ETHNIC MINORITIES, DISCRIMINATION AND WELL-BEING IN THE ESS

Transcription

1 ETHNIC MINORITIES, DISCRIMINATION AND WELL-BEING IN THE ESS Ko OUDHOF 1 Statistics Netherlands Division of Social and Spatial Statistics Development and Support Department PO Box JM Voorburg The Netherlands Summary: Does discrimination as perceived by ethnic minorities have a sizeable impact on their wellbeing, measured in terms of subjective life satisfaction? The data of the European Social Survey (ESS) are used to find answers to this and several related questions. Firstly, how many persons belong to the group of ethnic minorities according to various criteria? Secondly, to what degree do ethnic minorities experience discrimination as measured in the ESS? Thirdly, are disadvantageous effects of their experienced discrimination measurable? Although the ESS sample characteristics and the measurement of perceived discrimination give rise to caution, the analysis of the ESS data provides very promising results. Keywords: ethnic minorities, discrimination, well-being, satisfaction, survey, ESS Introduction Many people might at some time have experienced a situation in which they have found themselves excluded, harassed or treated differently from other persons without justification, only because of their biological, physiological or personal characteristics, their origin or language, their abilities, manifestations of belief or preferences. The risks of such experiences are however not randomly distributed among the population (Olli & Olsen, 2006). In every society specific groups can be distinguished who experience such discrimination very often. For victims of discrimination, the experience appears to have a large impact on their personal lives (Van den Berg & Evers, 2006). In this paper I shall pay special attention to ethnic minorities, as a group with a relatively high risk of experiencing ethnic or racial discrimination. The main question studied here is whether the assumed 1 The author wants to thank Ineke Stoop, Marieke Wageveld, Rik van der Vliet and Lieneke Hoeksema for their helpful comments. The views expressed in this paper are those of the author and do not necessarily reflect the policies of Statistics Netherlands

2 negative impact of ethnic or racial discrimination can be confirmed by survey data of a rather general nature from many European countries. First of all, the concept ethnic minority as a specific definition of groups in the population should be defined and measured; similarly discrimination : in which forms and under which conditions do people encounter it and how might the researcher assess this? And do ethnic minorities actually experience ethnic or racial discrimination more often? Compared with these conceptual questions, elementary questions like how many people? and to what extent? seem rather basic and mainly statistical by nature, but they are of course core questions. Some issues are more political, like is it really a major problem in all countries? and how should these problems be approached?, and will therefore be left out of account in this paper. Various questions at both the conceptual and theoretical level should be answered, and quite a lot of measurement problems should be considered before a decisive conclusion can be drawn about the empirical test of the assumed impact of discrimination. In this paper some of these research problems will only be touched upon. The cross-national survey data from the European Social Survey project (ESS) will be used for the empirical elements in the exploration of these problems. In the next section this data source will be introduced. Section 3 will focus particularly on empirical information on ethnic minorities. In section 4 the next step will be taken to define the concept of discrimination and its measurement in the ESS. Some survey results on perceived discrimination will be presented 2. After an introduction of the presented measure of well-being, the analysis and its outcomes on the impact of discrimination on well-being will presented and discussed in section 5. In the last section, the exploratory analyses of discrimination in this paper will be evaluated. ESS The European Social Survey (ESS) is a two-yearly multi-nation survey covering over thirty countries. The first round (R1) was conducted in 2002/2003, the second (R2) in 2004/ The project is funded jointly by the European Commission, the European Science Foundation and academic funding bodies in each participating country, and is designed and carried out to exceptionally high standards. The data are freely available for non-profit purposes. More information can be found on the ESS website The ESS questionnaire includes two main sections, each consisting of approximately 120 items; a 'core' module which will remain relatively constant from round to round, and two or more 'rotating' 2 The ESS definitions of both ethnic minorities and perceived discrimination were considered in more detail in Oudhof (2006b). In that paper, the gender dimension was one of the main topics, and therefore gender aspects will be largely left aside here. 3 Third round data were collected in 2006 and are scheduled to become available at the end of

3 modules, repeated at intervals. The core module aims to monitor change and continuity in a wide range of social variables, including media use, social and public trust; political interest and participation; socio-political orientations, governance and efficacy; moral, political and social values; social exclusion, national, ethnic and religious allegiances; well-being, health and security; demographics and socio-economics. Taking together R1 and R2, 27 countries have participated in the ESS, most of them in both years. Data are available for all EU member states except Cyprus, Latvia, Lithuania and Malta, and furthermore for Switzerland, Israel, Iceland, Norway, Ukraine and Turkey 4. Sampling and data collection in the ESS are carried out according to a standardised design. The target population of the ESS consists of the residential population aged 15 years and older. The data are weighted to correct for selectivity caused by the sampling design. Moreover, they are also weighted to produce a total data set for all countries in which each country is counted in proportion to its population size. For this reason 5 the weights of countries participating only in either R1 or R2 were doubled (including a small correction for population growth). Ethnic minorities Various criteria are used to define and classify ethnic minorities. Many explanations of differences between countries or regions in this respect can be traced back to their historical background. This might explain why, especially in western European countries, the differentiation according to ethnic or cultural background is often used as the equivalent of foreign background. The use of criteria like citizenship or country of birth might still be useful (for the time being) in several western European countries 6 when focusing on the social problems related to the situation of recently immigrated groups of the population (Bisogno, 2005; Franco, 2005), but it could be very inadequate in other countries, such as those in eastern Europe or the USA. Citizenship and country of birth are not adequate criteria to classify ethnic minorities in countries where they have been living for many generations (like the Roma in eastern Europe or the Afro-Americans in the USA) or where they might even be the original inhabitants (Morning, 2004). 4 Second round data for Italy and Turkey, which were collected in 2006 and thus were very late available, have not been used in the analysis, presented in this paper. 5 This step, in which the possible changes between R1 and R2 in the countries participating in only one of both years were ignored, was preferable to various alternatives. Without this weighting procedure more questionable selectivity in the totals might have been caused, because the situation in countries participating twice would contribute twice as much to the aggregated results. Another alternative would have been to use only the data of countries participating twice, which would have meant a loss of almost 15 percent of all records as well as some interesting countries. 6 Analysis of data on the social position of ethnic minorities in the Netherlands showed that citizenship has hardly any explanatory value in addition to foreign background (Oudhof, 2006a) 3

4 Up to now, hardly any comparable information has been available on the relative size of the population groups which are not defined according to citizenship or own country of birth. Unfortunately, this gap in statistical knowledge will not be filled adequately by the ESS. Apart from the limited possibilities to estimate small populations because of the ESS modest national sample sizes, specific sampling frames and selective non-response, another major obstacle arises from the population coverage by the ESS: in countries where any minority language is spoken as a first language by 5 percent or more of the population, the questionnaire has to be translated into that language. Conversely, it follows that an unknown number of persons from small minority groups who do not speak one of the main languages will probably not have participated in the ESS survey. Such systematic bias might be fatal for any objective to estimate the size of minority populations, depending on the language abilities of the minority groups. In spite of this, it is very tempting to focus on some related research questions while keeping in mind the limitations of any conclusions. The main reason for this attraction is the fact that the ESS is quite unique because it contains several questions which make it possible to confront three different classifications of ethnicity for these 26 countries in Europe. The first classification criterion is citizenship, which is the most traditional criterion in EU statistics. We shall differentiate between people who are citizens in the respective countries, and those who are not. The second classification criterion is native/foreign background, derived from country of birth of the people themselves in combination with country of birth of the parents. The three classes are native background, western foreign background, and non-western background 7. In addition, an extension is possible between first generation (foreign born) and second generation (born in the country). The third classification distinguishes between people who do and people who do not belong to an ethnic minority group in the country, as measured by a direct question. All three criteria are measured in the ESS, which makes it possible to describe their relative size (table 1). About 13 percent of the total number of respondents 8 in the ESS countries have a foreign background: 4 percent have a non-western and 9 percent a western background. One must be aware that each measure is relative to specific countries: having a Belgian background in the Netherlands 7 The classification is similar to the definition and method used in the Netherlands: persons are considered as having foreign background if one of their parents was born outside the country. One has a native background if both parents are born in the country of residence. After classifying someone as having a foreign background, the specific country (or continent) of origin is derived from the person s country of birth. The country of origin for the second generation is derived from the country of birth of the mother, if applicable; if not, the country of birth of the father is used to assign second generation persons to a specific country. Western background is defined as Europe, North America and Oceania as continent of birth, while non-western background is defined as Africa, Asia and Latin America. 8 One might add here: who are sufficiently able to express themselves in one of the main languages in the countries 4

5 counts as having a western background. Considering the criterion of citizenship: only 3 percent will be considered as foreigners, while 4 percent of the total of all people in these samples identify themselves as belonging to an ethnic minority group. The corresponding figures for the separate countries in table 1 show very large variation in the samples. A number of high percentages can be explained rather easily by recent historical developments, like the large shares of people with a foreign background (western as well as nonwestern) in Israel and the high percentages of persons with a western background and non-citizens in Estonia (Russians) 9. It is tempting to connect the high percentages of persons with a foreign background in Luxembourg and Switzerland at least partly to the relatively large number of international organisations established in these countries (considering also the high number of noncitizens). Likewise it is also somewhat speculative to relate the high percentages of non-westerners in western European countries like the UK and France to their colonial past, in addition, of course, to more recent immigration flows. 9 Sometimes respondents mention states that no longer exist as their own or their parents country of birth, e.g. former Soviet Union or former Yugoslavia. In the coding process these had to be treated in the same way as existing countries. Especially for countries like Estonia, which were part of these former states, this complicates the interpretation of data on foreign background. 5

6 Table 1. Percentages persons belonging to minority groups by various criteria, sex and country (round 1+2). % non-western % belonging to minority sample size % western background background % not citizen of country ethnic group (unweighted) male female male female male female male female male female % abs Total AT Austria BE Belgium CH Switzerland CZ Czech Republic DE Germany DK Denmark EE Estonia ² ES Spain FI Finland FR France GB United Kingdom GR Greece HU Hungary IE Ireland IL Israel ¹ IS Iceland ² IT Italy¹ LU Luxembourg NL Netherlands NO Norway PL Poland PT Portugal SE Sweden SI Slovenia SK Slowakia ² UA Ukraine ² ¹ only round 1 data ² only round 2 data 6

7 The variation should not only be explained at the population level. Another possible factor is the fact that in some countries, the UK and France for example, more people with a foreign background may participate in the survey because serious language problems may be less common there than in many other European countries 10. Eastern European countries in particular have few people with a non-western background. They do, however, have sizable percentages of people who feel they belong to a minority group in combination with zero percentages of non-citizens. Apart from some remarkable exceptions (Israel, Switzerland) no significant differences between women and men occur. Table 2. Total size of risk population groups of ethnic discrimination Belonging to minority ethnic group yes no citizenship citizenship yes no yes no Type of background: regional Type of background: generation n (weighted). Total Native background Western background non-western background first generation second generation first generation second generation total population Type of background: regional Type of background: generation row % column % Native background Western background non-western background first generation second generation first generation second generation total population This might be another, more indirect effect of the historical relations of some countries. 7

8 Even more interesting than comparing the relative size of the separate groupings according to the three criteria might be to inspect the crossings of those criteria, which would reveal the size of their overlap. These will be explained only on the weighted aggregate level of all weighted countries. Of course, these crossings would also be possible for each of the countries, but that would go beyond the scope of this paper 11 According to this sample of the population in the 26 countries, 87 percent of the total population is classified as having a native background, while 98 percent of the latter group are citizens of the country and do not consider themselves as belonging to an ethnic minority group; thus one could conclude that one in every seven people in the residential population aged over 15 years qualifies as a foreigner or belonging to an ethnic minority by any one of the three criteria. This conclusion should be qualified immediately as an underestimate because of the probable language-specific selectivity of the response. In this study we shall qualify this group as the risk population of ethnic discrimination The foreign background criterion in table 2 shows more detail than that in table 1 and also shows the difference between the first and second generations. The majority of people with a western background are born in the country of residence, and belong to the second generation. The largest group among those with a non-western background is formed by the first generation, i.e. the immigrants. For all population groups defined by the background criterion, the relative shares of those who consider themselves as belonging to an ethnic minority group in combination with citizenship were assessed. Among the largest group, i.e. people with a native background, 1.8 percent consider themselves as belonging to an ethnic minority group. Among people with a foreign background these shares are much higher, especially among the non-western group. It is also very interesting to observe that the first generation shows much higher percentages of self-perceived ethnic minority members than the second generation. Almost all people with a native background also have citizenship in the respective countries. Among the first generation foreigners - people born elsewhere - the shares of non-citizens are relatively high. Among the second generation people born in the country concerned - the noncitizen percentages are much lower, and more so among non-westerners. The fact that second generation people are more often citizens of the country of residence follows largely from the ius solis principle in a number of countries (France, UK) according to which being born in the country 11 The 2008 edition of the Labour Force Survey will provide a good opportunity to go deeper into the crossings of the dimensions foreign background and citizenship, but the dimension belonging to an ethnic minority group will not be measured. Ethnic minorities with a native background will not become visible in the LFS, unless individual countries themselves include such questions in the LFS questionnaire. 8

9 entitles people to citizenship of that country. Second generation children are generally also entitled to citizenship of the country of birth in countries that use the ius sanguinis principle if one parent was born in the country. Discrimination The elementary definition of discrimination is the violation of the basic principle that all persons should be treated equally in equal circumstances. Craig (2005) summarises the assessment of discrimination by four essential elements: 1) an individual or group is in comparison, treated or affected differently than the comparator, 2) the difference is disadvantageous to the individual or group, 3) the difference in treatment or effect is causally linked to a characteristic of the individual or group protected by antidiscrimination legislation, and 4) there is no exception or justification permitting the difference in treatment or effect. Several grounds of discrimination are explicitly mentioned in the EU legislation, particularly in the Article 13 Directives: ethnic origin, religion and belief, sexual orientation, disability, age and gender. Thus, whatever reason is used to justify different outcomes between persons in comparable situations, these grounds are explicitly unacceptable as such (Makkonen, 2005). Discrimination can manifest itself in various ways: 1) directly or indirectly, depending on the criteria causing different outcomes; and 2) culpable or systemic, depending on the possibility to assign responsibility (Olli & Olsen, 2005). Compared with other subjects in social science, the definitional problems of discrimination are complicated by the intrinsic relations of the concept with legal discussions. This might also continue to affect the discussions on the statistical measurement of discrimination, especially when this measurement focuses on the assessment of objective discrimination, considering the unjustified inequality of outcomes of specific groups (Olli & Olsen, 2005). Various research designs and measurement instruments have been used to assess discrimination, varying from experimental research in laboratory and field settings (often called situation testing) via analyses of data on inequality of outcomes, to collecting data on subjective experiences of discrimination (Olli & Olsen, 2006). Observational data might be just as useful as data sources like surveys or registers (Blank et al., 2004). The measurement instrument of discrimination in the ESS can be positioned somewhere between measuring discrimination experiences and attitude measurement. All respondents were asked the following question: Would you describe yourself as being a member of a group that is discriminated against in this country? If they answered affirmatively, the follow-up question was: 9

10 On what grounds is your group discriminated against? with the following answer categories: colour or race/ nationality/ religion/ language/ ethnic group/ age/ gender/ sexuality. After naming one category, the respondent was asked: What other grounds? No questions refer directly to the respondent s own experiences of discrimination. As these questions can justifiably be doubted as a valid measure of discrimination, it is not certain to what degree the outcome can be interpreted as indicators of the level of discrimination. Because the questions are referring to the own membership group of the respondent, I will denote the measured discrimination in this paper as perceived discrimination 12. The answers to the first question in the ESS questionnaire, on the self-assessment of whether the respondent belongs to a group that is subject to discrimination (table 3), show that 5.9 percent of the total population consider themselves as belonging to some target group of discrimination. These figures refer to discrimination in general, without any further specification of possible grounds. Breaking down the figures by the criteria used to define ethnic minorities brings large differences to light. Each of the three criteria raises the percentage of perceived membership of a discriminated group, but the self-assessment as to whether one belongs to an ethnic minority group turns out to be the most influential. The highest level of perceived discrimination is reported by persons with a second generation nonwestern background. But there are also high percentages of perceived membership of discrimination target groups among ethnic minorities with a native background and in possession of citizenship. 12 An alternative might have been assessed discrimination but this concept is generally associated with measurement instruments on more general views on the incidence of discrimination in social life, as used e.g. in the Eurobarometer (TNS, 2007) 10

11 Table 3. Perceived discrimination¹ by ethnic discrimination risk population groups Belonging to minority ethnic group yes no citizenship citizenship yes no yes no Type of background: regional Type of background: generation % Total Native background 26.5 * Western background first generation second generation non-western background first generation second generation 49.2 * total population ¹ Percentage persons (weighted) giving an affirmative answer to the question: "Would you describe yourself as being a member of a group that is discriminated against in this country?" The level of perceived discrimination can also be studied from the specification of grounds of discrimination. The answers referring to the mentioned grounds were used to construct two separate measures of discrimination: the first refers to ethnic discrimination and includes colour or race, ethnic group and nationality; the second includes the remaining grounds of discrimination: religion, language, age, gender, sexuality and other. Both measures allow multiple answers. Table 4. Perceived grounds of discrimination by ethnic discrimination risk population and number of ethnic and other grounds Number of discrimination grounds mentioned Total (=100%) Risk population ethnic discrimination? multiple discrimination ethnic grounds no yes total multiple discrimination other grounds no yes total About 10 percent of the ethnic risk population actually reports perceived ethnic discrimination as measured here, compared with 0.5 percent of the rest of the population (table 4). Multiple grounds 11

12 of ethnic discrimination are mentioned by 2 percent. However, the ethnically defined risk population also reports a higher level of perceived discrimination on the other grounds, although the difference between the two population categories is smaller. Religion and language were mentioned particularly more often by the population at risk of ethnic discrimination. Findings on the discrimination perceived by ethnic minority women can serve as an illustration of the cumulative discrimination on different grounds, gender and ethnicity, for two independent risk populations (women resp. ethnic minority). Thus, one might wonder if ethnic minority women are subject to discrimination both as women and as ethnic minority members, and if this would make a difference compared to the level of discrimination experienced by ethnic minority men (Oudhof, 2006b). In order to assess this cumulative effect of both grounds of discrimination the percentage of persons perceiving ethnic discrimination as well as gender discrimination has been assessed and compared with the percentages of the separate grounds on their own. The results of this exercise can be seen in figure 1 which shows the percentages of males and females perceiving discrimination on gender as well as ethnic grounds next to the total percentages of all males and females perceiving either gender discrimination or ethnic discrimination. Fewer women than men describe themselves as a member of an ethnic discriminated group This holds also for women in almost all ethnic minorities according to all three criteria (see also Oudhof, 2006b). However, the share of people perceiving gender as a ground of discrimination is rather small, also for women, compared to the percentages perceiving ethnic discrimination grounds. Ethnic minority women (and men) are generally more inclined to perceive gender discrimination than other women, but this does not apply to the citizenship criterion. Perceived gender discrimination among ethnic minority women is rather often joined by perceived ethnic discrimination, but the inverse proves not to be true: only a small part of women reporting perceived ethnic discrimination are reporting this in conjunction with perceived gender discrimination. How to explain this is not immediately clear, but the specific cultural background (including traditional views on gender roles) of many ethnic minorities might be a possible explanation. 12

13 Figure 1. Multiple discrimination on ethnic grounds as well as gender by ethnic discrimination risk populaton and sex. Type of background Native background Belonging to ethnic minority group not belong ethnic min yes belong ethnic min Citizenship citizen non-citizen citizen non-citizen* Western background 1st generation not belong ethnic min yes belong ethnic min citizen non-citizen citizen non-citizen Western background 2nd generation not belong ethnic min yes belong ethnic min citizen non-citizen citizen non-citizen (too low N) Non-Western background 1st generation not belong ethnic min yes belong ethnic min citizen non-citizen citizen non-citizen Non-Western background 2nd generation not belong ethnic min yes belong ethnic min citizen non-citizen (too low N) citizen non-citizen (too low N) Total ethnic discrimination male ethnic discrimination female gender discrimination male gender discrimination female ethnic + gender discrimination male ethnic + gender discrimination female 13

14 Well-being and discrimination The impact of discrimination on personal well-being has not been studied much in EU countries. Most research findings on the effects of discrimination experiences on psychological well-being, mental and physical health stems from studies in the USA. In a review on research evidence on this topic, Williams et al. (2003) conclude that consistent findings show that perceptions of discrimination tend to be associated with poorer health across a broad range of outcomes and across socially disadvantaged groups in different societies. Among the 53 empirical studies, six reported on the effects on subjective well-being measures like happiness and life satisfaction, and five of these assessed negative impact. Generally, the consequences of discrimination are similar to those of other stressors (See also Shields & Price, 2003; Wong et al., 2003). The level of well-being has been measured by the question on subjective life satisfaction. This measure has generally been used as an indicator of social well-being since the 1970s, although discussions on its reliability and validity are still ongoing (Diener, 1997; Kahneman & Krueger, 2006). The question in the ESS is formulated as follows: All things considered, how satisfied are you with your life as a whole nowadays? Please answer using this card, where 0 means extremely dissatisfied and 10 means extremely satisfied. 13. The preliminary answer to the question of whether perceived discrimination has an impact on wellbeing might be provided by the results in table 5a, which shows the mean scores on the subjective life satisfaction indicator by ethnic risk groupings, as well as perceived ethnic discrimination. The lower scores on the well-being indicator are indeed found for the persons reporting perceived discrimination compared with persons who do not consider themselves a member of a discriminated group. The mean values on the indicator are indeed still lower for those belonging to ethnic minority groups. The patterns of the differences by the other two characteristics, background and citizenship, is less clear-cut at first sight. Closer inspection of the differences by carrying out multiple classification analysis (MCA) 14 confirms the impressions concerning discrimination and belonging to minority groups. It appears that citizenship has a negative relation to life satisfaction, while the rather small differences on the 13 Many other surveys use the same question with sometimes minor variations. In the 2003 First European Quality of Life Survey (Alber et al., 2004) in EU25 and three 2003 candidate countries, the wording of the question was: All things considered, how satisfied would you say you are with your life these days, scaled from 1 very dissatisfied to 10 very satisfied. In spite of complete independence of EQLS and the ESS in questionnaire, design and data processing, the correlation of subjective life satisfaction measures between the two surveys on the aggregate level of 23 overlapping countries turned out to be.98. It is open to debate whether this correlation should be interpreted as reliability or validity. 14 MCA examines the relationships between several categorical independent variables and a single dependent variable and is thus closely related to the analysis of variance. MCA determines the effects of each predictor before and after adjustment for its 14

15 regional and generation background show larger disadvantageous effects of western than nonwestern origin. intercorrelations with other predictors in the analysis.. 15

16 Table 5a. Mean subjective life satisfaction by background, minority group, citizenship and perceived ethnic discrimination. Belonging to minority ethnic group Total yes no citizenship citizenship yes no yes no ethnic discrimination ethnic discrimination ethnic discrimination ethnic discrimination Type of background yes no yes no yes no yes no Regional Generation Native * * Western first generation Western second generation * non-western first generation non-western second generation * * total population Table 5b. Multiple classification analysis on subjective life satisfaction by background, minority group, citizenship and perceived ethnic discrimination. N Predicted Mean Deviation Unadjusted Adjusted for Factors Unadjusted Adjusted for Factors Belong to minority ethnic group in country Yes No Regional and generation background Native Western first generation Western second generation Non-western first generation Non-western second generation Citizen of country Yes No Ethnic discrimination No Yes

17 The results in table 5a en 5b might partly be explained by other differences between the groups on relevant characteristics, like different age distributions, socio-economic position or country of residence. For this reason several other variables were added to the MCA: country of residence, gender, level of education (high vs. low) and paid work status (no paid work, employee, selfemployed). Instead of the rather general dichotomous variable on perceived ethnic discrimination, the extra information in the ESS on the grounds of discrimination in the analyses was used by introducing multiple ethnic discrimination and multiple discrimination on other grounds. The results of this multivariate analysis can be found in table 6 (evaluating the variance components) and figure 1 (showing the effects from the MCA). The variables in the analysis explain 17 percent of the variance of life satisfaction. If we consider the values of beta, which can be interpreted as equivalent to the standardised regression coefficient, the largest share of the variance can be ascribed to the effect of the country of residence. The importance of multiple ethnic discrimination in relation to life satisfaction is much smaller (Eta=0.06) and is halved to 0.03 by adjusting for the effect of various background variables, among which the risk population characteristics. The role of discrimination on other grounds is more important, considering that beta is The role of the country level might be interpreted in different ways. It could be viewed as variation between countries at the macro level: could there be stable characteristics at country level which explain the differences? Or otherwise, should this be seen as variation at the individual level which should be interpreted as compositional or clustering effects of unspecified individual characteristics? Estimation of random variances according to the intercept-only model (Maas & Van Tubergen, 2006; Hox, 2002) shows that the variance of the country level errors equals 0.570, while the much larger variance of individual level errors is , thus the macro-level explains only 1 percent of the total variance. 17

18 Table 6. Summary of results of variance analysis on ' How satisfied are persons with life as a whole ' (total sample) Sum of Squares Degrees of Freedom Mean Square F Level of significance Beta (adjusted) Eta Main Effects with Covariates Country Gender Education two groups Status in paid work Belong to minority ethnic group in country Origin and generation Citizen of country Multiple discrimination ethnic grounds Multiple discrimination other grounds Year of birth (standardised covariate) Model Residual Total R square

19 Figure 2. Deviations from mean predicted value of satisfaction with life as a whole, adjusted for factors and covariates (total sample) Iceland Denmark Switzerland Finland Luxembourg Sweden Norway Netherlands Ireland Austria Belgium United Kingdom Spain Italy Slovenia Germany Israel France Greece Czech Republic Estonia Poland Portugal Hungary Slowakia Ukraine GENDER 1 Male 2 Female EDUCATION TWO GROUPS low high STATUS PAID WORK no paid work employee self-employed BELONG TO MINORITY GROUP IN COUNTRY Yes No ORIGIN AND GENERATION native western first generation western second generation non-western first generation non-western second generation CITIZEN OF COUNTRY Yes No MULTIPLE DISCRIMINATION ETHNIC GROUNDS MULTIPLE DISCRIMINATION OTHER GROUNDS

20 The adjusted effect sizes of the factors on subjective life satisfaction are visible in figure 2. Rather high satisfaction levels can be found in the northern European countries, while the lowest levels are measured in eastern European countries and Portugal. Socio-demographic characteristics have rather small effects 15 and although the effect sizes of the risk population characteristics are still visible and consistent with the effects in table 5, they are reduced. The impact of perceived discrimination on ethnic grounds as well as the stronger effect of discrimination on other grounds are clearly present in the results. But there is no consistent relation between the number of grounds of perceived ethnic discrimination and a lower level of life satisfaction. One way to account for the possible effect of interaction between various socio-demographic and risk population characteristics would have been to bring interaction variables into the analysis. This would have led to small numbers of observations in many cells. An alternative way to inspect the minority specific effects is to carry out the same analysis for only the risk population, as defined by the three defining characteristics mentioned above. The results are presented in table 7 and figure 3, and are very similar to the results for the total population. About 15 percent of the variance is now explained, and the significance of ethnic discrimination in particular is now somewhat larger (beta=0.08), while the role of country is somewhat smaller (0.35) but nevertheless dominant. The adjusted deviations of most countries from the predicted scores of life satisfaction of these ethnic minority groupings are strikingly similar to those of the total population: minor differences can be seen for the UK, Italy and Slovenia. The effect sizes of the socio-demographic and risk population characteristics are likewise hardly different from the results in the analysis above on the data of the total sample. The effects of ethnic discrimination on life satisfaction are almost identical to the effects reported on the total population, but the cumulative effects of more grounds within discrimination on other grounds are now better visible. As mentioned above, the effect of discrimination on other grounds among ethnic minority groups might be ascribed partly to the grounds religion and language. 15 The effect of year of birth, for which non-standardised regression coefficient = 0.07 (higher satisfaction when younger) should be mentioned here, too. 20

21 Table 7. Summary of results of variance analysis on ' How satisfied are persons with life as a whole' (minority population) Sum of Squares Degrees of Freedom Mean Square F Level of significance Beta (adjusted) Eta Main Effects with Covariates Country Gender Education two groups Status in paid work Belong to minority ethnic group in country Origin and generation Citizen of country Multiple discrimination ethnic grounds Multiple discrimination other grounds Year of birth (covariate) Model Residual Total R square

22 Figure 3. Deviations from mean predicted value of satisfaction with life as a whole, adjusted for factors and covariates (ethnic minority groups in sample) Iceland Denmark Finland Switzerland Luxembourg Sweden Italy Norway Netherlands Ireland Belgium Austria Spain United Kingdom Slovenia Germany Israel France Greece Czech Republic Poland Estonia Hungary Portugal Slowakia Ukraine GENDER 1 Male 2 Female EDUCATION TWO GROUPS low high STATUS PAID WORK no paid work employee self-employed BELONG TO MINORITY GROUP IN COUNTRY Yes No ORIGIN AND GENERATION native western first generation western second generation non-western first generation non-western second generation CITIZEN OF COUNTRY Yes No MULTIPLE DISCRIMINATION ETHNIC GROUNDS MULTIPLE DISCRIMINATION OTHER GROUNDS

23 Conclusion The European Social Survey is especially designed to measure internationally comparable attitudes and opinions on various policy issues, like prejudice on ethnic issues. Unfortunately, it contains only a few questions that might be related to discrimination experienced on the grounds of ethnic background. Nevertheless, the results of these few questions proved to be a rich source of information on the subject. The ESS questionnaire contains some questions on the socio-demographic background of the people in the sample, which made it possible to describe the relative size and the overlap of ethnic and foreign subgroups in the samples according to different definitional criteria in 26 countries, including almost all EU member states. Unfortunately, it is not possible to interpret these as valid estimates of the subpopulations in the countries, because particularly the language-related selectivity of the samples caused a systematic bias of the data. In spite of this, it is valuable to have some information on the relations among these criteria as background information to interpret the effects of grouping criteria. It would be very important to have data sources available that provide valid internationally comparable data on the various classification criteria. This would unveil very important information for discussions on policy issues with respect to the position of minorities. Such information would also be very useful to decide on the design of research projects that require internationally comparable measurement of such ethnic subpopulations (including not only immigrants and their descendents, but also national minorities). The ESS data also made it possible to give a general statistical description of ethnicity-related discrimination and to show the strength of the association between measures of discrimination on various grounds and various criteria used to define and classify minorities. Some doubts on the validity of the measure of discrimination as referring to actual discrimination experiences might be justified. The question on discrimination in the ESS questionnaire refers to selfassessment (...describing yourself as a member ) in combination with an opinion or attitude (...of some group that is discriminated against ). It does not refer directly to personal experiences with discrimination, but the self-identification element makes it more than just an opinion. This suggests that the answer might stem from very different sources, not only personal experiences or perceived experiences, but also fear, prejudice or only intensified attention to the issue in policy or media 16. Although the relatively high numbers of affirmative 16 This interpretation could be a plausible explanation for the high scores on gender discrimination in Sweden or Iceland (Oudhof, 2006b). 23

24 answers to this question have the advantage that fewer statistical-technical problems arise, this is at the same time its weakness, if it cannot be interpreted as a valid indicator of the risk of discrimination 17. For validation purposes it might be useful to add one or more small questions to respondents on the frequency of their own personal discrimination experiences. The main question in this paper concerned the impact of perceived discrimination, as measured in the ESS, on well-being. The question was inspired by a growing number of studies which report the negative impact of discrimination on health and well-being. In the analyses in this paper subjective life satisfaction was used as well-being indicator. In spite of the doubts on the measurement qualities of perceived discrimination, there was evidence of significant and consistent negative effects on the life satisfaction indicator. Although the results are presented in terms like effects and explanation, one should still be aware of the interpretational limits of survey data: the interpretations in this paper actually refer to the statistical relation between reported perceptions of discrimination and reported well-being, as measured at one point in time. For that reason, alternative interpretations that well-being has affected the perception of discrimination or that both phenomena can be explained by other neglected factors are difficult to contradict. But what remains is that the presented statistical relations are consistent with various other studies on the relation between discrimination and well-being. The assessed relations between perceived discrimination and life satisfaction were stronger than those of many socio-economic and achievement-related characteristics like education or paid work status. Large differences in life satisfaction also appeared between countries, which were consistent with results of quality of life surveys. This consistency diminishes the probability of design effects as an explanation for between-country variation 18. Still it is rather surprising that the country effects among minority groups were almost identical to those among the total population. Considering the results of these analyses, one might conclude that there is more than enough reason to include items on perceived or experienced discrimination in surveys on quality of life, health or (non-)participation on the labour market. To act on such intentions, one should also take initiatives to develop and introduce standardised tools to measure risk populations and discrimination (Olli & Olson, 2006). This will not happen automatically, but will demand 17 The same considerations are applicable to the type of questions used in the Eurobarometer 2006 on discrimination issues, which refer to assessed discrimination: all respondents are asked to indicate to which degree ethnic minorities are at risk of being the victim of discrimination (TNS, 2007; Olli & Olsen, 2006). 18 Other reasons to consider design effects as less plausible are the intensive efforts of the ESS organisation to monitor carefully the uniformity in design, data collection and processing in participating countries, as well as the assessed small share of variance between countries on the macro level. 24

25 common efforts of users and producers, policy officials, NGO s and researchers on an international scale. Literature Alber, Jens, Jan Delhey, Wolfgang Keck & Ricarda Nauenburg, 2004, Quality of Life in Europe, First Quality of Life Survey 2003, European Foundation for the Improvement of Work and Lving Conditions, Office for Official Publication of the European Communities, Luxemburg, Bisogno, Enrico, 2005, A more comprehensive approach to defining stocks relevant for international migration, paper presented at the 28th CEIES seminar Migration Statistics, Riga 9 and 10 June 2005 Blank, Rebecca M., Marilyn Dabady, and Constance F. Citro (ed.), 2004, Measuring Racial Discrimination.Panel on Methods for Assessing Discrimination. National Research Council, Committee on National Statistics, Division of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press. Craig, R.,2005, Systemic Discrimination in Employment and the Promotion of Ethnic Equality, University of Oslo, Oslo. Diener, Ed, Eunkook Suh and Shigehiro Oishi, 1997, Recent Findings on Subjective Well-Being, in: Indian Journal of Clinical Psychology, March Franco, Ana, 2005; Using the European Union Labour Force Survey to get information on migrants and their descendants, paper presented at the 28th CEIES seminar Migration Statistics, Riga 9 and 10 June 2005 Hox, Joop, 2002, Multilevel Analysis, Techniques and Applications, Lawrence Erlbaum Ass.,Inc., New Yersey. Kahneman, Daniel, & Alan B. Krueger, 2006, Developments in the Measurement of Subjective Well- Being, in: Journal of Economic Perspectives, vol.20, no.1, Maas, Ineke, & Frank van Tubergen (2006), Youth unemployment among immigrants and native background population in fifteen European countries (only on dutch), in: Van Tubergen & Maas, Allochtonen in Nederland in international perspectief, boekaflevering Mens en Maatschappij Makkonen, Timo, 2005, Data as an Opportunity to Promote Equal Rights, in: Simo Mannila (ed), Data to Promote Equality, Proceedings of the European Conference, Finnish Ministry of Labour, Helsinki 25