Communities in Roma Sampling Communities in Roma Sampling Slides for presentation to ISNMI debate on Methodological questions of Roma surveys, Cluj, July 13th 2008 D it S d U i Dumitru Sandu, University of Bucharest it f B h t
1
Is it good to cluster by Roma ethnic communities in Roma sampling, in the current Romania? •benefits •costs •alternatives What we know about community vs sparse residence of Roma as relevant for sampling? Experience basis to answer: p •PROROMA survey, World Bank, National Agency for Roma (NAR), 2005 •RIB Roma Inclusion Barometer, Soros Foundation Romania (SFR), 2006 •Work attitudes survey (WA), Roma sample of 999 adult persons of 18 years old and more Soros Foundation Romania and Gallup Organization, May 2008*. •Community Census of Migration, IOM, GOR, 2001 Structure of the presentation: •What are the Roma characteristics that gives the “sampling situation”? •RIB sampling design and its evaluation •How many Roma live in their ethnic communities •How are Roma clustered in their ethnic communities as compared with sparse ones *The The author of the presentation as not involved in designing the sample for WA but the sampling design was exactly author of the presentation as not involved in designing the sample for WA but the sampling design was exactly the same. The implementer of the sampling scheme was professor Mircea Comsa from Babes‐Boliay University. I thank SFR for allowing me the use of some data from WA data basis before publishing them.
2
High community inclusion
about 70% of Roma people live in communities that are larger than 20 hhds
High g heterogeneity by
residential pattern,
Roma R characteristics that are relevant for sampling and interviewing and interviewing
degree and type of Roma identification,
huma capital
low education stock
economic capital
low income,high unemployment , poor housing
i h b High poverty by
challenges
accessibility
network capital
Low visibility by
marginal location in localities, poor communication infrastructure
lower bridging vs bonding social capital
marginal location, low institutionalised marginal location low institutionalised identity, identity high mobility 3
Profile of RIB sample, 2006
Sampling design
a three stage probability sample with stratification in the first stage
Stages
1 locality with Roma 2. Roma communities of A or B type (see previous slide) within locality within locality 3. Roma self identified persons after screening within in A/B communities or in locality
Sample volume p
1500 by design and 1387 as implemented y g p
Spread of the sample in 97 localities out of 1735 localities recorded with at least 9 Roma in 2002 census
4
Basic stratification scheme for Roma Inclusion Barometer* •A stratification by locality size and residential type of locality as identified in 2002 national census and by residential pattern as derived from PROROMA survey •Stratification scheme applied for the 1735 localities having at least 9 Roma people in 2002 census
Locality type
Probable distribution of Roma by locality type, function of the results of PROROMA survey** C. without any A. well identified Roma B. ppoorlyy identified Roma communityy communities of at least Roma communities of identification in 20 hhds at least 20 hhds locality Total
Commune, under 200 Roma
3.3
3.3
6.9
13.5
commune, 201-600 Roma
9.2
3.4
8.1
20.7
communa, over 600 Roma
13.3
4.5
8.9
26.7
city,under 800 Roma
3.8
1.6
3.0
8.4
city, over 800 Roma
22.2
2.9
5.6
30.7
total
51.9
15.7
32.5
100.0
•For details see http://dumitru.sandu.googlepages.com/SelectiaesantionuluideromiFSD06b.pdf •** Dumitru Sandu, Roma Communities , Social Map, NAR‐WB, Bucharest p, , ,, July 2005, available y , also at http://www.anr.gov.ro/docs/statistici/Roma_Social_Mapping_187.pdf
5
Comparing PROROMI and RIB survey data with Roma population data at national level 2002
Degree of urbanisation % Cultural area % 11 BC NT SV VR 12 GL IS 13 BT VS 21 AG DB PH 22 BZ BR 23 G TL IL CL 31 DJ MH OT 32 GJ VL 41 DOBR 51 AB HD 52 BV SB 53 CJ MS 54 CV HG 55 BN SJ 61 MM SM 62 AD BH 71 BANAT 80 BUCURESTI Persons pe hhd A Average age off adults d lt %women
Roma population , 2002 census
PROROMA
BIR romi (1387 persoane de peste 17 ani)
39
41
37
6.3 4.3 1.5 7.9 3.8 10.5 9.3 1.9 1.6 3.9 6.6 11.3 1.8 4.4 4.2 89 8.9 4.5 7.2 100.0 4.66
8.1 6.5 3.9 11.3 4.4 11.1 5.8 2.3 2.7 5.6 4.4 8.8 3.2 2.4 3.4 10 3 10.3 0.4 5.5 100 4.67
8.6 2.7 1.3 8.5 3.4 10.4 11.0 2.5 2.3 6.0 3.6 11.9 3.0 1.3 3.8 10 3 10.3 4.7 4.6 100.0 5.64 40 6 40.6 53.0
6
Comments on RIB sampling scheme, PROROMI and census data 1.
In principle, cluster sampling with so differentiated units as Roma communities is less indicated as compared to to voting areas or constituencies as sampling units. But Roma communities are less heterogeneous than localities , communes or cities, and, consequently better units to cluster compared to localities cluster compared to localities.
2.
Exclusion error (zero or very low probability to include into the sample Roma from some territorial units) seems to be higher by using constituencies vs ethnic communities.
3.
RIB scheme from 2006, applied in the same way to WA survey from 2008, is based on a sampling frame that is in line with self‐identification principle (national census) and probable selfidentification principle of PROROMI enumeration. Possible exclusion errors in the census are compensated by the benefits of PROROMA survey that is independent from the census.
4.
Due to the fact that about 70% of Roma population live in own communities, Census distribution and PROROMA are consistent.
5.
RIB and WA data are highly consistent when analyzed – indirect proof of sampling design scheme validity. h lidit
6.
PROROMI data refers not to selfidentified or heteroidentified Roma but to “probable selfidentified Roma”, due to the fact that at least one of the key informants in the PROROMA survey were Roma from the reference community and the other two were city hall Roma expert and another person from locality as agreed by the Roma community person and the city hall representative. 7
Good to use triangulation in Roma social research, as in many other areas
1.
When working with populations like Roma ‐ that have different degrees of ethnic self‐identification, rather poor formal institutions identification and are highly mobile – the triangulation is the only solution: it is good to combine information from official standard censuses community censuses or solution: it is good to combine information from official standard censuses , community censuses or enumerations of the PROROMA type, standard surveys at individual level and anthropological approaches.
2.
Each of the above mentioned approaches have strong and weak points. The solution is not to refuse some of them by ideological criteria but to improve by informed criticism each of them and to f h b id l i l i i b i b i f d ii i h f h d combine their use. Transforming the limits of personal specialization in one method or another into value judgment criteria does not help too much.
3.
Roma people in the current Romania, and by long term tradition, live to a large degree in p p , y g , g g communities as clusters of neighbored households. Ignoring that basic fact in sampling design and/or in data analysis is a kind of de‐sociologyzation that could contribute to increasing the probability of producing statistical artefacts. Survey on Roma are conditioned by several factors related to sampling design, phrasing of the questionnaire ethnicity professionalism and strangeness (their quality of being local or nonlocal questionnaire, ethnicity, professionalism and strangeness (their quality of being local or nonlocal people) of the operators. All these factors should be considered when one assess a Roma survey.
4.
5.
National surveys on Roma could benefit from expanding the efforts to build a Roma social map for Romania. Census data could be considered as starting points . Improvements could be accomplished by building on PROROMA experience. 8
How many Roma are living in Roma ethnic communities: PROROMI community survey, 2005, % 100% 90% 80%
38 59
70%
67
68
60%
72
73
78
81
71
50% 38 14
27
20
16 Transilvvania
24
Mold dova
10%
16
4 13 24
15
12
9
6
13
13
17 ttotal
13
20%
Olttenia
30%
Cris.‐Maram mures
40%
sparsed
small groups (<20 hhds.)
Munttenia
Bucu uresti
Dobro ogea
B Banat
0%
Data source: PROROMA survey, National Agency for Roma (NAR), 2005, own computations D.S. 605849 Roma reported persons from 987 localities. Data basis is larger than the one reported in D.Sandu, Roma Communities Social Map, NAR, Bucharest, 2005 due to the fact that computations referred to a smaller number of items that were validate items that were validate separately. Figures for Bucharest are rather poor estimates. Their elimination from the total does not change too much the percentages for the country. percentages for the country.
large groups (>=20 hhds.)
•About 70% of the country Roma population lived in communities larger than 19 hhds, in 2005. •The others live in small communities (17%) or dissipated among majority population (12%). •Muntenia and Oltenia are the regions with the largest share of Roma people living in large communities.
9
How many Roma are living in Roma ethnic communities: Roma Inclusion Barometer (RIB), 2006, % 100% non‐answers
90% 33
80% 70%
55 large communities
60% 50% 40%
37
30% 20% 10%
small communities
14 9 31
21
sparse distribution
0% distribution without distribution including non‐answers non‐answers
The reference question in RIB survey of Soros Foundation Romania (SFR) was : “What is the percentage of Roma living in your area… ?”. Living in large Roma communities is the case of persons that declare to have more than 50% Roma neighbors. Small communities are, by convention, those of 25% to 49% Roma neighbors. If the person declares of having 0 to 24% % % g p g % Roma neighbors she or he is considered to live in an area with sparse distribution of Roma.
•“What is the percentage of Roma living in you area or zone?” is a question that is relevant for the structure and for the size of residence community. The larger the percentage of Roma perceived neighbors percentage, of Roma perceived neighbors , the larger the size of the reference community. The figures in this diagram are significant for the size of the perceived communities or everyday life interactions communities. •There is a highly significant correlation (r=0.38, p=0.001) between the percenteg of Roma neighbors and the percentage of perceived Roma in the locality. The finding support the view that the size of the perceived Roma community is larger in i dR it i l i localities with more Roma. •More than half of Roma people feel that they live in large Roma communities, tha is to say , having more than 50% Roma neighbors neighbors. •About 15% Roma people feel that they live in small Roma communities. •About one third of Roma people consider that they do not live in a Roma community due to the fact that that their neighbors are Roma in a low percentage.
10
“How many of your neighbors from the street or from the block where you live are Roma or Gipsy?” 100% 90% 80% 70%
55 70
60% rather all Roma
50%
about three quarters Roma
40%
about half Roma
30%
about one quarter Roma
20%
no/rather no Roma
10%
15 10
0%
SMALL ROMA COMMUNITY
14
10
20
6 % Roma on the street/in the block
% Roma by community type
Data source: Work attitudes survey, Roma sample of 999 adult persons of 18 years old and more Soros Foundation Romania and Gallup Organization, May 2008. The survey used the same sampling design as for RIB 2006. Reading example: about 55% of Roma people live in neighborhoods (on streets or in Blocks) where rather all neighbors are Roma. Adding to these the about 15% Roma that lives on streets or in blocks where about three quarters people are Roma one gets an estimation of about 70% R Roma living in large communities. The estimation is largely consistent with the ones derived from RIB 2006 and PROROMI li i i l iti Th ti ti i l l i t t ith th d i df RIB 2006 d PROROMI 2005. 11
Education structure by perceived size of Roma communities 100% 90%
23
17
14
12
higher
80% 25 70% 37 60%
37
gymnasium
34
50%
27
40% 30%
primary 27
27
30
20% 10%
35 17
19
19
sparse distribution
small communities
large communities
0% non‐answer
illiterate
Data source: RIB – SFR, 2007, own computations, DS
•Perceived large Roma communities ,that are the main residential environment for Roma people, are constituted mainly by very low educated people. • It is very likely that the large share of people that did not answer the question on the percentage of Roma in the neighborhood do live also in very large Roma communities. In fact they have a very similar education structure as people that declare of having lots of Roma neighbors. •If the previous hypothesis is true one can conclude that RIB data support the PROROMI data that indicate a proportion of about 70% Roma people living in large communities (of more than 20 hhds) •The pattern of living in a sparse distribution is specific for more educated Roma. 12
Identity profiles by perceived Roma community type, %
sparse distribution "What is your ethnicity/nationality"
Mother tongue
small large unspeci communities communities fied Total
Roma
69
64
64
59
63
Gipsy
31
36
36
41
37
Romanian
57
51
49
38
47
Romani
39
47
43
59
48
Data source:RIB Data source:RIB – SFR, 2007, own computations, DS. Reading example: 69% of the Roma living out of SFR, 2007, own computations, DS. Reading example: 69% of the Roma living out of Roma communities define themselves as Roma, not as Gipsy.
•Roma people in sparse residential pattern, as compared to Roma in small or large communities, d fi th define themselves in a lower identity way, more as l i l id tit •Roma vs Gypsy •Having Romanian as mother tongue, vs Romani
13
A much lower network capital for Roma as compared to the values on a national representative sample RIB 2006 national representative sample, RIB , 2006
Having useful connections in business in justice to get a credit to get a job abroad at police at co nt center at county center for health problems at city hall
%national sample 14 17 12 17
%Roma %Roma sample 3 4 3 5
%national sample/ %Roma sample 4.7 4.5 3.7 3.2
23 21 7 36 22
9 8 3 15 13
2.6 2.6 26 2.6 2.4 1.7
Reading example: only 3% Roma have useful connections to get a job as compared Reading example: only 3% Roma have useful connections to get a job as compared with 17%, the corresponding figure for the national sample. It is especially in economic and law area that Roma have a lower network capital as compared to majority population.
14