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
dierenJals
Bonin et al. 2007, Manning and Swaeld 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, Crdenas et al. 2011

Impacts of culture and context

Massai vs Khasi
Patriarchal vs matrilineal society

Running
Dierence between GR04 and DvER11

Single-sex vs coeducaJonal schools

SelecJon issues?

Colombia vs Sweden
Unexpected ndings: No gender gaps in Colombia in all
tasks, but some gender gaps in Sweden

Gneezy & RusJchini 2004, Booth & Nolen 2011, Dreber et al. 2011, Crdenas 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, Crdenas et al. 2011,

Impact of tasks
Some inuence
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, Crdenas et al. 2011

QuesJons and ContribuJon

o What causes the gender gap in these preferences?
When does it emerge?
What is the inuence of the country in which the study
is performed (culture and context)?
What is the inuence of tasks?
Extrinsic vs intrinsic moJvaJon?
o How do these preferences develop?
InteresJng to compare dierent 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,
signicantly more compeJJve in running task

1 Individual
andrope,
competitive
running (seconds)
Girls aTable
re beber
in skipping
but no performance
signicant indierence
in performance
and skipping rope (number of jumps)
change reacJon to compeJJon

Age Group

Running task
Gender

Skipping rope task

Round 1
(Ind.)

Round 2
(Comp.)

SR testa
(pvalue)

Round 1
(Ind.)

Round 2
(Comp.)

SR test
(pvalue)

Nb

Grades 2 to 6

Boys
Girls

17.39
18.95

16.98
18.32

< 0.001
< 0.001

5.92
8.38

6.39
7.12

0.340
0.692

154/153
147/145

Grades 7 to 9

Boys
Girls

15.13
16.94

14.92
16.27

< 0.001
< 0.001

67.10
70.92

72.27
79.45

< 0.001
< 0.001

100/96
91/88

All Grades

Boys
Girls

16.50
18.19

16.17
17.54

< 0.001
< 0.001

29.66
32.50

31.79
34.44

0.002
0.016

254/249
238/233

a
b

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

Average Performance Change: PE Tasks

By Task, Age Group and Gender

Running

Seconds

-.8
-.6
-.4

p=0.001

p=0.023

-0.66

-0.63
-0.41

-.2

-0.21

0
Grades 2 to 6

Grades 7 to 9

Skipping Rope
p=0.527

10
Jumps

8
6

8.93

p=0.800

5.51

4
2
0

0.47

-1.26

Grades 2 to 6

Grades 7 to 9
Boys

Girls

Running: Gender of the opponent eect?

Change in performance in the compeJJve environment according to the
gender composiJon of the pairs
Both boys and girls compete more with girls!

Table 2 Impact of opponents gender on competitiveness
Running againsta

Skipping rope againstb

Sample of

Boys

Girls

pvalue

Boys

Girls

pvalue

Boys
Girls

-0.25
-0.46

-0.49
-0.76

0.008
0.006

162/91
94/144

2.12
2.57

2.74
2.60

0.558
0.996

129/118
118/118

a
b

Average performance change (in seconds)

Average performance change (number of jumps)

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

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

he level of diculty of the math task was the same throughout all the stages of th

Math task

xperiment 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 10

1 + 12 = . . .
3 + 5 = ...
11 + 4 = . . .
17 + 18 = . . .
10 + 23 = . . .

82 + 18 = . . .
48 + 10 = . . .
47 + 14 = . . .
39 + 6 = . . .
68 + 16 = . . .

93 + 67 = . . .
63 38 = . . .
2 38 = . . .
71 + 52 = . . .
89 47 = . . .

96 + 93 + 3 = . . .
33 9 85 = . . .
83 + 97 + 14 = . . .
31 39 + 28 = . . .
9 41 75 = . . .

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

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

ord search puzzles were not generated by a computer there might have been slig

Word search task

$

&

'

&

&

25

25

<

25

>

&

&

&

25

&

<

25

25

&

>

25

25

<

&

25

&

CompeJJveness
i
n
M
ath
a
nd
V
erbal
t
asks
Table 3 Individual and competitive performance in the mathematical
and verbal tasks

Math taska
Age Group

Gender

Verbal taskb

Stage 1
(Ind.)

Stage 2
(Comp.)

SR test
(pvalue)

Stage 1
(Ind.)

Stage 2
(Comp.)

SR test
(pvalue)

Gr. 2 to 6

Boys
Girls

17.44
16.36

20.83
20.12

< 0.001
< 0.001

5.31
5.80

5.56
6.09

0.071
0.137

224c
204

Gr. 7 to 10

Boys
Girls

8.12
7.96

10.32
8.29

< 0.001
< 0.001

9.20
10.61

9.74
11.61

0.112
0.003

164
173

All Grades

Boys
Girls

13.50
12.51

16.18
15.62

< 0.001
< 0.001

6.95
8.01

7.33
8.62

0.018
0.001

388
377

Average number of correctly solved math exercises

Average number of correct words found in the word search puzzle
c
One boy did not participate in the second stage of the verbal task
b

for all pair-wise comparisons). However, while boys solve slightly more exercises in the math

Average Performance Change: Classroom Tasks

By Task, Age Group and Gender

p=0.322

5
4

p=0.327

3.76

3.39

3
2

2.36
1.71

1
0

Grades 2 to 6

Grades 7 to 10

Verbal Task
p=0.195
Number of Words

Number of Exercises

Math Task

.5

0.99
p=0.920

0.54
0.25

0.29

Grades 2 to 6
Boys

Grades 7 to 10
Girls

No signicant gender gap in selecJon into compeJJon

Share of Individuals Willing to Compete
By Task, Age Group and Gender
Math Task

Compete

.8
.6

p=0.855
0.61

p=0.924

0.60

.4

0.43

0.43

.2
0

Grades 2 to 6

Grades 7 to 10

Verbal Task

Compete

.8

p=0.942

p=0.519

.6
.4

0.56

0.60

0.59

0.53

.2
0

Grades 2 to 6

Grades 7 to 10
Boys

Girls

SelecJon into compeJJon

No gender eect in the regression analysis (OLS, logit, probit)
The only signicant 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

Distribution of Risk Preferences

By Risk Measure, Age Group, and Gender

Risk Preferences

Number of risky options choses

Sure option in the switching row

Boys are more risk taking than Grades

girls 2 to 6
.4

Girls

4
6
7.5
Sure amount in the switching row

10

.1

Fraction
.2
.3

Boys

.1

Fraction
.2
.3

.4

Grades 2 to 6

2
3
4
Number of risky choices

Fraction
.2
.3
.1
0

.1

Fraction
.2
.3

.4

Grades 7 to 10

.4

Grades 7 to 10

2
3
4
Number of risky choices

10

Fraction
.1
.2
0

Fraction
.1
.2

.3

All Grades

.3

All Grades

4
6
7.5
Sure amount in the switching row

2
3
4
Number of risky choices

4
6
7.5
Sure amount in the switching row

10

Risk Preferences
Boys are more risk taking than girls
Risk Preferences
Average number of risky choices

By Age Group and Gender

4
p<0.001

p=0.116
3.24

3
2.86
2.52

2.48

Grades 2 to 6

Grades 7 to 10

Boys

Girls

Figure 4. Gender differences in risk preferences

3.2. Risk preferences

Altruism from DG

Girls donate 9% more on average then boys do, but doesnt vary with age
Altruistic Preferences
Preferences
ByAltruistic
Age Group
and Gender

70

60

50

Average points given to charity

By Age Group and Gender

Average points given to charity

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

70
p=0.014

p=0.007
p=0.007

p=0.014

60

61.66
58.88

50

50.58

61.66

58.88

52.70

52.70

40

50.58

30

40

Grades 2 to 6

Grades 7 to 10
Boys

Girls

Figure 5. Gender differences in altruism

30

Grades 2 to 6

Grades 7 to 10

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

Figure 5. Gender differences in altruism

3.3. Altruism

In this section we look at gender differences in altruism as measured via donations in a dictator

findpreferences
that girls give significantly
more with
than boys
(p < 0.001),
and this is the
These resultsgame.
on We
risk
are in line
those
of Cardenas
et case
al. for(2012), who find
both age groups, see Figure 5. Girls donate on average 60.16 points whereas boys donate 51.48

Summary of Results
o Girls are signicantly 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!