WPS 16-02-1
Working Paper Series
Individual attitudes towards
migration: reconciling opposing
views
Tobias Müller
Silvio H. T. Tai
February 2016
Individual attitudes towards migration:
reconciling opposing views∗
Tobias Müller† and Silvio H. T. Tai‡
February 10, 2016
Abstract
There is disagreement in the literature about the determinants of attitudes toward immigration. Some authors emphasize the role of economic motivations, others argue that attitudes
are mostly determined by individual beliefs. This paper reconciles the two positions. We
estimate a structural model of individual attitudes towards immigration, accounting for unobserved individual beliefs, and use this model to carry out a decomposition analysis of attitudes
in 20 European countries. We find that economic mechanisms are significant determinants
of attitudes but that other (non-economic) factors play a more decisive role in the relation
between individual education levels and attitudes to immigration.
JEL classification: F22, J61.
Keywords: Attitudes Towards Immigrants, Political Economy, Welfare State, Labor Market.
∗
A preliminary version of this paper circulated under the title “Individual attitudes towards migration: a reexamination of the evidence”. Financial support by the TOM (Transnationality of Migrants) Marie-Curie research
and training network is gratefully acknowledged. We would like to thank Michel Beine, Frédéric Docquier, Jaime de
Melo, Marcelo Olarreaga and Frédéric Robert-Nicoud for very useful comments and suggestions on a previous draft
of the paper.
†
Geneva School of Economics and Management (GSEM), University of Geneva, 40 boul. du Pont-d’Arve, 1211
Geneva 4, Switzerland. Email: [email protected]
‡
PPGE/PUCRS, Pontifical Catholic University of Rio Grande do Sul and RITM, University Paris-Sud 11. Email:
[email protected]
1
Introduction
Although migration has been the neglected dimension of globalization, its importance is rising
fast. In Europe, many countries have experienced important immigration flows in recent years
and a large share of new jobs is occupied by immigrants. The foreign-born share is now higher in
many European countries than in the United States.1 These trends can be expected to continue
in the future, with growing migration pressure on the supply side and increasing needs for young
workers in ageing societies. However, public opinion is not very favorable to further immigration
in many European countries. For policy makers, it is crucial to understand the underlying causes
of individual attitudes towards immigration. Are they mainly due to fears about labor market
competition? Are natives seeing immigration as a threat to the welfare state? Or are these attitudes
predominantly based on individual beliefs and cultural values that are unrelated to the economic
consequences of immigration?
These questions are the subject of a lively controversy in the literature about the determinants
of attitudes towards immigration. One strand of the literature emphasizes the role of economic
motivations and distinguishes labor market and welfare state channels. According to the first
channel, the skill level of immigrants relative to that of natives influences the natives’ receptiveness
toward immigration (Scheve and Slaughter, 2001; Mayda, 2006; O’Rourke and Sinnott, 2006).
Natives are more receptive to immigrants whose skills are complementary to their own (e.g., highskill natives are in favor of low-skill immigration). The second channel is analyzed by Hanson et
al. (2007) and Facchini and Mayda (2009) who argue that individual attitudes also depend on the
expected impact of immigration on the tax-benefit system in modern welfare states. In particular,
low-skill immigrants represent a burden especially for high-income natives in a redistributive system
where a certain level of assistance is guaranteed by the state.
By contrast, another strand of the literature argues that attitudes toward immigration are mostly
determined by individual values and beliefs (Citrin et al., 1997; Hainmueller and Hiscox, 2007, 2010).
According to this view, the correlation between education and individual attitudes is predominantly
1
According to OECD (2015), the foreign-born share in 2013 was higher in Luxembourg (42.6%), Switzerland
(28.3%), Austria (16.7%), Ireland (16.4%), Slovenia (16.1%), Sweden (16.0%), Belgium (15.6%), Norway (13.9%),
Spain (13.4%), Germany (13.3%) than in the United States (13.1%).
1
explained by the fact that more educated individuals value cultural diversity and tolerance and
exhibit lower levels of ethnocentrism. Using new survey data for the U.S., Hainmueller et al. (2015)
argue that the labor market mechanism does not seem to play a role in the explanation of attitudes
towards immigration since high-skilled immigration is preferred over low-skilled immigration by
natives of all skill levels. In a recent survey of the literature, Hainmueller and Hopkins (2014, p.
227) state more generally that “there is little accumulated evidence that citizens primarily form
attitudes about immigration based on its effects on their personal economic situation”.
Our paper proposes a novel approach that sheds new light on this debate. We estimate a
structural model of individual attitudes towards immigration, accounting for unobserved individual
beliefs about the desirability of immigration. The model is then used to carry out a decomposition
analysis of attitudes towards immigration and evaluate the relative importance of economic motivations. The paper contributes to the literature in three respects. First, we address the concerns of the
second strand of the literature by controlling for unobserved individual-specific beliefs and cultural
values in our empirical analysis. Thus we ensure that our estimates of the relation between education
levels and attitudes to immigration do not simply reflect the fact that more educated individuals
are more tolerant toward other cultures. Second, our econometric estimates rely on a structural
model that accounts for labor market and welfare state effects of immigration. Our measures of
immigrants’ relative skill levels and marginal tax rates are consistent with the model and are based
on better data than past contributions. This approach enables us to recover a structural parameter
(the elasticity of substitution between human capital and raw labor) which governs the labor market effects of immigration. Comparing the estimated substitution elasticity with other estimates
from the empirical literature on the labor market effects of immigration provides an important
consistency check of our results. Third, we use the structural model to decompose the relationship
between education levels and attitudes towards immigration into three components: labor market
effects, welfare state effects and other (non-economic) effects. This decomposition method enables
us to evaluate the relative importance of the economic and non-economic determinants of attitudes.
We use data for 20 European countries from the first wave of the European Social Survey (2002)
which included a special module on immigration. The economic model predicts that the relation
2
between an individual’s level of education and her attitudes depends on the relative skill level of
immigrants (labor market mechanism) and its interaction with the marginal tax rate (welfare state
mechanism). Our empirical test of these economic mechanisms exploits the fact that in our dataset,
each individual answers four questions on the desirability of immigration from different regions of
origin (rich or poor countries in Europe or outside Europe). This particular feature of the data
enables us to account for unobserved individual-specific beliefs about the desirability of immigration
by using random-effects and fixed-effects models. In the latter, the economic mechanisms are
identified solely by the within-individual variations of attitudes towards immigrants from different
origins. This test of the relevance of the economic mechanisms is much more demanding than the
tests carried out in other contributions.2
Our results reconcile the two strands of the literature and rehabilitate the role of economic
mechanisms. On the one hand, we find that labor market threats and worries about the welfare
state both matter for individual attitudes towards immigration, even if we account for unobserved
individual beliefs in a fixed-effects logit model. Moreover, our estimate of the elasticity of substitution between human capital and raw labor is consistent with other estimates found in the labor
economics literature. These results lead us to conclude that economic mechanisms matter for attitudes towards immigration, especially when individuals are asked to choose between different types
of immigration.
On the other hand, the results of our decomposition analysis also lend support to the argument
that other (non-economic) factors play a more decisive role in the relation between individual education levels and attitudes to immigration. Three factors seem to be at play. First, there is a strong
relationship between the level of education and individual beliefs about immigration that is not
explained by economic mechanisms but strongly correlated with beliefs about the cultural impact
of immigration. Quantitatively this effect seems to dominate. Second, the labor market mechanism
turns out to have a small impact on attitudes since human capital and labor are estimated to be
2
Our paper differs from past contributions also in another respect. We use OECD data (OECD, 2008a) about
the education levels of immigrants and natives in European countries, which allows us to measure the relative skill
levels of immigrants (for the four immigrant groups in each destination country) with greater precision. Moreover,
the relative skill ratios are defined in a way that is consistent with the theoretical model. Therefore our relative skill
ratios are less subject to measurement error than the proxies used in previous contributions. Similarly, marginal tax
rates are measured in a manner that is consistent with the linear tax-benefit scheme of the model.
3
close substitutes, a result which is consistent with the empirical evidence on the (small) impact of
immigration on wages. Third, our decomposition analysis shows that the labor market and welfare state mechanisms tend to compensate each other. This phenomenon might explain why many
papers in the recent literature fail to identify these economic mechanisms.
Why do labor market and welfare state mechanisms tend to compensate each other? Our
economic model allows for the possibility that governments adjust either the benefit level or the
marginal tax rate when new immigrants arrive (as in Facchini and Mayda, 2009). The empirical
test clearly rejects the former adjustment mechanism in favor of the latter. As a result, the welfare
state channel tends to attenuate labor market effects: the arrival of low-skilled immigrants increases
the return to human capital and leads the government to raise the marginal tax rate. Hence the
net return to human capital varies relatively little.
Our paper is also related to Card et al. (2012) who find that economic motivations explain
a smaller share of variation in attitudes than worries about potential externalities related to the
composition of the population (“compositional amenities”). However, their approach differs from
ours since they consider an individual’s opinion about the effect of immigration on the economy as a
whole (“sociotropic” concerns) rather than the impact of immigration on the individual’s personal
economic situation. Hainmueller and Hopkins (2014) also stress the importance of sociotropic
concerns about the cultural and economic effects of immigration on the destination country.
The remainder of the paper is structured as follows. Section 2 presents the theoretical model by
introducing progressively the different mechanisms at work. Section 3 describes the data and gives
some descriptive evidence. Section 4 reports on the empirical findings and Section 5 presents the
conclusion.
2
The Model
This section describes the model that will help us to determine how concerns about the labor market
and the welfare state, on the one hand, and cultural beliefs, on the other hand, influence attitudes
towards immigrants. We develop the model in three steps. First, we present a simple model of
4
the labor market with heterogeneous human capital, where immigrants and natives can be either
substitutes or complements. Second, we consider the welfare-state channel by introducing a linear
tax-benefit schedule in the model. Because of the assumption of a balanced government budget,
the tax-benefit schedule has to adapt to the arrival of new immigrants. We will consider two polar
cases: either the benefit changes (at constant marginal tax rates) or the marginal tax rate varies (at
constant per capita benefits). Third, we add cultural values and beliefs with respect to immigration
to the model. We account for the role of immigration policies and cultural norms at the national
level and, most importantly, for individual heterogeneity in beliefs about immigrants.
2.1
The Labor Market
In our model, the population of each country is divided into natives and different groups of immigrants, m (characterized by their region of origin and their skill level). In each country c, there are
m
LN
c natives and Lc immigrants of group m. Each individual i living in country c supplies one unit
of “raw” labor and hic units of human capital. Aggregate output is given by Yc = F (Hc , Lc ), where
P
P
m
Lc = LN
c +
m Lc and Hc =
i hic and F is an aggregate production function exhibiting constant
returns to scale. Per capita output can be written as yc ≡ Yc /Lc = F (Hc /Lc , 1) ≡ f (hc ), where
hc = Hc /Lc is the per capita human capital stock in country c.3
With perfectly competitive factor markets and profit maximization by the representative firm,
prices and marginal products of production factors are equalized. Marginal products are given by
f 0 (hc ) (human capital) and f (hc ) − hc f 0 (hc ) (raw labor). Earnings of individual i (holding hic units
of human capital and 1 unit of raw labor) can therefore be written as
yic = f (hc ) − hc f 0 (hc ) + hic f 0 (hic ) = f (hc ) + (hic − hc )f 0 (hc ).
3
(1)
Physical capital can be added to the model without changing the qualitative conclusions if perfect international
mobility of capital is assumed. To see this, define aggregate output as Yc = G(Kc , Hc , Lc ), where G is an aggregate
production function with constant returns to scale. A factor-price constrained revenue function (Neary, 1985) can be
defined as G̃(rc , Hc , Lc ) = maxKc {G(Kc , Hc , Lc ) − rc Kc }. With the world rental rate of capital r∗ given, the optimal
stock of physical capital is defined implicitly by ∂G/∂Kc = r∗ and G̃ has the same properties as an unconstrained
revenue (or aggregate production) function, as shown by Neary (1985). Moreover, G̃ is linearly homogeneous with
respect to Hc and Lc . Therefore, if we assume that r∗ does not change with immigration, we can redefine f as
follows: f (hc ) = G̃c (r∗ , Hc /Lc , 1).
5
We assume that individuals consider small changes in the average human capital hc of their
country when they are asked about their immigration preferences. A small change in human capital
has the following impact on an individual’s income: dyic = (hic − hc )f 00 (hc )dhc . The country’s
average human capital stock hc increases (decreases) with immigration if immigrants are on average
more (less) skilled than current residents. Denote Hcm the total human capital of immigrants of
m
m
group m and hm
c = Hc /Lc the average human capital of this group. Assuming that new immigrants
of group m hold on average the same level of human capital than “old” immigrants of that group,
4
m
Combining these elements, we can express the influence of
we have dhc = (hm
c − hc )(dLc /Lc ).
immigration (from group m) on individual i’s income as follows:
m
zic
dyic /yc
≡
=
dLm
c /Lc
hic
hm
1
c
−1
1−
θH θL ,
hc
hc σ
(2)
where σ is the elasticity of substitution between the inputs raw labor and human capital and θH
and θL are the share of human capital and of raw labor in aggregate income.5
zi c
θ Hθ L
hm
1− c
σ
hc
1
−
hic
hc
θ Hθ L
hm
1− c
σ
hc
Figure 1: Labor Market Mechanism (Low-Skill Immigration, hm < h)
In view of the interpretation of our empirical results, it is useful to represent the relation between individual human capital and attitudes towards immigration as defined by equation (2).
Figure 1 depicts the case where immigrants are on average less educated than the resident population (hm
c /hc < 1). Due to labor market competition, immigration reduces earnings of low-skilled
4
This result is obtained as follows. Denoting HcN the total human capital of natives, we have
X
X
X
X
m
N
hc = (1/Lc )
hic = (HcN +
Hcm )/Lc = (HcN +
hm
Lm
c Lc )/(Lc +
c ).
i
m
m
m
Deriving the last expression with respect to Lm
c yields this intermediate result.
5
Note that [−hc f 00 (hc )f (hc )]/[f 0 (hc )[f (hc ) − hc f 0 (hc ))] equals the inverse of the elasticity of substitution σ.
6
natives and increases earnings of high-skilled natives.
2.2
Adding the Welfare State
The labor-market model can be extended to incorporate welfare state considerations by introducing
income redistribution. This is the other major determinant of attitudes according to the recent
economic literature (Facchini and Mayda, 2009; Hanson et al., 2007). Redistribution is accomplished
using a linear tax-benefit schedule. A constant marginal tax rate tc is applied to each individual’s
income and each individual in country c receives an identical benefit bc . We require that the
government’s budget in country c is balanced, which implies: tc f (hc ) = bc . Earnings of an individual
i can now be rewritten as: yic = (1 − tc )[f (hc ) + (hic − hc )f 0 (hc )] + bc .
With immigration, the tax-benefit schedule has to be adjusted in order to ensure a balanced
budget of the government. Following Facchini and Mayda (2009), we focus on the two extreme
cases where either the taxation level tc remains constant and the benefit bc adjusts, or the benefit
remains constant and the marginal tax rate adjusts. The next paragraphs detail these two cases.
If we consider a constant marginal tax rate, a shock in tax revenues would lead to an adjustment
in the level of the benefit, bc . Therefore we have tc f 0 (hc )dhc = dbc and equation (2) becomes:
m
zic
dyic /yc
=
=
dLm
c /Lc
hic
hm
hm
1
c
c
−1
1−
θH θL (1 − tc ) − 1 −
tc θH .
hc
hc σ
hc
(3)
How does the introduction of the welfare state change the relation between individual human
capital and attitudes towards immigration? In Figure 2, we consider the case of low-skill immigration where the benefit level adjusts to ensure a balanced government budget. This figure compares
the pure labor market model (dashed line) with the complete model which includes income redistribution. Two changes stand out. First, low-skill immigration represents a net cost for the tax-benefit
system and entails therefore a decrease in the income of all natives. This is reflected by a parallel
downward shift of the schedule in figure 2. Second, taxation lowers the return to human capital
and decreases therefore the slope in figure 2. It should be emphasized however, that the slope does
not change sign, compared to the pure labor market model, if benefits adjust and the marginal tax
7
zi c
hm
θ Hθ L
1 − c (1 − t c )
hc
σ
1+
−
hm
θ Hθ L
1− c
σ
hc
hic
hc
tc σ
1 − tc θ L
1 − tc +
t cσ
θL
Figure 2: Welfare Mechanism - Benefit Adjustment (Low-Skill Immigration, hm < h)
rate is constant.
In view of the estimation, we can rewrite equation (3) as
m
zic
where ωcm = 1 −
hm
c
hc
hic
=
hc
hm
1− c
hc
tc
θ θ
σ H L
1
hic
θH θL − tc
σ
hc
hm
1
c
1−
θH θL + ωcm ,
hc σ
(4)
− σ1 θH θL − tc θH collects all terms that are specific by country and
by immigrant group.
Turn now to the alternative case where the marginal tax rate tc adjusts to compensate a variation
in government revenues. If the benefit bc is constant, the change in the marginal tax rate tc is given
by tc f 0 (hc )dhc + f (hc )dtc = 0, and equation (2) becomes:
m
zic
dyic /yc
=
=
dLm
c /Lc
hic
hm
1
hm
c
c
2
−1
1−
θH θL (1 − tc ) − tc θH − 1 −
tc θH .
hc
hc
σ
hc
(5)
In the case of low-skill immigration, the marginal tax rate has to increase in order to ensure
a balanced government budget. As a consequence, highly skilled natives have to bear a greater
share of the welfare cost from immigration than unskilled natives. This adjustment is reflected by a
large change in the slope in figure 3. As the analytical expression makes clear, the rotation is much
larger than in the previous case and individual human capital and attitudes towards immigration
may even become negatively related if the fiscal costs of low-skill immigration are higher than the
complementarity advantages in the labor market. The latter outcome will be observed in countries
with a large welfare state (i.e. a large initial tc ). As the benefit level is kept constant in this case,
8
low-skill natives are better protected than in the benefit-adjustment case (the downward shift in
figure 3 is less pronounced).
z ic
hm
θ Hθ L
1− c
hc
σ
1 − tc −
t cσθ H
θL
hi c
hc
tslope=0 =
Intercept = −
hm
θ Hθ L
1− c
hc
σ
1 − tc +
t cσ
θL
θL
θ L + σθ H
(1 − θ H )
Figure 3: Welfare Mechanism - Tax Adjustment (Low-Skill Immigration, hm < h)
In view of the estimation, we can rewrite equation (5) as
m
zic
where
κm
c
=
hic
=
hc
1−
hm
1
hic
hm
1
c
c
2
1−
θH θL − tc
1−
θH + θH θL + κm
c ,
hc σ
hc
hc
σ
hm
c
hc
tc
θ θ
σ H L
(6)
2
− σ1 θH θL − tc θH + tc θH
collects all terms that are specific by
country and by immigrant group.
2.3
Cultural Values and Beliefs
In the economic model spelled out above, worries about labor market competition and the welfare
state are the only determinants of natives’ attitudes toward immigrants. In contrast, Hainmueller
and Hiscox (2007) argue that noneconomic factors play a more important role in the explanation of
these attitudes. Although there is no generally accepted theory of the formation of values and beliefs,
the following three issues have been pointed out in the literature (Card et al., 2012; Hainmueller
and Hiscox, 2007 and 2010). First, individual values and beliefs are shaped by education. More
educated individuals tend to be more open and tolerant towards other cultures and ethnic groups.
9
Second, there are differences in beliefs and values between countries since beliefs are influenced
by cultural norms and immigration policies that differ between countries. Third, there is a large
idiosyncratic component in beliefs about immigration (which includes racial and ethnic prejudice).
Our model accounts for these three factors. Beliefs of individual i in country c with respect to
immigrant group m are described by
r(m)
m
= gc (hic ) + vcm + µic
ψic
,
(7)
where the first term on the right hand side takes account of Hainmueller and Hiscox’s (2007 and
2010) argument that more educated respondents tend to be more open and tolerant towards other
cultures and ethnic groups. Cultural norms and immigration policies are captured by the term vcm
r(m)
which varies by country and by immigration group. Finally, the term µic
accounts for idiosyncratic
factors, capturing those determinants of values and beliefs that are not observable in the survey.
For this individual effect, our identifying assumption is that each individual might hold different
(unobservable) views about immigration from different regions of origin r (‘Europe’ or ‘Rest of the
world’) but that these views do not depend on the fact whether the immigrant comes from a rich
or poor country of that region.6
2.4
Estimating Equation and Parameter Identification
Our complete model is obtained by integrating cultural values and beliefs, as defined in (7), into
the economic framework spelled out in equations (4) and (6). We define a latent variable capturing
attitudes towards immigration and we assume that this latent variable contains an “economic”
component, a “belief” component and a term um
ic which acocunts for individual heterogeneity in
attitudes. We assume furthermore that the economic component of attitudes is proportional to the
impact of immigration on income, by a factor β.7 Attitudes of individual i towards immigration of
6
In our dataset, we have two regions (‘Europe’ or ‘Rest of the world’) and four groups of immigrants (from poor
and rich countries in each of the two regions).
7
This proportionality assumption is introduced in order to clarify the problem of identification of the structural
parameters, in particular σ. Our survey data contains qualitative information about attitudes towards immigration
whereas the economic framework spelled out above defines the effect of immigration on an individual’s income.
Because of the normalization assumption in econometric models of discrete dependent variables, the structural
10
group m are therefore given by
m
m
m
+ um
+ ψic
= βzic
z̃ic
ic .
(8)
It is useful to distinguish again the two polar cases of budget adjustment. If the benefit level adjusts
to balance the government’s budget, attitudes towards immigrants of group m are summarized by
m
z̃ic
hic
=
h
|c
hm
1− c
hc
β
hic
hm
β
r(m)
c
θH θL − tc
1−
θH θL + βωcm + gc (hic ) + vcm + µic +um
.
|
{z
} ic
σ
hc
hc σ
{z
}
beliefs
economic factors
(9)
By contrast, if the marginal tax rate adjusts to balance the government’s budget, the equation is
m
z̃ic
hic
=
h
|c
hic
hm
β
hm
β
r(m)
c
c
2
m
θH θL − tc
1−
1−
βθH + θH θL + βκm
+um .
c + gc (hic ) + vc + µic
|
{z
} ic
hc σ
hc
hc
σ
{z
}
beliefs
economic factors
(10)
We can now define an estimating equation that captures both cases, (9) and (10), of the theoretical model. We assume that individual heterogeneity has an observed and an unobserved component
0
m
and can be written as: um
ic = δ Xic + ic , where Xic is a vector of personal characteristics. The
latent variable capturing attitudes towards immigrants is given by
r(m)
m
z̃ic
= λ0 + λ1 Aic + λ2 Aic Rcm + λ3 tc Aic Rcm + δ 0 Xic + ζcm + µic
+ m
ic ,
(11)
where Aic = hic /hc and Rcm = 1 − hm
c /hc . Furthermore, we assume that gc is a linear function of an
individual’s relative stock of human capital: gc (hic ) = λ1 Aic .8 Our observed variable is Zicm which
m
is equal to 1 if z̃ic
> 0 and equal to 0 otherwise.9
The two versions of the economic model can be distinguished as follows. If the benefit level b is
parameters can can only be defined up to a factor of proportionality in the estimations.
8
To ensure consistency with our measures of human capital in the economic part of the model, we chose to measure
educational status in relative terms (the individual’s human capital relative to the country’s average human capital).
As the model also includes fixed country-immigrant group effects, the fact that we use a relative (rather than an
absolute) measure of education does not make a difference in terms of the empirical results.
9
m
In the survey, the observed variable Zic
is qualitative and can take four (ordered) values expressing individual
m
i’s opinion about the desired level of immigration of immigrant group m. In most estimations, we recode Zic
as a
dichotomic dependent variable.
11
endogenous, the theoretical model predicts that
λ2 = −λ3 = θH θL β/σ,
ζcm = βωcm + vcm .
The restriction λ2 = −λ3 can be easily tested. By contrast, if the marginal tax rate t is endogenous,
the theoretical model predicts that
λ2 = θH θL β/σ,
2
+ θH θL /σ),
λ3 = −β(θH
m
ζcm = βκm
c + vc
To choose the relevant version of the model, we proceed as follows. First, we test the restriction
λ2 + λ3 = 0. If this restriction cannot be rejected, we conclude that the benefit level bc adjusts
endogenously. Note that the elasticity of substitution σ between raw labor and human capital
cannot be identified in this case (even if θH and θL are known) because of the normalization
assumption which is necessary in a model with a qualitative dependent variable.10 By contrast, if
the restriction λ2 + λ3 = 0 is rejected and if λ2 and λ3 have the signs predicted by the theoretical
model, we conclude that the marginal tax rate tc adjusts endogenously. In this case, the elasticity
of substitution σ between raw labor and human capital can be identified assuming that θH and θL
are known:11
θL
σ=−
θH
3
λ3
+1 .
λ2
(12)
Data
In this section, we first provide an overview of the data on attitudes towards migration. In particular,
we give descriptive evidence that motivates our assumption that the idiosyncratic component of
beliefs depends on the region of origin of immigrants but not on the fact whether the origin country
is rich or poor. Then we describe the construction of our indicators of (relative) human capital
levels. These indicators are defined in a way that is consistent with our model. We use detailed
10
In the case where the benefit adjusts to balance the government’s budget constraint, only the ratio β/σ can be
identified from the estimations of λi .
11
The cost share parameters θH and θL can be calculated using the raw data and assumptions on the return to
human capital (see section 3).
12
OECD data to measure the relative skill levels of the four immigrant groups in each destination
country. Finally, we use two different methods to estimate the average marginal tax rates in all
destination countries.
3.1
Attitudes Towards Immigrants
Data on attitudes are taken from the first round of the European Social Survey (ESS) which covers
the period 2002-2003.12 This round of the ESS included a rotating module with detailed questions
about attitudes to immigration, according to the location and the wealth of the immigrant’s origin
country. Using a scale from 1 (few) to 4 (many)13 , a respondent living in country c answers different
versions of the question: “to what extent do you think country c should allow people from [region
of origin] to come and live here?”. The four regions of origin (and the corresponding answers) are
the following:
Europe rich: allow many/few immigrants from richer countries in Europe
Europe poor : allow many/few immigrants from poorer countries in Europe
RoW rich: allow many/few immigrants from richer countries outside Europe
RoW poor : allow many/few immigrants from poorer countries outside Europe
Figure 4 indicates the average opinions expressed by natives14 in each destination country. These
attitudes exhibit little variability with respect to the origin of the immigrants. It may seem that
respondents are either receptive or hostile to immigration regardless of the immigrants’ origin (from
Europe or not, from a rich country or not). At first glance this observation gives some support to
the arguments of Hainmueller and Hiscox (2007) who point out that individuals base their attitudes
more on cultural values and beliefs than on economic factors. While the economic factors depend
on the characteristics of immigrants relative to the characteristics of natives, the cultural factors
and beliefs depend solely on the education level of the native. Moreover, as these beliefs relate to
immigration in general, an individual tends to answer in the same way the four questions above,
12
We use edition 6.1 of ESS Round 1. Table A.1 in the appendix lists destination countries and their respective
frequency in the survey. For more information, see http://www.europeansocialsurvey.org/
13
Original questions use an inverted scale.
14
We consider in this study only the individuals that were born in the survey’s country
13
disregarding the origin of the immigrant.
4.00
Allow Many Immigrants = 4
Allow Few Immigrants= 1
3.50
3.00
2.50
2.00
1.50
From Rich European Countries
From Poor European Countries
From Rich Non-European Countries
en
ed
nd
rla
Sw
ly
Ita
itz
e
Sw
Po
la
nd
Sp
Re
ai
pu
n
bl
ic
Cz
ec
h
G
er
m
an
y
N
or
w
ay
D
en
m
ar
k
Ire
la
nd
Fr
an
ce
et
he
rla
nd
s
Be
lg
iu
m
N
in
gd
om
m
bo
ur
g
Fi
nl
an
d
Lu
xe
us
tri
a
U
ni
te
d
K
A
ry
tu
ga
l
Po
r
un
ga
re
G
H
ec
e
1.00
From Poor Non-European Countries
Figure 4: Average Attitudes by Destination Country
A crucial assumption in our estimation strategy is that the idiosyncratic component of beliefs
depends on the region of origin of immigrants but not on the fact whether the origin country is rich
or poor. The large number of questions contained in the ESS enables us to check whether cultural
values and beliefs are mainly related to the regions of origin of immigrants, regardless of the income
level of their origin countries. Thus, we decompose the answer to each of the four questions into a
regional component (Europe or Rest of the world) and a component which is specific to the income
level of the immigrants’ origin country (rich or poor countries). The general component of attitudes
is measured as the average attitude toward immigrants regardless whether they come from poor or
rich countries. The specific attitude is the deviation of the attitudes regarding each category of immigrant (poor or rich) from the average. If the argument of Hainmueller and Hiscox (2007) stands,
one would expect that individual cultural beliefs are more correlated with the regional component
than with specific attitudes. The ESS survey provides some questions that refer to cultural values
and beliefs (e.g.“Is the country’s cultural life undermined or enriched by immigrants?”). We decom14
pose the correlation between a question on cultural beliefs (Culture) and a question on attitudes
towards immigrants (e.g. from poor countries in Europe, Europe poor) as follows:
Cov(Culture, Europe poor) = Cov(Culture, Europe avg) + Cov(Culture, ∆Europe poor),
where Europe avg = (Europe poor + Europe rich)/2
and ∆Europe poor = Europe poor − Europe avg.
Table 1 presents the decomposition of the covariances between the four questions on attitudes
towards immigrants and several questions that refer to cultural values and beliefs. One can see that
these beliefs are mostly correlated with the regional component of attitudes. Specific attitudes to
immigrants from poor countries (or from rich countries) are only weakly correlated to these individual beliefs. More than 90% of the covariance between the opinion that “immigrants undermine
a country’s culture” and attitudes toward immigrants from poor countries can be attributed to the
regional component of attitudes. This result, and the other decompositions in table 1, lend support to our assumption that the idiosyncratic component of individual beliefs and cultural values
is related to the regional dimension of immigration.
Table 1: Decomposition of the Covariances: Some Native’s Individual Characteristics
Europe
RoW
allow poor immig?
allow rich immig.?
allow poor immig?
allow rich immig?
Immigrants:
average deviation average
deviation
average
deviation average
deviation
1. contribute to taxes?
89.2%
10.8%
113.8%
-13.8%
87.4%
12.6%
116.9%
-16.9%
2. bring down wages?
89.9%
10.1%
112.7%
-12.7%
89.9%
10.1%
112.7%
-12.7%
3. should belong to the majority’s race?
96.4%
3.6%
103.9%
-3.9%
96.8%
3.2%
103.4%
-3.4%
90.7%
9.3%
111.5%
-11.5%
90.1%
9.9%
112.4%
-12.4%
4. undermine country’s culture?
5. get crime problem worse?
89.2%
10.8%
113.8%
-13.8%
87.2%
12.8%
117.2%
-17.2%
6. should be christian?
88.4%
11.6%
115.1%
-15.1%
86.6%
13.4%
118.3%
-18.3%
7. should be white?
86.9%
13.1%
117.7%
-17.7%
84.6%
15.4%
122.2%
-22.2%
Note: Original questions are: 1. taxes and services: immigrants take out more than they put in or less, 2. average wages/salaries generally brought down by immigrants, 3. allow many/few immigrants of same race/ethnic group as majority, 4. country’s cultural
life undermined or enriched by immigrants, 5. immigrants make country’s crime problems worse or better, 6. qualification for
immigration: christian background, 7. qualification for immigration: be white.
Individual Native’s Opinions
3.2
Measure of Human Capital
In our model, two indicators play a crucial role: the ratio between a native’s human capital and his
country’s average human capital (hic /hc ), and the ratio between immigrants’ human capital and
the host country’s average human capital (hm
c /hc ). To ensure consistent measurement, we will use a
single data source for each of the two ratios (ESS for the former, OECD (2008a) for the latter) and
15
define a measure of human capital that is consistent with our theoretical framework. Our measure
of human capital is inspired by the empirical growth literature (Klenow and Rodriguez-Clare, 2005)
where human capital per capita is defined as a Mincerian function of schooling.15
Our model differs from the aggregate production function used in these growth models because
we distinguish “raw” labor from human capital. Therefore, our measure of human capital should
exclude the return to raw labor. In our model, individual income is given by yi = FL +FH hi whereas
the Mincer model states that yi = ceρsi , where ρ is the return to schooling and si denotes years of
schooling attainment. To ensure consistency between the two, we define individual human capital
as hi = (ceρsi − FL0 )/(FH0 ) where superscript 0 denotes values at the initial equilibrium. Defining
the marginal productivity of “raw labor” as FL0 = ceρsmin (and assuming that FH0 = FL0 by choice
of units) yields the following measure of individual human capital:
hi = eρ(si −smin ) − 1,
(13)
where smin denotes “minimum” years of schooling which correspond to our definition of raw labor.
The return to schooling ρ is set to 8.5%, following Klenow and Rodriguez-Clare (2005) who rely
on the returns estimated by Psacharopoulos and Patrinos (2004) for a large set of countries. To
identify the elasticity of substitution σ, we estimate the cost share of raw labor (θL = 1 − θH )
as follows. For an individual i, we have θLi = eρ(smin −si ) . The average cost share of raw labor is
obtained as the average of θLi on the entire sample: θL = 0.5421.
Now turn to the measure of the ratio hic /hc . The ESS includes two main variables concerning
native individual education. The variable edulvl provides the level of education according to seven
categories.16 The variable eduyrs provides the years of education for each individual. We want to
translate the different education levels into years of schooling attainment, regardless of how many
15
A more complete version of the Mincer model would include individuals’ work experience in addition to schooling.
We do not include years of experience in our measure of human capital. First, experience could only be measured as
potential experience using data on age (e.g., experience=age-schooling years-6), involving important measurement
errors especially for women. Second, the literature agrees on the fact that substitution across experience groups
is much easier than substitution between education levels. Our measure of human capital should reflect primarily
differences between education levels since in our model, human capital and raw labor are imperfect substitutes.
16
The seven education categories are: not completed primary education, primary or first stage of basic, lower
secondary or second stage of basic, upper secondary, post secondary non-tertiary, first stage of tertiary, second stage
of tertiary.
16
years it takes an individual to reach a given education level. Therefore our measure of the individual
years of schooling si of natives is defined as the median (in the entire sample) of eduyrs within each
education level (edulvl ). Individual human capital is then calculated using (13) and hc is obtained
by averaging over all individuals of country c in the ESS. As the lowest education level in our sample
corresponds to 4 years of schooling, we set smin to 4.
Data on immigrants’ education level are obtained from OECD (2008a).17 The level of education
is provided for natives and immigrants by categories following the International Standard Classification of Education (ISCED) 1997.18 In the data, four categories gather the six levels of ISCED
classification, namely: primary level (ISCED 0/1/2), secondary level (ISCED 3/4), tertiary level 1
(ISCED 5A/5B) and tertiary level 2 (ISCED 6). Following the ISCED definitions and according
to educational system the European countries, we attributed a certain number of years to each
education category.19
To define the four group of immigrants that appear in the survey questions, we have to distinguish
“poorer countries” from “richer countries” in Europe and in the Rest of the world. In both regions,
we classify countries with a GDP per capita higher than $10,000 as “rich countries” and all others
as “poor countries” (source: World Development Indicators for the year 2003). This classification
yields country groups that seem to correspond to the general perception of rich and poor countries.
For example, Hungary and Gabon are considered to be poor countries (for a complete listing, see
figures A.1 and A.2 in appendix I). Our classification of rich countries is also very close to the
category of “high income” countries established by the World Bank.20
Is it true that immigrants from poor countries are generally less educated than natives, as many
authors assume? The database helps to shed light on this question. Figure 5 plots the relative
level of human capital (hm
c /hc − 1) of immigrants from poor countries of Europe against the relative
level of immigrants from rich countries of Europe (Figure 6 does the same for countries outside
17
Docquier et al. (2009) provide another, widely used database on stocks of immigrants and natives by education
level. As the disaggregation of education levels is finer in OECD (2008a), we chose to use the latter.
18
Available at http://www.unesco.org/education/information/nfsunesco/doc/isced 1997.htm
19
We attribute 8 years to the primary level, 12 years to the secondary level, 15 years to the tertiary level 1 and 17
years to the tertiary level 2.
20
The World Bank divides countries into four groups according to their GNI per capita. For the year 2003 the
categories were defined as follows: low income, $735 or less; lower middle income, $736 - $2,935; upper middle income,
$2,936 - $9,075; and high income, $9,076 or more.
17
Europe). In both figures the first quadrant includes destination countries where immigrants from
rich and poor countries have a higher level of education than the total population. Here we find
countries as diverse as Great Britain, Ireland, Hungary, Italy, Portugal and Spain. In the second
quadrant immigrants from rich countries are more educated than total population while immigrants
from poor countries are less educated than total population. Finally, the third quadrant indicates
destination countries where immigrants from rich and poor countries have a lower level of education
than the total population. The only clear pattern that seems to emerge from these two figures is that
most countries can be found above the 45 degree line. This indicates that in most host countries,
Relative Human Capital (Rich)
−.2
0
.2
.4
.6
immigrants from rich countries are more educated than immigrants from poor countries.
PT
GR
HU
ES
IT
DK
NO
IE
GB
AT
NL
LU
SE
−.4
CH CZ
FI
BE
FR
PL
DE
−.4
−.2
0
.2
Relative Human Capital (Poor)
GDP per capita threshold = 10k
.4
.6
Figure 5: Immigrants’ Human Capital from European Countries (threshold=10k)
3.3
Marginal Tax Rates
In our model, the welfare state is represented by a simple linear tax-benefit system. To measure the
degree of redistribution in all destination countries, we rely on indicators published by the OECD
(2008b) in the “Taxing Wages” series. For all 20 destination countries, we estimate marginal tax
rates that are representative of the real income tax paid by wage earners. The OECD provides
average and marginal tax rates at four different points of the wage distribution for adult, full-time
18
.6
NL
ES
FR
PT
Relative Human Capital (Rich)
0
.2
.4
CH
DE
GB
BE
LU
IT
IE
CZ
GR
HU
DK
PL
−.2
FI
AT
SE
NO
−.2
0
.2
Relative Human Capital (Poor)
GDP per capita threshold = 10k
.4
Figure 6: Immigrant’s Human Capital from RoW countries (threshold=10k)
workers in manufacturing sectors: at 67%, 100%, 133% and 167% of average earnings.
We use two simple methods to estimate a marginal tax rate for each country, based on the
tax schedule for single wage earners.21 First, we calculate a simple average of marginal tax rates
at the four points of the income distribution. Second, we adjust a linear tax-benefit schedule to
the average tax rates at the four points of the wage distribution.22 Reassuringly, the two simple
methods yield very similar results (see Table A.2). The only noticeable differences between the
two methods appear when there is a large jump in marginal tax rates at one point of the income
distribution (Greece, United Kingdom). In the following section, we report estimation results using
the tax data obtained from the first method but none of our results change significantly if we use
the alternative data set.
21
For Switzerland, tax rates are differentiated further at the regional level since direct tax rates vary strongly
from Canton to Canton. We use information on marginal tax rates at the Canton level (AFC, 2003) to calculate
(population-weighted) average marginal tax rates for the six regions that are distinguished in the ESS data for
Switzerland. These regional tax rates are then used to differentiate at the regional level the marginal tax rate
estimated by OECD (2008b) for Switzerland.
22
Denote the tax paid by the individual by T = tY − b, where Y is the individual’s income. The average tax rate
is T /Y = t − b(1/Y ). Therefore, we regress the average tax rate on the inverse of income. The constant term of the
regression is the estimated unique marginal tax rate.
19
4
Results
In this section, we report results from the estimation of our structural model and use the model
to decompose predicted attitudes into economic (labor-market and welfare-state effects) and non
economic components. In the estimation, we use random-effects and fixed-effects models to identify
labor-market and welfare-state effects, taking unobserved individual beliefs into account.
4.1
Estimation methods
The descriptive evidence given in section 3.1 points to the existence of cultural values and beliefs
that influence individual attitudes towards immigration in general. In the estimation of the model,
we will first focus on the question whether economic factors matter in the relationship between
education (or human capital) and attitudes towards immigration even if unobserved beliefs about
immigration in general are taken into account in the estimation procedure.
The estimating equation (11) that corresponds to the complete version of the model is reproduced
for convenience here:
r(m)
m
z̃ic
= λ0 + λ1 Aic + λ2 Aic Rcm + λ3 tc Aic Rcm + δ 0 Xic + ζcm + µic
+ m
ic ,
Economic factors (labor-market and welfare-state mechanisms) are captured by the two interaction
terms Aic Rcm and tc Aic Rcm , whereas the other terms capture the observed and unobserved components of individual beliefs and values.23 The observed variable Zicm is qualitative and can take
four (ordered) values expressing individual i’s opinion about the desired level of immigration of
immigrant group m. In some estimations (random-effects and fixed-effects logit), we recode Zicm as
a dichotomic dependent variable.24
Estimation of equation (11) by (ordered) probit is problematic if the unobserved individual
23
In fact, the term ζcm contains an “economic” element, as spelled out in Section 2.4. This element is taken into
account in the decomposition analysis below.
24
m
The variable Zic
is coded such that higher values indicate more open attitudes towards immigration. In the
random-effects and fixed-effects logit specifications, the dependent variable is recoded as follows: the answers “allow
many to come and live here” and “allow some” are coded as 1, whereas “ allow a few” and “allow none” are coded
as 0.
20
r(m)
effect µic
is an important determinant of attitudes towards immigration. First, Hainmueller and
Hiscox (2007) argue forcefully that individual beliefs are correlated with education and therefore
with some of the explanatory variables in (11). In this case, these variables are endogenous and
r(m)
the (ordered) probit model leads to inconsistent estimates. Second, even if µic
is not correlated
with the explanatory variables, the probit maximum likelihood estimator is not consistent (Greene,
2003, p.679). In this omitted variable case, the inconsistency of the ML-estimator stems from the
non-linearity of the model. In practice, this might be a smaller problem than the first one.
To address these problems, we estimate the model using four different methods. First, we
estimate equation (11) separately for each question involving one of the four immigrant groups
(rich/poor countries in Europe/Rest of the World) using ordered probit models (specifications (1)
to (4) in Tables 2 and 3). This is the approach used in past research (Scheve and Slaughter, 2001;
Mayda, 2006; O’Rourke and Sinnott, 2006; Hanson et al., 2007; Facchini and Mayda, 2009) and it
r(m)
fails to address the problems of omitted variables and endogeneity by ignoring µic
.
Second, we estimate equation (11) jointly for all four immigrant groups using a random-effects
r(m)
logit model (specification (5) in Table 3). We assume that there is a common random effect µic
for rich and poor countries within a region r (Europe or RoW) but that the random effect may differ
between regions. We also estimate this model separately for each region (specifications (6) and (7)
r(m)
in Table 3). This model accounts for omitted individual factors by treating µic
as an unobserved
random variable which is assumed to follow a normal distribution. Note that the random-effects
r(m)
logit estimator is consistent only if the individual-specific effect µic
is not correlated with the
regressors.
Third, we allow for correlation between random effects and regressors by applying a procedure
proposed by Chamberlain (1984). We estimate the “Chamberlain version” of the random-effects
logit model assuming that the correlation between unobserved effects and explanatory variables is
r
given by µric = ν1 Aic Rcr,poor + ν2 Aic Rcr,rich + ξ1 tc Aic Rcr,poor + ξ2 tc Aic Rcr,rich + ηic
, where r = r(m) is the
origin region of migrant group m. The results from the estimation of (11) with this parameterization
of µric are given in column (8) of Tables 2 and 3. Specifications (9) and (10) report the results of
separate estimations for each region.
21
r(m)
In our fourth method, we allow for the possibility that µic
is arbitrarily correlated with
explanatory variables by using a fixed-effects logit model. The estimation of this model relies
on conditional maximum likelihood, where the incidental parameters problem does not arise. Only
observations for individuals whose attitudes differ between immigration from poor countries, on the
one hand, and immigration from rich countries, on the other hand, are taken into account for the
estimation. It should be emphasized that this method (and the preceding) fully address Hainmueller
and Hiscox’s (2007) criticisms of the economic literature since the estimated relationship between
human capital and immigration preferences is purged from all unobserved beliefs about immigration
from a given region.
4.2
Labor Market Mechanism
As many papers in the literature limit their empirical analysis of economic mechanisms to the labor
market channel, it is useful to consider first a model where the welfare state mechanism is excluded.
Table 2 presents estimation results of model (11) under the restriction λ3 = 0. The results are quite
instructive because they depend on the estimation method used: our results reproduce those found
in the economic literature (Mayda, 2006) but show also that these results are not robust to the use
of more general estimation methods that account for unobserved individual beliefs, confirming the
criticism of Hainmueller et al. (2015).
First, the traditional econometric approach used in the literature is the (ordered) probit. In
specifications (1) to (4) of Table 2, the labor market mechanism is found to be a significant determinant of attitudes towards immigration (λ2 is significantly positive for three out of four groups of
immigrants). These results are similar to those found by Mayda (2006) although our definition of
the relative skill composition is slightly different.
Second, these results mostly hold up when a random-effects logit estimator is used (specifications
(5) to (7) in Table 2). This estimator crucially relies on the assumption that the unobserved
individual effects are uncorrelated with the regressors. When we relax this assumption by modeling
the correlation between regressors and individual effects using our third estimator (Chamberlain
random-effects logit), we find no significant effect of the labor market mechanism on attitudes
22
toward immigrants. Interestingly, the coefficient of individual education (λ1 ) is highly significant
in all specifications, suggesting that non-economic factors play an important role in the formation
of attitudes.
Finally, the fixed-effects logit estimator takes into account unobserved individual effects without
any constraints on correlation with explanatory variables (specifications (11) to (13) in Table 2).
The signs of the estimated coefficients λ2 of the labor market mechanism are now reversed and (in
part) significant. These results are not consistent with the predictions of the theoretical model and
are reminiscent of the findings of Hainmueller et al. (2015) for the U.S. regarding attitudes towards
immigrants with different skill levels.
23
24
Europe
Rich
(1)
0.47***
(0.032)
0.56***
(0.13)
32719
Ordered
RoW
Rich
(2)
0.46***
(0.055)
0.16
(0.15)
32719
RoW
Poor
(4)
0.41***
(0.029)
0.48***
(0.10)
32719
Table 2: Determinants of Attitudes - Labor Market Model
Probit
Europe
Poor
(3)
0.37***
(0.025)
0.26***
(0.093)
32719
R.E. Logit
R.E. Logit Chamberlain
F.E. Logit
Eur+RoW
Europe
RoW
Eur+RoW
Europe
RoW
Eur+RoW
Europe
RoW
R+P
R+P
R+P
R+P
R+P
R+P
R+P
R+P
R+P
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
1.78***
1.70***
1.86***
1.90***
1.85***
2.19***
(0.041)
(0.052)
(0.066)
(0.052)
(0.063)
(0.10)
Aic Rcm
λ2
0.49***
1.13***
0.20
-0.20
-0.04
-0.27
-0.40*
-0.60
-0.36**
(0.11)
(0.19)
(0.14)
(0.13)
(0.28)
(0.16)
(0.21)
(0.63)
(0.16)
Observations
130876
65438
65438
130876
65438
65438
24204
12818
11386
N groups (id-ctry)
65438
32719
32719
65438
32719
32719
ρ
0.80
0.77
0.83
0.80
0.77
0.83
(0.0031)
(0.0049)
(0.0039)
(0.0031)
(0.0049)
(0.0039)
σu
3.67
3.36
4.04
3.67
3.36
4.03
(0.037)
(0.047)
(0.057)
(0.037)
(0.047)
(0.057)
log likelihood
-69401.87
-35115.09
-34242.47
-69355.43
-35096.22 -34205.25
Notes: The dependent variable is the answer to the question “to what extent do you think country c should allow people from [origin region/poor or rich country] to come
and live here?”. In specifications (1) to (4), the answer is coded as an ordinal variable taking 4 values (higher values indicate attitudes favorable to immigration). In
specifications (5) to (13), the answer is coded as a binary variable.
Regressions (5), (8) and (11) include fixed effects for country-immigrant’s origin location (either Europe or Rest of the World). All the other regressions include simple
country fixed effects. In both case these fixed effects are interacted with group of immigrants (poor or rich countries). Dummies variables control for gender and political
orientation. Continuous variables control for individual age and individual age squared. Robust standard errors are country clustered in regressions (1)-(4) and (11)-(13).
∗∗∗ , ∗∗ , ∗ denote significance at the 1%, 5%, 10% levels.
Specification
Origin Region
Poor/Rich/Pooled
Variable
Coeff.
Aic
λ1
4.3
Economic Mechanisms
The preceding results establish that the labor market mechanism alone does not provide a satisfactory explanation of attitudes towards immigration. It remains to be seen whether the complete
model, which contains also the welfare state mechanism, performs better from an empirical point
of view. Estimation results for all four econometric methods are presented in Table 3.25 If we use
an ordered probit estimator for all four questions separately (specifications (1) to (4)), the main
finding is that individual education plays a significant role in all specifications, as predicted by those
authors who stress the role of cultural values. Economic mechanisms seem to matter only for the
question about immigrants from rich countries outside Europe: in specification (2) the parameters
λ2 and λ3 have the expected signs and are significantly different from zero. By contrast, the ordered probit estimates in regressions (1),(3) and (4) do not yield significant results for these two
parameters which are relevant from an economic point of view.
If we take unobserved individual heterogeneity into account using the random effects logit estimator, the parameters capturing economic mechanisms become highly significant and have the
expected signs (specifications (5) to (7)). The results also point to the importance of unobserved
individual beliefs in the formation of attitudes towards immigration from a given region: a large
proportion of the residual variance can be attributed to the individual-specific unobserved effect (ρ
is close to 80% and highly significant).
The results change slightly if we relax the assumption that the individual-specific effects are
uncorrelated with the independent variables in the regression, using either the Chamberlain version
of the random-effects logit (specifications (8) to (10)) or fixed-effects logit (specifications (11) to
(13) in Table 3). The main coefficients of the model remain highly significant and the estimated
values of the ratio −λ3 /λ2 (which matters from a structural viewpoint) increase by about 20%. The
hypothesis of absence of correlation between random-effects and other regressors is clearly rejected.26
Therefore we retain the RE-Chamberlain and fixed-effects models as our preferred specifications.
25
In all estimations (except random-effects logit), standard errors are corrected for heteroscedasticity and clustering
at the country level using White’s (1980) method.
26
We test this hypothesis by testing jointly the hypotheses ν1 = ν2 = 0 and ξ1 = ξ2 = 0. Moreover, Hausmann
test of the standard logit model (vs. fixed-effects logit model) clearly rejects the former.
25
What do these results tell us about the way the government budget adjusts to immigration?
The restriction λ2 + λ3 = 0 is rejected in specifications (5) to (10) at the 1 percent level. Moreover,
λ2 and λ3 have the signs predicted by the second version of the model where the tax rate adjusts
endogenously. Hence, the impact of immigration on government revenues seems to be predominantly
absorbed by a rise in marginal tax rates instead of a reduction in the benefit level.
As discussed in section 2.4, we are able to identify the elasticity of substitution (σ) between
raw labor and human capital in this version of the model. Remarkably, the ratio −λ3 /λ2 does
not vary much across the different regressions. This implies that our model yields a rather robust
estimate of the elasticity of substitution between raw labor and human capital. For our preferred
estimation methods (random-effects-Chamberlain logit and fixed effects logit), the values for σ vary
between 2.86 and 3.25. In the context of our theoretical framework, these rather high elasticities
indicate that natives perceive a small impact of immigration on relative wages, a result which seems
consistent with the empirical literature on the labor market consequences of immigration.
To put our estimates of the elasticity of substitution into perspective, it is useful to compare the
implied wage effects of immigration with other estimates in the literature. A first comparison can be
made in terms of direct partial wage elasticities. Following Borjas (2003, section IV), this elasticity
is measured as the percent change in wages of native workers (of a given skill group) associated
with a percent change in labor supply of the same skill group.27 In our model, the wage elasticity
of an unskilled individual who does not hold any human capital is −θH /σ = −0.15 (taking the
fixed-effects specification (11) in Table 3 as a reference).28 According to this estimate, if the arrival
of unskilled immigrants increases total labor supply of raw labor by 10%, the wage of unskilled
workers decreases by 1.5%. At the other extreme of the skill spectrum, the direct partial wage
elasticity for university graduates is −0.098. These elasticities are well in the range of estimates
27
Borjas (2014) gives an overview of recent estimates of this elasticity for different countries. The terminology
direct partial wage elasticities is due to Ottaviano and Peri (2012).
28
In equation (2), the partial wage elasticity for an unskilled individual is
θ H θ L yc
θH
dyc0 /yc0
m yc
= zic
=−
=−
0
0
dLm
/L
y
σ
y
σ
c
c
c
c
where yc0 denotes income of an individual with zero human capital (hic = 0), immigrants are assumed to be unskilled
0
(hm
c = 0), and yc /yc = θL is the share of raw labor in average income.
26
for Germany (around −0.1) reported by Borjas (2014) but they are lower than the estimate for
Norway (−0.27), the only other European country included in Borjas’ (2014) survey. In our model,
partial wage elasticities are lower (in absolute value) for medium-skilled individuals, especially if
their skill level is close to the average skill level of the destination country. However, these small
wage elasticities are not in contradiction with estimates obtained using other empirical approaches,
as argued below.
These direct partial wage effects only give a very incomplete picture of the reaction of wages to
immigration, as has been emphasized by Ottaviano and Peri (2012). A more meaningful measure
of the impact of immigration can be obtained by simulating the wage effects of immigration in a
structural model whose parameters have been estimated econometrically. This approach, which
was initiated by Borjas (2003, section VII), emphasizes the role of complementarity effects between
workers with different skill levels. Moreover, recent contributions in this strand of the literature
argue that natives and immigrants are imperfect substitutes within a given skill group (Ottaviano
and Peri, 2012). As a consequence of the latter assumption, natives are partially shielded from
the competition with immigrants and the impact of immigration on natives’ wages is strongly
attenuated. Seen in this light, the small wage effects for natives obtained in our model do not seem
unrealistic although our simple model does not account explicitly for the imperfect substitution
between immigrants and natives.
Recent studies on the UK and Germany confirm that the impact of immigration on the relative
wage of high to low-skilled natives is very small. In the UK, the share of immigrants in the total
workforce increased from 7% in 1975 to 12% in 2005 while the share of immigrants among highskilled workers rose more than proportionally. According to the simulations of Manacorda et al.
(2012), immigration caused the relative wage of university graduates vs. secondary educated to
decline by only 0.04% per year for natives workers (or 1.2% over the period of 30 years).29 The case
of Germany is different since immigrants are less skilled than natives. According to Brücker and
Jahn (2011), a one percent increase in the labor force due to immigration decreases the relative wage
29
See Manacorda et al. (2012, Table 8). Their simulation results show clearly to what extent the imperfect
substitutability between natives and migrants (within a skill cell) shields native workers (but not “older” immigrants
from labor market competition by immigrants: immigration caused the relative wage of university graduates vs.
secondary educated to decline by 0.89% per year for migrants (a decline of almost 24% over 30 years).
27
of natives (university graduates vs. high-school graduates) by 0.14% in the long run. Felbermayr
et al. (2010) simulate the effect of immigration flows that would have been observed if Germany
had not restricted immigration from Eastern Europe during a transition period. These immigration
flows would have resulted in an increase of the German labor force by 2.1% and an increase in the
relative wage of high-skilled vs. low-skilled natives by 0.18% in the long run (or by only 0.03%
in a version of the model that accounts for rigid wages and unemployment). The results of these
studies confirm that the impact of immigration on wages of native workers seems to be very small
in European countries. Our estimate of the perceived elasticity of substitution between raw labor
and human capital is consistent with this evidence.
28
29
R.E. Logit Chamberlain
F.E. Logit
RoW Eur+RoW Europe
RoW Eur+RoW Europe
RoW
R+P
R+P
R+P
R+P
R+P
R+P
R+P
(7)
(8)
(9)
(10)
(11)
(12)
(13)
1.87***
1.83***
1.79*** 2.05***
(0.066)
(0.054)
(0.068) (0.107)
2.91***
2.90***
6.12*** 1.93***
2.78***
7.20*** 1.71**
(0.571)
(0.604)
(1.247) (0.709)
(0.96)
(1.38)
(0.73)
-8.82*** -10.12*** -20.90*** -7.08*** -9.92*** -24.96*** -6.40***
(1.803)
(1.917)
(4.102) (2.228)
(3.04)
(4.90)
(2.14)
3.03
3.48
3.41
3.67
3.57
3.47
3.74
2.40
2.95
2.86
3.16
3.04
2.92
3.25
65438
130876
65438
65438
24204
12818
11386
32719
65438
32719
32719
0.832
0.804
0.775
0.832
(0.0039) (0.0031) (0.0049) (0.0039)
4.039
3.68
3.37
4.034
(0.057)
(0.037)
(0.048) (0.057)
-34230 -71224.31
-35081
-34190
Notes: The dependent variable is the answer to the question “to what extent do you think country c should allow people from [origin region/poor or rich country] to come and live here?”. In specifications (1) to
(4), the answer is coded as an ordinal variable taking 4 values (higher values indicate attitudes favorable to immigration). In specifications (5) to (13), the answer is coded as a binary variable.
Regressions (5), (8) and (11) include fixed effects for country-immigrant’s origin location (either Europe or Rest of the World). All the other regressions include simple country fixed effects. In both
case these fixed effects are interacted with group of immigrants (poor or rich countries). Dummy variables control for gender and political orientation. Continuous variables control for individual age and
individual age squared. Robust standard errors are country clustered in regressions (1)-(4) and (11)-(13). ∗∗∗ , ∗∗ , ∗ denote significance at the 1%, 5%, 10% levels.
The hypothesis of absence of correlation between random-effects and other regressors is clearly rejected in equations (8) to (10), we test this hypothesis by testing jointly the hypotheses ν1 = ν2 = 0 and
ξ1 = ξ2 = 0. Moreover, Hausmann test of the standard logit model (vs. fixed-effects logit model) clearly rejects the former. nHausman tests for regressions (11) to (13) reject H0: zero individual effects at
the 1% level.
Specification
Ordered Probit
R.E. Logit
Origin Region
Europe RoW Europe RoW Eur+RoW Europe
Poor/Rich/Pooled Rich
Rich
Poor
Poor
R+P
R+P
Variable
Coeff.
(1)
(2)
(3)
(4)
(5)
(6)
Aic
λ1 0.46*** 0.42*** 0.37*** 0.41*** 1.80***
1.73***
(0.031) (0.060) (0.027) (0.029)
(0.041)
(0.052)
λ2 0.73
Aic Rcm
0.90** -0.31
0.76*
3.64***
4.82***
(0.64) (0.39) (0.36) (0.45)
(0.452)
(0.799)
tC Aic Rcm
λ3 -0.68 -2.71** 2.07
-0.83
-10.44*** -13.04***
(2.21) (1.21) (1.49) (1.27)
(1.459)
(2.732)
−λ3 /λ2
3.01
2.68
2.87
2.71
σ
2.38
2.21
2.02
Observations
32719 32719 32719 32719
130876
65438
N groups (id-ctry)
65438
32719
ρ
0.804
0.775
(0.0031) (0.0049)
σu
3.67
3.37
(0.037)
(0.048)
log likelihood
-69376
-35103
Table 3: Determinants of Attitudes - Complete Model
4.4
Economic Mechanisms and Individual Beliefs: a Decomposition
The econometric evidence discussed so far allows us to confirm that the economic channels play
a significant role in the formation of attitudes. In this section, we go a step further and compare
the relative importance of economic mechanisms and individual values and beliefs by simulating
the econometric model with different configurations of parameters. This simulation exercise relies
on the complete model (11) which includes the labor market and welfare state mechanisms. We
use the Chamberlain version of the random-effects logit corresponding to regressions (9) and (10)
from Table 3 to predict (latent) attitudes towards immigrants.30 These predicted attitudes are the
starting point for our decomposition (“total attitudes”).31 The contribution of the “welfare state”
mechanism to attitudes can then be calculated as follows. Setting marginal tax rates equals to zero,
we recalculate predicted attitudes of the model. This provides a measure of attitudes in the absence
of a welfare state, including only the two mechanisms of “labor market competition” and “individual
values and beliefs”.32 The difference between “total attitudes” and the latter predicted attitudes
represents the contribution of the welfare state mechanism. Analogously, we obtain the predicted
values of attitudes determined by the labor market mechanism. In the absence of the labor market
mechanism, raw labor and human capital would be perfect substitutes and immigration would have
no impact on relatives wages (σ = ∞ and λ2 = 0). Finally, the difference between “total attitudes”
(obtained from the complete model) and the prediction determined by the economic factors is
attributed to “individual values and beliefs”.33
30
The (Chamberlain) RE logit enables us to account for the role of individual education, which is part of the
“individual values and beliefs” component of attitudes. We prefer to use the Chamberlain RE logit rather than
the fixed effects estimates because the impact of individual education on attitudes cannot be estimated in the fixed
effects specification. However, the role of the economic mechanisms in the decomposition should not be affected by
that choice since both specifications yield very similar estimates of the impact of the economic variables on attitudes.
31
As our decomposition analysis focuses on the link between individual education (human capital) and attitudes
towards immigration, we “neutralize” the influence of other variables (age, gender, political orientation) by calculating
predictions at the mean of those variables.
32
In practice, this amounts to setting λ3 = 0. We also correct the fixed country effects in order to account for
the fact that tc = 0 changes the value of κm
c in equation (6). A detailed description of this procedure is provided in
Appendix II.
33
This is arguably a strong assumption which is discussed further below.
30
1
0
−1
−2
1
0
−1
.5
1
1.5
2
.5
1
1.5
2
1
2
3
1
2
.5
1
1.5
2
1
2
3
1
2
3
1
1.5
2
.5
1
1.5
2
0
0
2
3
2
3
0
1
2
1
2
.5
1
1.5
2
1
2
3
0
1
2
3
4
PRT: Rmc = −.576, tc = .24
0
ITA: Rmc = −.097, tc = .297
0
FRA: Rmc = .112, tc = .263
0
CZE: Rmc = .103, tc = .175
L.M. + Welfare Eur/Poor
3
POL: Rmc = .095, tc = .091
1
IRL: Rmc = −.201, tc = .31
1
FIN: Rmc = .157, tc = .409
0
CHE: Rmc = .143, tc = .2263
Labor Market Eur/Poor
.5
NOR: Rmc = .018, tc = .381
0
HUN: Rmc = −.085, tc = .405
0
ESP: Rmc = −.085, tc = .255
0
BEL: Rmc = .182, tc = .435
1
1.5
2
1
1.5
2
2.5
1
2
3
0
.5
1
1.5
2
SWE: Rmc = .032, tc = .399
0
LUX: Rmc = .014, tc = .298
.5
GBR: Rmc = −.173, tc = .31
.5
DEU: Rmc = .201, tc = .428
Figure 7: Simulation - Economic Determinants, Immigrants from Poor European Countries
x−axis corresponds to individual education hi/h, and y−axis corresponds to lattent attitudes towards immigration
0
NLD: Rmc = .293, tc = .356
0
GRC: Rmc = .051, tc = .242
0
DNK: Rmc = .094, tc = .462
0
AUT: Rmc = .171, tc = .281
1
−.5
2
0
−2
.1
0
.5
−2
1
0
−3 −2 −1
1
.5
0
−1 −.5
2
−1
0
1
−4
1
.5
0
−.5
0
−.3 −.2 −.1
1
0
−1
1
0
2
−2 −1
0
1
−3 −2 −1
1
.5
0
−1 −.5
1
.5
0
−1 −.5
1
.5
0
−1 −.5
5
0
−5
−10
2
0
−2
−4
2
1
0
−1
.1
0
−.1
−.2
.2
0
−.6 −.4 −.2
31
We consider first the role of economic determinants. Figure 7 shows the impact of the economic
determinants on (predicted) attitudes regarding immigrants from poor European countries. To
remain consistent with the theoretical model, we plot these predicted values against the relative
education of natives (hi /h). Figure 7 depicts the predicted values determined by the labor market mechanism (black dots) and the sum of the predicted values determined by the labor market
mechanism and the tax-benefit mechanism (red crosses). The theoretical mechanisms can be illustrated using the example of Belgium, where immigrants are less educated than the average resident
(Rcm > 0). In this case, the labor market mechanism is harmful to low skilled natives and beneficial
for high skilled natives (captured in Figure 7 by the black dots with a positive slope and a negative
intercept). From the tax-benefit point of view, less educated immigrants would represent a burden
for all natives, reducing the slope according to the level of the taxes (tc ). We expect that the slope
changes sign if the marginal tax rate is higher than 29%34 , which is indeed the case for Belgium: the
cumulated effect of economic mechanisms is that natives are against immigration, and this negative
attitude is stronger for skilled natives. This exercise can be carried out for all countries of our
sample, giving a detailed panorama of attitudes according to the economic determinants.
Turn now to the role of individual values and beliefs. Figure 8 depicts predicted individual values
and beliefs regarding immigrants from poor European countries and compares them to the prediction
of the complete model. Two observations are in order regarding figure 8. First, attitudes towards
immigration seem to be mostly determined by individual values and beliefs whereas economic
determinants, taken together, play a minor role. Second, individual values and beliefs are positively
correlated with the level of the respondent’s education: better educated natives are more open
to immigration. This result, which is robust across all groups of immigrants, seems to confirm
Hainmueller and Hiscox’s (2007) argument that noneconomic factors are predominant in explaining
attitudes towards immigration.35
34
This critical value for tax is obtained from figure 3, where tslope=0 = θL /(θL + σθH ) = 0.5421/(0.5421 + 2.86 ∗
0.4579) = 0.29.
35
Figures A.3 to A.8 in Appendix II reproduce this analysis for the other groups of immigrants, respectively for
immigrants from rich European countries, for immigrants from poor non-European countries and for immigrants
from rich non-European countries. The qualitative conclusions that can be drawn from these figures are the same as
those for immigrants from poor European countries.
32
5 10
−10 −5 0
5 10
−10 −5 0
5 10
−10 −5 0
5 10
.5
0
0
0
.5
0
1
1
NLD
GRC
1
DNK
1
AUT
2
3
2
2
.5
0
0
0
.5
1
1
1
NOR
HUN
ESP
1
1.5
2
2
1.5
2
3
3
2
0
0
0
0
Individual Values, Eur/Poor
2
1.5
1.5
BEL
.5
1
1
1
POL
IRL
FIN
1
CHE
2
2
2
1.5
2
3
3
3
1
.5
1
2
PRT
ITA
1
FRA
1
2
3
1.5
2
2
4
3
0
0
.5
.5
.5
1
1
Total Attitudes, Eur/Poor
0
0
0
0
CZE
1
1
SWE
LUX
1.5
GBR
DEU
2
1.5
2
1.5
2
3
2.5
2
Figure 8: Simulation - Individual Beliefs and Predicted Values, Immigrants from Poor European Countries
x−axis corresponds to individual education hi/h, and y−axis corresponds to total attitudes towards immigration
−10 −5 0
33
A more systematic view of the relative importance of each factor can be gained by decomposing
the variance of predicted attitudes towards immigrants into its three components: labor market
channel, welfare state effect, individual values and beliefs (see table A.3 in the appendix).36 Two
observations stand out. First, the economic determinants play an important role but the labor
market and the welfare state channels tend to compensate each other. Taken together, the two economic channels taken together are much less correlated with predicted attitudes than each channel
taken separately. Second, individual values and beliefs explain a much larger share of predicted
attitudes than the economic mechanisms.
How robust are these results with respect to the omission of other economic or non-economic
determinants? It could be argued that the model only accounts for the two main economic channels
that have been discussed in the literature but neglects other economic mechanisms that possibly
influence attitudes towards immigrants. For example, natives might fear that rental costs or housing
prices increase with immigration. If the housing market is segmented by income (or skill), natives
tend to oppose the arrival of immigrants that have similar earnings and compete for the same type
of apartments or houses. In this case, the fear of increasing housing costs would be captured by
the labor market mechanism in our empirical analysis.37 More generally, the labor market channel
captures all phenomena where immigrants compete with similarly skilled natives in the demand for
goods or the supply of factors.38
How confident can we be about the importance of individual values and beliefs in our decomposition? We attribute to this category those components of attitudes that are idiosyncratic or vary
with the education level of the native respondent but do not depend on the skill level of immigrants.39 To conclude, we carry out an indirect consistency check of this assumption by comparing
36
We use the decomposition described above where total predicted attitudes (A) are the sum of the labor market
channel (L), welfare state channel (W ) and values and beliefs (B). The variance of total predicted attitudes is then
decomposed as follows: V ar(A) = Cov(A, L + W + B) = Cov(A, L) + Cov(A, W ) + Cov(A, B). This decomposition
is carried out for each country separately in order to neutralize the influence of unobserved factors at the country
level.
37
Obviously, current house- or land-owners would generally benefit from immigration but this factor is controlled
for by the individual fixed effects in our empirical analysis (specifications (11) to (13) in Table 3).
38
On the other hand, phenomena where high-skill (or high-income) natives are disproportionately affected by the
arrival of low-skill immigrants are captured by the welfare state mechanism (with an exogenous benefit level).
39
More precisely, we assume that the idiosyncratic beliefs of natives can depend on the origin of the migrant
r(m)
(Europe or Rest of the world) but not on the migrant’s skill level, as captured by the term µic in equation (7).
34
attitudes towards immigration with the respondent’s evaluation of the influence of immigration
on the host country’s culture. The ESS includes an item that asks respondents whether the host
country’s “cultural life is generally undermined or enriched by people coming to live here from other
countries?” If, as we assume, the education level in specifications (1) to (9) in Table 3 captures
the influence of individual values and beliefs on attitudes towards immigration, then the respondent’s education level should play a similar role in her evaluation of the influence of immigration
on the host country’s culture. Moreover, if we estimate simultaneously the equations explaining an
individual’s attitudes toward immigration and her evaluation of the impact of immigration on the
host country’s culture, we expect the error terms of the two equations to be correlated due to the
idiosyncratic component of beliefs.
Table 4 reports the results of the comparison between cultural beliefs and attitudes towards
immigrants from rich European countries. The marginal impact of education on attitudes and on
cultural beliefs is quasi identical, whether the two equations are estimated separately (specifications
(1) to (4)) or jointly using a bivariate probit (specifications (5) and (6)). Both dependent variables
are coded as binary variables. As both dependent variables have a very similar distribution (the
mean of attitudes towards immigrants from rich European countries is 0.563 and the mean of
cultural beliefs is 0.526), the fact that the coefficients are very similar in the two probit equations
implies that marginal effects of education on attitudes or beliefs are also quasi identical.
The comparison of the different specifications in Table 4 shows that the impact of education on
attitudes does not depend on the details of the estimated equation. Results for attitudes to immigration are almost identical whether the original sample is used (specification (1)) or the sample
is restricted to those individuals who also answered the question on cultural beliefs (specification
(2)).40 The marginal impact of education on cultural beliefs also hardly changes whether all interaction terms are included in the equation (specification (3)) or a reduced specification without these
interaction terms is used (specification (4)). Incidentally, it is not surprising that these interaction
terms (which capture the economic mechanisms) are not a significant determinant of cultural beliefs
in specification (3).
40
Note that specification (1) in Table 4 is estimated using a probit based on a binary dependent variable whereas
specification (1) in Table 3 is estimated using ordered probit with an ordinal dependent variable taking four values.
35
These results are consistent with our assumption that the education term in equation (11) captures individual values and beliefs. Moreover, the correlation ρ in the bivariate probit specification
((5) and (6)) is positive and highly significant, indicating that individuals who believe that immigrants enrich the host country’s culture tend to be more open to immigration. This positive
correlation confirms the important role of the idiosyncratic component of beliefs in attitudes to
immigration.
Table 4: Comparison of Cultural Beliefs and Attitudes towards Immigrants
Specification
Question
Variable
Aic
Aic Rcm
tC Aic Rcm
Observations
ρ
Europe
Rich
(1)
0.52***
(0.03)
1.36**
(0.64)
-2.43
(2.49)
32719
Probit
Europe Culture
Rich
(2)
(3)
0.51*** 0.54***
(0.03)
(0.05)
1.35**
1.21
(0.66)
(1.01)
-2.45
-2.54
(2.58)
(3.17)
31430
31430
Bivariate Probit
Culture Culture
Europe
Rich
(4)
(5)
(6)
0.49*** 0.54***
0.51***
(0.05)
(0.05)
(0.03)
1.21
1.32**
(1.01)
(0.66)
-2.54
-2.39
(3.17)
(2.57)
31430
31430
31430
0.30
(0.01)
Notes: Culture denotes the question: “would you say that [country’s] cultural life is generally undermined
or enriched by people coming to live here from other countries?” Original answers are given on
a scale between 0 (undermined) and 10 (enriched). The variable Culture is coded as 0 for all
answers below (and including) 5, and as 1 for answers between 6 and 10. All regressions include
country fixed effects. Dummy variables control for gender and political orientation. Continuous
variables control for individual age and individual age squared. Robust standard errors are country
clustered in regressions (1)-(4) and (9)-(10). ∗∗∗ , ∗∗ , ∗ denote significance at the 1%, 5%, 10%
levels.
5
Conclusion
This paper develops and estimates a structural model of attitudes toward immigration and assesses
the relative importance of economic and non-economic determinants. We consider two economic
mechanisms. According to the labor market mechanism, natives have positive attitudes toward
immigrants if they are complements in the labor market. High-skilled natives are predicted to
prefer low-skilled to high-skilled immigrants. The welfare state mechanism operates through the
36
adjustment of the tax-benefit system to immigration. With a guaranteed benefit level, the marginal
tax rate has to adjust and low-skilled immigrants represent a burden especially for high-skilled
natives. Among the non-economic determinants of attitudes, we include all motivations that shape
an individual’s general attitude toward immigration. In particular, it is well known that natives
with higher education are more open to diversity and other cultures.
Using data for 20 European countries in 2002, we find a significant impact of the labor market
and the welfare state mechanisms on attitudes towards migration. By contrast to previous contributions, our results are obtained by controlling for unobserved individual beliefs about immigration
in general. This is a much more demanding test of the relevance of economic mechanisms than in
Facchini and Mayda (2009) or Hanson et al. (2007) since (in the fixed effects specification of our
model) parameters are identified solely by within-individual variations of attitudes towards different
types of immigrants. Finally, simulations with our structural model indicate that although these
two economic mechanisms matter, their net effect is much smaller than the impact of non-economic
factors on attitudes towards immigration.
Our results reconcile the two strands of the literature that are represented by Facchini and
Mayda (2009) and Hainmueller and Hopkins (2014). On the one hand, most respondents express
identical attitudes towards immigration from different origin countries. This indicates that there
are individual-specific beliefs and values that determine overall attitudes towards all types of immigration. Here our results are in agreement with Hainmueller and Hopkins (2014) and confirm that
this individual component of attitudes strongly depends on education levels and is highly correlated
with beliefs about the impact of immigration on cultural values.
On the other hand, our results lend support to the importance of economic mechanisms in
explaining attitudes towards different types of immigration. Our estimates of the role of economic
mechanisms are driven by individuals who express different attitudes towards immigration from rich
or poor countries of a same region. Although this minority of respondents only play a small role
in the determination of overall attitudes towards immigration, their preferences are crucial when
decisions about the type of immigration are taken. In this context we find a significant effect of
economic mechanisms, confirming the results originally obtained by Facchini and Mayda (2009).
37
By rehabilitating the role of economic mechanisms, our results are in disagreement with Hainmueller and Hopkins’ (2014) statement that personal economic circumstances do not matter for the
formation of attitudes towards immigration. How can these divergent conclusions be explained?
First, the estimates of our structural model show that the labor market mechanism has to be combined with the welfare state mechanism in order to provide a satisfactory explanation of individual
attitudes toward immigration. Our results confirm that, taken alone, the labor market mechanism
is not a significant determinant of attitudes.
Second, the two economic mechanisms tend to neutralize each other. Our theoretical model
shows that the welfare state mechanism tends to attenuate (or even reverse) the impact of human
capital on attitudes predicted by the labor market mechanism. It might therefore not be surprising
that many studies fail to identify the labor market channel if they do not take the welfare state
mechanism into account. For example, Hainmueller and Hopkins’ (2015) finding that high-skilled
natives tend to prefer high-skilled to low-skilled immigrants would in principle be consistent with
a tax-adjustment model, as spelled out in section 2.2 above.
Third, the combined effect of the two economic mechanisms can only be properly identified in a
comparative setting with cross-country variation in immigrants’ relative human capital and in the
degree of taxation and redistribution. The specific predictions of economic theory in this context
provide a more complete and meaningful test of economic mechanisms than what is possible in
single-country studies. Moreover, our estimates of the elasticity of substitution between raw labor
and human capital seem consistent with the empirical literature on the labor market effects of
immigration.
References
AFC (Administration fédérale des contributions) (2003), “Charge fiscale: Chefs-lieux des cantons
2002”, Berne.
Borjas, G. J., 2003, “The Labor Demand Curve Is Downward Sloping: Reexamining the Impact
of Immigration on the Labor Market,” The Quarterly Journal of Economics, 118(4): 13351374.
Borjas, G. J., 2014, “Immigration Economics”, Cambridge MA: Harvard University Press.
38
Brücker, H. and E. J. Jahn, 2011, “Migration and Wage-setting: Reassessing the Labor
Market Effects of Migration,” The Scandinavian Journal of Economics, 113: 286-317.
Card, D., Dustmann, C. and I. Preston, 2012, “Immigration, Wages and Compositional
Amenities,” Journal of the European Economic Association, 10: 78-119.
Citrin,J., D. P. Green, C. Muste and C. Wong, 1997, “Public Opinion Toward Immigration
Reform: The Role of Economic Motivations,” The Journal of Politics, 59:858-881.
Chamberlain ,G., 1984, “Panel Data,” Chapter 22 in Z. Griliches and M. Intrilligator, eds.,
Handbook of Econometrics, North-Holland, Amsterdam, 1247-1318.
Docquier, F., B. L. Lowell and A. Marfouk, 2009, “A gendered assessment of highly
skilled emigration,” Population and Development Review, 35(2): 297-322.
Dustmann, C. and I. P. Preston, 2007, “Racial and Economic Factors in Attitudes to Immigration,” The B.E. Journal of Economic Analysis & Policy, 7(1): Advances, Article 62.
Felbermayr, G., W. Geis and W. Kohler, 2010, “Restrictive immigration policy in Germany: pains and gains foregone?”, Review of World Economics, 146:1-21.
Facchini, G. and A. M. Mayda, 2009, “Individual attitudes towards immigrants: Welfare-state
determinants across countries,” Review of Economics and Statistics 91(2): 295-314.
Greene, W. P., 2003, Econometric Analysis - Fifth Edition, Englewood Cliffs: Prentice Hall.
Hainmueller, J. and M. J. Hiscox, 2007, “Educated Preferences: Explaining Attitudes Toward Immigration in Europe,” International Organization, 61: 399-442.
Hainmueller, J., M. J. Hiscox, Y. Margalit, 2015, “Do concerns about labor market
competition shape attitudes toward immigration? New evidence,” Journal of International
Economics, 97(1): 193-207.
Hainmueller, J. and D. J. Hopkins, 2014, “Public Attitudes toward Immigration,” Annual
Review of Political Science, 17: 225-249
Hanson, G. and K. Scheve and M. Slaughter, 2007, “Local Public Finance and Individual
Preferences over Globalization Strategies,” Economics and Politics, 19: 1-33.
Klenow, P., A. Rodriguez-Clare, 2005, “Externalities and Growth,” Handbook of Economic
Growth, Volume 1A, Aghion, P. and S. N. Durlauf (eds), North-Holland.
Manacorda, M., A. Manning and J. Wadsworth (2012), “ The Impact of Immigration
on the Structure of Wages: Theory and Evidence from Britain,” Journal of the European
Economic Association, 10:120-151.
Mayda, A. M., 2006, “Who is against immigration? A cross-country investigation of individual
attitudes toward immigrants,” Review of Economics and Statistics, 88(3): 510-530.
Neary, J. P., 1985, “International Factor Mobility, Minimum Wage Rates, and Factor-Price
Equalization: A Synthesis,” The Quarterly Journal of Economics, 100(3): 551-570.
Ottaviano, G. I. P. and G. Peri, 2012, “Rethinking the Effect of Immigration on Wages,”
Journal of the European Economic Association, 10: 152-197.
Psacharopoulos, G. and H. A. Patrinos, 2004, “Returns to investment in education: a
further update,” Education Economics, 12(2): 111-134.
OECD, 2008a, “A Profile of Immigrant Populations in the 21st Century: Data from OECD
Countries,” OECD Publishing, Paris.
39
OECD, 2008b, “SourceOECD Taxing Wages Statistics,” SourceOECD Vol 2008 Release 01.
OECD, 2015, “International Migration Outlook 2015,” OECD Publishing, Paris
O’Rourke, K., R. Sinnott, 2006, “The Determinants of Individual Attitudes Towards Immigration,” European Journal of Political Economy, 22 :838-861.
Scheve, K. F., M. J. Slaughter, 2001, “Labor Market Competition And Individual Preferences Over Immigration Policy,” Review of Economics and Statistics, 83(1): 133-145.
White, H., 1980, “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct
Test for Heteroskedasticity,” Econometrica, 48(4): 817-838.
40
Appendix I
Table A.1: Sample Size by Destination Country
Country
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Luxembourg
Netherlands
Norway
Poland
Portugal
Republic Czech
Spain
Sweden
Switzerland
United Kingdom
Frequency
1808
1588
1310
1863
1298
2588
2193
1453
1782
1090
917
2143
1878
1886
1311
1089
1431
1682
1606
1803
Table A.2: Marginal Income Tax Rates (Percentages)
Country
Austria
Belgium
Czech Republic
Denmark
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Luxembourg
Netherlands
Norway
Poland
Portugal
Spain
Sweden
Switzerland
United Kingdom
Average of
Marginal Rates
28.1
43.5
17.5
46.2
40.9
26.3
42.8
24.2
40.5
31.0
29.7
29.8
35.6
38.1
9.1
24.0
25.5
39.9
20.7
31.0
Regression on
Average Rates
29.2
43.6
16.4
47.8
41.7
26.0
42.0
20.8
41.0
30.3
30.8
30.7
35.9
37.0
9.2
23.8
24.7
39.7
20.0
26.6
Note: Reference year 2002. See the main text for details.
41
TUN
FJI
BRA
RUS
JAM
VCT
VEN
URY
ARG
ZAF
DMA
BLZ
PAN
LBY
GRD
CRI
MYS
MUS
LCA
BWA
GAB
CHL
LBN
GNQ
MEX
PLW
KNA
TTO
SYC
OMN
SAU
ATG
BRB
42
ISR
BHS
KWT
NZL
ARE
SGP
HKG
AUS
CAN
QAT
JPN
USA
ESP
ITA
26
20
BEL
DEU
FRA
GBR
FIN
AUT
NLD
SWE
ISL
IRL
DNK
GDP per capita ’000 $
MLT
SVN
PRT
GRC
CYP
40
GDP per capita ’000 $
KOR
BHR
MDA
UKR
BLR
BIH
ALB
MKD
YUG
BGR
ROM
TUR
LVA
LTU
POL
SVK
HRV
EST
HUN
CZE
CHE
NOR
60
15
10
Figure A.1: Thresholds of GDP per capita for European countries
40
10
Note: For the sake of clarity this figure excludes 91 other RoW countries used in the
analysis with GDP per capita lower than $2300.
Figure A.2: Threshold of GDP per capita for RoW countries
LUX
Appendix II: Simulations
This appendix explains the simulation procedure which is based on equation (11):
r(m)
m
z̃ic
= λ0 + λ1 Aic + λ2 Aic Rcm + λ3 tc Aic Rcm + δ 0 Xic + ζcm + µic
+ m
ic ,
where the fixed effects are related to the structural parameters as follows:
ζcm
=
βκm
c
+
vcm ,
κm
c
hm
tc
1
c
2
= 1−
θH θL − θH θL − tc θH + tc θH
hc
σ
σ
“Total” attitudes are based on the model
r(m)
m
z̃ic
= λ0 + λ1 Aic + λ2 Aic Rcm + λ3 tc Aic Rcm + δ 0 Xic + ζcm + µic
hm
tc
1
c
m
2
θH θL − θH θL − tc θH + tc θH + vcm
ζc = β 1 −
hc
σ
σ
+ m
ic ,
(A.1)
In the absence of a welfare state (tc = 0) the model becomes
r(m)
m
z̃ic
= λ0 + λ1 Aic + λ2 Aic Rcm + δ 0 Xic + ζcm + µic
1
hm
c
m
− θH θL + vcm
ζc = β 1 −
hc
σ
+ m
ic ,
(A.2)
The part of attitudes due to the welfare state (WS) mechanism is obtained by taking the difference
between models (A.1) and (A.2):
m
z̃ic
|W S =
λ3 tc Aic Rcm
hm
+β 1− c
hc
tc
2
θH θL − tc θH + tc θH
σ
(A.3)
If the labor market mechanism does not operate because human capital and raw labor are perfect
substitutes (σ = ∞), we have λ2 = 0 and the model becomes
r(m)
m
= λ0 + λ1 Aic + δ 0 Xic + ζcm + µic
z̃ic
ζcm = vcm
43
+ m
ic ,
(A.4)
The part of attitudes due to the labor market (LM) mechanism is obtained by taking the difference
between models (A.2) and (A.4)
m
z̃ic
|LM =
λ2 Aic Rcm
1
hm
c
− θH θL
+β 1−
hc
σ
(A.5)
The total contribution of economic mechanisms to attitudes is given by the sum of (A.3) and (A.5).
Non-economic factors (NE) are then measured as the difference between “total” attitudes and the
contribution of economic mechanisms:
r(m)
m
z̃ic
|N E = λ0 + λ1 Aic + δ 0 Xic + ζcm + µic
(A.6)
where the error term m
ic is omitted. As our analysis focuses on the relation between human capital
and attitudes, we want to neutralize the influence of other personal characteristics (age, gender,
politics) contained in Xic . Therefore we replace Xic by country averages X̄c for the simulations.
All structural parameters in the decomposition equations (A.3),(A.5) and (A.6) can be identified
from the estimation of equation (11):41
r(m)
m
ẑic
= λ̂0 + λ̂1 Aic + λ̂2 Aic Rcm + λ̂3 tc Aic Rcm + δ̂ 0 X̄c + ζ̂cm + µ̂ic
r(m)
where µ̂ic
,
are the individual effects predicted by the Chamberlain random-effects model.42 The
remaining parameters are estimated as follows: σ̂ can be estimated from equation (12), β̂ = −(λ̂2 +
2
λ̂3 )/θH
and vcm is obtained as a residual (vcm = ζ̂cm − β̂κm
c ).
Finally, in terms of predicted variables our decomposition is given by:
m
m
m
m
|N E
= ẑic
|LM +ẑic
|W S +ẑic
ẑic
41
(A.7)
m
The model is estimated using individual characteristics Xic but for the prediction of total attitudes ẑic
we replace
Xic by country averages X̄c as indicated above.
42
More precisely, they are the predicted effects from the specification µric = ν1 Aic Rcr,poor + ν2 Aic Rcr,rich +
r
ξ1 tc Aic Rcr,poor + ξ2 tc Aic Rcr,rich + ηic
, where r = r(m) is the origin region of migrant group m.
44
1.5
2
0
.05
1
1.5
2
−.05
−3 −2
0
.5
0
ESP: Rmc = −.17, tc = .255
.5
1
1.5
2
0
FIN: Rmc = .037, tc = .409
1
2
.5
FRA: Rmc = .117, tc = .263
1
1.5
2
GBR: Rmc = −.144, tc = .31
.5
2
1
3
−1
0
0
1
1.5
2
0
2
3
0
0
1
2
3
−1
−1 −.5
0
0
IRL: Rmc = −.165, tc = .31
.5
1
1.5
2
.5
ITA: Rmc = −.145, tc = .297
1
1.5
2
2.5
LUX: Rmc = .005, tc = .298
2
3
2
3
0
1
2
3
0
POL: Rmc = .134, tc = .091
1
2
3
0
PRT: Rmc = −.398, tc = .24
1
2
3
SWE: Rmc = .028, tc = .399
0
1
2
0
0
.5
1
1.5
2
−.2
−.4
−5
−10
−1
−1
−.5
0
0
0
1
.5
1
2
1
1
−.05
−1
0
NOR: Rmc = −.131, tc = .381
.2
1
NLD: Rmc = −.07, tc = .356
5
0
−2
−1
−2
−4 −2
0
0
0
0
0
2
1
2
1
4
4
1
HUN: Rmc = −.21, tc = .405
.05
.5
GRC: Rmc = −.338, tc = .242
2
0
−.5
−1
0
−2 −1
1
0
0
1
2
DEU: Rmc = .153, tc = .428
2
1
1
.5
DNK: Rmc = −.171, tc = .462
.5
0
−.05
−.5
0
−1.5 −1 −.5
.5
0
1
CZE: Rmc = −.007, tc = .175
1
CHE: Rmc = .007, tc = .2263
.05
BEL: Rmc = .074, tc = .435
.5
AUT: Rmc = −.093, tc = .281
0
1
2
Labor Market Eur/Rich
3
0
1
2
3
4
0
.5
1
1.5
2
L.M. + Welfare Eur/Rich
x−axis corresponds to individual education hi/h, and y−axis corresponds to lattent attitudes towards immigration
Figure A.3: Simulation - Economic Determinants, Immigrants from Rich European Countries
BEL
CHE
CZE
DEU
−5
0
5
AUT
0
.5
1
1.5
2
0
.5
1
2
0
.5
1
ESP
1.5
2
0
1
FIN
2
.5
1
FRA
1.5
2
GBR
−5
0
5
DNK
1.5
0
.5
1
1.5
2
0
1
3
0
1
HUN
2
3
0
.5
1
IRL
1.5
2
.5
1
1.5
ITA
2
2.5
LUX
−5
0
5
GRC
2
0
1
2
3
0
1
3
0
1
NOR
2
3
0
1
POL
2
3
0
1
PRT
2
3
SWE
−5
0
5
NLD
2
0
1
2
.5
1
1.5
2
0
1
Individual Values, Eur/Rich
2
3
0
1
2
3
4
0
.5
1
1.5
2
Total Attitudes, Eur/Rich
x−axis corresponds to individual education hi/h, and y−axis corresponds to total attitudes towards immigration
Figure A.4: Simulation - Individual Beliefs and Predicted Values, Immigrants from Rich European
Countries
45
1.5
2
.2
−.6 −.4 −.2
.2
0
.5
1
1.5
2
−.2
0
.2
ESP: Rmc = −.048, tc = .255
.5
1
1.5
2
0
FIN: Rmc = .189, tc = .409
1
2
.5
FRA: Rmc = .011, tc = .263
1
1.5
2
GBR: Rmc = −.156, tc = .31
3
0
1
2
3
1
1
0
1
2
3
2
3
1
1.5
2
0
−.5
.5
0
1
2
Labor Market RoW/Poor
3
1
1.5
2
2.5
LUX: Rmc = −.05, tc = .298
.2
0
1
2
3
PRT: Rmc = −.313, tc = .24
1
.05
0
−.05
.5
2
−.2 −.1 0
2
.1
.01
−.02 −.01
1
1
POL: Rmc = .027, tc = .091
0
.2
0
−.2
−.4
0
0
NOR: Rmc = .004, tc = .381
1.5
0
3
1
−1
2
NLD: Rmc = .062, tc = .356
.5
ITA: Rmc = −.06, tc = .297
−2
1
−1
−1
0
0
0
−.5
0
0
IRL: Rmc = −.406, tc = .31
.1
2
0
1
2
.5
−.04 −.02
−.5
0
HUN: Rmc = −.347, tc = .405
−.2 −.1
2
0
1
2
3
0
1
2
3
SWE: Rmc = −.001, tc = .399
−.002 0 .002 .004 .006
1.5
.2
1
.1
.5
GRC: Rmc = −.116, tc = .242
2
0
−1
−.2
−.2 −.1
−.1
0
0
0
0
.1
.1
0
.2
.1
0
0
DEU: Rmc = .083, tc = .428
.5
1
CZE: Rmc = −.106, tc = .175
.02
.5
DNK: Rmc = .027, tc = .462
.5
0
−.2 −.1
−.02 −.01
0
−.06−.04−.02 0 .02
CHE: Rmc = −.069, tc = .2263
.4
BEL: Rmc = .01, tc = .435
.01
AUT: Rmc = .004, tc = .281
4
0
.5
1
1.5
2
L.M. + Welfare RoW/Poor
x−axis corresponds to individual education hi/h, and y−axis corresponds to lattent attitudes towards immigration
Figure A.5: Simulation - Economic Determinants, Immigrants from Poor R.o.W. Countries
BEL
CHE
CZE
DEU
−10 −5 0
5 10
AUT
0
.5
1
1.5
2
0
.5
1
2
0
.5
1
ESP
1.5
2
0
1
FIN
2
.5
1
FRA
1.5
2
GBR
−10 −5 0
5 10
DNK
1.5
0
.5
1
1.5
2
0
1
3
0
1
HUN
2
3
0
.5
1
IRL
1.5
2
.5
1
1.5
ITA
2
2.5
LUX
−10 −5 0
5 10
GRC
2
0
1
2
3
0
1
3
0
1
NOR
2
3
0
1
POL
2
3
0
1
PRT
2
3
SWE
−10 −5 0
5 10
NLD
2
0
1
2
.5
1
1.5
2
0
1
Individual Values, RoW/Poor
2
3
0
1
2
3
4
0
.5
1
1.5
2
Total Attitudes, RoW/Poor
x−axis corresponds to individual education hi/h, and y−axis corresponds to total attitudes towards immigration
Figure A.6: Simulation - Individual Beliefs and Predicted Values, Immigrants from Poor R.o.W
Countries
46
1
2
0
2
0
1
1.5
2
0
.5
1
1.5
2
2
0
−1
0
FIN: Rmc = .143, tc = .409
1
2
.5
FRA: Rmc = −.523, tc = .263
1
1.5
2
GBR: Rmc = −.435, tc = .31
3
3
1
2
3
1
1
0
0
−1
−1
.5
1
1.5
2
.5
ITA: Rmc = −.388, tc = .297
2
0
NOR: Rmc = −.167, tc = .381
1
2
3
1
1.5
2
2.5
LUX: Rmc = −.405, tc = .298
0
POL: Rmc = .123, tc = .091
1
2
3
0
PRT: Rmc = −.509, tc = .24
1
2
3
SWE: Rmc = −.212, tc = .399
0
1
2
.5
1
1.5
2
.5
0
−.5
0
−2
−4
−.2
0
−2
−.5
0
0
2
.5
.2
1
1
0
1
0
0
4
3
.4
2
2
−1
0
−1
1
NLD: Rmc = −.642, tc = .356
1
IRL: Rmc = −.418, tc = .31
1
2
1
1
0
−1
−2
0
0
2
2
1
1
0
0
HUN: Rmc = −.313, tc = .405
−1
2
0
1.5
−1
1
2
.5
GRC: Rmc = −.346, tc = .242
2
0
−1
−.5
0
−2 −1
−.5
0
0
.5
1
1
.5
ESP: Rmc = −.603, tc = .255
2
1.5
2
1
.5
.5
−1 −.5
0
−1
−1
0
0
−.5
0
DNK: Rmc = −.141, tc = .462
DEU: Rmc = −.44, tc = .428
1
1
.5
2
1
.5
CZE: Rmc = −.391, tc = .175
3
CHE: Rmc = −.482, tc = .2263
2
BEL: Rmc = −.367, tc = .435
1
AUT: Rmc = −.217, tc = .281
0
1
2
Labor Market RoW/Rich
3
0
1
2
3
4
0
.5
1
1.5
2
L.M. + Welfare RoW/Rich
x−axis corresponds to individual education hi/h, and y−axis corresponds to lattent attitudes towards immigration
Figure A.7: Simulation - Economic Determinants, Immigrants from Rich R.o.W. Countries
BEL
CHE
CZE
DEU
−5
0
5
AUT
0
.5
1
1.5
2
0
.5
1
2
0
.5
1
ESP
1.5
2
0
1
FIN
2
.5
1
FRA
1.5
2
GBR
−5
0
5
DNK
1.5
0
.5
1
1.5
2
0
1
3
0
1
HUN
2
3
0
.5
1
IRL
1.5
2
.5
1
1.5
ITA
2
2.5
LUX
−5
0
5
GRC
2
0
1
2
3
0
1
3
0
1
NOR
2
3
0
1
POL
2
3
0
1
PRT
2
3
SWE
−5
0
5
NLD
2
0
1
2
.5
1
1.5
2
0
1
Individual Values, RoW/Rich
2
3
0
1
2
3
4
0
.5
1
1.5
2
Total Attitudes, RoW/Rich
x−axis corresponds to individual education hi/h, and y−axis corresponds to total attitudes towards immigration
Figure A.8: Simulation - Individual Beliefs and Predicted Values, Immigrants from Rich R.o.W.
Countries
47
Table A.3: Decomposition of the Variances: Labor, Tax and Individual Effects
Location Labor
Tax
Individual
Country
Austria
Europe
14.5%
19.0%
66.5%
Row
-10.4% 16.0%
94.3%
(obs=3616)
Belgium
Europe
35.6% -55.8%
120.2%
Row
-14.6% 19.0%
95.7%
(obs=3176)
Czech Republic
Europe
8.4%
22.8%
68.9%
Row
-8.1%
31.9%
76.2%
(obs=2178)
Germany
Europe
49.4% -68.3%
119.0%
Row
-13.6% 31.1%
82.5%
(obs=5176)
Denmark
Europe
-11.1% 156.6%
-45.5%
(obs=2620)
Row
-2.9%
30.5%
72.4%
Spain
Europe
-53.0% 51.9%
101.1%
Row
-40.0% 53.0%
86.9%
(obs=2862)
Finland
Europe
27.0% -25.3%
98.3%
(obs=3726)
Row
12.3% -18.0%
105.7%
France
Europe
34.4% -30.8%
96.4%
(obs=2596)
Row
-28.5% 17.2%
111.4%
United Kingdom Europe
-63.5% 62.1%
101.3%
Row
-34.3% 52.1%
82.2%
(obs=3606)
Greece
Europe
-16.2% 106.6%
9.6%
(obs=4386)
Row
-12.5% 25.7%
86.8%
Hungary
Europe
-37.2% 89.5%
47.7%
(obs=2906)
Row
-21.0% 26.2%
94.7%
Ireland
Europe
-80.0% 85.6%
94.4%
(obs=3564)
Row
-70.8% 79.7%
91.1%
Italy
Europe
-46.2% 54.7%
91.5%
(obs=2180)
Row
-21.7% 43.5%
78.3%
Luxembourg
Europe
3.4%
-3.5%
100.1%
(obs=1834)
Row
-24.1% 35.7%
88.4%
Netherlands
Europe
25.7% -80.3%
154.6%
(obs=4286)
Row
-22.9%
-6.9%
129.8%
Norway
Europe
-20.3% -16.9%
137.2%
(obs=3756)
Row
-6.2%
-8.1%
114.4%
Poland
Europe
28.6% -10.3%
81.7%
(obs=3772)
Row
5.5%
-3.0%
97.5%
Portugal
Europe -265.5% 175.9%
189.6%
(obs=2622)
Row
-54.3% 60.9%
93.4%
Sweden
Europe
7.8%
-11.6%
103.8%
(obs=3364)
Row
-7.6%
-2.6%
110.2%
Switzerland
Europe
26.9% -30.4%
103.6%
(obs=3212)
Row
-35.4% 19.0%
116.4%
48