RC33 Aug Lambert1 Ethnicity and the Comparative Analysis of Contemporary Survey Data Paul S. Lambert Stirling University, UK Paper prepared for the VIth International RC33 conference, Amsterdam, August 2004
RC33 Aug Lambert2 Ethnicity and the Comparative Analysis of Contemporary Survey Data This paper discusses issues that arise in the cross-nationally comparative analysis of social survey data when interest concerns ‘ethnicity’ and related concepts. A number of both practical and theoretical problems arise, and a review of data in a selection of cross-nationally harmonised surveys {ESS; ISSP; WVS; LIS} reveals that current resources are a long way from satisfying the requirements of most social science analysts. However, it is argued that such problems are not an excuse for abandoning the comparative analysis of ethnic differences, and examples of alternative analytical strategies are reviewed for their adequacy.
RC33 Aug Lambert3 Ethnicity and comparative analysis 4 well-known problems: Are there more; what are our options? A steps-in-survey-methods approach: 1)Data Collection 2)Variable Operationalisations 3)Data Analysis ‘Referents’ Sparsity Dynamisms National traditions
RC33 Aug Lambert4 1) Data Collection [ Bias / prejudice / interview context –nothing to make ethnicity a special case? ] Survey design: pre-harmonised v’s ex-post Population representation / sparsity: –Absolute numbers - minorities of eg < 5% –Substructures within minorities –Missing data / non-response
RC33 Aug Lambert5 Sparse representation of minority gps: Some common solutions: pool/boost data; merge / abandon categories; missing data models;.. – but national specific and/or primary adjustments ESSISSPWVSLIS approx: average over countries % majority # minority % missing1213 / 0
RC33 Aug Lambert6 2) Variable operationalisations 2a) Competing referents Real data: many sparse and/or uninteresting categories.. Can achieve conceptual clarity (eg H-Z 03), but choices must balance theoretical prefs v’s practical options ESSISSPWVSLIS Citizenship {}{}{}{} Ctry of birth {}{} {}{} Time in ctry {}{} Parents ctry Language {}{}{}{}{}{} Subjective (dichot){}{}{}{}{}{} Religion
RC33 Aug Lambert7 2b) Dynamism Changing migration inflows; mixed identities; intermarriage; etc… Outcome: categories frequently revised / expanded Solutions: none in existing survey treatment as categorical data V. powerful political / operational forces 2c) National Traditions Britain : ethnic / racial groupGermany : citizenship N. Irel : religious denom.Switzerland : canton / language
RC33 Aug Lambert8 3) Variable Analysis Often neglected element of research Pressure to collapse categories / reduce data Issues: Pooled or separate cross-national analyses Ethnic differences as focus v’s background Often our only sphere of influence Typically: rich data collected; collapsed to minority/majority dichotomy in analysis
RC33 Aug Lambert9 So: Diversity, sparsity, dynamism & national context are recognised, & celebrated in sociological theories, but problematic for X-N survey research Retain all categorical boundaries: –Abandon (& attack) surveys for ethnicity research –Restriction to country-by-country survey comparisons Not always viable (eg sub-projects) Still leads to same problems eventually (eg regions) Manipulate categorical boundaries: –Absolutist: prioritise some divisions only –Relativistic: weight / assess categorical distinctions
RC33 Aug Lambert10 i) An ‘absolutist’ solution? Traditional: immigrant v’s not-immigrant –But ignores many other differentiations A summary ‘EC9’ measure (ESS valid %) : 4 dichot differentiations:CPLV – 78.6Cplv – 0.3 C - Born in host countryCPLv – 2.2 cLV – 3.2 P - Parents from host ctryCPlv – 1.5cLv – 3.3 L - Speaks host languageCpLV – 5.5 clv – 3.0 V - ‘Visible’ minority (id; rel)CpLv – 2.3Missing – 9.2
RC33 Aug Lambert11 ‘EC9’ properties: –Captures most but not all differentiations –Not always fully operable & no official derivations –Decent numbers in most categories Relative attractions: –High associations with many other vars –Categorical nature - communicable
RC33 Aug Lambert12 ii) A ‘relativist’ solution? Existing problems centre on categorical boundaries and distinctions (too many..) Can we limit role of boundaries but keep info? How? Single source metric(s) – eg years in country; Summary function (eg, theoretical placement; empirical derivation - ‘SORs’) Proposal : quantitative scores to indicate relative locations of ethnic categories in a low dimension space of ethnic differences
UK example : ‘SOR’ scores describe ‘persistent diversity’ (Lambert 2002)
RC33 Aug Lambert14 Cross-national examples? Reference to a consistent metric brings conceptual comparability? Derivation of metrics can be national specific, using large data sources and expertise Some examples (eg LIS SORs) sensible metrics which explain most ethnic difference –(could further combine with specific dummies) ESS examples – underwhelming patterns
RC33 Aug Lambert15 Conclusions Major and obvious problems with X-national ethnicity research But can’t ignore : comparisons will still be made Need for: Prescriptions on data constructions and categorical formulations Options for handling awkward categorical data – eg quantitative scores