patterns of prejudice the forest behind the bar charts ... · the forest behind the bar charts:...

32
PLEASE SCROLL DOWN FOR ARTICLE This article was downloaded by: [Rughinis, Cosima] On: 28 September 2010 Access details: Access Details: [subscription number 926982887] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37- 41 Mortimer Street, London W1T 3JH, UK Patterns of Prejudice Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713395163 The forest behind the bar charts: bridging quantitative and qualitative research on Roma/Ţigani in contemporary Romania Cosima Rughiniş Online publication date: 15 September 2010 To cite this Article Rughiniş, Cosima(2010) 'The forest behind the bar charts: bridging quantitative and qualitative research on Roma/Ţigani in contemporary Romania', Patterns of Prejudice, 44: 4, 337 — 367 To link to this Article: DOI: 10.1080/0031322X.2010.510716 URL: http://dx.doi.org/10.1080/0031322X.2010.510716 Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Upload: others

Post on 25-Jan-2020

10 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [Rughinis, Cosima]On: 28 September 2010Access details: Access Details: [subscription number 926982887]Publisher RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Patterns of PrejudicePublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t713395163

The forest behind the bar charts: bridging quantitative and qualitativeresearch on Roma/Ţigani in contemporary RomaniaCosima Rughiniş

Online publication date: 15 September 2010

To cite this Article Rughiniş, Cosima(2010) 'The forest behind the bar charts: bridging quantitative and qualitativeresearch on Roma/Ţigani in contemporary Romania', Patterns of Prejudice, 44: 4, 337 — 367To link to this Article: DOI: 10.1080/0031322X.2010.510716URL: http://dx.doi.org/10.1080/0031322X.2010.510716

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

Page 2: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

The forest behind the bar charts: bridging

quantitative and qualitative research on

Roma/T,igani in contemporary Romania

COSIMA RUGHINIS,

ABSTRACT Rughinis,discusses three controversial issues with regard to surveys of

the Romani population: ethnonym use, self-identification versus hetero-attribution

of Romani ethnicity, and the use of variables in reference to Romani settlements. She

uses data sets from ten surveys of Romanian Roma between 2000 and 2008 as well as

the 2002 Romanian Census to compare two types of samples, and to explore the

consequences of several research choices for the quality of the data. In addition to

specific methodological issues, Rughinis,addresses the relevance to such surveys of

qualitative research in Romani communities.

KEYWORDS ethnicity, ethnonyms, prejudice, sampling, surveys, Roma, Romania,stereotypes, T

,igani

The Roma, known in Romania by the ethnonyms ‘Romi’ or ‘T,igani’, are

one of the largest minorities in Europe. Romania probably has the largestRomani population in absolute numbers, while significant minorities arealso found in Bulgaria, Hungary, Serbia and Montenegro, the SlovakRepublic, Turkey, FYR Macedonia, Spain, France, Italy and Germany.1

Within Romania, the Romani minority comprises 2.5 per cent of thepopulation according to the 2002 Census. Its documented history goes backto the fourteenth century, when Roma were recorded as slaves. That slaverywas abolished in 1856, but lack of land ownership meant that these newlyfreed men and women remained in a socially marginal position. During theSecond World War many Roma were forcibly deported to Transnistria where

I am grateful to Vlad Achimescu, Fabrice Bardet, Sorin Cace, Vlad Grigoras,, Dumitru

Sandu, Alina Silian, Irina Tomescu-Dubrow, Paula Tufis,, Bogdan Voicu and Malina Voicu

for helping me with commentaries and data. I would also like to thank the twoanonymous reviewers who offered valuable observations and advice. My work on thisarticle has been supported by a Junior Visiting Research Fellowship at the CentralEuropean University (CEU) in Budapest in January�/February 2007. The opinionsexpressed are my own and do not necessarily express the views of the CEU.1 Dena Ringold, Mitchell A. Orenstein and Erika Wilkens, Roma in an Expanding Europe:

Breaking the Poverty Cycle (Washington, D. C.: The World Bank 2005), available on theWorld Bank website at http://siteresources.worldbank.org/EXTROMA/Resources/roma_in_expanding_europe.pdf (viewed 26 May 2010).

Patterns of Prejudice, Vol. 44, No. 4, 2010

ISSN 0031-322X print/ISSN 1461-7331 online/10/040337-31 # 2010 Taylor & Francis

DOI: 10.1080/0031322X.2010.510716

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 3: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

thousands died.2 After 1945 the Communist regime forced many Roma tosettle in fixed residences, to enrol their children in schools and to work in

factories or agricultural co-operatives. The T,igan ethnic identity was denied:

even use of the ethnonym was forbidden. Such policies led to the creation ofa low-skilled Romani workforce that was quickly pushed out of the labourmarket after the fall of the centrally planned economic system in 1989. Theyears since 1990 have seen the appearance of severely impoverished andethnically homogeneous Romani communities, largely excluded from socialservices, but also the emergence of a small number of wealthy Romanifamilies and Romani intellectual and political elites. The first years ofdemocracy in Romania witnessed increased ethnic conflict, including violentclashes between communities. Since 1996 outbreaks of violence against the

Roma have gradually subsided. Indicators of social distance show increasedethnic tolerance or, at least, the awareness that such tolerance is sociallydesirable. The proportion of respondents that would not accept a Romanineighbour has declined from 72 per cent in 1993, to 60 per cent in 1997, 49per cent in 1999 and 37 per cent in 2006.3 Nonetheless, exclusion fromschools and the labour market means that many Roma still live at themargins of contemporary Romanian society.4

The Roma are a lively presence not only in the Romanian socialimaginary, haunted by the stigmatized T

,igani, but also in current socio-

logical research. Coincidentally, the Romanian Communist regime censoredboth the Roma and sociological research to the extent that both becameinvisible. In 1990 the Romanian Roma and Romanian sociology began theirpublic re-emergence and, not too long afterwards, in 1992, the firstRomanian survey of the Romani population was conducted by CatalinZamfir and Elena Zamfir.5 Eighteen years have since elapsed and, in thisarticle, I will explore several surveys that included Romanian Roma

samples or subsamples conducted between 2000 and 2008, and focus onmeasures of ethnicity and sampling strategies. I am particularly interestedin finding better ways to integrate quantitative data about the Roma withqualitative research conducted in Romani settlements.The article is divided into six sections. After presenting the data sources

and the main conceptual distinctions used in the analysis, I will discuss the

2 Achim estimates that around 25,000 Roma were deported to Transnistria and aroundhalf of them died there; Viorel Achim, T

,iganii ın istoria Romaniei (Bucharest: Editura

Enciclopedica 1998).3 Malina Voicu, ‘Tolerance and perceived discrimination’, in Gabriel Badescu, VladGrigoras

,, Cosima Rughinis

,, Malina Voicu and Ovidiu Voicu, Roma Inclusion Barometer

(Bucharest: Open Society Foundation 2007), 56, available on the Soros FoundationRomania website at www.osf.ro/en/publicatii.php?cat�16 (viewed 26 May 2010).

4 See, for example, Gabor Fleck and Cosima Rughinis,(eds), Come Closer: Inclusion and

Exclusion of Roma in Present-day Romanian Society (Bucharest: Human Dynamics 2008).5 Catalin Zamfir and Elena Zamfir (eds), T

,iganii: Intre ignorare s

,i ıngrijorare (Bucharest:

Alternative 1993).

338 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 4: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

ethnonyms used in referring to the Romanian Roma. Given widespreadawareness of the reluctance of Romani people to state their ethnicity, I willthen consider alternative strategies for measuring Romani ethnicity insurveys, at the individual and the community level. In the next section Iwill analyse current survey information about Romani settlements. Thearticle concludes with a presentation of the main findings and a discussionof their significance for the possibility of bringing together quantitative andqualitative research on Romani communities.

While my analysis is focused on Romanian society, covering surveys from2000 to 2008, the article is also relevant for surveys of Roma in otherEuropean countries with sizeable Romani minorities*/such as Hungary,Bulgaria, Slovakia, the Czech Republic and so on*/given that the mainchallenges to quantitative research on Roma are the same in all these places:respondents’ reluctance to self-identify as Roma, and an unknown distribu-tion across settlement types. Of course, these challenges may be more or lesssalient in different social contexts, as indicated by comparative internationalstudies of the Roma, such as that by Janos Ladanyi and Ivan Szelenyi,6 or theUnited Nations Development Program (UNDP) report.7 My aim is toillustrate the uncertainties facing researchers when surveying the Romanipopulation, and the subsequent consequences for data analysis, with a focuson contemporary Romanian Roma.

Concepts and data

The construction of a Roma sample is a matter of science and also amatter of choice, since there is no agreed-upon sampling frame. Oneimportant choice refers to the method used to identify Romani respon-dents. In the following discussion I will distinguish between three formsof ethnic identification. First, by ‘ethnic self-affiliation’, I mean anindividual’s private conception of his/her ethnicity: ethnic self-affiliationcannot be directly observed. Second, the term ‘self-identification’ refers tothe answer that respondents give to survey questions regarding theirethnic identity, questions that are asked within specific social settings: inan official context, such as the Census or by a public institution; in apublic but non-governmental context, such as a survey conducted by ascientific or commercial organization; or as part of other interactions, suchas conversations with friends, media interviews and so on. Third, I use

6 Janos Ladanyi and Ivan Szelenyi, ‘The social construction of Roma ethnicity inBulgaria, Romania and Hungary during market transition’, Review of Sociology, vol. 7,no. 2, 2001, 71�/89.

7 UNDP, Avoiding the Dependency Trap: The Roma in Central and Eastern Europe (Bratislava:UNDP 2002), available on the UNDP website at http://roma.undp.sk (viewed 26 May2010).

COSIMA RUGHINIS,

339

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 5: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

the term ‘ethnic hetero-attribution’ to refer to the ethnicity attributed to anindividual by an observer. These three forms of ethnic classification mayor may not coincide. For example, a respondent may think of herself asHungarian, while declaring a Romanian ethnic identity to a surveyinterviewer, who in turn thinks of her as being Romani.A second choice in sampling relates to the strategy used for finding

respondents. One can distinguish two main types of Roma samples. On theone hand, there are custom-built Roma samples, designed to be representativeof the Romani population, in which respondents are found by a targetedsearch of selected settlements. On the other, there are Roma subsamples thatare comprised of Romani respondents from a larger (usually nationallyrepresentative) sample, such as that used by the Family Budget Survey(Ancheta Bugetelor de Familie) of the National Institute of Statistics (INSSE,Institutul Nat

,ional de Statistica). Given the few self-identified Romani

respondents in national samples comprising around 1,000 respondents, avariation on the Roma subsample consists in the aggregation of the self-identified Romani respondents frommore than one nationally representativesample, such as the ‘ethnic oversample’ used by Ladanyi and Szelenyi.8

In the following sections I will discuss the consequences of these samplingchoices for survey data, and the theoretical and empirical assumptions onwhich they are based. Among the survey findings, as a case in point, I willfocus on those regarding educational attainment level because this variable is apredictor for many other social phenomena; furthermore, it has considerablepolicy relevance and is available in all surveys.

Data sources

I have relied on secondary analyses of post-2000 surveys of RomanianRomani respondents in order to achieve some degree of comparability amongthem. Therefore, I have excluded from most comparative analyses at leasttwo important surveys of Romanian Roma: the Research Institute for Qualityof Life (Institutul de Cercetare a Calitat

,ii Viet

,ii) surveys from 1992 (conducted

by Catalin Zamfir and Elena Zamfir) and 1998 (conducted by Catalin ZamfirandMarian Preda). The data sets used are publicly available from a variety ofsources,9 and their main features are summarized in Table 1.At this point it would be useful to point out the various difficulties that

arise when conducting a comparative analysis. In some cases, socio-

8 Ladanyi and Szelenyi, ‘The social construction of Roma ethnicity in Bulgaria, Romaniaand Hungary during market transition’.

9 I have worked in the research teams for the Public Opinion Barometers, the RomaInclusion Barometer 2006, the Inclusion 2007 survey and the Work Attitudes 2008survey. Therefore, I have more familiarity with these databases. For the others, thisarticle was an opportunity to use them in greater depth for the first time.

340 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 6: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Table 1 Post-2000 surveys of Romanian Roma (by acronym)

IPUMSI 2002 Integrated Public Use Microdata Series International, based at the

University of Minnesota.10 The IPUMSI database is a random sample of

10% of the Romanian national 2002 Census entries, thus differing in

several ways from survey databases, including the field sampling

methodology and the relationship between interviewer and respondents.

Romani individuals are identified in the database using the ‘ethnicity’

variable, and therefore analyses are conducted on a subsample of the

national sample.

PEGTS 2000 Poverty, Ethnicity and Gender in Transitional Societies 2000 survey, co-

ordinated by Janos Ladanyi and Ivan Szelenyi. The data set (available

online) used in analyses includes a national random sample and also a

subsample of hetero-identified Romani respondents.

ERBA 2001,

2002

Ethnic Relations Barometers, conducted by the Ethnocultural Diversity

Resource Center (EDRC, Centrul de Resurse pentru Diversitate Etnocul-

turala) in Cluj.11 I have included the national and Roma samples in analyses.

UNDP 2002 United Nations Development Program 2002 survey includes Roma samples

for the Czech Republic, Slovakia, Hungary, Romania and Bulgaria.12

POB

2001�/2004Public Opinion Barometers, carried out twice a year by the Soros

Foundation Romania (formerly the Open Society Foundation).13 Each

survey is based on a national sample. I have created a Roma subsample

by aggregating all the Romani respondents from the national samples.

HPS 2004 Health Policy Survey 2004, carried out by the Center for Health Policies

and Services (Centrul pentru Politici s,i Servicii de Sanatate) in

Bucharest.14 It includes a Roma sample.

FBS 2005 Family Budget Survey 2005, carried out by INSSE, on a national sample.15

RIB 2006 Roma Inclusion Barometer 2006, carried out by the Soros Foundation

Romania.16 It includes a national sample and a Roma sample.

Inclusion

2007

The Inclusion 2007 survey, carried out by Human Dynamics as part of the

EU’s PHARE 2004�/2006 programme ‘Strengthening Capacity and

Partnership Building to Improve Roma Condition and Perception’.17 It

includes a Roma sample and a comparative sample of the population

living in the same localities as the respondents in the Roma sample.

WA 2008 Work Attitudes 2008 survey, carried out by the Soros Foundation

Romania.18 It includes a Roma sample and a national sample.

10 The data set is available on the IPUMSI website at https://international.ipums.org/international

(viewed 9 June 2010).11 The data sets are available on the EDRC website at www.edrc.ro/resources.jsp (viewed 9 June 2010).12 For the data set, see UNDP, Avoiding the Dependency Trap.13 The data sets are available on the Soros Foundation Romania website at www.osf.ro (viewed 9

June 2010).14 The data set has been made available courtesy of the survey co-ordinator Sorin Cace.15 Although INSSE is a publicly funded institution, the Family Budget Survey micro-data are not public,

and I could not access the database. Therefore, I rely on the analyses available in Ringold,

Orenstein and Wilkens, Roma in an Expanding Europe and its background materials.16 The data set is available from the Soros Foundation Romania.17 Initial results of thesurveywerepublished inFleckandRughinis

,(eds),ComeCloser, 202�/14; thesecond

phase was conducted by the Institute for the Study of National Minority Problems (ISPMN, Institutul

pentru Studierea Problemelor Minoritat,ilor Nat

,ionale), and details are available on the ISPMN website

at http://ispmn.gov.ro/ancheta-sociologica-proiect-phare-2004-intarirea-capacitatii-si-construirea-de-

parteneriate-pentru-a-imbunatati-situatia-romilor-si-perceptia-lor-publica (viewed 15 June 2010).18 The data set is available from the Soros Foundation Romania.

COSIMA RUGHINIS,

341

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 7: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

demographic information is missing. For example, the ERBA 2002 survey ismissing information about the ethnic identity of 163 respondents in the‘Roma sample’ (26 per cent of the sample), whom I have reluctantly assumedto be Roma. The UNDP database has age codified only by age intervals.Since the first interval is 15�/19, it is impossible to select an adult populationaged 18 and over. Therefore, I have not included the UNDP database in theanalysis of levels of educational attainment (Figure 2). In addition, the ERBAcodification of education differs from that of the other databases (as detailedbelow) and I couldn’t therefore use it for the sample comparison in Figure 2.In all of the following analyses, unless otherwise specified, I have looked

only at adult (aged 18 and over) self-identified Romani respondents.

Aggregating self-identified Roma subsamples

There are many estimates of the total Romani population of Romania.Census-based estimates have ranged from a minimum of 0.3 per cent in1966*/proof of intense assimilation pressures under the Communistregime*/to 2.5 per cent in 2002. In 1993 the Research Institute for Qualityof Life proposed a larger estimate of around 800,000 to one million peoplecomprising the ‘Roma population identified as such by others, according totheir way of life’,19 around 4 per cent of the total population. In subsequentresearch, in 1998, the Institute proposed a figure of around 1.5 million Roma,35 per cent of whom were hetero-identified. Based on a community survey,Dumitru Sandu proposed an estimate of 730,000�/970,000 persons ‘with ahigh probability of Roma self-affiliation’.20 These larger estimates includedRomani people who declared another ethnicity in survey contexts, whilepresumably considering themselves Roma in private contexts.In order to achieve a meaningful subsample of self-identified Romani

individuals from a national sample, the sample must be rather large, sinceself-declared Roma comprise only roughly between 2.0 and 2.5 per cent ofthe total number of respondents (2.2 per cent in the FBS 2005 database, forexample). In the Census, self-identified Roma represented 1.8 per cent of thepopulation in 1992 and 2.5 per cent in the IPUMSI 2002 database.21 Thismeans that private organizations doing research on nationally representativesamples of around 1,000�/1,500 individuals cannot use self-identificationsince, for every 1,500 people in the national sample, there would be onlyaround 35 Roma. Larger samples are necessary in order to use this method:

19 Zamfir and Zamfir (eds), T,iganii, 60.

20 Dumitru Sandu, Comunitat,ile de romi din Romania: O harta a saraciei comunitare prin

sondajul PROROMI (Bucharest: World Bank 2005), available online at www.anr.gov.ro/docs/statistici/PROROMI__Comunitatile_de_Romi_din_Romania_187.pdf (viewed 9June 2010).

21 Labelled as ‘Gypsy’ in the IPUMSI database.

342 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 8: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Table

2ComparisonofsurveyswithRomasamplesorsubsamples,includingonly

respondents

providinginform

ationabouteducationallevel

IPUMSI

2002

PEGTS

2000

ERBA

2001

ERBA

2002

UNDP

2002

POB

2001�/2004

HPS

2004

FBS

2005

RIB

2006

Inclusion

2007

WA

2008

Custom-built

Romasample

(no.of

respondents)

�/

�/

543

540

947

�/

1,109

(members

ofreference

couples)

�/

1,387

1,104

(household

members)

996

Roma

subsample

(no.of

respondents)

29,683

113

�/

�/

�/

325

�/

1,498

�/

�/

�/

%self-

identifiedin

Romasample

Not

specified

31%

91%

Notestimated

dueto

missing

data

for163

respondents

95%

Not

specified

91%

Not

specified

93%

100%

for

headsof

households

100%

Includes

inform

ation

on

All

household

members

Randomly

selected

respondents

Randomly

selected

respondents

Randomly

selected

respondents

Randomly

selected

respondents

Randomly

selected

respondents

All

household

members

ofrandomly

selected

couples

All

household

members

Randomly

selected

respondents

Randomly

selected

headsof

households

and

household

members

Randomly

selected

respondents

Otherdata

(no.of

respondents)

National

sample

(1,660,215)

National

sample

(1,051);

poverty

subsample

(505)

National

sample

(802);

Transylvanian

and

Hungarian

samples

National

sample

(800);

Transylvanian

and

Hungarian

samples

Roma

samplesfrom

5European

countries

Multiple

national

samples

(14,854)

National

sample

National

sample

(69,324)

National

sample

(1,215)

Non-R

oma

comparative

sample

(1,142

households)

National

sample

(2,391)

UrbanRoma22

39%

59%

Notavailable

49%

38%

�/

36%

37%

41%

43%

42%

Urbannational

53%

59%

Notavailable

54%

�/

�/

�/

56%

54%

�/

59%

22Proportionsofurbanrespondents

inboth

theRomaandnationalsamplesare

givenhere

toillustrate

thedegreeofsim

ilarity

ofsample

structures.TheRoma

samplesare

more

diversein

theirurban/ruraldistributionthanthenationalsamples.

COSIMA RUGHINIS,

343

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 9: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

INSSE conducts annual surveys on 36,000 households as part of the FBS,

thus including approximately 1,500 self-declared Romani persons.The PEGTS 2000 survey relied on a Roma subsample aggregating hetero-

identified Romani respondents from many national surveys conducted by

Romanian market research companies across one year. These companies

were asked to instruct their researchers to note the perceived ethnicity of

their respondents. Of the resulting subsample of 368 individuals, a total of

113 respondents self-identified as Roma when asked about their ethnicity in

the subsequent PEGTS survey.Following the PEGTS 2000 strategy I have built a Roma subsample

including all self-identified Romani individuals from the national samples

included in the consolidated POB 2001�/2004 database of the Soros Founda-

tion Romania, arriving at a total of 325 Romani respondents. This approach

can only serve an exploratory purpose, given its methodological limitations:

that is, aggregating very small subsamples produced in different years may

not lead to a good overall sample.

Ethnonyms used in measuring Romani ethnicity

The first challenge in surveying Roma is deciding how Romani ethnicity

can be empirically measured. There are several ethnonyms that are usually

understood to indicate Romani ethnicity. For example, in Romania, the

main present-day alternatives are ‘T,igan’ (singular male) and ‘T

,iganca’

(singular female), or ‘Rom’ (singular male) and ‘Romnie’ or ‘Roma’

(singular female). ‘T,igan’ and ‘T

,igani’ (plural) have been used historically,

by Roma, and especially by the non-Roma, as an umbrella term for the

various groups (neamuri) of Roma. ‘Rom’ and ‘Romi’ (plural) are also

umbrella terms in widespread use, and promoted after 1990 by the Roma

NGO community and by the Council of Europe,23 in reaction to the

pejorative connotations of ‘T,igan’. Also worth noting is that, historically,

the people collectively seen by non-Roma as ‘T,igani’ identify themselves

primarily in terms of affiliation to a smaller group, namely, the neam (tribe).

In the following sections I will discuss the measurement practices used in

the surveys of Roma listed above, and the relevance of these practices for

data analysis. In particular, I will look for differences in the levels of

educational attainment associated with the various indicators of Romani

ethnicity.

23 Council of Europe, ‘Roma and Travellers glossary’, 11 December 2006, available on theCouncil of Europe website at www.coe.int/t/dg3/romatravellers/Source/documents/GlossaryRoma.doc (viewed 10 June 2010).

344 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 10: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Rom versus T,igan

The ‘Rom’ ethnonym entered official, academic and popular usage in

Romania after the fall of Communism as a result of the reaction by Romani

elites to the negative stereotyping of T,igani.24 Both ‘Rom’ and ‘T

,igan’ are still

used in popular speech, and the ‘Rom’ ethnonym has often been contested inpublic discourse because of its similarity to ‘Romania’, the name of the

country, and its corresponding ethnonym ‘Romanian’.25 Such debates have

offered a vivid illustration of the ongoing processes of ethnic boundary

maintenance.Even today, after more than a decade of use, the Romanian term ‘Rom’ is

mired in linguistic uncertainty. The term is sometimes spelled as ‘Rrom’, and

the conventions for spelling it with one or two initial r’s are not clearly

established.26 Dictionaries such as the Dict,ionarul explicativ al limbii romane

(1998), Dict,ionar de sinonime (2002), Dict

,ionar ortografic al limbii romane (2002)

and Noul dict,ionar explicativ al limbii romane (2002) only list the term ‘Rom’/

plural ‘Romi’, with no mention of a feminine form or the spelling ‘Rrom’.27

The Dict,ionar ortografic, ortoepic s

,i morfologic al limbii romane (2005) includes

both the nouns ‘Rom’ and ‘Rrom’, with the feminine forms ‘Roma’ and

‘Rroma’. The feminine noun ‘Romnie’, while occasionally used, is not listed

in these dictionaries or discussed in the public debate, which is exclusively

focused on the masculine noun ‘Rom’.There are several solutions to the problem of how to ask for ethnic

affiliation in surveys, including:

. the use of open-ended questions: the 2002 Census asked an open-endedquestion on ethnicity, and the IPUMSI data set includes the recodedvariable, unifying ‘Rrom’ and ‘T

,igan’ with the same value;

24 See, for example, Shannon Woodcock, ‘Romania and Europe: Roma, Rroma andT,igani as sites for the contestation of ethno-national identities’, Patterns of Prejudice,

vol. 41, no. 5, 2007, 493�/515.25 Most recently in a campaign initiated by Gabriela Antoniu’s article ‘Propunere

Jurnalul National: ‘‘T,igan’’ ın loc de ‘‘rom’’’, Jurnalul National, 17 March 2009.

26 According to Delia Grigore, in the Romani language the ‘Rrom’ spelling is used tomark the distinctive nasal pronunciation of the initial ‘r’; since the pronunciation inthe Romanian language is not nasal, ‘Rom’ should be the correct spelling (DeliaGrigore, ‘De ce rrom s

,i nu t

,igan’, Cotidianul, 10 November 2007).

27 I have consulted the dictionaries in the versions available online in the virtual libraryDexOnline (http://dexonline.ro, viewed 10 June 2010). The following dictionarieslisted the terms ‘Rom’ and/or ‘Rrom’: Academia Romana, Institutul de Lingvistica‘Iorgu Iordan’, Dict

,ionarul explicativ al limbii romane (Bucharest: Editura Univers

Enciclopedic 1998); Mircea and Luiza Seche, Dict,ionar de sinonime (Bucharest: Editura

Litera Internat,ional 2002);Dict

,ionar ortografic al limbii romane (Bucharest: Editura Litera

Internat,ional 2002); Noul dict

,ionar explicativ al limbii romane (Bucharest: Editura Litera

Internat,ional 2002); Academia Romana, Institutul de Lingvistica ‘Iorgu Iordan�/Al.

Rosetti’, Dict,ionar ortografic, ortoepic s

,i morfologic al limbii romane (Bucharest: Editura

Univers Enciclopedic 2005).

COSIMA RUGHINIS,

345

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 11: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

. the use of both ‘Rom’ and ‘T,igan’ in a closed ethnicity question as distinct

options (RIB 2006, WA 2008);. the use of the combined ‘Rom/T

,igan’ in a closed ethnicity question as a

single option (HPS 2004);. the use of only the ‘Rom’ ethnonym in a closed ethnicity question

(Inclusion 2007).

Data from RIB 2006 indicate that around 37 per cent of Romani respondentsself-identified as ‘T

,igani’, while data from WA 2008 indicate that 16 per cent

did so. It seems appropriate, therefore, to use both ‘Rom’ and ‘T,igan’ as

distinctive categories.

The neam affiliation

Traditionally, Roma belong to different ethnic subgroups, the so-calledneamuri, such as Caldarari, Ursari, Lingurari, Rudari, Lautari, T

,igani de

vatra and so on. Such neamuri used to have distinctive traditional occupa-tions and lifestyles transmitted through family and endogamy but inRomania their specificity has been in decline. Still, there are neamuri, suchas, for example, the Caldarari or the Lautari,28 that still preserve traditionaloccupations as well as a more traditional Romani lifestyle.According to Delia Grigore, the same tension between self-identification

and hetero-attribution can be seen in the classification of Romani neamuri.29

A neam is a group based on common social institutions and a certain degreeof consanguinity, maintained by intermarriage. Members of a particularneam often share similar occupations, a mother tongue, religion or socialposition, which has led to many attempts at scholarly classifications ofneamuri based on such criteria. Still, as in the case of all ethnic communities,the neamuri have been primarily differentiated by patterns of inter- andintra-group relationships in which occupation, religion, language etc. haveplayed important but not defining roles.30

Studies of communities of Romanian Roma/T,igani illustrate and dis-

cuss their diversity.31 They are often socially isolated from neighbouringcommunities, including neighbouring Romani communities. While many of

28 Margaret Beissinger, ‘Occupation and ethnicity: constructing identity amongprofessional Romani (Gypsy) musicians in Romania’, Slavic Review, vol. 60, no. 1,2001, 24�/49.

29 Delia Grigore, Curs de antropologie s,i folclor rrom (Bucharest: Editura Credis 2001),

51�/62, available on the Romanian Ministry of Education website at http://administraresite.edu.ro/index.php/articles/c426 (viewed 10 June 2010).

30 Ibid., 5, 53, 60.31 See Beissinger, ‘Occupation and ethnicity’; Denes Kiss, ‘Romii din Herculian s

,i

rolul religiei penticostale ın viat,a lor comunitara’, in Tamas Kiss, Laszlo Foszto and

Gabor Fleck (eds), Incluziune s,i excluziune: Studii de caz asupra comunitat

,ilor de

romi din Romania (Cluj-Napoca: Editura Institutului pentru Studierea Problemelor

346 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 12: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

them, especially in rural areas, can be associated with a dominant neam

identity, it is not clear in the above-mentioned literature if and to what extent

such neamuri integrate several communities, and what role they play in

community life. In some cases, such as the Lautari or the Caldarari, marriage

within the neam is strongly preferred, but in other cases there is no such

practice. Overall, the institution of the neam in contemporary Romania

remains rather mysterious.Many surveys ask Roma for their neam affiliation, although it is not clear

what its social relevance is. Data from the 1992 Research Institute for Quality

of Life survey indicate that 40 per cent of self-identified Romani respondents

did not state any neam affiliation;32 a proportion of 23 per cent of the RIB 2006

respondents considered themselves ‘just Roma/T,igani’, but only 5 per cent of

the WA 2008 respondents did so. Quantitative analysis of neamuri is difficult

because there are only a few cases of any particular neam, and there are no

theoretically grounded rules for aggregation into broader categories.Table 3 summarizes the proportion of those Romani survey respondents

who both supplied information about their level of education (see Table 2)

and declared themselves ‘Romanized Roma’*/‘T,igan romanizat’ or ‘Rom

romanizat’*/or Rudari. These two categories are often considered the most

assimilated neamuri, with little if any connection to traditional Romani

occupations or customs. The table indicates that the proportion of ‘Roma-

nized Roma’ varies widely between samples. Still, in those samples that

included the largest number, WA 2008 and RIB 2006, the ‘Romanized Roma’

and the Rudari were, on average, significantly better educated than other

Roma,33 thus supporting the assumption of higher rates of assimilation.

Minoritat,ilor Nat

,ionale/Editura Kriterion 2009); Laszlo Foszto, Colect

,ie de studii despre

romii din Romania (Cluj-Napoca: Editura Institutului pentru Studierea ProblemelorMinoritat

,ilor Nat

,ionale/Editura Kriterion 2009); Raluca Pas

,ca, ‘Specificite du terrain

tsigane: notes d’enquetes aupres des Chaudronniers de Sarules,ti’, SACR Annual

Review (Yearbook of the Romanian Society of Cultural Anthropology) (Bucharest:Paideia 1999), 90�/100; Cosima Rughinis

,, ‘Comment s’organise pour un meilleur

environnement dans un quartier d’exclus a Bucarest?’, in Francois Hainard andChristine Verschuur (eds), Femmes dans les crises urbaines: relations de genre etenvironnements precaire (Paris: Karthala 2001), 231�/70; Monica S

,erban, ‘Aspecte ale

atas,amentului fat

,a de comunitatea locala: Studiu comparativ asupra a doua

comunitat,i de rromi caldarari bogat

,i’, Revista de Cercetari Sociale, no. 3�/4, 1998,

151�/79; Alina Silian, ‘Elemente pentru o regandire a incluziunii sociale*/Comunitatearomilor din Veseus

,-Jidvei’, in Kiss, Foszto and Fleck (eds), Incluziune s

,i excluziune;

Gabriel Troc, ‘A state of despair: Roma (Gypsy) population during transition’, StudiaEuropaea, vol. 47, no. 1�/2, 2002, 49�/90; Malina Voicu (ed.), Nevoi s

,i resurse ın comunitat

,ile

de romi (Bucharest: Fundat,ia Soros Romania 2007); Malina Voicu and Claudiu D. Tufis

,

(eds), Romii: Poves,ti de viat

,a (Bucharest: Fundat

,ia Soros Romania 2008); and Cerasela

Voiculescu, ‘Construct,ii identitare la rromii din Sangerogiu de Mures

,’, Sociologie

Romaneasca , no. 1�/2, 2002, 100�/25.32 Zamfir and Zamfir (eds), T

,iganii, 243.

33 Statistically significant for Chi Square Text, p�0.01.

COSIMA RUGHINIS,

347

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 13: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Nevertheless, the sociological or policy relevance of the neam variable isuncertain. There is no clear understanding of the current processes of socialdifferentiation between the Romani neamuri, if there are any. While someneamuri seem actively to maintain their distinctiveness, this degree ofcommunal traditionalism, self-reliance or social distance from others couldbe measured independently of a neam variable. Lacking any theoreticalunderstanding of the relevance of the neam for shaping life trajectories inpresent-day Romania, this variable should be used with caution.

Self-identification versus hetero-attribution of Romani ethnicity

The sociological conception of ethnicity has gradually moved away from itstaken-for-granted understanding as a classification of social groups withclear cultural or even biological borders. This shift became more visiblewithin the discipline following the publication of a seminal work by FredrikBarth, Ethnic Groups and Boundaries.35 Barth challenged the definition ofethnic groups as ‘culture-bearing units’ based on biological reproduction,and focused on the maintenance of social borders as their key definingfeature.36 This perspective has been further developed by other scholars.37

Ethnicity has come to be understood as a process of group creation and

Table 3 ‘Romanized Roma’ or Rudari among Romani survey respondents who provided

information about education attainment level

ERBA 200234 WA 2008 RIB 2006 Inclusion 2007

Romanizat and Rudar (%) 12 51 56 14

Other neam or ‘don’t know’ (%) 88 49 44 86

Total (%) 100 100 100 100

Total Romani respondents 397 991 1,366 864

Source: secondary analysis of data sets

34 Only self-identified Romani respondents in the Roma sample are included in theanalysis. Of the 618 respondents in the Roma sample, 401 self-declared as Roma,54 self-declared as another ethnic identity and, for 163 respondents, their affiliationwas unknown.

35 Fredrik Barth, ‘Introduction’, in Fredrick Barth (ed.), Ethnic Groups and Boundaries: TheSocial Organization of Culture Difference (Boston: Little Brown 1969).

36 Ibid., 10�/15.37 See, particularly, AndreasWimmer, ‘Themaking and unmaking of ethnic boundaries: a

multilevel process theory’, American Journal of Sociology, vol. 113, no. 4, 2008, 970�/1022; Roger Ballard, ‘Race, ethnicity and culture’, in Martin Holborn (ed.), NewDirections in Sociology (Ormskirk: Causeway 2002); LoıcWacquant, ‘Race as civil felony’,International Social Science Journal, vol. 57, no. 183, 2005, 127�/42; Michael Omi andHowardWinant, ‘Racial formations’, in Paula S. Rothenberg (ed.), Race, Class, Gender inthe United States, 7th edn (New York: Worth Publishers 2007), 13�/22; Rogers Brubaker,

348 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 14: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

differentiation that invokes various histories, practices or symbols, and that

is closely intertwined with power relations. For example, Roger Ballard

writes:

Ethnic consolidation is not a product of cultural distinctiveness per se, but is best

understood as the outcome of the articulation of cultural distinctiveness in

situations of political and/or economic competition. As a result it normally erupts

in response to patterns of inequality of one form or another.38

This perspective means, for Rogers Brubaker, ‘taking as a basic analytical

category not the ‘‘group’’ as an entity but groupness as a contextually

fluctuating conceptual variable’.39

Individuals participate in the processes that create ethnic classifica-

tions*/such as defining ethnic labels, shaping stereotypes of ethnic group

members, and attributing ethnic identities to people*/including their own.

There are many strategies for producing collective identities, and ‘Gypsy-

ness’ has been a particularly elusive category.40 Identification with a certain

ethnic category may be contested: people may or may not agree with the

ethnic self-affiliation of other individuals, or with the ethnic affiliation that

others attribute to them. In this context, Ivan Szelenyi writes about

‘classificatory struggles to create Roma identity’,41 referring to the varying

ethnic attributions of survey respondents, survey interviewers and ‘local

experts’. Patricia Ahmed, Cynthia Feliciano and Rebecca Emigh also observe

that ‘social classification is a more dynamic process, taking place within the

context of struggles between those who classify and those who are

classified’.42

‘Ethnicity without groups’, Archives Europeenes de Sociologie/ European Journal ofSociology, vol. 43, no. 2, 2002, 163�/89; Rogers Brubaker, Mara Loveman and PeterStamatov, ‘Ethnicity as cognition’, Theory and Society, vol. 33, no. 1, 2004, 31�/64; andHenry Hale, ‘Explaining ethnicity’, Comparative Political Studies, vol. 37, no. 4, 2004,458�/85.

38 Ballard, ‘Race, ethnicity and culture’.39 Brubaker, ‘Ethnicity without groups’, 167�/8.40 See, for example, the discussion in Paloma Gay y Blasco, ‘‘‘We don’t know our

descent’’: how the Gitanos of Jarana manage the past’, Journal of the RoyalAnthropological Institute, vol. 7, no. 4, 2001, 631�/47.

41 Ivan Szelenyi, ‘Poverty under post-Communist capitalism: the effects of class andethnicity in a cross-national comparison’, paper presented at the conference ‘Unityand diversity’, Bruges, 29�/30 October 2001, 51, draft available on the Cordis website atftp://ftp.cordis.europa.eu/pub/belgium/docs/szelenyi_paper.pdf (viewed 15 June2010).

42 Patricia Ahmed, Cynthia Feliciano and Rebecca Jean Emigh, ‘Internal and externalethnic assessments in Eastern Europe’, California Center for Population Research,On-Line Working Paper Series, CCPR-055-06, December 2006, 3, available on theCCPR website at http://papers.ccpr.ucla.edu/papers/PWP-CCPR-2006-055/PWP-CCPR-2006-055.pdf (viewed 15 June 2010).

COSIMA RUGHINIS,

349

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 15: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

While groupness and processes of classification may be theoretically therelevant starting-points in understanding ethnicities, surveys must copewith the challenge of finding Romani settlements and communities, andmeasuring ethnicity of particular, individual respondents. How can this beachieved?

The reluctance error

Given the stigma associated with Romani identity, there is widespreadagreement that a significant proportion of people who conceive ofthemselves as Roma do not declare a Romani identity in official contexts,such as the Census. Debates about collecting ethnic information fromRoma, such as the Project on Ethnic Relations roundtable in Strasbourg inMay 2000,43 have revolved around the tension between the lack of reliabledata, due to respondents’ reluctance to self-identify, and the human rightsissues involved in the hetero-attribution of ethnicity. The European Frame-work Convention for the Protection of National Minorities specifies inArticle 3: ‘Every person belonging to a national minority shall have theright freely to choose to be treated or not to be treated as such.’44 Still, theExplanatory Report for the Convention adds a gloss to Article 3 (Para. 35):‘This paragraph does not imply a right for an individual to choosearbitrarily to belong to any national minority. The individual’s subjectivechoice is inseparably linked to objective criteria relevant to the person’sidentity.’45 Accordingly, the tension between self-identification and thehetero-attribution of ethnicity is underwritten by the requirement for‘objective criteria’ with regard to ethnic identity. Overall, the tendency indata collection has been towards greater use of self-identification since thehetero-attribution of ethnicity at individual level is seen to contradict theFramework Convention.The human rights debate has turned into a scientific debate as well, as

illustrated, in the Hungarian case, by Ferenc Babusik.46 In Hungary, the 1993Minorities Act made self-identification the only legitimate way to measureethnicity. While sociologists such as Istvan Kemeny, Gabor Kertesi and

43 For the report on the meeting, see Project on Ethnic Relations (PER), Roma andStatistics. Strasbourg, France. May 22�/23, 2000 (Princeton, NJ: PER 2000), available onthe PER website at www.per-usa.org/reports/PERStrasbourg.pdf (viewed 15 June2010).

44 Framework Convention for the Protection of Minorities, 1 February 1995, available onthe Council of Europe website at http://conventions.coe.int/Treaty/EN/Treaties/Html/157.htm (viewed 15 June 2010).

45 Explanatory Report, available on the Council of Europe website at http://conventions.coe.int/Treaty/EN/Reports/Html/157.htm (viewed 15 June 2010).

46 Ferenc Babusik, ‘Legitimacy, statistics and research methodology: who is Romani inHungary today and what are we (not) allowed to know about Roma’, Roma RightsQuarterly, no. 1, 2004, 14�/18.

350 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 16: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Gabor Kezdi defended the hetero-attribution of ethnicity for individualrespondents as a solution to respondents’ reluctance, Janos Ladanyi andIvan Szelenyi argued that hetero-attribution measures something differentthan self-identification.47 Based on their analysis, we may say that hetero-attribution may be a partial solution to the reluctance error, but only at therisk of introducing additional systematic errors (for a discussion, see below).

The debate on self-identification has developed, to a large extent, withouta documented estimate of the size and variability of the reluctance error. Onthe one hand, the willingness to self-identify in censuses is different from thewillingness to self-identify in surveys. On the other hand, such willingness isvariable across regions and time. And, last but not least, hetero-attribution atthe individual level is different from hetero-attribution at the communitylevel.

Variability of the reluctance error

There is considerable evidence, especially from qualitative studies, ofcommunities in which Romani identity is publicly accepted and individualshave also chosen to self-identify as non-Roma in censuses. In Romania,where censuses were conducted in 1992 and 2002, several scholars havestudied such communities.48 In some cases, a majority of the communitymembers declared Romani identity in the Census while in other cases only afew did. Of course, such case studies cannot give a quantitative estimate ofthe overall bias. Quantitative information from the UNDP 2002 data setindicates that, in Romania, 80 per cent of those who declared Romaniethnicity in the survey also declared it in the 1992 Census.49

The same data set demonstrates considerable variability at the nationallevel: rates of self-identified Roma who also declared Romani ethnicity in theCensus range from 35 per cent in the Czech Republic to 80 per cent inRomania and Bulgaria (see Table 4).

47 Ibid., 16.48 See, for example, Cosmin Briciu, ‘Cartierul Dodeni-Bicaz’, in Voicu (ed.), Nevoi s

,i resurse

ın comunitat,ile de romi, 27; Vlad Grigoras

,, ‘Buhus

,i-Mocirla’, in Voicu (ed.),Nevoi s

,i resurse

ın comunitat,ile de romi, 51;Marian Preda, ‘Conflictul din localitateaMihail Kogalniceanu’,

in Zamfir and Zamfir (eds), T,iganii, 178; S

,erban, ‘Aspecte ale atas

,amentului fat

,a de

comunitatea locala’, 153; Voiculescu, ‘Construct,ii identitare la rromii din Sangerogiu de

Mures,’, 102; Levente Salat and Valer Veres, ‘Ethnic minorities and local public

administration in Romania: cases of ethnocultural tension and segregation’, paperpresented at the 10th NISPAcee Annual Conference, ‘Delivering public services in CEEcountries: trends and developments’, Cracow, 25�/7 April 2002, available online athttp://adatbank.transindex.ro/vendeg/htmlk/pdf4098.pdf (viewed 15 June 2010).

49 Unfortunately, since the UNDP survey was conducted in Romania in December 2001,respondents’ answers refer to the Census in 1992 and not to the 2002 Census, andtherefore may be influenced by failures of memory. For information about the UNDPsurvey period, see UNDP, Avoiding the Dependency Trap, Annexes.

COSIMA RUGHINIS,

351

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 17: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

While Census under-estimates are well documented, there is little if anyevidence of survey under-estimates. Discrepancies between interviewerhetero-attributions of ethnicity and self-identification cannot be interpretedto indicate under-representation, because survey interviewers do not havepersonal knowledge of the respondent and thus cannot be sure that therespondent self-identifies as Roma in other contexts.The Inclusion 2007 survey asked for a primary and a secondary affiliation,

attempting to see what proportion of respondents would declare aRomanian ethnicity and simultaneously assert a secondary Romani identity.The difference it made in estimating the Romani population was small:in the comparative sample, out of 956 respondents, 4.7 per cent declareda primary Romani ethnicity while only an additional 0.4 per cent(4 respondents) declared a secondary Romani ethnicity. Of course, atnational level, this difference could become significant. In March 2008 thepresident of the Institute for the Study of National Minority Problems(ISPMN, Institutul pentru Studierea Problemelor Minoritat

,ilor Nat

,ionale),

Horvath Istvan, proposed that the next Census, scheduled for 2011, shouldallow for multiple ethnic identifications, in order to obtain better estimates ofthe Romani population.TheUNDP survey, assuming a reluctance error in survey self-identification,

proposed an ‘implicit endorsement’ identification strategy (see Table 5). Theinterviewer approached potential respondents, hetero-identified as Roma,with the opening question: ‘Good morning/day, we are conducting a surveyamong the Roma population. Would you mind being interviewed?’ Explicitrejection led to the cancellation of the interview, but acceptance wasinterpreted to mean that the respondent was Roma.50 This method requiresa leap of faith, since it is not clear whether acceptance of the interview reallydoes indicate an implicit declaration of Romani ethnicity or is given for otherreasons, such as out of politeness or because of a misunderstanding. This dataset also indicates variation between the national contexts in which the surveywas conducted. The proportion of selected respondents who explicitly

Table 4 Census declaration of ethnicity by respondents who ‘feel Roma’

What ethnic affiliation did you

declare in the last Census?

Bulgaria

(%)

Czech

Republic (%)

Romania

(%)

Slovakia

(%)

None 8 5 11 0

Roma 80 35 80 48

Majority 6 55 8 43

Other minority 5 6 0 8

Total 100 100 100 100

Number of respondents 712 708 840 868

Source: secondary analysis of UNDP 2002 database

50 UNDP, Avoiding the Dependency Trap, 87.

352 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 18: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

rejected a Romani identity was 5 per cent in Romania but 14 per cent inBulgaria. The research report included all respondents in the analysis but it isnot clear what is gained from mixing the 5�/15 per cent proportion of hetero-identified respondents with the clearly defined category of self-identifiedrespondents.51

One important observation at this point is that the proportion of self-declaredRomanian Roma in nationally representative samples of various survey organizationsis lower than the proportion in the 2002 Census or in the FBS 2005 sample. In thePOB surveys there were 457 cases in all databases aggregated for the period1998�/2004, a mere 1.6 per cent of the total number of cases.52 In other nationalsurveys as well the proportion of Romani respondents ranged from 0.9 percent to 1.9 per cent.53 Since the Census and FBS are conducted by officialinstitutions, and given the reticence of some Roma to identify themselves asRoma or T

,igani openly to public authorities,54 I anticipated that the relation-

ship would be reversed. This asymmetry indicates a bias in the samplingmethodology of nationally representative surveys, a bias that may be due, forexample, to the under-representation of Romani residents in electoral lists.Electoral lists, compiled by local population bureaux, have been used as asampling frame by the POB and many other national-level surveys. It ispossible that proportionally more Roma than other ethnic groups do notregister their residence with the bureaux: if, for example, they live ininformally rented accommodation or in houses for which they are not thedocumented owners. If electoral lists are the source of the bias, this wouldexplain why the FBS survey is unaffected, since it uses Census lists as asampling frame.

Table 5 Self-identification of respondents selected by ‘implicit endorsement’ identification

strategy

Do you feel

Roma?

Bulgaria

(%)

Czech Republic

(%)

Romania

(%)

Slovakia

(%)

Hungary

(%)

Yes 85 84 95 90 100

No 14 13 5 9 �/

No response 1 3 0 0 �/

Total 100 100 100 100 100

Source: secondary analysis of UNDP 2002 database

51 The Hungarian sample is an exception because the Hungarian researchers decided toinclude only self-identified Romani respondents.

52 With a 95 per cent degree of confidence, the proportion is in the interval 1.55�/1.71per cent.

53 Cosima Rughinis,, ‘Introduction’, in Fleck and Rughinis

,(eds), Come Closer, 11.

54 A proportion of 80.5 per cent of self-affiliated Roma in the UNDP survey said thatthey declared a Romani identity in the previous Census; 8.5 per cent declaredRomanian ethnicity, 0.5 per cent other ethnicity, and the remaining 10.5 per cent saidthat they had declared no affiliation in the Census.

COSIMA RUGHINIS,

353

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 19: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

The national-level random sample of the PEGTS 2000 survey also includeda mere 1.1 per cent Romani respondents, a clear under-representation inrelation to the Census. As in the case of the POB surveys, this may be due tounder-representation of Romani respondents on electoral lists, perhapscombined with a general reluctance of survey interviewers to enter Romanicommunities.Strictly speaking, it is impossible to identify the reluctance bias in surveys

using survey data since this information falls into a blind spot of theinstrument. A combination of survey data and qualitative information frominterviewers with in-depth knowledge of the respondents and theircommunity would be needed. The blind spot refers not only to the averagedegree of under-representation of self-affiliated Roma in surveys, but also tothe variability of this under-representation and the factors that mayinfluence it. For example, it is assumed*/but not documented*/that surveysthat use Romani interviewers increase the probability of ethnic self-identification,55 provided that the interviewers look both Romani andtrustworthy to the respondents.

Hetero-attribution at the individual level

The first difficulty in using hetero-attribution as a proxy for self-affiliation atthe individual level is the fact that each measures something distinct andthere are no theoretical reasons to assume a strong correlation between them.Why should we expect the ethnic self-conception of a person to betransparent to an observer? More to the point: why should we expect thatthe ethnic self-conception of a person who chooses to hide her ethnicaffiliation in a survey would be transparent to an observer? Are theremarkers of ethnic affiliation that can be legitimately used by observers toinfer ethnicity? The main risk of hetero-attribution is that observers identifypeople who correspond to a stereotypical portrait of Roma; therefore,including these respondents in the ‘Roma sample’ would only erroneouslyconfirm the stereotype.56

For example, in Figure 1, we can distinguish five possible areas ofdivergence between family origins, self-affiliation, self-identification andhetero-attribution. Area A indicates the outcome of the process of ethnicassimilation in which descendants of Roma affiliate to non-Romani ethniccommunities. Area B together with Area C indicates the error of under-representation of ethnic affiliation when using self-identification: notall people who privately self-affiliate as Roma identify themselves assuch in public contexts. Area C indicates the correction introduced by using

55 Babusik, ‘Legitimacy, statistics and research methodology’, 18.56 Project on Ethnic Relations, Roma and Statistics, 14; Ladanyi and Szelenyi, ‘The social

construction of Roma ethnicity in Bulgaria, Romania and Hungary during markettransition’, 82.

354 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 20: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

hetero-attribution: a proportion of self-affiliated Roma who do not self-

identify as such in surveys are indicated by observers to be of Romani

ethnicity. Area D indicates the respondents who are of Romani origin and

who no longer consider themselves Roma, but are deemed to have Romani

ethnicity by observers. Finally, Area E indicates the most serious errors

introduced by hetero-attribution: people who do not consider themselves

Roma and have no Romani origins, but are considered by observers to be

Roma. Of course, the relative size of these errors and corrections may be very

different than the one in Figure 1. Still, it is important to observe that hetero-

attribution introduces its own, systematic, errors. For example, respondents

in Area E probably correspond visually more to the Romani stereotype*/

having darker skin and living in poverty57*/than respondents in Area A or

Area B.The lack of theoretical justification for the use of hetero-attribution is

underlined by empirical evidence that there is considerable divergence in

perceived ethnic membership. The study conducted by Ladanyi and Szelenyi

explores the differences between self-identification and hetero-attribution by

interviewers using an ‘ethnic oversample’. The ethnic oversample consists of

respondents who were hetero-identified as Roma by first-stage interviewers

Figure 1 Errors in measuring ethnic affiliation by self-identification and hetero-attribution

57 Ladanyi and Szelenyi, ‘The social construction of Roma ethnicity in Bulgaria, Romaniaand Hungary during market transition’, 81�/2.

COSIMA RUGHINIS,

355

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 21: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

while conducting various samples at national levels. From these people, hetero-identified as Roma by the first-stage interviewer, only 71.7 per cent were also hetero-identified as Roma by the second-stage PEGTS interviewer.58 The same dataindicate considerable divergence between self-identification and hetero-attribution by an interviewer: from the 368 respondents selected as Romaby survey interviewers, only 30.7 per cent self-identified as Roma in thesubsequent PEGTS survey.Surveys, such as UNDP 2002, or the community survey ProRomi 2005,59

also sometimes make use of hetero-attribution by individuals who havespecial professional experience, and are thus considered ‘experts’*/such asteachers, mediators, NGO activists, social workers and so on. Ethnicattribution by experts differs from ethnic attribution by survey interviewersbecause access to information about respondents is professionally mediated.As Ladanyi and Szelenyi point out, in comparison to interviewers, expertshave detailed knowledge of ‘social problem’ cases, which may bias theirattribution towards a Romani ethnicization of poverty; they may also havemore detailed knowledge of ancestry, but it is unclear what their criteria arefor inferring current ethnicities from information about the past.60 Overall,we may say that professional expertise in the fields of social work, educationor community development is not expertise in ethnic affiliation, especiallythe ethnic affiliation of individual respondents. This observation is particu-larly relevant for those individuals who do not self-identify as Roma inpublic interactions. If ethnicity is to be conceptually different from lifestyle orstandard of living, it follows then that it is not directly observable, and noamount of expertise can qualify a person to detect a person’s ‘true’ ethnicaffiliation against his or her declarations. The same can be said of observers’ethnicity as well, since Romani interviewers or activists are not, by virtue oftheir own affiliation, better observers of others’ ethnic affiliations.

Hetero-attribution at the community level

Experts and local interviewers may, however, have access to a certain type ofknowledge relevant to understanding local communities or settlements. Itmay not be professional expertise per se, but direct contact and communica-tion with the sampled population that gives them access to local ethnicclassifications. If ethnicity is to be understood as a process of the creationand maintenance of borders between communities, local observers haverelevant knowledge about these social borders, including intermarriagenorms, friendship and trade or neighbourhood patterns. Two caveats apply.First, such local knowledge is useful only when there are distinctive

58 Ibid., 86�/7.59 Sandu, Comunitat

,ile de romi din Romania.

60 Ladanyi and Szelenyi, ‘The social construction of Roma ethnicity in Bulgaria, Romaniaand Hungary during market transition’, 81�/2.

356 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 22: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

processes of border maintenance at settlement level, differentiating one

settlement from neighbouring settlements. If there are no such patterns of

selective closure and openness, enforced in the community, there is little to

be known about the ethnic affiliation of community members. Second, such

local knowledge is more relevant at community level than at individual

level. A local observer can identify community norms and practices more

easily than private self-conceptions of individuals.Therefore, informed observers can contribute to a settlement-level survey

of Romani communities. Such surveys were conducted, for example, in

Romania in 2005,61 and 2007�/8,62 and in Slovakia in 1997 and 2003�/4.63

When surveys like these are completed by local authorities, a main problem

that occurs is that the latter are constrained, by virtue of their official

responsibilities, to comply with Census information,64 or other official

information on ethnic groups. Therefore, the survey reproduces to some

extent the ‘reluctance error’ of the Census. A solution to this problem is to

address questionnaires to private observers who have the freedom to gather

and interpret information, including official data, without being constrained

by strict adherence to that data.65 Of course, special training must be

provided to these observers in order to make sure that they properly identify

communities of people who self-identify as Roma in unofficial contexts.Community surveys offer only probable data concerning individual

respondents. For example, Dumitru Sandu interprets the ProRomi survey as

referring to a population ‘which probably self-identifies as Roma’.66

Some members of this population with a high probability of Roma self-

identification may not actually self-identify as Roma and, for all intents and

purposes, may not be Roma, even if they live together with Romani people.

For example, in the PEGTS 2000 survey, 13 per cent of self-identified Romani

respondents stated that there was, in their household, a person who was not

Romani. The Inclusion 2007 survey interviewed self-identified Romani heads

of households but also additional, randomly selected members of the

households. While in some cases the randomly selected member was the

same as the head of the household, there were 548 respondents who were

61 The ProRomi survey presented in Sandu, Comunitat,ile de romi din Romania.

62 Inclusion 2007.63 Eva Haviarova, ‘Demographic and socioeconomic data on the Roma in Slovakia and

their use by public administration’, paper presented at the 12th NISPAcee AnnualConference, ‘Central and Eastern European countries inside and outside the EuropeanUnion: avoiding a new divide’, Vilnius, 13�/15 May 2004, available on the UN PublicAdministration Network website at http://unpan1.un.org/intradoc/groups/public/documents/nispacee/unpan018442.pdf (viewed 15 June 2010).

64 As a field supervisor for the Inclusion 2007 local authority survey, I often receivedtelephone calls from local authorities asking how should they report a given Romanicommunity whose members declared other ethnic identities in the Census.

65 Such as in the ProRomi 2005 survey and the Slovakia 2003�/4 survey mentioned above.66 Sandu, Comunitat

,ile de romi din Romania, 42.

COSIMA RUGHINIS,

357

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 23: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

different members of the household: 53 of them, around 10 per cent, self-identified as a non-Romani ethnicity (Romanian, Hungarian, Other). Sincethey lived in a household that had amemberwho self-identified as Roma, theybelonged to a population with a high probability of Romani self-identificationalthough they did not actually self-identify as Roma. Still, interviewers hetero-attributed Romani ethnicity to 32 of the 53 individuals, amounting to aproportion of 60 per cent. Therefore, as we have seen before, interviewersmayuse contextual information to override individual self-identification, and it isunclear whether this reduces or amplifies measurement error.

Self-identification and hetero-attribution in sampling

When designing a custom-built sample of the Romani population, there aretwo different stages that require information about ethnicity: first, theselection of localities and neighbourhoods, and, second, the selection ofindividual respondents. The selection of localities usually starts with Censusinformation on the proportion of self-declared Roma at locality level. Giventhe negative bias of Census estimates, which would distort samplerepresentativeness,67 this information is usually complemented by addi-tional data at aggregate level. For example, the sampling method for RIB2006, Inclusion 2007 and WA 2008 used information from the ProRomicommunity survey;68 the UNDP survey used information from individualswho were considered experts; and Fredrik Babusik used a combination ofdata based on hetero-attribution at local level.69

This first stage is followed by respondent selection, which may rely onself-identification only or may also use hetero-attribution. Two of the firstsurveys to be conducted in Romania on the Romani population used bothself-identification and hetero-attribution to select respondents.70 Hetero-identified respondents were considered to be Roma, and the variable ofethnic self-affiliation was used in analysis to differentiate among differenttypes of respondents (the ones who self-affiliated and the ones who did not).Following legal and political debates concerning the use of self-

identification as the only acceptable method for selecting individual Romanirespondents, subsequent surveys in Romania have usually opted to includein their analyses only those respondents who explicitly declared theirethnicity to be Romani/T

,igani. The RIB 2006,71 WA 2008 and Inclusion

67 Susanne Milcher and Andrey Ivanov, ‘The United Nations DevelopmentProgramme’s vulnerability projects: Roma and ethnic data’, Roma Rights Quarterly,no. 1, 2004, 7�/13.

68 Sandu, Comunitat,ile de romi din Romania, 6.

69 Babusik, ‘Legitimacy, statistics and research methodology’, 17�/18.70 Zamfir and Zamfir (eds), T

,iganii; Catalin Zamfir and Marian Preda (eds), Romii ın

Romania (Bucharest: Editura Expert 2002).71 Badescu et al., Roma Inclusion Barometer.

358 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 24: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

200772 samples included only self-identified Romani respondents in theiranalyses.

Settlement types in Roma samples

Custom-built samples depend on information about the national distributionof Romani residents at the settlement (neighbourhood) level.73 However,because of the bias in Census information about ethnic affiliation, it isunclear how many Roma live in homogeneous Romani communities, andhow many live in relatively integrated neighbourhoods.

The distribution of respondents across settlement types is importantnot only for methodological reasons, but also from a theoreticalperspective. If ethnicity is understood as a process of border creationand maintenance, residential and community interaction patterns areessential factors in ethnic differentiation. Segregated and isolated neigh-bourhoods or schools enhance differentiation, and increase the relevanceof ethnicity in local cultures. At the same time, qualitative researchersfocusing on several Romani communities within the same localitiesindicate that neighbouring Romani communities may display widelydivergent life trajectories for Romani people.74 Community features areimportant sources, and possibly predictors, of differentiation. None-theless, the measurement of community traits in surveys is occasional,rather than systematic.

Various scholars have pointed out that the survey-based, sociologicalunderstanding of Roma focuses on the average Romani individual, whosemost salient feature in the literature is social exclusion.75 The variationin individual life trajectories and choices, and, in particular, the diversityof Romani communities are explored by qualitative, anthropological

72 Fleck and Rughinis,(eds), Come Closer.

73 By ‘settlement’, I mean neighbourhoods or communities that have a commonlyknown name and that are distinguishable from other settlements in the localunderstanding, though not necessarily according to municipal or other administrativeinstitutions.

74 See, for example, Mariana Goina, ‘Intre ‘‘t,iganii de matase’’ s

,i cei ce ‘‘fierb ın suc

propriu’’: Studiu de caz al comunitat,ii de romi din Curtici, judet

,ul Arad’, in Kiss,

Foszto and Fleck (eds), Incluziune s,i excluziune; S

,erban, ‘Aspecte ale atas

,amentului fat

,a

de comunitatea locala’; Voiculescu, ‘Construct,ii identitare la rromii din Sangerogiu de

Mures,’; Stefania Toma, ‘‘‘T

,iganul meu’’ s

,i ıncrederea*/relat

,ii economice informale

ıntr-o comunitate multietnica din Transilvania’, in Kiss, Foszto and Fleck (eds),Incluziune s

,i excluziune; and Gabriel Troc, ‘A state of despair: Roma (Gypsy)

population during transition’.75 See, for example, Gyorgy Csepeli and David Simon, ‘Construction of Roma identity in

Eastern and Central Europe: perception and self-identification’, Journal of Ethnic andMigration Studies, vol. 30, no. 1, 2004, 129�/50 (134�/5); and Gabor Fleck, Ioana Floreaand Cosima Rughinis

,, ‘Conclusions’, in Fleck and Rughinis

,(eds), Come Closer, 217.

COSIMA RUGHINIS,

359

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 25: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

researchers from quite a different angle.76 There are few connection pointsbetween the two perspectives, reminding us of the parable of the blind menand the elephant.In the following sections I will explore the data on community character-

istics available in surveys on Romanian Roma. By ‘Romani community’ or‘Romani settlement’ I refer to neighbourhoods that have a sizeable Romanipopulation, a name commonly known and used by local people, andtherefore a certain public identity. Some of these settlements are tightlybound communities, while others are conflict-ridden neighbourhoods orsomething in between. I will refer to all of them under the generic terms‘communities’ or ‘settlements’.

The settlement-type blind spot

The RIB 2006 and WA 2008 surveys followed the same sampling proce-dure,77 designed by Dumitru Sandu;78 the Inclusion 2007 survey followedsimilar sampling recommendations. In this procedure localities are selectedon the basis of Census information and also on the basis of ProRomi 2005community survey data in order to represent various settlement types. TheProRomi 2005 survey was a postal survey in which a respondent team,consisting of a representative of the mayor’s office, a Romani resident and alocal expert, filled in a questionnaire with information about Romanicommunities in each locality. The survey indicated that 70 per cent of thepopulation identified as Roma by the respondent team lived in ‘wellidentified’ communities of twenty households or more, while 12 per centlived in communities that were ‘poorly identified’, and 17 per cent livedoutside of identified communities. Sandu used this information to concludethat 70 per cent of the Romani population lived in large groups,12 per cent lived in small groups and 17 per cent lived scattered throughoutlargely non-Romani neighbourhoods, as well as to distribute the sampleacross the different settlement types.79 Still, while uncertainty in identifying

76 Michael Stewart, ‘Approaches to Roma and Gypsies from within social anthropologywith particular reference to the Anglo-Saxon traditions’, paper presented at theconference ‘Az etnicitas empirikus kutatasanak dilemmai’, Research Institute ofEthnic and National Minorities, Hungarian Academy of Science, 11�/12 June 2008,draft available on the Hungarian Academy of Science website at www.mtaki.hu/hirek/080611_12.html (viewed 16 June 2010).

77 Badescu et al., Roma Inclusion Barometer, 5�/6.78 Dumitru Sandu, ‘Schema de es

,antionare pentru Barometrul Incluziunii Romilor, 2006

(BIR)’, available on the author’s website at http://dumitru.sandu.googlepages.com/SelectiaesantionuluideromiFSD06b.pdf (viewed 16 June 2010).

79 Dumitru Sandu, ‘Communities in Roma sampling’, slides for presentation at ISNMIconference, ‘Methodological questions of Roma surveys’, Cluj, 13 July 2008, availablefrom a link on the author’s webpage at http://dumitru.sandu.googlepages.com/didactice (viewed 16 June 2010).

360 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 26: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

a Romani community in a locality may indeed reflect the ethnic hetero-geneity of the community, it may also reflect a survey respondent team’slack of knowledge or interest. As is often the case with surveys of Roma, acertain degree of trust is needed to compensate for the ambiguity of theinformation.

For cost-efficiency reasons, custom-built Roma samples search forrespondents mainly in neighbourhoods with a large proportion of Romaniresidents. Sampling Roma in Romani neighbourhoods is less costly thansampling geographically dispersed Roma. According to the design of thesample, some resources may be allocated to the search for these morescattered Romani residents. With the exception of ERBA 2001 and HPS2004, which reported a sampling method that aimed to include Romanirespondents living in non-Romani neighbourhoods, the other samplesrelied on a first stage of finding Romani communities, and then selectingrespondents within those communities. The ERBA 2001 presentation of itssample methodology specified the use of hetero-attribution by threeneighbours who selected Roma living in the non-Romani communities.80

Overall, given the lack of clear methodological oversight on the spatialdistribution of Romani respondents, we can expect that custom-built Romasamples will over-represent respondents in homogeneous Romani com-munities.

Subsamples of Romani respondents presumably have an opposite bias,since interviewers in national surveys are likely to avoid, to the extent thatthey have a choice, homogeneous Romani communities that are oftenisolated, perceived as dangerous and generally more difficult to navigate.This difference between subsamples and custom-built samples is consonantwith the under-representation of self-identified Roma in subsamplescompared to the Census. As we have seen above, national samples includea lower proportion of self-identified respondents than the Census, thusindicating a certain bias in respondent selection. Starting from the assump-tion that a larger proportion of Romanian Roma live in rather homogeneoussettlements, an under-representation of such areas would lead to an under-estimate of the Romani population.

Indeed, there seems to be a clear difference in the degree of residentialsegregation between the Roma subsample of PEGTS 2000, and all thecustom-built Roma samples (see Table 6).81 For self-identified Roma,

80 Metro Media Transilvania, Barometrul Relat,iilor Etnice (Cluj: EDRC 2001), 3, available

on the EDRC website at www.edrc.ro/programe/Prezentare-BARE-2001.pdf (viewed15 July 2010). While this was a formal methodological framework, it would be aninteresting topic in itself to see how it was put into practice by researchers. Therelationship between official methodological procedures and researchers’ fielddecisions may offer surprising insights into the construction of Roma samples in allkinds of surveys.

81 For the subsamples in POB and FBS, there is no information about communityheterogeneity in the data set.

COSIMA RUGHINIS,

361

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 27: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Table

6Ethnic

homogeneity

ofrespondents’neighbourhoods

insurveys

ofthe

Romanipopulation,according

tothe

specific

indicators

of

homogeneityused

PEGTS

2000(Interviewer)

Inclusion2007

(Respondent)

Inclusion2007

(Interviewer)

UNDP

2002

(Interviewer)

WA

2008

(Respondent)

RIB

2006

(Respondent)82

Total‘ethnic

over-sample’

Self-identified

Roma

Romanisettlement(%

)11

15

83

75

55

70

45

‘Gypsysettlement’

‘Members

ofwhich

ethnic

grouplivemostly

intheneighbourhood

where

youpresently

live?’Answer:‘Roma’

‘‘Only

oralm

ostonly

Romanifamilies’’

‘Typeofareawhere

therespondentlives’

Answer:‘Principally

Roma’

‘Inyourstreet/blockof

flats,abouthow

many

neighbours

are

Roma/

T ,igani?’

‘All/

alm

ostall’and

‘Threequarters’

‘Ifyouthinkofthepeople

wholivein

thesamearea

asyou,whatpercentageof

thoseare

Roma/T ,igani?

(Openanswer)’

Answer:70%

ormore

Majority

Roma(%

)17

31

�/

16

34

10

15

‘Majority

ofthepopulationare

Roma’

‘There

are

Romaand

non-R

omafamiliesin

equalnumbers’

‘Mixed’

‘Abouthalf’

Answer:41�/69%

Majority

non-R

oma(%

)72

54

17

911

20

40

‘Majority

ofthepopulation

are

poornon-R

oma’or

‘Noconcentrationofeither

poororRoma’

Otheranswers:

Romanian,Hungarian,

other

‘There

are

some

Romanifamilies,butthe

majority

are

non-R

oma’

and‘There

are

nomore

Romanifamiliesnearby’

‘Principally

non-R

oma’

‘Aroundaquarter’,

‘None/Alm

ostnone’

Answer:30%

orless

Total(%

)100

100

100

100

100

100

100

Statistically

significantdifference

ineducationallevel

Yes(p�0.01)

No

No

No

Yes(p�0.01)

Yes(p�0.01)

No83

Source:secondary

analysis

ofPEGTS

2000,Inclusion2007,UNDP

2002andWA

2008data

sets

8233percentoftherespondents

did

notanswertheethnic

homogeneityquestion,andtheyare

excludedfrom

analysis.

83There

wasnostatisticaldifferencein

educationallevelbetweenrespondents

from

neighbourhoodswhere

Romaconstituted70percentormore

ofthepopulationandthosefrom

neigh-

bourhoodswithfewerRoma.There

washoweverastatisticaldifferencebetweenthosewhoestimatedtheRomanipopulationoftheirneighbourhoodsandthosewhodid

not.

362 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 28: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

interviewer estimates of residents in Romani settlements was 15 per cent inPEGTS 2000, while in the Roma custom-built sample all estimates werearound 45 per cent and above. If there is indeed a positive bias in custom-built samples in favour of the homogeneous settlements, and a negative biasin the subsamples in favour of scattered residents, it follows that the actualdistribution of Romani residents across various types of settlements falls in ablind spot of the survey method, much like the issue of the under-representationof Roma.

A second tentative conclusion, after reviewing the measurement anddistribution of community type across several samples, is that variablesrelated to community type are not standardized: each survey hasintroduced its own variables and items (see Table 7 for a general overview).This makes comparisons difficult, if not impossible. Information may becollected from the respondent, from the interviewer, or both. The mainindicators that we find in these surveys refer to the ethnic heterogeneity ofthe neighbourhood and the local schools. Other relevant data, such as thename of the community*/which would allow some reference to additionalresearch conducted in the area, the size of the community or theaccessibility of various local institutions*/are consistently missing. Overall,community-level variables are weakly measured and rarely used in datainterpretation.

Influence of sampling methodology on ‘educational level’survey results

We have seen that, apparently, Roma samples differ from Roma subsamplesof national samples because the former sample Roma living in Romanicommunities, therefore under-representing those that live in non-Romanineighbourhoods. The question arises as to whether this difference leads todifferent results on basic indicators between the Roma samples and Romasubsamples. Table 5 indicates that community type may be associated withthe educational level of the respondent. Indeed, the 1992 survey conductedby Zamfir and Zamfir indicated a strong correlation between communitytype and school attendance by children, although it is not clear whethersuch a correlation would remain if only self-identified Roma wereanalysed.84

We can formulate the following hypothesis, based on the generalassumption that people living in Romani communities are more sociallyexcluded than dispersed Roma: Roma samples under-represent the moreeducated Roma in comparison to Roma subsamples. In order to put thishypothesis to the test, I have compared the distribution of educationalachievement across various samples and subsamples in the years

84 Zamfir and Zamfir (eds), T,iganii, 97.

COSIMA RUGHINIS,

363

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 29: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

2004�/2008.85 The comparison is based on the assumption that Romanieducational achievement is relatively stable across a several-year span. Theactual distribution can be seen in Figure 2. On the one hand, we have threesubsamples of Romani respondents: the data sets from PEGTS 2000, theexploratory POB 2002�/2004 and the FBS 2005. On the other hand, we havefour custom-built Roma samples that include information on the sameindicator: HPS 2004, RIB 2006, Inclusion 2007 and WA 2008.If this comparison points to any conclusion, it might be that there is no

difference between the three Roma subsamples and the four Romasamples.86 Therefore, the hypothesis does not seem to be supported by data.

Table 7 Availability and source of information on respondent’s settlement in various

surveys (I�interviewer, R�respondent)

PEGTS

2000

Inclusion

2007

UNDP

2002

RIB

2006

WA

2008

ERBA

2001,

2002

Settlement name �/ �/ �/ �/ �/ �/

Size �/ �/ �/ �/ �/ �/

Ethnic heterogeneity in community I R, I I R R R

Ethnic heterogeneity in children’s

school

�/ R R �/ R �/

Poverty heterogeneity in community I R �/ R �/

Community separation (clear

distinction from neighbouring

areas)

�/ �/ I �/ �/ �/

Community isolation (main locality,

school, employment etc.

accessible/ distant)

�/ R I R

I

�/ �/

Evaluation of local inter-ethnic

relations

�/ �/ �/ �/ �/ R

Source: secondary data analysis

85 Of course, there are different codifications for questions on education. For alldatabases except ERBA 2001�/2, I could recode educational achievement into fourcategories: 1 no schooling at all; 2 complete or incomplete primary school (grades 1 to4); 3 complete or incomplete middle school, vocational or apprentice school,sometimes also including special schools (this category always has only a handfulof cases and most often is missing); 4 complete or incomplete high school (grades9�/12) or any complete or incomplete higher education (college, university,postgraduate study etc.). The ERBA questionnaires asked for the last schoolgraduated from, and therefore ‘no schooling’ is grouped together with graduates ofthe first 3 grades, and the primary school graduates are grouped together withincomplete middle school graduates.

86 Post-stratification of all Roma samples on residence, gender and age categories(according to the structure of FBS 2005) leads to very similar results, probably becauseeducational achievement in Roma samples has a weak correlation with residence,gender and age.

364 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 30: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

This may be the result of several processes. On the one hand, it is possiblethat respondents’ education is not correlated to present-day community typesince many respondents presumably lived in other neighbourhoods duringtheir childhood and youth. On the other hand, it is possible that thecorrelation is weakened by measurement errors, such as the randomvariability component of respondents’ estimates.

Conclusions and recommendations

The analysis of surveys of Romanian Roma conducted between 2000 and2008 suggests several conclusions. The survey measurement of Romaniethnicity in Romania is confronted by quite a few challenges, but theirimportance is not necessarily proportional with the attention they receive incurrent debates.

Ethnonyms themselves are not yet standardized across surveys, or indeedin the spoken language. For example, ‘Rom’ is also spelled as ‘Rrom’, and itsfeminine forms are not recognized in current Romanian dictionaries. Survey

Figure 2 Educational level of Romanian Roma in sociological surveys

0%

20%

40%

60%

80%

100%

5.00 Higher education,

college or more (13+ yrs)

4.00 Vocational, high school

(8–12 yrs)

3.00 Middle school (5–8 yrs)

2.00 Primary (1–4 yrs)

1.00 No schooling

2 4 0 1 1 1 1

16 26 13 19 15 17 18

31 30 31 29 33 33 37

35 20 32 34 28 31 30

17 20 23 18 23 18 14

PEGTS

2000

POB

2002-4

FBS

2005

HPS

2004

RIB

2006

Inclusion

2007

WA

2008

Source: secondary analysis of data sets

COSIMA RUGHINIS,

365

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 31: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

respondents identify themselves as both ‘Rom’ and ‘T,igan’ when given the

option in a survey question.The neam identification is commonly asked for but it has no clear

theoretical relevance, and neither has it been used systematically inquantitative analyses because of the few number of cases of most neamurias well as the lack of rules for aggregation.The reluctance error is the most intensely debated measurement issue

for surveys of Roma. Still, while reluctance in Census declarations isdocumented by qualitative and quantitative research alike, there is nosuch evidence for survey reluctance. This is not to say, of course, that thereare no systematic errors in survey self-identification for self-affiliated Roma.Still, the seriousness of the assumed bias is unknown. On the other hand, wecan better identify a different bias: the proportion of Romani respondents innational samples is systematically lower than in the Census. Since it isunlikely that respondents are more reluctant in unofficial surveys than in theCensus, it is plausible that this under-representation occurs because ofsampling procedures.Hetero-attribution of ethnicity at the individual level is not an adequate

substitute for self-identification because it causes additional systematicerrors of unknown magnitude. Of course, hetero-identified ethnicity mayprovide valuable information on its own regarding processes of ethnicclassification or factors that shape the experience of discrimination. Hetero-attribution at the community level is theoretically justified when it relies onlocal information about community norms and practices, provided that it isnot automatically translated as implying an individual’s ethnicity. Surveydata indicate that a significant proportion of non-Romani respondents live inRomani households, presumably as a result of mixed marriages. Therefore,the connection between community attribution and individual ethnicity isentirely theoretical. Information from the hetero-attribution of ethnicity atcommunity level is useful for sampling procedures in order to correct for theselective under-representation of Romani communities in the Census.Although the relevance of community configurations for understanding

life trajectories of Roma and their variability has been consistently supportedby qualitative, community studies, such variables are weakly measured andrarely used in quantitative analysis. The distribution of the Romanipopulation across community-level variables is difficult to study becauseof biases in Roma samples. Specifically, it seems that subsamples of Romanirespondents over-represent Roma who are sparsely dispersed, whilecustom-built samples over-represent homogeneous community residents.This bias does not seem to affect the comparability of sample results oneducational achievement, but it may affect other variables.Starting from the results of survey analysis, it seems that there is some

space for improving ethnicity measurement in quantitative studies of Roma.At the individual level, there seems to be no reason to replace self-identification as the main indicator of ethnic affiliation, at least in the

366 Patterns of Prejudice

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010

Page 32: Patterns of Prejudice The forest behind the bar charts ... · The forest behind the bar charts: bridging quantitative and qualitative research on Roma/T,igani in contemporary Romania

Romanian context. If hetero-attribution is used, the analysis of self-identifiedand hetero-identified respondents together as constituting a Roma sample isnot theoretically justified. Separate analyses should be conducted.

In addition to the ISPMN proposal of introducing in the Census multipleethnic affiliations, survey research focused on measuring the reluctance errorin the Census would also allow for a meaningful extrapolation of the Censusinformation, in order to infer the number of Roma in Romania. Extrapola-tions based on hetero-attribution risk including an unknown proportion ofnon-Roma who correspond to observers’ stereotypes of Roma. In effect, theyare not estimates of the Romani population, but of the Roma-like populationfrom the interviewers’ point of view.

Surveys of Romani communities can also provide valuable information ataggregate levels, which could further inform sampling procedures. Suchsurveys should not rely only on official informants since they are legallybound to respect Census declarations and therefore fail to compensate forreluctance errors. Community surveys should aim to obtain unofficial,expert estimates grounded in local knowledge and informed by official data.

Connecting survey research to qualitative research is notoriously challeng-ing. With regard to Roma/T

,igani, community focused research could be

an important bridge between the two approaches. While qualitative,anthropologically oriented research often takes into account Romani com-munities and their variability, this approach is all but missing from surveyresearch, which is almost exclusively focused on individuals and individualvariability. It is possible that improving the quality of settlement-level datain surveys, so as to allow for more detailed, even multilevel analysis, wouldfacilitate the integration of quantitative and community studies. Surveyinformation concerning settlements should include: neighbourhood nameand size, information on ethnic heterogeneity in the neighbourhood and inlocal schools, information on the salience of the settlement and its isolation,as well as evaluations of the quality of local inter-ethnic relationships. Suchinformation should be elicited from the respondents but also frominterviewers and, in the best of all possible worlds, it should be standardizedto allow for comparisons across samples.

A stronger bond between survey and anthropological research on Romawould lead to significant improvements. Survey blind spots in evaluatingthe reluctance error and the distribution of community types could bepartially overcome. The variability of Romani personal and communalcharacteristics would be more meaningfully measured, overcoming thesurvey tendency to reinforce a stereotypical view of Roma as social workcases. Last but not least, anthropological, quantitative and policy-focusedresearchers who study Roma would come together more frequently as anetwork of shared interests.

Cosima Rughinis,is Associate Professor at the Department of Sociology and

Social Work at the University of Bucharest.

COSIMA RUGHINIS,

367

Downloaded By: [Rughinis, Cosima] At: 19:05 28 September 2010