New Research on Migration, 6 March 2012

Immigration and Public Opinion in Europe 2002-2012 Ian Preston University College London, Centre for Research and Analysis of Migration @IanPPreston @CReAM Research (drawing on joint work with David Card and Christian Dustmann) 1

Introduction Use large cross-European attitudinal dataset to assess the factors driving immigration Evaluate the nature of changes in attitudes over the last decade Partly based on: • D. Card, C. Dustmann and I. Preston, ”Immigration, Wages, and Compositional Amenities”, Journal of the European Economic Association, 2012

Partly work in progress on more recent data

2

European Social Survey: Round 1, Immigration policy Conducted in 21 countries with about 40,000 respondents in 2002 Special immigration module, design team leaders based at CReAM Biggest issues are cross-cultural comparability and neutrality of phrasing Asks opinion on immigration policy: ”To what extent do you think [this country] should allow people to come and live her?” separately with respect to immigrants from • rich/poor European/non-European countries • same/different ethnicity as majority 3

Table 1: Distributions of Opinions on Immigration Policy How Many Allowed to Immigrate Mean

Std Dev

None

Few

Some

Many

Rich European Countries

2.72

0.85

9.9

32.6

43.3

14.2

Poor European Countries

2.57

0.81

9.0

36.8

42.8

11.4

Rich non-European Countries

2.53

0.84

11.3

36.2

41.1

11.4

Poor non-European Countries

2.49

0.82

10.8

39.0

40.1

10.0

Same Race/Ethnicity as Majority

2.73

0.80

6.3

30.8

47.1

15.8

Different Race/Ethnicity from Majority

2.48

0.82

11.3

38.9

40.0

9.7

People from:

Source: ESS Answers on scale 1-4, higher values more open to immigration

4

Distributions of Opinions on Immigration Policy Opinions more open to richer immigrants, European immigrants and immigrants of same ethnicity but not dramatically so

5

European Social Survey: Effects of immigration Also asks about broad effects of immigration: • Would you say it is generally bad or good for [this country’s] economy that people come to live here from other countries? • Is [this country] made a worse or better place to live by people coming to live here from other countries?

6

Table 2: Distributions of Opinions on Effects of Immigration Responses on 10 point scale Mean

Deviation

0-1

2-4

5

6-8

9-10

Good/Bad for the Economy

4.97

2.36

9.3

25.0

28.1

32.0

5.5

Makes Country Better/Worse Place to Live

4.77

2.18

8.3

28.1

34.6

24.8

4.3

Immigration:

Source: ESS Answers on scale 0-10, higher values more positive to immigration

7

Table 3: Cross Country Differences in Opinions Allow Immigrants Good/Bad Make Better/Worse for Economy Place to Live All Countries 0.52 0.50 0.48 Germany 0.55 0.51 0.48 Spain 0.54 0.54 0.47 France 0.49 0.51 0.45 Greece 0.36 0.36 0.34 Italy 0.59 0.53 0.45 Ireland 0.57 0.50 0.53 Poland 0.57 0.45 0.52 Sweden 0.69 0.55 0.62 United Kingdom 0.49 0.44 0.46 Source: ESS

8

9

Cross Country Differences in Opinions, 2002 Greece , Hungary particularly hostile to immigration Sweden, Switzerland particularly liberal Attitudes to impact on economy and quality of life strongly correlated across (as well as within) countries

10

European Social Survey: Economic concerns • Do you agree or disagree that wages and salaries are brought down by immigration? • Do you agree or disagree that immigrants harm the economic prospects of the poor? • Do you agree or disagree that immigrants help to fill jobs where there are shortages of workers?

• Would you say that immigrants generally take jobs away from natives or help create new jobs?

• On balance do you think that immigrants take out more (in health benefits or welfare services) than they put in (in taxes)?

11

Table 4: Correlations with Economic Concerns Correlation Mean

Immig Policy

Good/Bad for Economy

Better/Worse Place

Wages brought down

0.49

0.33

0.34

0.34

Harm poor

0.43

0.35

0.37

0.38

Fill jobs where shortage

0.63

0.17

0.26

0.19

Take away jobs

0.45

0.36

0.52

0.47

Take out more than put in

0.42

0.32

0.52

0.45

Indicators of Economic Concerns:

Source: ESS All answers scaled to lie between 0-1, higher values more positive to immigration

12

Correlations with Economic Concerns Differing dimensions of economic concern mainly fairly similarly associated with openness to immigration Effect on skill gaps seen as positive but the one aspect very weakly associated with attitudes to policy Effects on jobs and public finances most strongly associated with overall opinion on effect on economy and quality of life

13

European Social Survey: Sociocultural concerns • Do

you agree or disagree that it is better for a country if everyone shares the same customs and

traditions?

• Do you agree or disagree that it is better for a country if there is a variety of different religions? • Do you agree or disagree that it is better for a country if everyone can speak one common language? • Would you say that a country’s cultural life is undermined or enriched by the presence of immigrants? • Do you agree or disagree that a country should stop immigration if it wants to reduce social tension?

14

Table 5: Correlations with Sociocultural Concerns Correlation Mean

Immig Policy

Good/Bad for Economy

Better/Worse Place

Common customs/traditions

0.41

0.33

0.31

0.35

Common religion

0.51

0.25

0.25

0.29

Common language

0.17

0.10

0.08

0.12

Immigrants undermine cultural life

0.58

0.42

0.56

0.61

Immigrants create social tensions

0.47

0.46

0.43

0.45

Immigrants worsen crime

0.31

0.31

0.38

0.47

Indicators of Sociocultural Concerns: Better to have:

Source: ESS All answers scaled to lie between 0-1, higher values more positive to immigration

15

Correlations with Sociocultural Concerns Homogeneity of culture and religion more important than language and also more associated with attitudes to policy Concerns about cultural effects and social tensions fairly strongly associated with hostility to immigration

16

Factor model for opinions Opinions y and answers to indicator questions z hypothesised to be driven by common factors f of smaller dimension than z, given characteristics X y = Λf + αX + µ z = Mf + CX + ν f = BX + ω

17

Identification Assuming • correlation between indicators and between opinions and indicators, given X, is channeled solely through the factors • we can identify enough indicators exclusively with specific channels then we can estimate • the importance of these channels to opinions • the correlation between the factors

18

Table 6: Summary of Baseline Estimates λ1 Economic concerns λ2 Sociocultural concerns Fraction of V ar[y|X] explained by factors:

Immig Policy

Good/Bad for Economy

Better/Worse Place

0.027

0.122

0.048

(0.003)

(0.003)

(0.003)

0.102

0.038

0.101

(0.003)

(0.003)

(0.003)

0.315

0.504

0.496

Estimated correlation of f1 and f2 : 0.79 Source: ESS Standard errors in parentheses Sample size is 29,036

19

Summary of Baseline Estimates Both economic and sociocultural channels appear relevant to opinions on policy but the latter more so One standard deviation difference along sociocultural dimension associated with about four times the effect on openness to immigration of one standard deviation difference in economic dimension Economic dimension (reassuringly) more strongly associated with opinion on economic effect while sociocultural dimension more strongly associated with effect on quality of life Opinions in the two dimensions strongly correlated

20

Table 7: Explaining Variation in Attitudes toward Different Potential Immigrant Groups Estimates of λ Economic Sociocultural Average of 4 Country Groups 0.027 0.102 (0.003) (0.003) People from: Rich European Countries 0.029 0.079 (0.004) (0.004) Poor European Countries 0.030 0.111 (0.004) (0.004) Rich non-European Countries 0.023 0.095 (0.004) (0.004) Poor non-European Countries 0.024 0.124 (0.004) (0.004) Source: ESS Standard errors in parentheses Sample size is 29,036

21

Explaining Variation in Attitudes toward Different Potential Immigrant Groups Economic concerns fairly similarly related to attitudes to immigration from all sources - slightly stronger for Europeans but barely significantly so Sociocultural concerns stronger influence for non-European sources and particularly for immigration from poorer countries

22

Table 8: Summary of Baseline Estimates, by Immigrant Status Immig Policy

Good/Bad for Economy

Better/Worse Place

Estimates for Non-immigrants Only (n=26,914) λ1 Economic concerns λ2 Sociocultural concerns

0.026

0.120

0.048

(0.004)

(0.003)

(0.003)

0.103

0.039

0.100

(0.004)

(0.003)

(0.003)

0.030

0.137

0.040

(0.013)

(0.014)

(0.012)

0.085

0.026

0.108

(0.013)

(0.015)

(0.013)

Estimates for Immigrants Only (n=2,122) λ1 Economic concerns λ2 Sociocultural concerns

Source: ESS Standard errors in parentheses

23

Summary of Baseline Estimates, by Immigrant Status Estimating separately for immigrants and non-immigrants produces strikingly similar pattern of associations Importance of economic channel comparable for the two groups Sociocultural concerns possibly a little less important in determining attitudes of immigrants themselves, though difference not particularly well determined

24

Table 9: Summary of Baseline Estimates, by Country Economic

Sociocultural

France

-0.029

(0.026)

0.156

(0.027)

Germany

0.047

(0.011)

0.088

(0.011)

Greece

0.021

(0.021)

0.089

(0.022)

Ireland

0.043

(0.015)

0.082

(0.015)

Italy

0.035

(0.018)

0.113

(0.018)

Poland

0.063

(0.015)

0.053

(0.015)

Spain

-0.012

(0.021)

0.154

(0.022)

Sweden

0.022

(0.014)

0.099

(0.013)

Sweden

0.007

(0.012)

0.105

(0.012)

United Kingdom

0.050

(0.013)

0.097

(0.013)

Source: ESS Standard errors in parentheses

25

Cross-Country Comparison of Weightings

26

Summary of Baseline Estimates, by Country Estimating separately by country shows the economic dimension to be more important more or less everywhere Poland is the only country where the two weightings are estimated to be comparable

27

Decomposition of attitudinal differences If we are prepared to assume that • influence of observed characteristics on the indicators is channelled wholly through the factors then we can break down differences in immigration opinion associated with observed characteristics Plausibility of arguments regarding the source of such associations can be assessed

28

Table 10: Decomposition of Attitudinal Differences Immig Policy Good/Bad for Economy Decomposition of Age Gap (Age > 60 vs. Age < 30) Total estimated gap -0.071 -0.019 Gap attributed to economic concerns -0.006 -0.025 Gap attributed to sociocultural concerns -0.047 -0.018 Decomposition of Education Gap (tertiary vs. lower secondary) Total estimated gap 0.134 0.124 Gap attributed to economic concerns 0.020 0.091 Gap attributed to sociocultural concerns 0.095 0.035 Decomposition of Unemployment Gap (unemp vs. employed) Total estimated gap -0.030 -0.035 Gap attributed to economic concerns -0.009 -0.040 Gap attributed to sociocultural concerns -0.014 -0.005 Decomposition of Urban Gap (large city vs. rural) Total estimated gap 0.028 0.030 Gap attributed to economic concerns 0.003 0.013 Gap attributed to sociocultural concerns 0.021 0.008

Better/Worse Place -0.043 -0.010 -0.047 0.102 0.036 0.094 -0.030 -0.016 -0.014

0.023 0.005 0.021 Source: ESS Standard errors in parentheses Sample size is 29,036

29

Decomposition of Attitudinal Differences The greater hostility to immigration of the less educated is sometimes argued to be evidence of the importance of economic explanations of attitudes But educational differences in sociocultural attitudes seem a more empirically compelling explanation The same is true of, for example, differences associated with age and between urban and rural respondents

30

Three factor model, international concerns • Do you agree or disagree that richer countries have a responsibility to accept people from poorer countries?

• Do you agree or disagree that all countries benefit if people can move where their skills are most needed?

• Do you think that immigration is good or bad for the sending countries in the long run?

31

Table 11: Summary of Baseline Estimates, Three Factor Model λ1 Economic concerns λ2 Sociocultural concerns λ3 International concerns

Immig Policy

Good/Bad for Economy

Better/Worse Place

0.023

0.123

0.047

(0.004)

(0.004)

(0.003)

0.088

0.034

0.096

(0.004)

(0.004)

(0.003)

0.034

0.005

0.011

(0.004)

(0.002)

(0.002)

Estimated correlation of f1 and f2 : 0.79 f1 and f3 : 0.40 f2 and f3 : 0.47 Source: ESS Standard errors in parentheses Sample size is 27,302

32

Table 12: Decomposition of Attitudinal Differences, Three Factor Model Immig Policy Good/Bad for Economy Better/Worse Place Decomposition of Age Gap (Age > 60 vs. Age < 30) Total estimated gap -0.071 -0.020 -0.043 Gap attributed to economic concerns -0.005 -0.026 -0.010 Gap attributed to sociocultural concerns -0.042 -0.016 -0.045 Gap attributed to international concerns 0.005 0.001 0.002 Decomposition of Education Gap (tertiary vs. lower secondary) Total estimated gap 0.131 0.120 0.098 Gap attributed to economic concerns 0.017 0.089 0.034 Gap attributed to sociocultural concerns 0.081 0.032 0.088 Gap attributed to international concerns 0.003 0.001 0.010 Source: ESS Standard errors in parentheses Sample size is 27,302

33

Summary of Baseline Estimates, Three Factor Model Adding a channel capturing international altruism suggests that this may be an important influence A one standard deviation difference on attitudes in such a dimension is associated with a greater difference in openness to immigration than a similar difference in opinions on economic effects Opinions in this dimension seem nonetheless very weakly associated with overall opinions on effects on the economy and quality of life in the receiving country Also unimportant in explaining differences by age or education

34

European Social Survey: Rounds 2-6 Retains a subset of immigration questions For example, questions on support for immigration but only for • poor non-European countries • same/different ethnicity to majority Also questions on effects but only on • economy • quality of life • culture

35

Cohort comparisons Populations are changing We can group respondents into cohorts by date of birth (ten year bands) to track average responses within consistently defined populations Tempting to connect these up into lifetime attitudinal paths but need to recognise role of between-cohort differences in characteristics Even after controlling for characteristics, time, age and cohort effects are difficult to separate

36

37

38

39

Attitudes in the United Kingdom 2002-2012 Openness to immigration declines in all cohorts except the youngest (born in the 70s) Older generations are more hostile but a large part of this is explicable by differences in, for example, education

40

Cross country comparisons We focus on six countries with different experiences • GB: United Kingdom • DE: Germany • CH: Switzerland • IE: Ireland • ES: Spain • PL: Poland

41

42

43

44

45

Cross Country Differences Trends in opinion are often very similar across generations within countries but very different across countries Openness to immigration declines in GB, CH and IE but improves, particularly in later years, in DE, ES, and across the whole period in PL After financial crisis, opinion becomes • more favourable on economic effects in DE but less so in IE and ES • more favourable on cultural effects in DE and ES but less so in IE

46

Conclusion Hostility to immigration has economic and sociocultural aspects but the latter is more important This is true across many European countries Differences associated with education, age and so on have more to do with differences in sociocultural than economic opinion Complicated cross-country pattern in paths of opinion over the last decade

47