Identity & Role Models

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/323972833

Social Identity and Role Models

Article  in  SSRN Electronic Journal · January 2016

DOI: 10.2139/ssrn.2783502

CITATIONS

0 READS

206

2 authors, including:

Some of the authors of this publication are also working on these related projects:

Spatial Distribution of Poverty in the European Union View project

Income Stratification and Slovak Society Polarisation Perspectives until 2030 View project

Tomas Zelinsky

Technical University of Kosice - Technicka univerzita v Kosiciach

35 PUBLICATIONS   126 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Tomas Zelinsky on 14 May 2019.

The user has requested enhancement of the downloaded file.

Department of Economics

School of Business, Economics and Law at University of Gothenburg

Vasagatan 1, PO Box 640, SE 405 30 Göteborg, Sweden

+46 31 786 0000, +46 31 786 1326 (fax)

www.handels.gu.se [email protected]

WORKING PAPERS IN ECONOMICS

No 672

Social Identity and Role Models

Joseph Vecci and Tomás Zelinský

September 2016

ISSN 1403-2473 (print) ISSN 1403-2465 (online)

Social Identity and Role Models ∗

Joseph Vecci† Tomáš Želinský‡

September 17, 2016

Abstract

We present a lab-in-the-field experiment and surveys of marginalised Roma chil- dren in Slovakia to examine whether reminding Roma of their ethnicity reduces their performance in a cognitive task. Research on social identity and stereotypes has documented that when individuals feel their social group is negatively stereo- typed in a domain their performance declines, which can reinforce discrimination. In an effort to break the cycle of negative stereotypes we remind Roma of either Roma or non-Roma role models. We find that the activation of a Roma’s ethnicity reduces cognitive performance. In contrast Roma exposed to Roma role models outperform those reminded of their ethnicity and also non-Roma role models. We then attempt to understand the channels through which social identity and role models effect performance. We show that priming a Roma’s identity has a direct effect on confidence, decreasing performance.

JEL Codes: C93, J15.

Keywords: Social Identity, artefactual field experiments, discrimination, role

models, Roma.

∗We are grateful for comments and suggestions from Julie Chytilova, Lata Gangadharan, Vilas Gobin, Andreas Leibrandt, Pushkar Maitra, Dillip Mookerjee, Paulo Santos, Julia Talbot-Jones and Kevin Wu. This paper benefited from comments by participants at the University of Gothenburg, Australian and New Zealand Workshop on Experimental Economics, Monash University, the Joint Annual Meeting of the Slovak Economic Association and the Austrian Economic Association, The Economic Science Association Annual Meetings and the Southern Economic Association Conference. We thank Slávka Mačáková of ETP Slovakia and her colleagues for assistance with organizing pilots and providing contacts in communities, and Juliána Lachváčová, Kamila Kurincová, Katka Jacková, Ľubica Štiblárová, Marek Řihák, Martina Žudelová, Silvia Mrázková, Slávka Martinková, Štefan Kováč, Tomáš Štofa and social community workers for assistance in the field. This work was supported by the Slovak Research and Development Agency under the contract No. APVV-0125-12. IRB clearance obtained from the Technical University of Košice. †[email protected], Department of Economics, University of Gothenburg, Gothenburg, Sweden ‡[email protected], Faculty of Economics, Technical University of Košice, Košice, Slovakia

1 Introduction

Social identity commonly refers to an individual’s own perception of self, based on

his or her membership of a group such as ethnicity, race, or gender (Tajfel, 1973). A

person’s social identity provides a set of rules that govern group behaviour, as such social

identity can explain behavioural differences across groups (Hoff and Pandey, 2006; Chen

and Li, 2009; Benjamin et al., 2010; Akerlof and Kranton, 2000; Akerlof, 2002; Morita

and Servátka, 2013). An important aspect of one’s social identity are the stereotypes;

the physical, mental and psychological characteristics attributed to a typical member

of a given social group. Once a set of characteristics are used to describe a social

group, those characteristics influence the behaviour of people who are associated with

the group (Shih et al., 1995) 1. A negative stereotype associated with one’s social group

can generate negative perceptions about one’s self, leading people to perform worse than

their abilities would suggest. This is commonly known as stereotype threat. 2

Stereotype threat is especially prevalent in populations that suffer from discrimination.

The perceived negative characteristics of these groups in a particular domain, tends to

influence the group members’ self-beliefs (DellaVigna, 2010). As such, negative stereo-

types have been found to explain race-based performance differences on academic tests

(Steele, 1997; Steele and Aronson, 1995; Shih et al., 1995).3 For instance, in India

when the associated stereotypes of low caste, who have been historically discriminated,

is made salient, the group’s performance declined compared to the non-discriminated

high caste (Hoff and Pandey, 2006, 2014) .

Negative stereotypes can be particularly harmful as they can re-enforce discrimination

1In recent work Bordalo et al. (2016) show that stereotypes predominately highlight the largest differences between groups as such many stereotypes are exaggerated.

2 For instance, African Americans are often stereotyped as low academic-ability, giving them reduced incentives to work hard to be rewarded for their effort (DellaVigna, 2010).

3Shih et al. (1995) find that Asian American women perform better on a maths test when their ethnic identity is activated, but worse when their gender identity is activated. A similar stereotype threat was also found when comparing a task named “sport intelligence” in one treatment and “natural athletic ability” in the second treatment on a sample of black and white college students. Black participants performed significantly worse in “sports intelligence” diagnostic and in contrast, white participants performed worse in the “natural athletic ability” diagnostic (Stone et al., 1999).

1

across groups. A negative stereotype increases negative thoughts about oneself and can

reinforce a negative identity. In turn, this can directly affect confidence, aspirations

and levels of effort expended by individuals or groups, ultimately affecting their learn-

ing and performance, re-enforcing negative attitudes towards the group, and terminally

discrimination (Loury, 2002). A key component of this cycle is that negative stereo-

types not only influence how a group member perceives themselves but also how others

perceive the social group. For example, research has found that negatively stereotyped

names can reduce an employer’s effort to inspect resumes (Bartoš et al., 2016), stereo-

types also influence wages and employment opportunities (Bertrand and Mullainathan,

2004).4 Thus, understanding the factors that influence group stereotypes can be impor-

tant in countering discrimination directed towards social groups.

In this paper we examine the affect of role models on social identity and in particular

stereotypes associated with a social identity. We begin by investigating whether remind-

ing a negatively stereotyped ethnic group of their ethnic identity affects their academic

performance. Then in an effort to break the cycle that negative stereotype may per-

petuate we examine the activation of ethnic and non-ethnic role models on academic

performance. Reminding people of their role models, individuals who are perceived as

worthy of emulation (Pleiss and Feldhusen, 1995), may remind subjects of a positive

aspect of their social identity, offsetting negative self perception, improving performance

and as such nullifying the effects of negative stereotypes (Akerlof and Kranton, 2000).5

By comparing ethnic role models to non-ethnic role models we can also identify if role

models, who are similar to participants in terms of ethnicity and perhaps upbringing,

have a different effect on behaviour than role models whom students may not be able

to relate. Lastly, we attempt to understand the channel through which negative stereo-

types and role models affects performance. We hypothesise that priming subjects ethnic

identity may have a direct effect on confidence, resulting in a decrease in performance.

4See (Guryan and Charles, 2013) for an excellent discussion on statistical discrimination while List and Rasul (2011) and Altonji and Blank (1999) provide a summary of the related literature.

5Research on social identity confirms that similarity between self and others increase the likelihood of social comparison (Tajfel, 1973).

2

As such we test whether confidence changes when a subject’s ethnicity is made salient

and whether role models can offset any reduction in confidence.

We employ an artefactual field experiment with 396 Roma children in Eastern Slovakia.

Roma are the largest minority in Europe and according to the EU they suffer from

widespread discrimination. In a recent report Roma were found to be the minority

discriminated against most often with 47% of Roma feeling discriminated, followed by

Sub-Saharan (41%) and North (41%) Africans, Central and Eastern Europeans (23%),

Turkish (23%) and Russians (14%) (FRA, 2009). This translates into Roma’s belief

that a negative attitude towards them exists in education, housing, and employment

(European Commission, 2014). This context provides a unique opportunity to study the

affects of social identity and whether role models can influence the negative stereotypes

associated with ones social group.

Similar to Shih et al. (1995), Benjamin et al. (2010), Cadsby et al. (2013),Yan Chen and

Shih (2014) and Cohn et al. (2015) we use a pre-experiment background questionnaire

to make Roma’s ethnicity salient. Participants then take part in a simple math task

to gage cognitive performance. In two additional treatments using the background

questionnaire participants are either reminded of Roma or non-Roma role models. We

find that the activation of a Roma’s ethnicity reduces cognitive performance compared

to the control of no ethnic or role model information. In contrast, students reminded

of Roma role models outperform students reminded of their ethnicity and non-Roma

role models, but they perform similar to the control. We then find that the reduction

in performance when children are reminded of their identity can at least partially be

attributed to a reduction in confidence. Our finding suggest that Roma role models can

potentially reverse this decrease in confidence.

The effect of role models on students’ outcomes has been investigated in a number

of empirical studies. This research suggests that role models significantly influence

outcomes as they have the potential to motivate individuals and serve as a source

3

of inspiration (Hurd et al., 2011). In this regard Dee (2004, 2005, 2007); Ehrenberg

et al. (1995) estimate the race interactions between students and teachers. They find

a positive effect of similar race teachers on various student outcomes. Fairlie et al.

(2014) use survey data to examine if minority students benefit from taking courses with

a minority lecturer. They show that the performance gap in terms of class drop-out

rates and grade performance between white and under-represented minority students

was lower when taught by minority instructors. Bettinger and Long (2005) investigate

the impacts of policies designed to increase female representation on college faculties.

Their research suggests that female instructors positively influence course selection and

major choice in some disciplines. While DellaVigna (2010) examines whether Barack

Obama had a positive effect on the achievements of African Americans, he finds little

difference in a range of outcomes including crime rates and labour force participation

between African Americans and whites. Further, role models have been shown to have

an affect on adolescent consumer purchase intentions and purchase behaviour (Martin

and Bush, 2000) as well as adolescents’ attitudes toward violence and violent behaviour

(Hurd et al., 2011).

There is a large psychology and growing economic literature on social identity and

separately on role models. Our work differs from this research in three important ways.

First, as discussed above, the economic literature has largely focused on the effect of role

models on education and wage outcomes. It is largely mute on the effect of role models on

social identity. We bridge the literature on social identity and in particular stereotypes,

with that of role models. Role models may have a direct effect on one’s self perception

and as such identity. Akerlof and Kranton (2000) hypothesise that role models may

help individuals define their self-concept or sense of self as such role models may play

an important role in social identity. This paper provides empirical analysis of this

hypothesis. In this respect the paper most similar to ours is the novel work by Olivetti

and Zenou (2013), who show that mothers and friends’ mothers shape the work choices

of their children later in life. However, they do not explicitly examine social identity

4

or the effect of role models on negative stereotypes, creating a gap in the literature.

Second, our experimental approach allows us to explicitly observe and identify the

effects of role models from other social and environmental factors. This is particularly

difficult in empirical studies because many factors that influence student and role model

interactions tend to be unobserved. For example without an experiment it is difficult to

isolate whether effects are due to students or teachers behaving differently. Third, we

add to the literature by investigating confidence as a possible channel through which

negative stereotypes and importantly role models affect behaviour. By influencing belief

in one’s social group, role models may directly affect confidence. Credible empirical

evidence on the effect of role models on self-confidence is rare due to the difficulties in

measuring and collecting data on confidence. Improved self-confidence has been found

to impact: motivation (Bénabou and Tirole, 2002); firm outcomes (Camerer and Dan,

1999); labour market outcomes (Koszegi, 2006); wage rates (Fang and Moscarini, 2005);

and the persistence in intergenerational income and educational inequality (Filippin and

Paccagnella, 2012).

2 Experiment Design

Upon the commencement of the experiment subjects responded to a “background ques-

tionnaire” that varied by experimental treatment (see A.1). The experiment consisted

of a control whereby neither ethnicity nor role models were made salient and three treat-

ments: 1) Roma ethnicity salient (Roma salient); 2. Roma role model; 3. non-Roma

role model. Following Shih et al. (1995), Benjamin et al. (2010) 6, we use a background

questionnaire because it makes identity and role models salient without explicitly re-

minding the subjects of their ethnicity, this reduces the probability of an experimenter

demand effect.

In the control neither stereotype or role models were made salient. The background

6See also (Cadsby et al., 2013),(Yan Chen and Shih, 2014), and (Cohn et al., 2015)

5

questionnaire consisted of 5 non-de-script background questions such as favourite food,

favourite colour etc. The questions were unrelated to identity or role models. In the

Roma ethnicity salient treatment, the background questionnaire included the same five

simple questions asking subjects their favourite food, favourite colour followed by three

questions on their ethnicity, these were: their self-classified ethnicity, language spoken

most frequently at home and whether their grandparents spoke any language other than

Roma.

The background questionnaire in the Roma role and non-Roma role model treatments

did not include questions on ethnicity but consisted of six simple questions regarding

famous Roma/non-Roma icons. In this study, we define role models as nonparental

adults who adolescents look up to and want to emulate. To gather information on

the people Roma children look up to we surveyed and informally interviewed 50 Roma

adults and children as well as teachers and social community workers two weeks prior

to the experiment. Each interviewee was asked to list the people they most look up to

or they think Roma look up to. A list was then compiled of the most popular Roma

and non-Roma. In order to cover a broad group of role models we, selected popular

male and female role models from a range of occupations. As Roma role models the

following people were selected: Vladimir Olah, a poet who established a Roma cultural

association in Slovakia; Dr Jan Cibula, an activist who was nominated for a Nobel

Peace Prize and was a past president of the International Roma Union; and four Slovak

or Czech popular Roma singers/performers. These were: Silvia Sarkoziova, the lead

female vocalist in the Gypsy Devils band; Igor Kmeto, the front man of a popular band

Kmetoband; female vocalist Vera Bila; and the members of an all male Slovak band

Gipsy Kajkos. The participants were asked questions such as the maiden name of Vera

Bila, the number of band members in Gipsy Kajkos, the name of Silvia Sarkoziovas’

group, and the association established by Vladimir Olah. The survey did not contain

reference to the role models ethnicity.

To ensure Role models are comparable across treatments we selected the most pop-

6

ular four non-Roma musicians and two non-Roma people from other fields, similarly

we ensured there were two female non-Roma role models. The non-Roma role models

consisted of two football players Cristiano Ronaldo and Marek Hamsik from Portugal

and Slovakia respectively and four singers/performers Justin Bieber (Canada), Shakira

(Columbia), Helena Vondrackova (Czech) and Karel Gott (Czech) – the latter two hav-

ing been popular in Slovakia (and formerly Czechoslovakia) for decades. Questions

included “the soccer club of the sportsmen, the age of Justin Bieber and whether the

singer Karel Gott has been awarded “The Golden Nightingale” prize more than 20 times.

The ethnicity of the role models was never made salient.

To ensure the icons selected are actually the subjects’ role models, as part of the post

experiment survey we asked subjects whether they admire the people mentioned in the

background questionnaire. In the Roma role model treatment 90% of subjects and 94%

in the non-Roma role model treatment look up to the people/groups mentioned. This

suggests that the experiment is measuring who Roma children look up to.7

After participating in the background questionnaire, subjects took part in a numeracy

task. Participants were given 65 strings of numbers, each string contained between

10 and 20 digits. Subjects were given 3 minutes to count the digit of zero in each

string. For each correctly solved puzzle children received 5c. We selected this task as

it required mathematical ability, yet it was simple enough for children who have not

completed primary school to understand. Roma children are perceived to lack academic

ability, and as such the prevailing stereotype is that their numeracy skills are lower than

the average population.8 Further, the task was labeled as a “numeric maths task” to

emphasise the tasks mathematics nature. The instructions included an example practice

string of numbers that subjects completed before they were able to commence the task.

7This differs to a large part of the Role model literature such as Marx et al. (2009) who select people that may act as a role model such as teachers and then test if a role model effect exists.

8According to UNDP (2012) approximately 10% of adult Roma’s attained higher than primary education and 38% (of those older than 16) self-report difficulties with reading and/or writing. While in Europe Roma children are 5 times less likely to attend compulsory primary education compared to the majority (FRA, 2014a).

7

After understanding of the instructions had been checked but prior to the commence-

ment of the numeric task a measure of confidence was elicited. We examine confidence

as a possible channel through which identity and role models effect behaviour. Similar

to Niederle and Vesterlund (2007) and Dasgupta et al. (2015) subjects were asked to

predict their own performance in the task. Subjects were asked to provide an estimate

of the number of numeric puzzles they expected to complete in 1 minute and 3 min-

utes. Then in the post experiment survey two further questions were asked to elicit

relative confidence : 1) If you were to compare your performance with everyone else in

this settlement, how would you rate your performance in comparison to other people?

Possible answers were: “among the best, better than average, same as average, worse

than average and among the worst”; 2) The second question was identical to the first

except settlement was replaced with Slovakia. The latter two questions were asked in

the post experiment survey in order to avoid priming ethnicity.

3 Setting and Village Selection

Roma children are the subject of our study. We specifically select early adolescent

children, as researchers have found that as children enter adolescence, they increasingly

focus their attention on nonparental adults to identify models of who they want to

emulate (Erikson, 1968; Scales and Gibbons, 1996). We study Roma as they are the

largest minority in Europe and according to the EU Commission on Justice, they suffer

from pervasive historic discrimination, which has further risen during the economic

crisis. In EU Member States 85% of Italians and 66% of French hold an unfavourable

views of Roma. While in Greece, Britain and Poland, about half hold a negative view of

Roma, similarly 40% hold this view in Germany and Spain (PRC, 2014). In the context

of education around 60% of Slovak pupils reported an objection when asked to share

the same desk with a Roma pupil and almost 50% report bad, or very bad experiences

with Roma people (Slovikova, 2012). Similarly, according to Bielikova (2010) Slovak

8

adolescents are the least tolerant towards Roma – 39% of respondents consider Roma

to be “inferior people”, while 5.9% of pupils considered Africans and 6.5% Asians to

be “inferior people”. These experiences are confirmed in our post experiment survey

which asks subjects “How often do you hear your class mates saying bad things about

Romas?”, 41% report every day, 13% of respondents report at least once a week, while

22% report at least once a month. This suggests perceived negative perceptions towards

Roma children exist within these communities.

The experiment was conducted in November 2015 in 15 municipalities in Eastern Slo-

vakia with Roma children living in segregated settlements (on the edge or outskirts of

municipality). It’s estimated that Eastern Slovakia accounts for 85% of the total Roma

population living on the edge or outskirts of municipalities in Slovakia. In Slovakia

around 91% of Roma people are at risk of poverty, while only 21% are in paid work

(FRA, 2014b), 87% of the Roma population aged 18-24 leave school without complet-

ing secondary education (FRA, 2014a). The situation is similar in other EU countries

where: 87% of Roma are at risk of poverty, 35% of men and 21% of women are in a

paid work (FRA, 2014b), and early school leavers rates of Roma range between 72 –

98% (FRA, 2014a).

Using the Atlas of Roma Communities in Slovakia 2013 (Musinka et al., 2014) we ran-

domly select 15 villages in four districts in Eastern Slovakia. The Atlas is a census of

the Roma community initiated by the Slovak government and UNDP aimed at moni-

toring the living conditions of the Roma population. Information was collected on all

settlements in Slovakia with 30 or more Roma people. Column 1 in Table 1 presents

the average characteristics for the sample settlements while column 2 presents average

characteristics for Eastern Slovakia municipalities (with Roma settlements on the edge

or outskirts). The results show that there is no statistically significant difference be-

tween the sample settlements and other Eastern Slovakia settlements in characteristics

such as percentage of households with access to water, public sewerage system, using

electricity/gas and distance to kindergarten/primary school or physician. Distance to

9

the closest general practitioner and pediatrician show a small statistical difference. All

selected settlements are not further than 30 km from the regional capitals: cities of

Košice and Prešov (see Fig. 1).9

3.1 Participant Recruitment and Procedure

To recruit participants for the experiment, village representatives were contacted by

telephone and/or email, asking for permission to conduct research in their community

centre. In each village one local social community worker was then hired to invite and

recruit individual households.10 Each social worker was trained in how to select and

invite participants.11 The social community workers were directed to randomly select

households from different parts of the village the day before the actual experiment.

Community workers approached households informing them that they could participate

in research the following day with the possibility of earning some money and that a

session would last 60-90 minutes. Within each household unit we invited both parents

(or just the mother/father if a single parent household) and two randomly selected

children above the age of 8.12 Upon agreeing to participate a contact number for each

household was recorded. On the morning of the experiment participants were informed

of the time of their session.13 Only children who were able to read, write and understand

basic math were eligible to participate.

Each session consisted of four children from at least 2 households. Between four and

six sessions took place in each village, cumulating in between 32 – 48 participants per

9To aid in comparing community characteristics in Table A1 we report the average characteristics of our sample villages (including Roma and non-Roma) taken from the 2011 census and compare them to Eastern Slovakia and Slovakia. As expected our sample village contains a much higher rate of Roma than villages in either Eastern Slovakia or Slovakia in general.

10The social workers were either employed with an NGO (ETP Slovakia – www.etp.sk) working on development programmes in the Roma settlements or employed by municipalities. Social community workers were responsible for visiting Roma families and providing counselling services to them on a daily basis. Because of this relationship a high level of trust towards them exists.

11Social workers were never informed about the topic of the research. 12Adults participated in a separate experiment. There were 19 such cases that more than two children

from the household participated. 13Some of the participants were advised they would serve as alternates.

10

village and 396 subjects in total. We do not believe that information spillovers were

a significant issue due to the task being explained to participants as a simple maths

quiz reducing the likelihood of participants identifying the research questions being

investigated.

Experiments took place in the village community centre. Upon arrival the participants

were screened for eligibility and the consent form read aloud, and a hard copy then signed

by the parents. Adults were then moved to a separate room where they participated

in a separate experiment.14 Upon entering the room each child was placed on a desk

and assigned a unique ID in order to ensure anonymity of the participants. Each

participant received their own envelope with answer sheets with their unique participant

ID. Decision sheets were handed to and collected from participants simultaneously. The

average earning of participants was e5. Data was entered and checked by two separate

research assistants. Three pilot rounds were conducted prior to the experiment (pilot

data is not included in the sample).

3.2 Treatment Balance

Table 2 lists the demographics of the average subject in our experiment. Column 1

presents the full sample means, columns 2 - 5 report the averages by treatment where

“C” refers to the control, “T1” the Roma salient treatment, “T2” Roma role model

treatment, and “T3” non-Roma role model. Participants in the sample are on average

13 years old and are currently attending primary school (80%). Within participating

households, 71% of children’s parents are married and 62% have at least one parent that

completed primary school. Almost 9% of households have a parent who is employed and

45% have a parent that is actively unemployed. Finally, the mean household income is

e383.44 per month.

Columns 6-11 present the mean differences in demographics across treatments. To com-

14The experiment consisted of a discrete choice experiment exmaining employment preferences.

11

pare treatments we use a Mann-Whitney test and well as the normalised difference

between treatments, reported in square brackets (Imbens and Wooldridge, 2009). We

find little systematic difference in average characteristic across treatments barring gen-

der. According to the Mann-Whitney test there are a greater number of male children in

the Roma role model treatment compared with the control and Roma salient treament.

As a robustness test, we also examine behaviour by gender. On the other hand the

normalised difference shows that only a single pairwise comparison (Gender: Control -

Roma Role Model)is above the 0.25 rule of thumb that would indicate differences in the

distribution across treatments (Imbens and Wooldridge, 2009). This suggests that on

average subjects are similar in terms of observable characteristics.

4 Results

In this section we provide details of the experimental results. In order to understand

the effect of a subject’s ethnicity on performance we compare the control to the Roma

salient treatment. To examine if subjects reminded of role models have better outcomes

we compare the control to the role model treatments. We compare the role model

treatments to the ethnicity salient treatment to identify if reminding children of role

models with the same (different) ethnicity reduces negative self perceptions.

Table 3 displays the average number of numeric puzzles solved by treatment. Column

4 reports the differences in means and the associated level of statistical significance. In

the Roma identity revealed treatment Roma children solve 8.0% less numeric puzzles

than when identity is not primed (control). This result is consistent with negative

stereotype threat identified in the literature and suggests that Roma children hold a

negative stereotype about their ability in this math task. A number of other key results

are evident: Children in the Roma role model treatment solve 8.7% more puzzles than in

the ethnicity salient treatment. Making Roma role models salient reduces the negative

self-perception of Roma children that appears when subjects are reminded of their

12

ethnicity. However, this role model affect only appears when the role model is part

of the Roma community and not when a role model is non-Roma. We also find little

statistical difference between priming Roma role models and the control. As a robustness

test we compare the number of numeric puzzles solved by gender. On average males

solve 26 puzzles compared to 27.26 for females (p-value=0.17). This suggests that the

marginally larger proportion of males in the Roma role model treatment is unlikely to

be driving this result.15

Figure 2 presents the distribution of the number of math puzzles solved by treatment.

Using a two-sided Kolmogorov-Smirnov test, the null hypothesis of equality of distribu-

tions between the Roma salient and the control is rejected (p–value = 0.014). The mass

of the distribution of the math puzzles solved by those in the Roma salient treatment

lies to the left of that when identity is not revealed. We also find that the distribution

of the Roma role model treatment lies to the right of the Roma salient treatment (K-S

test, p=0.092). While the distribution of maths puzzles solved by subjects in the Roma

role model treatment is similar to the control treatment (K-S test, p=0.293). Finally,

we find that the number of math puzzles solved by those in the non-Roma role model

treatment lies to the left of the Roma role model treatment (p-value-0.022). In sum-

mary, the negative stereotypes related to Roma’s identity can have a significant negative

effect on achievement, however, in comparison reminding children of role models associ-

ated with their Roma identity but not non-Roma role models can significantly improve

achievement.

15 These results use the full sample, despite a small proportion of subjects not considering the Roma and non-Roma icons as actual role models. Excluding these subjects may bias the result. When we re-estimate our main results in Table 3 excluding those who do not believe the role models are actually role models we find the number of maths puzzles solved by children in the Roma role model treatment increases slightly to 28.27 while the non Roma role model group decreases slightly to 25.09. The statistical significance of the results are quantitatively unchanged.

13

4.1 Channels

The decline in academic performance of Roma children when reminded of their ethnic

identity raises an important question - What are the channels through which negative

stereotypes affect performance and can role models influence these channels? One hy-

pothesis is that priming a subject’s identity may have a direct influence on confidence.

This could result in a decrease in performance if participants are primed positively by

providing them with an example of someone similar to them who has been successful.

Table 4 (columns 1 and 5) presents our first measure of confidence - participants’ ex-

ante guess of the number of numeric puzzles they expect to solve in one minute. The

results demonstrate that subjects’ ex-ante expectation of performance is lowest when

they are reminded of their ethnicity. We find that children reminded of their Roma

ethnicity expect to perform worse in the task compared to when their ethnicity is not

made salient. In turn subjects in the Roma role model treatment expect to perform

better than those in the Roma salient treatment. We also find that the expectation of

performance associated with the activation of a Roma role model is similar to that in

the non-Roma role model and the control treatments. 16

In columns 2 and 6 (Table 4) we examine students expectation of relative performance.

We create a dummy variable equal to 1 if a subject rated their performance as at least

better than average compared to other people within their settlement. This is our

second measure of confidence. Subjects were asked to compare their performance with

everyone else in their village. We find that children reminded of Roma role models are

more likely to believe they will perform better than others in their village compared

to children in the Roma salient treatment. There is little difference across the other

treatments.

16Participants were also asked to guess how many numeric puzzles they expected to solve in 3 minutes. Results based on this response are quantitatively unchanged.

14

5 Conclusion

Social identity is an important and growing field of research in economics, in part

because identity has a significant influence on behaviour. Its effect on behaviour can

be welfare reducing when social groups are negatively stereotyped such as the Roma

group studied here. Roma are one of the most socially excluded groups in Europe, they

face significant rates of poverty and material deprivation, low levels of education and

high levels of long-term unemployment (FRA, 2014a,b). In order to overcome Roma

deprivation and spark Roma inclusion, it is estimated that the EU and local governments

spent 3 billion EUR on activities targeting Roma during the 2000-2006 programming

period and a further 4.7 billion EUR between 2007-2013 on inclusion and anti-poverty

governmental programs and initiatives.17 Role models may influence social identity

and improve confidence and inspire young people to achieve goals, as such role models

may reduce discrimination and provide a cost effective tool to aid inclusion. This paper

breaks new ground by experimentally examining whether different role models affect self

perception and as such whether role models reduce the effects of a negative stereotype,

a small step in breaking the cycle of others negative perception towards Roma.

The results of this study indicate that reminding Roma children of their ethnicity reduces

performance. This result is consistent with the theory of stereotype threat. Reminding

children of their identity can be detrimental to their academic performance, this may

be particularly true if a negative stereotype associated with their social group exists.

We find that reminding children of role models from their own social group can improve

achievement compared to reminding children of their ethnicity. The desirability of a

policy that emphasizes Roma role models is complicated by our finding that reminding

children of Roma role models has a similar effect on performance as when nothing is

highlighted, as in the control treatment. This suggests that not emphasising a child’s

ethnicity may be the important policy option. Although when discussing a child’s

ethnicity it is far from determental to remind children of others from their social group

17Estimates based on European Commission (2010) report.

15

that have been successful.

Despite this it is important to note that Roma are often in contact and exposed to

non-Roma (who are the majority), for example the vast majority of teachers are non-

Roma. According to theories of context dependence (Tversky and Kahneman, 1983),

the settings in which decisions are made influence behaviour. In this context the envi-

ronment Roma are most often exposed in schools and throughout societal interactions

may be more reflective of the Roma salient treatment rather than the control-where

Roma are not reminded of their ethnicity. This would imply the effect of Roma role

models on academic performance maybe underestimated when comparisons are made

to the control.

We also find that reminding Roma children of non-Roma role models - a group that

is not associated with their social group may actually decrease performance in the

maths task. It may be possible that role models that show little resemblance to oneself

may increase their belief that these positions are unattainable, which may lead to self-

deflation decreasing their performance. To understand this mechanism would require

further research.

Finally, we investigate confidence as a possible channel through which stereotypes and

role models affect behaviour. We find that confidence decreases when subjects are re-

minded of their ethnicity relative to the Roma role model treatment and the control.

This result suggests that the effects of negative stereotypes and role models on achieve-

ment at least partially operates through changed confidence. Because confidence plays

an important role in behaviour, the effect of negative stereotypes can indirectly impact

the outcomes of children in other areas, such as aspirations and employment decisions,

which may continue to effect their decisions later in life.

16

References

Akerlof, G. A. and R. E. Kranton (2000). Economics and identity. Quarterly Journal of Economics 115(3), 715–753.

Akerlof, George, K. R. (2002). Identity and schooling: some lessons for the economics of education. Journal of Economic Literature 40(4), 1167–1201.

Altonji, J. and R. Blank (1999). Race and gender in the labor market. In O. C. Ashenfelter and D. Card (Eds.), Handbook of Labor Economics. Elsevier.

Bartoš, V., M. Bauer, J. Chytilová, and F. Matějka (2016). Attention discrimination: Theory and field experiments with monitoring information acquisition. American Economic Review 106(6), 1437–1475.

Bénabou, R. and J. Tirole (2002). Self confidence and personal motivation. Quarterly Journal of Economics 117(3), 871 – 915.

Benjamin, D., J. Choi, and A. Strickland (2010). Social identity and preferences. American Economic Review 100(4), 1913–1928.

Bertrand, M. and S. Mullainathan (2004). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review 94(4), 991–1013.

Bettinger, E. P. and B. T. Long (2005). Do faculty serve as role models? The impact of instructor gender on female students. American Economic Review 95(2), 152–157.

Bielikova, M. (2010). Expressions of intolerance, violence and extremism by young people aged 13 to 18. Bratislava: Iuventa.

Bordalo, P., K. Coffman, N. Gennaioli, and A. Shleifer (2016). Stereotypes. Quarterly Journal of Economics, forthcoming.

Cadsby, C. B., M. Servátka, and F. Song (2013). How competitive are female profession- als? A tale of identity conflict. Journal of Economic Behavior and Organization 92, 284 – 303.

Camerer, C. and L. Dan (1999). Overconfidence and Excess Entry: An Experimental Approach. American Economic Review 89(1), 306–318.

Chen, Y. and S. X. Li (2009). Group identity and social preferences. American Economic Review 99(1), 33 – 50.

Cohn, A., M. A. Maréchal, and T. Noll (2015). Bad boys: How criminal identity salience affects rule violation. The Review of Economic Studies 82(4), 1289–1308.

Dasgupta, U., L. Gangadharan, P. Maitra, S. Mani, and S. Subramanian (2015). Choos- ing to be trained: Do behavioral traits matter? Journal of Economic Behavior & Organization 110, 145 – 159.

Dee, T. (2004). Teachers, race and student achievement in a randomized experiment. The Review of Economics and Statistics 86(1), 195—210.

17

Dee, T. (2005). A teacher like me: Does race, ethnicity, or gender matter? American Economic Review 95(2), 158–165.

Dee, T. (2007). Teachers and the gender gaps in student achievement. Journal of Human Resources 42(3), 528–554.

DellaVigna, S. (2010). The Obama Effect on Economic Outcomes: Evidence From Event Studies. Technical report, University of California, Berkeley.

Ehrenberg, R. G., D. D. Goldhaber, and D. J. Brewer (1995). Do teachers’ race, gender and ethnicity matter? Evidence from the nels 1988. Industrial and Labor Relations Review 48(3), 547–561.

Erikson, E. (1968). Identity: Youth and crisis. New York: Norton.

European Commission (2010). The European Social Fund and Roma. Brussels, Bel- gium: European Commission.

European Commission (2014). Report on the implementation of the EU Framework for National Roma Integration Strategies: Communication from the Commission to the European Parliament, the Council, the European Economic and Social Commitee and the Committee of the Regions. Brussels, Belgium: European Commission - Directorate-General for Justice.

Fairlie, R. W., F. Hoffmann, and P. Oreopoulos (2014). A community college instruc- tor like me: Race and ethnicity interactions in the classroom. American Economic Review 104(8), 2567–91.

Fang, H. and G. Moscarini (2005). Morale hazard. Journal of Monetary Economics 52(4), 749 – 777.

Filippin, A. and M. Paccagnella (2012). Family background, self-confidence and eco- nomic outcomes. Economics of Education Review 31(5), 824 – 834.

FRA (2009). European Union Minorities and Discrimination Survey (EU-MIDIS): Main Results Report. Vienna, Austria: European Union Agency for Fundamental Rights.

FRA (2014a). Education: the situation of Roma in 11 EU Member States (Roma Survey - Data in Focus). Vienna, Austria: European Union Agency for Fundamental Rights.

FRA (2014b). Poverty and employment: The situation of Roma in 11 EU Member States (Roma Survey - Data in Focus). Vienna, Austria: European Union Agency for Fundamental Rights.

Guryan, J. and K. Charles (2013). Taste-based or statistical discrimination: The eco- nomics of discrimination returns to its roots. Economic Journal 123(572), 433–468.

Hoff, K. and P. Pandey (2006). Discrimination, social identity, and durable inequalities. American Economic Review 96(2), 206–211.

Hoff, K. and P. Pandey (2014). Making up people: The effect of identity on performance in a modernizing society. Journal of Development Economics 106(1), 118 – 131.

18

Hurd, N. M., M. A. Zimmerman, and T. M. Reischl (2011). Role model behavior and youth violence: A study of positive and negative effects. The Journal of Early Adolescence 31(2), 323–354.

Imbens, G. W. and J. M. Wooldridge (2009). Recent developments in the econometrics of program evaluation. Technical report, NBER Working Paper 14251.

Koszegi, B. (2006). Egoutility, overconfidence, and task choice. Journal of the European Economic Association 4(4), 673 – 707.

List, J. A. and I. Rasul (2011). Field experiments in labor economics. In A. Orley and C. David (Eds.), Handbook of Labor Economics, pp. 103–228. North Holland.

Loury, G. (2002). The anatomy of racial inequality. Harvard University Press.

Martin, C. A. and A. J. Bush (2000). Do Role Models Influence Teenagers’ Purchase Intensions and Behavior? The Journal of Consumer Marketing 17(5), 441–454.

Marx, D. M., S. J. Ko, and R. A. Friedman (2009). The “Obama Effect”: How a salient role model reduces race-based performance differences. Journal of Experimental Social Psychology 45(4), 953–956.

Morita, H. and M. Servátka (2013). Group identity and relation-specific investment: An experimental investigation. European Economic Review 58, 95 – 109.

Musinka, A., D. Skobla, J. Hurrle, K. Matlovicova, and J. Kling (2014). Atlas of Roma Communities in Slovakia 2013. Bratislava, Slovakia: UNDP Europe and the CIS, Bratislava Regional Centre.

Niederle, M. and L. Vesterlund (2007). Do women shy away from competition? Do men compete too much? The Quarterly Journal of Economics 122(3), 1067–1101.

Olivetti, Claudia, P. E. and Y. Zenou (2013). Mothers, friends and gender identity. Technical report, NBER Working Paper, 19610.

Pleiss, M. K. and J. F. Feldhusen (1995). Mentors, role models, and heroes in the lives of gifted children. Educational Psychologist 30(3), 159–169.

PRC (2014). A Fragile Rebound for EU Image on Eve of European Parliament Elections. Washington, DC, USA: Pew Research Center.

Scales, P. C. and J. L. Gibbons (1996). Extended family members and unrelated adults in the lives of young adolescents: A research agenda. The Journal of Early Adolescence 16(4), 365–389.

Shih, M., P. Todd, and A. Nalini (1995). Stereotype susceptibility: Identity salience and shifts in quantitative performance. Psychological Science 10(1), 80–83.

Slovikova, M. (2012). Tolerant Behaviour Model of Youth in Family and School. Bratislava, Slovakia: Institute of Information and Prognoses of Education.

Steele, C. and J. Aronson (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology 69(5), 797–811.

19

Steele, C. M. (1997). A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist 52(6), 613–629.

Stone, J., C. I. Lynch, M. Sjomeling, and J. M. Darley (1999). Stereotype Threat Effects on Black and White Athletic Performance. Journal of Personality and Social Psychology 77(6), 1213–1227.

Tajfel, H. (1973). Human groups and social categories. Cambridge University Press.

Tversky, A. and D. Kahneman (1983). Extensional versus intuitive reasoning: The conjunction fallacy in probabilistic reasoning. Psychological Review 90(4), 293–315.

UNDP (2012). Report on the Living Conditions of Roma households in Slovakia 2010. Bratislava, Slovakia: UNDP Europe and the CIS, Bratislava Regional Centre.

Yan Chen, Sherry X. Li, T. X. L. and M. Shih (2014). Which Hat to Wear? Im- pact of Natural Identities on Coordination and Cooperation. Games and Economic Behavior 84, 58–86.

20

Figure 1: Experimental Districts

Note: The thematic map shows the distribution of the proportion of people with attributed Roma ethnicity (in %). The ‘x’ symbols indicate geographic locations of the experimental villages. Source: (Musinka et al., 2014)

21

Table 1: Randomization at the settlement/village level

Characteristic Sample Eastern Slovakia Difference Norm. Diff (1) (2) (3) (4)

Access to water main (%) 52.0 52.5 −0.5 −0.009 (7.2) (1.9) (7.4)

Access to other source of water (%) 25.0 21.3 3.7 0.076 (6.2) (1.6) (6.4)

Without access to water (%) 13.9 10.9 3.8∗ 0.083 (4.5) (1.2) (4.7)

Public sewerage system usage (%) 30.5 20.5 10.0 0.191 (7.0) (1.6) (7.1)

Drain wells usage (%) 24.0 24.3 −0.3 −0.007 (4.9) (1.6) (5.1)

No access to sewerage system (%) 44.9 54.3 −9.4 −0.173 (6.3) (1.9) (6.6)

Electricity Usage (%) 91.7 92.5 −0.8 −0.029 (3.6) (0.9) (3.7)

Gas Usage (%) 20.6 13.6 7.0 0.162 (6.1) (1.3) (6.2)

Public street lights availability (%) 91.4 89.7 1.7 0.050 (3.7) (1.2) (3.9)

Heating by gas (%) 5.9 5.8 0.1 0.005 (2.2) (0.7) (2.3)

Heating by wood coal (%) 97.3 93.1 4.2 0.217 (1.4) (0.8) (1.6)

Kindergardens in the village (No.) 1.0 0.9 0.1∗ 0.333 (0.0) (0.0) (0.0)

Dist. to 1-4 grades primary school (km) 4.0 3.3 0.7 0.170 (0.3) (0.2) (0.3)

Dist. to 1-9 grades primary school (km) 5.0 5.1 −0.1 −0.025 (0.3) (0.2) (0.3)

Special primary schools in the village (No.) 0.3 0.2 0.1 0.156 (0.1) (0.0) (0.1)

Dist. to special primary school (km) 7.9 9.3 −1.4 −0.125 (1.5) (0.3) (1.5)

Dist. to the train stop (km) 9.2 10.1 −0.9 −0.054 (2.3) (0.5) (2.4)

Dist. to the general practitioner (km) 0.8 3.9 −3.1∗∗∗ −0.665 (0.3) (0.2) (0.4)

Dist. to the pediatrician (km) 1.2 5.2 −4.0∗∗∗ −0.662 (0.5) (0.2) (0.5)

Dist. to th gynecologist (km) 11.4 9.9 1.5 0.098 (2.4) (0.3) (2.4)

Notes: This table shows the ex ante balance in the characteristics of villages chosen for experiments. The upper part

of the table reports characteristics at settlement level, the lower part at the village level. Approximative (permutation)

Wilcoxon-Mann-Whitney Test significance level indications: ∗∗∗p < 0.01,∗∗ p < 0.05,∗ p < 0.10.

22

Figure 2: Distribution of numeric puzzles solved by treatments

Note: The figure presents comparison of numeric puzzles solved by control and all treatments (kernel density estimations).

23

T a b le

2 :

D em

o g ra

p h ic

ch a ra

ct er

is ti

cs o f

su b

je ct

s

C h

a ra

ct er

is ti

cs S

a m

p le

C T

1 T

2 T

3 C

- T

1 C

- T

2 C

- T

3 T

1 -

T 2

T 1

- T

3 T

2 -

T 3

(1 )

(2 )

(3 )

(4 )

(5 )

(6 )

(7 )

(8 )

(9 )

(1 0 )

(1 1 )

C h il d re n ’s

c h a ra c te ri st ic s

A g e

1 2 .9

9 5

1 3 .2

4 7

1 2 .7

5 5

1 3 .2

6 3

1 2 .7

6 0

0 .4

9 3

− 0 .0

1 6

0 .4

8 7

− 0 .5

0 8 ∗

− 0 .0

0 5

0 .5

0 3

(0 .1

1 9 )

(0 .2

4 9 )

(0 .2

1 0 )

(0 .2

5 2 )

(0 .2

4 1 )

(0 .3

2 5 )

(0 .3

5 4 )

(0 .3

4 7 )

(0 .3

2 8 )

(0 .3

2 0 )

(0 .3

4 9 )

[0 .1

5 3 ]

[− 0 .0

0 5 ]

[0 .1

4 3 ]

[− 0 .1

5 6 ]

[− 0 .0

0 2 ]

[0 .1

4 6 ]

G en

d er

= m

a le

0 .4

9 5

0 .4

1 9

0 .4

5 7

0 .5

9 6

0 .5

1 0

− 0 .0

3 8

− 0 .1

7 6 ∗∗

− 0 .0

9 1

− 0 .1

3 9 ∗

− 0 .0

5 3

0 .0

8 6

(0 .0

2 5 )

(0 .0

5 1 )

(0 .0

4 9 )

(0 .0

5 1 )

(0 .0

5 0 )

(0 .0

7 1 )

(0 .0

7 2 )

(0 .0

7 2 )

(0 .0

7 0 )

(0 .0

7 0 )

(0 .0

7 1 )

[− 0 .0

5 4 ]

[− 0 .2

5 3 ]

[− 0 .1

2 9 ]

[− 0 .1

9 8 ]

[− 0 .0

7 5 ]

[0 .1

2 2 ]

P ri

m a ry

sc h o o l

0 .8

4 8

0 .8

4 9

0 .8

7 7

0 .8

4 2

0 .8

2 0

− 0 .0

2 8

0 .0

0 7

0 .0

2 9

0 .0

3 5

0 .0

5 7

0 .0

2 2

(0 .0

1 8 )

(0 .0

3 7 )

(0 .0

3 2 )

(0 .0

3 7 )

(0 .0

3 8 )

(0 .0

4 9 )

(0 .0

5 3 )

(0 .0

5 3 )

(0 .0

4 9 )

(0 .0

5 0 )

(0 .0

5 4 )

[− 0 .0

5 7 ]

[0 .0

1 4 ]

[0 .0

5 6 ]

[0 .0

7 2 ]

[0 .1

1 4 ]

[0 .0

4 2 ]

P a re n ts ’ c h a ra c te ri st ic s

M a rr

ie d

0 .7

1 0

0 .6

9 8

0 .7

0 0

0 .7

1 7

0 .7

2 4

− 0 .0

0 2

− 0 .0

1 9

− 0 .0

2 6

− 0 .0

1 7

− 0 .0

2 4

− 0 .0

0 7

(0 .0

2 2 )

(0 .0

4 5 )

(0 .0

4 4 )

(0 .0

4 7 )

(0 .0

4 1 )

(0 .0

6 2 )

(0 .0

6 5 )

(0 .0

6 1 )

(0 .0

6 4 )

(0 .0

6 0 )

(0 .0

6 3 )

[− 0 .0

0 3 ]

[− 0 .0

3 0 ]

[− 0 .0

4 1 ]

[− 0 .0

2 7 ]

[− 0 .0

3 8 ]

[− 0 .0

1 1 ]

P ri

m a ry

ed u

ca ti

o n

0 .6

2 7

0 .5

7 5

0 .6

8 2

0 .6

6 3

0 .5

9 5

− 0 .1

0 6

− 0 .0

8 8

− 0 .0

1 9

0 .0

1 9

0 .0

8 7

0 .0

6 8

(0 .0

2 3 )

(0 .0

4 8 )

(0 .0

4 4 )

(0 .0

4 9 )

(0 .0

4 6 )

(0 .0

6 5 )

(0 .0

6 9 )

(0 .0

6 6 )

(0 .0

6 6 )

(0 .0

6 4 )

(0 .0

6 7 )

[− 0 .1

5 7 ]

[− 0 .1

2 8 ]

[− 0 .0

2 8 ]

[0 .0

2 8 ]

[0 .1

2 9 ]

[0 .1

0 0 ]

U n

em p

l+ a ct

w o rk

0 .4

3 6

0 .3

6 8

0 .5

0 0

0 .5

0 0

0 .3

8 8

− 0 .1

3 2 ∗

− 0 .1

3 2 ∗

− 0 .0

2 0

0 .0

0 0

0 .1

1 2

0 .1

1 2

(0 .0

2 4 )

(0 .0

4 7 )

(0 .0

4 8 )

(0 .0

5 2 )

(0 .0

4 5 )

(0 .0

6 7 )

(0 .0

7 0 )

(0 .0

6 5 )

(0 .0

7 1 )

(0 .0

6 6 )

(0 .0

6 9 )

[− 0 .1

9 0 ]

[− 0 .1

9 0 ]

[− 0 .0

2 9 ]

[0 .0

0 0 ]

[0 .1

6 1 ]

[0 .1

6 1 ]

E m

p lo

y ed

0 .0

9 0

0 .1

0 4

0 .1

0 0

0 .0

7 6

0 .0

7 8

0 .0

0 4

0 .0

2 8

0 .0

2 6

0 .0

2 4

0 .0

2 2

− 0 .0

0 1

(0 .0

1 4 )

(0 .0

3 0 )

(0 .0

2 9 )

(0 .0

2 8 )

(0 .0

2 5 )

(0 .0

4 1 )

(0 .0

4 1 )

(0 .0

3 9 )

(0 .0

4 0 )

(0 .0

3 8 )

(0 .0

3 7 )

[0 .0

0 9 ]

[0 .0

6 9 ]

[0 .0

6 5 ]

[0 .0

6 0 ]

[0 .0

5 6 ]

[− 0 .0

0 4 ]

M ea

n H

H in

co m

e 3 8 3 .4

3 3

3 7 6 .4

4 2

3 7 2 .0

4 6

3 5 9 .0

6 6

4 2 0 .5

7 5

4 .3

9 7

1 7 .3

7 6

− 4 4 .1

3 3

1 2 .9

8 0

− 4 8 .5

3 0

− 6 1 .5

0 9 ∗

(9 .7

1 3 )

(1 9 .6

0 5 )

(1 6 .4

0 0 )

(1 9 .4

7 3 )

(2 1 .3

5 0 )

(2 5 .5

6 0 )

(2 7 .6

3 3 )

(2 8 .9

8 6 )

(2 5 .4

5 9 )

(2 6 .9

2 2 )

(2 8 .8

9 7 )

[0 .0

1 7 ]

[0 .0

6 4 ]

[− 0 .1

4 6 ]

[0 .0

5 1 ]

[− 0 .1

7 0 ]

[− 0 .2

1 0 ]

N o te

s: T

h is

ta b

le sh

o w

s th

e e x p o st

b a la

n c e

in th

e d

e m

o g ra

p h

ic ch

a ra

c te

ri st

ic s

o f

su b

je c ts

a c ro

ss tr

e a tm

e n ts

. “ C

” re

fe rs

to th

e c o n tr

o l,

“ T

1 ”

th e

R o m

a sa

li e n t

tr e a tm

e n t,

“ T

2 ”

R o m

a ro

le m

o d

e l

tr e a tm

e n t

a n

d “ T

3 ”

n o n

-R o m

a ro

le m

o d

e l tr

e a tm

e n t.

M e a n

s (A

g e

a n

d H

o u

se h

o ld

in c o m

e )

a n

d p

ro p

o rt

io n

s (a

ll o th

e r

ch a ra

c te

ri st

ic s)

re p

o rt

e d

in c o lu

m n

s (1

) -

(5 ),

d iff

e re

n c e s

in m

e a n

s/ p

ro p

o rt

io n

s in

c lu

d in

g M

-W te

st in

c o lu

m n

s (6

) -

(1 1 ).

S E

e st

im a te

s in

p a re

n th

e se

s, n

o rm

a li z e d

d iff

e re

n c e s

in sq

u a re

b ra

ck e ts

. N

o rm

a li z e d

d iff

e re

n c e s

a re

c a lc

u la

te d

u si

n g

th e

fo rm

u la

a s

in Im

b e n

s a n

d W

o o ld

ri d

g e

(2 0 0 9 ).

A ru

le o f

th u

m b

is th

a t

w h

e n

n o rm

a li z e d

d iff

e re

n c e

e x c e e d

s 0 .2

5 in

a b

so lu

te v a lu

e ,

li n

e a r

re g re

ss io

n m

e th

o d

s te

n d

to b

e se

n si

ti v e

to th

e sp

e c ifi

c a ti

o n

. S

ig n

ifi c a n

c e

le v e l

in d

ic a ti

o n

s: ∗ ∗ ∗ p <

0 .0

1 ,∗

∗ p <

0 .0

5 ,∗

p <

0 .1

0 .

24

Table 3: Number of Numeric Puzzles Solved by Treatment

No. of Numeric Obs. Difference Treatment Puzzles Solved

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

C – T1 2.220∗

(1.270) C Identity not Revealed 27.76 93 C – T2 0.237

(1.277) T1 Roma identity Revealed 25.54 106 C – T3 2.403∗

(1.431) T2 Roma Role Model 28.00 96 T1 – T2 2.453∗∗

(1.160) T3 Non Roma Role Model 25.36 100 T1 – T3 0.565

(1.309) T2 – T3 2.640∗∗

(1.322)

Notes: This table shows absolute performance of subjects across treatments (1), and the differences among treatments (4). SE estimates in parentheses. T-test significance level indications: ∗∗∗p < 0.01,∗∗ p < 0.05,∗ p < 0.10.

25

T a b le

4 :

C o n fi d en

ce b y

T re

a tm

en t

A b

so lu

te C

o n

fi d

en ce

R el

a ti

v e

C o n

fi d

en ce

O b

s. D

iff er

en ce

A b

s. C

o n

fi d

en ce

R el

. C

o n

fi d

en ce

(1 )

(2 )

(3 )

(4 )

(5 )

(6 )

C –

T 1

6 .0

6 1 ∗∗ ∗

0 .1

1 4

(1 .7

4 2 )

(0 .0

7 2 )

C Id

en ti

ty n

o t

R ev

ea le

d 1 5 .2

7 5

0 .5

6 9 1

C –

T 2

1 .7

4 8

0 .0

8 1

(2 .1

7 9 )

(0 .0

7 3 )

1 R

o m

a id

en ti

ty R

ev ea

le d

9 .2

1 4

0 .4

5 1 0 3

C –

T 3

1 .5

4 5

0 .0

1 0

(1 .9

2 8 )

(0 .0

7 2 )

2 R

o m

a R

o le

M o d

el 1 3 .5

2 7

0 .6

4 9 2

T 1

– T

2 4 .3

1 3 ∗∗

0 .1

9 5 ∗∗ ∗

(1 .6

6 7 )

(0 .0

7 1 )

3 N

o n

R o m

a R

o le

M o d

el 1 3 .7

3 0

0 .5

5 1 0 0

T 1

– T

3 4 .5

1 6 ∗∗ ∗

0 .1

0 3

(1 .4

1 3 )

(0 .0

7 0 )

T 2

– T

3 0 .2

0 3

0 .0

9 1

(1 .8

6 3 )

(0 .0

7 1 )

N o te

s: T

h is

ta b

le sh

o w

s th

e e x a n te

e x p

e c ta

ti o n

o f

p e rf

o rm

a n

c e

in th

e ta

sk .

S E

e st

im a te

s in

p a re

n th

e se

s. T

-t e st

si g n

ifi c a n

c e

le v e l in

d ic

a ti

o n

s: ∗ ∗ ∗ p <

0 .0

1 ,∗

∗ p <

0 .0

5 ,∗

p <

0 .1

0 .

26

Table 5: Selected Census 2011 Village Characteristics

Characteristics Sample Villages Eastern Slovakia Slovakia (1) (2) (3)

Average age (years) 33.5 37.1 38.9 Single (%) 50.0 44.4 42.3 Married (%) 36.3 40.2 41.0

Slovak nationality (%) 78.4 85.9 86.8 Roma nationality∗ (%) 12.4 5.5 2.1 Slovak mother tongue (%) 70.5 81.6 84.9 Roma mother tongue (%) 17.8 7.2 2.5

Roman Catholics (%) 76.8 64.4 69.4 People with no religion (%) 8.8 9.5 15.0

Primary education level (%) 20.6 16.9 15.4 Secondary education level (%) 43.5 51.3 54.2 Tertiary education level (%) 11.2 12.8 14.3 No education (%) 24.7 19.0 16.1

Emplyed / self-employed (%) 68.2 75.9 82.1 Unemployed (%) 30.6 23.0 16.8 Retired (%) 13.8 18.0 19.7 Children + students (%) 30.9 26.3 23.3

*Nationality in context of ethnicity. Self-reported Roma ethnicity based on Census 2011 data (2.1 % at national level) is significantly lower than share of people with attributed Roma ethnicity (Musinka et al., 2014).

27

A.1 Appendix

A.1.1 Background Questionnaire

(C) Identity not revealed (control)

No.

QUESTION

RESPONSE

① FAVOURITE FOOD ————―――→

② PREFER WALKING UP OR DOWN HILLS?

UP

DOWN

③ FAVOURITE COLOUR ———―――→

④ DO YOU PREFER WINTER OR SUMMER?

WINTER

SUMMER

⑤ WHAT IS THE CAPITAL OF FRANCE?

LONDON

PARIS

DON’T KNOW

28

(T1) Roma Salient

No.

QUESTION

RESPONSE

① FAVOURITE FOOD ————―――→

② PREFER WALKING UP OR DOWN HILLS?

UP

DOWN

③ FAVOURITE COLOUR ———―――→

④ DO YOU PREFER WINTER OF SUMMER?

WINTER

SUMMER

⑤ WHAT IS THE CAPITAL OF FRANCE?

LONDON

PARIS

DON’T KNOW

⑥ YOUR NATIONALITY

ROMA

SLOVAK

HUNGARIAN

OTHER: .............................

⑦ LANGUAGE SPOKEN MOST FREQUENTLY AT HOME ——―――→

⑧ DO YOUR GRANDPARENTS SPEAK ANY OTHER LANGUAGE OTHER THAN ROMA?

YES

NO

29

(T2) Roma Role Model

No.

QUESTION

RESPONSE

① A POET VLADIMÍR OLÁH ESTABLISHED WHICH ASSOCIATION?

ROMA BEE MOTHER

SLOVAK BEE MOTHER

DON’T KNOW

②

AN ACTIVIST DR JÁN CIBUĽA WHO WAS NOMINATED FOR A NOBEL PEACE PRIZE AND WAS ALSO A PRESIDENT OF INTERNATIONAL ROMA UNION STUDIED AT HIGH SCHOOL IN:

RIMAVSKÁ SOBOTA

TRNAVA

DON’T KNOW

③ SILVIA ŠARKÖZI IS A MEMBERS OF GROUP

GIPSY KINGS

GYPSY DEVILS

DON’T KNOW

④ IGOR KMEŤO, MEMBER OF KMEŤOBAND SANG A SONG:

BUBA MARA

Ó MAŇO

DON’T KNOW

⑤ MAIDEN NAME OF THE SINGER VĚRA BÍLÁ WAS:

VĚRA GIŇOVÁ

VĚRA OLÁHOVÁ

NEVIEM

⑥ HOW MANY MEMBERS DOES THE BAND GIPSY KAJKOS HAVE?

FEWER THAN 4

MORE THAN 4

DON’T KNOW

30

(T3) Non-Roma Role Model

No.

QUESTION

RESPONSE

① IS JUSTIN BIEBER MORE THAN 18 YEARS OLD?

YES

NO

DON’T KNOW

② WHAT SOCCER CLUB DOES CRISTIANO RONALDO PLAY FOR?

REAL MADRID

FC BARCELONA

DON’T KNOW

③ WHAT SOCCER CLUB DOES MAREK HAMŠÍK PLAY FOR?

MANCHESTER UNITED

SSC NAPOLI

DON’T KNOW

④ IS SHAKIRA A PART OF A GROUP OR IS SHE A SOLO SINGER?

PART OF A GROUP

SOLO SINGER

DON’T KNOW

⑤ THE SINGER HELENA VONDRÁČKOVÁ PERFORMED A HIT

VYZNANIE (Declaration)

DLOUHÁ NOC (Long Night)

DON’T KNOW

⑥

HAS THE SINGER KAREL GOTT BEEN AWARDED „THE GOLDEN NIGHTINGALE“ PRIZE MORE THAN 20 TIMES?

YES

NO

DON’T KNOW

31

View publication statsView publication stats