix.18

Like many others, e.g. Niederle and Vesterlund (2007); Niederle and Yestrumskas (2008),we find a positive relationship between risk taking and willingness to compete in math and

18We have also tested whether the distributions of donations di�er between those who choose to compete andthose who do not, across tasks, age and gender. For both tasks none of these (pairwise) comparisons is significant,all p > 0.20.

15

Summary of Results

o  Girls are significantly more compeJJve than boys in a running task

o  In all other tasks and measures, girls and boys are equally compeJJve o  Boys and girls choose to compete at equal rates in either task

o  Boys are more risk taking than girls and the gender gap appears around the age of puberty

o  Boys are less altruisJc than girls o  Culture and context play even a bigger role than iniJally

expected, and not always in predictable ways

Thank You!

Documents

Gender and Preferences at a Young Age: Evidence from Armenia