Wages, Amenities and Negative Attitudes

Waisman, Gisela; Larsen, Birthe Working Paper Wages, Amenities and Negative Attitudes Working paper, No. 4-2012 Provided in Cooperation with: Department of Economics, Copenhagen Business School (CBS) Suggested Citation: Waisman, Gisela; Larsen, Birthe (2012) : Wages, Amenities and Negative Attitudes, Working paper, No. 4-2012, Copenhagen Business School (CBS), Department of Economics, Frederiksberg, https://hdl.handle.net/10398/8519 This Version is available at: https://hdl.handle.net/10419/208572 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/ Department of Economics Copenhagen Business School Working paper 4-2012 Department of Economics – Porcelænshaven 16A, 1. DK-2000 Frederiksberg Wages, Amenities and Negative Attitudes Gisela Waisman and Birthe Larsen Wages, Amenities and Negative Attitudes Gisela Waismanand Birthe Larseny z January 25, 2012 Abstract We exploit the regional variation in negative attitudes towards immigrants to Sweden in order to analyse the consequences of the attitudes on immigrants’welfare. We …nd that attitudes towards immigrants are of importance: they both a¤ect their labour market outcomes and their quality of life. We interpret the negative e¤ect on wages as evidence of labour market discrimination. We estimate the welfare e¤ects of negative attitudes, through their wage and local amenities, for immigrants with di¤erent levels of skills, origin, gender and age. Keywords: Attitudes towards immigration, Geographical Mobility, Wages, Amenities. JEL classi…cations: J15, J31, J61, J71 1 Introduction Sweden has gone from being a land of emigration to a land of immigration. Immigration was insigni…cant until World War II. During the …rst post-war decades, there was a sharp increase in demand for labour and workers were recruited from other European countries. These early labour immigrants adapted fairly well and gradually became accepted in the cities where they settled. Since the 1970s, when the need for labour shrank substantially, immigration to Sweden has become increasingly restricted to political refugees and their families. No other auent nation in recent decades has accepted as many political refugees, per capita, as Sweden has. The share of foreign-born reached 15% in 2010, about half of them from non developed countries. Stockholm University, SULCIS and Institut d’Anàlisi Econòmica, CSIC. Campus UAB, Spain. E-mail address: [email protected] yCEBR, Insead and Copenhagen Business School, Porcelænshaven 16A, 2000 Cph. F, email: [email protected] zWe are grateful to Torsten Persson, Mahmood Arai, Morten Bennedsen, Lars Calmfors, Ethan Kaplan, Ian King, Anna Larsson, Åsa Rosén, Lise Duedal Vesterlund, Fabrizio Zilibotti and seminar participants at the IIES, Copenhagen Business School, The Nordic Workshop- Sønderborg, EEA-ESEM, EALE, University of Melbourne, Integrationskonferensen, IFAU, LACEA/LAMES, IFN, NCoE Welfare REASSESS and Sciences Po, Paris, for helpful comments and to Christina Lönnblad for editorial assistance. All errors are ours. 1 Recently, many studies have detected the existence of negative attitudes towards immigrants.1For example, the SOM-institute (Gothemburg University) has investigated attitudes towards immigration and refugees since 1986 and found growing resistance against receiving refugees until 1992, while thereafter the attitude has slowly been more generous. But, in 2005, still nearly half the population thought that it was a good suggestion to receive fewer refugees in Sweden. Studies making a comparison across European countries, for example Card et. al. (2005), …nd that Sweden is one of the countries with the most generous attitudes towards immigrants. Therefore, if we …nd any e¤ect, then immigrants’welfare potentially is even more a¤ected in other countries. Do these attitudes matter? Or is it just something people say but never act upon? We explore if actual discrimination is related to negative attitudes towards immigrants. If attitudes do not in‡uence the immigrants’welfare, then they are no indication of discrimination and may be less of a concern. The aim of this paper is to exploit the regional variation in negative attitudes towards immigrants in order to analyse whether the mobility decisions and the labour market outcomes of immigrants are a¤ected by such attitudes. We recognize that not every native with negative attitudes may discriminate, but we conjecture negative attitudes to be systematically related to discrimination. We develop a simple model that describes how discrimination a¤ects immigrants when they are capable of forming networks. In this model, …rms choose their optimal number of employees using two methods of search: they either advertise or …nd workers through networking. We …nd that more severe negative attitudes reduce immigrant wages and amenities, and that the impact on wages is weaker the more immigrants present in a region, through the networking effect. We also …nd that more immigration directly increases immigrants’wages through a networking e¤ect, as well as their amenities. In the empirical analysis, we can disregard the immigrants’initial geographical sorting by concentrating on a group of immigrants for which there is an exogenous source of variation in their …rst location in Sweden, given by a refugee settlement policy pursued by the government. We study the movements from this …rst location as indication of better labour market conditions and/or better quality of life. We take into account the fact that the immigrants are not a homogeneous group by considering various kinds of heterogeneity, by origin, by level of education, by gender and by age. The placement of refugees in a region may exacerbate negative attitudes towards them, therefore we consider data on attitudes measured prior to the refugee settlement policy. Identi…cation fails if some other factor that we are not considering determines both the level of attitudes, the share of immigrants and the di¤erences in wages and quality of life in the region. We test this by including a placebo group in 1Some examples are the Intolerance Report (Intolerans 2004) and Westin (2000). 2 our analysis, immigrants from developed countries, that we expect be very little a¤ected by attitudes. The idea is that if our estimation of the e¤ect of attitudes on wages and amenities is the result of some other factor that produces lower wages, we should estimate the same e¤ect on this placebo group. In a nutshell, we …nd that attitudes towards immigrants are of importance for the refugees, but they have no e¤ect on the wages or quality of life of immigrants from developed countries. The location pattern of refugees shows that their quality of life is lower when attitudes are more negative towards them. A reduction of negative attitudes from the average value to zero increases their quality of life by an equivalent of 10% of their wages. Immigrants also receive lower wages when attitudes are more negative. The same reduction in negative attitudes would allow them to obtain 5% higher wages. We begin by forcing the coe¢ cients in the wage equation to be the same for those refugees that stayed where they had been placed (stayers) and those who moved. Then we allow these coe¢ cients to vary and observe that the e¤ect on wages is only present for the movers. Some individual characteristics, for example the type of education or occupation, may determine that some individuals are more vulnerable to discrimination than others. More vulnerable individuals, being more a¤ected by negative attitudes, are also more likely to move if they have been placed in municipalities with high negative attitudes. Reducing negative attitudes towards immigrants from the mean value to zero would increase the movers’welfare by an equivalent of 17% of their wages, while the stayers only bene…t from an equivalent of 10% of their wages via amenities. When we take heterogeneity into account, we see that the e¤ect on wages is present for well educated, female, older refugees and for those coming from Eastern Europe and Asia. Well educated movers would have 18% higher wages if negative attitudes decreased from the mean value to zero. Our interpretation is that some of these well educated immigrants may be performing jobs below their skill levels if they live in a municipality with very negative attitudes. We acknowledge the fact that the residuals of the amenities and wage equations in the municipality of placement are positively correlated to the residuals of the amenities and wages equations in the target municipality if the refugee moves. If an individual has high wages given his or her observed characteristics at placement, the individual is very likely to have a high wage after moving. We incorporate various alternative correlations in residuals and the e¤ect of attitudes on wages and amenities is very little a¤ected. Related Research Our paper relates to research on the discrimination of immigrants in the labour market, migration decisions and, in particular, empirical research about Sweden. In a companion paper, Larsen and Waisman (2008), we introduce labour 3 market discrimination in a search model where …rms cannot direct their search to natives or immigrants. Discrimination in that paper is supposed to take place on entry, while it takes place on exit in this paper. Our setting relates both to research on individuals’ migration decisions (Sjaastad (1962)) and self-selection (Roy (1951)). Nakosteen and Zimmer (1980) and Borjas et. al. (1992) apply Roy’s self-selection framework to internal migration. Other studies analyse the internal migration decision in Scandinavia. Åslund (2001) …nds that immigrants to Sweden are attracted to regions with many immigrants, better labour market opportunities and many welfare recipients. Damm and Rosholm (2005) …nd that the hazard rate into the …rst job of refugee immigrants to Denmark is decreasing in the local population size and the local share of immigrants and that geographical mobility had large positive e¤ects on the hazard rate into …rst job thus suggesting that restrictions on placed refugees’subsequent out-migration would hamper the labour market integration of refugees. None of these studies considers the e¤ect of di¤erent attitudes towards immigrants on their migration decision. Henry (2008) shows that the probability of African American migrants choosing a city in the US is signi…cantly reduced by the level of race-based crimes against them and by racially intolerant attitudes held by whites and the poor evolution of the feelings of whites about racial diversity. In her analysis, she does not study how attitudes a¤ect labour market outcomes. Knabe et al (2009) analyse the e¤ects of right-wing extremism on the wellbeing of immigrants in Germany. They …nd that the higher vote shares for the extreme right are associated with a lower subjective well-being of immigrants. Moreover, educated immigrants are more strongly a¤ected by right-wing attitudes of the host population than low-skilled immigrants. As compared to our paper, this study uses a di¤erent measure of right-wing attitudes in the native population and a subjective measure of life satisfaction as they cannot infer quality of life from migration decisions. Several empirical studies (for example Bevelander and Skyt Nielsen (1999) and Arai et. al. (1999)) have found lower income and employment rates for immigrants than for comparable natives in Sweden. These studies cannot tell us if the di¤erences are caused by ethnic discrimination or di¤erences in unobserved characteristics of the two populations. By analysing the di¤erence in labour market outcomes in regions with di¤erent attitudes towards immigrants, we intend to test discrimination in a more direct way. Other studies perform di¤erent types of more direct tests of discrimination in Sweden (Rooth (2001), Åslund and Rooth (2005)). These studies focus on the labour market outcomes of certain groups of immigrants, while we consider that attitudes may a¤ect their migration decision as well. In the next section, we present a simple model guiding our empirical analysis. 4 2 The Model We consider a search and matching model where natives and immigrants search for jobs and …rms search for workers. For simplicity, we assume that …rms may supply vacancies directed towards immigrants or natives.2When describing a job, it is possible to indicate the preferences against immigrant workers, for example, by demanding excellent native language knowledge. On the contrary, some job descriptions explicitly stress the appreciation of cultural diversity. We will describe the full model for jobs directed to immigrants only. We incorporate two additional features into the model. First, workers can use not only formal methods of search, but also their social networks (friends and acquaintances) to get a job. Second, immigrants are subject to discrimination in the labour market by individuals with negative attitudes towards immigration. We assume the presence of negative attitudes towards immigrants in a region increases the separation rate of immigrant workers from the …rm. The …rm opening a vacancy does not know if discrimination will take place, it only knows that immigrants have a higher separation rate caused by random negative shocks to the preferences of co-workers, clients, etc. The shock is thus considered a sudden irrational behaviour, in the sense that it is not a decision which requires any optimization from the …rm or the workers point of view. As a consequence, the worker may be …red or may voluntarily quit when the discomfort caused by discrimination is strong enough.3 2.1 Matching We follow Fontaine (2007) setting up a simple search and matching model including social networks. We assume that …rms advertise vacancies VIdirected to immigrants, unemployment is given by uI;there are LIemployees, and the labour market tightness faced by immigrants is given by I= (VI+ILI)=uI. The transition rate for an unemployed immigrant is given by f(I)=1eI, and for the …rm it is q(I) = 1eI=Iwhere f0(I) = eI>0,f00 (I) = eI<0, q0(I)<0and q00 (I)>0: 2In Larsen & Waisman (2008) we assume that it is not possible for …rms to supply vacancies directed towards immigrants or natives. Therefore, any negative impact on immigrants through vacancy supply, will also a¤ect natives. The simplifying assumption in this paper allows us to ignore e¤ects on natives’wages which is not the focus of the present analysis. 3In Larsen & Waisman (2008), we assume discrimination takes place in the matching process instead (both alternatives are simpli…cations) and provide a justi…cation for the mechanism through which discrimination is assumed to a¤ect immigrants. 5 2.2 The Firm The …rm chooses the number of vacancies o¤ered to immigrants so as to maximize pro…ts subject to negative attitudes towards immigrants and networking e¤ects. Each immigrant worker produces yand receives the bargained wage, wI. A …rm chooses the optimal number of vacancies to advertise, VItaking into account that its employees also produce new applicants. Each …rm hin municipality jfacing immigrants therefore solves the following Bellman equation j ILj I= max Vn [yLj ih wj nLj ih kyV j ih +_ j ILj ih](1) st _ Lj Ih =j ILih +Vj Ihqj Is1 + ajLj Ih;(2) Networking happens at the rate j ILihf(I);where we assume that j I= mIj.is the discount rate, sis the rate by which jobs are destroyed and ajis the rate determining how negative people in a region are against immigrants. Matches between immigrants and the …rm are dissolved more often the higher negative attitudes towards immigration are. With identical …rms, using (1)-(2) and Kuhn-Tucker conditions, we obtain the non-trivial solution in steady state determining labour market tightness, I: ky qj I=ywj I +s(1 + aj)j Iqj I: 2.3 The worker Let Uj Ibe the present discounted value facing an unemployed immigrant and Ej I be the present discounted value facing an employed immigrant, where j=t; p denotes either the target municipality the worker considers moving to, t; or the municipality where he or she has been placed, p: Uj I=Qj I+fj IEj IUj I; j =t; p; (3) Ej I=Qj I+wj I+s1 + ajUj IEj I; j =t; p; (4) where Qj Iis the quality of life or amenities the immigrant enjoys in a certain municipality. Negative attitudes towards immigration induce discrimination in housing, schools, hospitals, streets that reduce the quality of life of immigrants, so dQj I=daj<0:Living in a region with a large share of immigrant may o¤er bene…ts such as a larger availability of services and goods oriented towards immigrants (for example food), networks that could help recent immigrants …nd housing, etc. These factors increase the quality of life of immigrants living in a region implying that dQj I=dIj>0. 6 Wages are determined by Nash bargaining and we assume that the bargaining power is a half, so that Xj I=Ej IUj I;where Xj I=ky=q j I= ywj I +sj Iq(j I)giving that ky =Xj Iqj Iand thereby we obtain wj I= 0:51 + j I+j Iky: (5) The labour market tightness faced by immigrants j Iis determined by: 2k+s1 + aj=1j Ik+j Ikqj I;(6) More severe negative attitudes, aj; j =t; p reduces the attractiveness of opening a vacancy such that labour market tightness falls and thereby immigrants face lower wages. The opposite is true for more networking, j I:4 2.4 Mobility We assume for simplicity that only unemployed immigrants make a migration choice. When individuals decide whether to stay in the region of placement or not, they compare the value of staying as an unemployed worker to the value of moving, taking into account mobility costs. Immigrants have heterogeneous mobility costs5that are assumed to be uniformly distributed, cI2(0;1). Workers with high mobility costs …nd it too costly to move, whereas low mobility costs workers …nd it more than worthwhile to do so. The marginal immigrant is de…ned as having mobility costs ^cIwhich makes him or her just indi¤erent between moving or staying where the worker has been placed. The condition determining the moving costs of the marginal worker is Ut I^cI=Up I:(7) As wages are endogenous we can use equations (3)-(4), (7), the wage equation (5) and the free entry condition which gives the following condition Qt IQp I=ky +t Ip I= ^cI=(ky):(8) Equation (8) gives ^cIas a function of the endogenous variable t Iand p I. The higher the di¤erence in labour market tightness, which captures both wages and employment probability di¤erences, the more people will move. This is captured 4The same equations for natives would be wj N= 0:51 + j N+j Nkyand 2k(+s) = 1j Nk+j Nkqj N:These equations would also be valid for any group that is not much a¤ected by negative attitudes towards immigration, such as the immigrants from developed countries. 5Mobility costs depend on factors such as family situation, age, education level, etc. 7 individual variables a¤ect mainly the cost of moving, while the di¤erence in municipal characteristics between the placement and the target municipality a¤ect mainly the di¤erence in quality of life. From the model, we furthermore have that networking is directly increasing in immigration if the worker is an immigrant, whereby we include the fraction of immigrants in the analysis. The wage function at placement is assumed to have the form: wp i=0Xp+0Zi+ui; 0Xp=1ap+2Ip+3(apIp) + n P l=4 0 l v Xp l; where Xpare municipal characteristics at placement, including negative attitudes (ap), the share of immigrants from non developed countries (NDC) (Ip) and other municipal covariates v Xpand Ziare the individual characteristics. uiis an error term. Similarly, the wage function at the target municipality has the form: wt i=0Xt+0Zi+vi; 0Xt=1at+2It+3atIt+ n P l=4 0 l v Xt l; where Xtare the same characteristics in the target municipality. The municipal covariates v Xjthat characterize the labour market conditions in municipality jare open unemployment, the share of income originating in the private sector (market support), the share of …rms with less than 50 employees (share of small …rms) and the share of individuals with more than high-school education living in the municipality (% well educated). Municipal tax rates are also related to the economic conditions in the municipality, wherefore they are also included as controls. We include …xed e¤ects at the labour market area level to capture additional labour market di¤erences across regions that are constant during the period of analysis. We additionally control for the number of asylum seekers that came from the same country during the period 1985 - 1994. We control for the following individual characteristics: education, age, age squared, gender and civil status. The change in amenities and cost of moving is assumed to have the form: (Qp iQt ici) = 0(XpXt) + 0Zi+wi; 0(XpXt) = 1(apat) + 2(IpIt) + n P l=4 0 l( v Xl p  v Xl t ) where (XpXt)is the di¤erence between the municipal characteristics at the placement and those at the target municipality. 14 The amenities depend on the same factors as wages plus additional geographic controls: latitude (that in‡uences how dark it becomes in winter) and the ten-year average minimum temperature in the winter (January to March). In the literature on amenities, it is common to hypothesize that people prefer moderate climates. The cost of moving is assumed to depend on individual characteristics: education, age, age squared, gender and civil status. By maximum likelihood we minimize the error term: p i=wp i0Xp+0Zi+0(XpXt) + 0Zi(9) in the observations where the immigrant is a stayer and t i=wt i0Xt+0Zi0(XpXt) + 0Zi(10) in the observations where she is a mover. We begin our analysis assuming that the coe¢ cients in the wage equations are the same for at the municipality of placement and at the target municipality, that is, 0=0and 0=0in (9) and (10). In practise, this is equivalent to minimizing the error term j i=wj i0Xj+0Zi+ (2s1) 0(XpXt) + 0Zi;(11) where jis the municipality where the refugee lives (por t) and sis an indicator equal to one in the observations where the refugee is a stayer. When we instead allow for separate coe¢ cients at both municipalities, providing in practise di¤erent coe¢ cients for stayers and movers, we control for …xed e¤ects at the county level (25 counties) in order to increase the degrees of freedom. In our base equations we assume that the residuals in the wage and amenities regressions at the placement and target municipality are independent of each other. This assumption may not be realistic. High ability immigrants that have positive residuals upon placement are likely to also have positive residuals after moving. We incorporate three alternative positive correlations in residuals (0.25, 0.50 and 0.75) in the estimation to see how results are a¤ected. First, we run regressions for all immigrants, then we allow for heterogeneity across immigrants with di¤erent origins, educational level, gender and age. Identi…cation rests on the assumption that the e¤ect of the variables of interest on the wages and quality of life is independent of the residual terms. Identi…cation fails if some other factor determines both the level of attitudes and the di¤erences in wages and quality of life in the region, through its effect on the residual terms. It could be imagined, for example, that a generally bad labour market causes poor outcomes for recent immigrants as well as negative attitudes among natives. The attitudes we capture in our measure were displayed more than ten years before the period of analysis, but a bad labour 15 market may be persistent over time. We include several covariates to control for the labour market conditions, but acknowledging that this is not su¢ cient we do the following. To check whether some other factor determines both the level of attitudes and the di¤erences in wages and quality of life in the region, we include another group in our analysis, immigrants from developed countries, that we expect not to be signicantly a¤ected by attitudes. The idea is that if our estimation of the e¤ect of attitudes on wages and amenities is the result of some other factor that produces lower wages, we should estimate the same e¤ect on this placebo group. There is no considerable di¤erence between these groups of immigrants with respect to individual characteristics. They have on average a similar age (39.5 for immigrants from developed countries versus 38 for the refugees), gender composition (58% versus 55% are women), civil status (64% versus 68% are married). Most importantly, their educational level is not that di¤erent. In a measure that scales from 0 (no education at all) to 6 (Ph.D. level), a value of 3 corresponds to high-school education, so that the variable "well educated" in our study refers to values 4 to 6. The average level of education of immigrants from developed countries is 3.9 (with a standard deviation of 1.5), while it is 3.5 (with a standard deviation of 1.4) for the immigrants in our sample. 6 Results 6.1 Probit estimation of the probability of staying where placed and e¤ect of the variables of interest on wages The …rst column in Table III shows the marginal e¤ects in the probit estimation of the probability that a refugee stays where he or she has been placed.15 The variables of interest are important in the migration decision. Immigrants are less likely to stay in a municipality with more negative attitudes and lower share of immigrants from non developed countries. A small increase in negative attitudes decreases the probability that a refugee stays by almost 35%. The e¤ect of a small increase in the share of NDC immigrants is even stronger, it increases this probability by almost 175%. These marginal e¤ects are large, but so are the e¤ects of better labour market conditions in general. The open unemployment does not seem to be important compared to having a large share of small …rms in the economy, a large private sector and a large share of well educated individuals in the population. Refugees are more likely to stay when taxes are high, probably re‡ecting the appreciation of public services …nanced by the taxes. 15 The coe¢ cients indicate the change in this probability for an in…nitesimal change in a continous explanatory variable and the discrete change when dummies change from 0 to 1. 16 Low educated immigrants are 12% more likely to stay than low educated, women are 2% more likely to stay than men and married refugees are 5% more likely to stay than those that are unmarried. Latin Americans are more likely to stay than East Europeans while Asians and Africans are more likely to move. This pattern by continent con…rms the mean comparisons. Age reduces the probability that a refugee stays. The geographic variables do not seem to be very important once all other variables are considered. The results of the estimation of the e¤ect on the wages of stayers correcting for sample selection bias are also presented in table III. We …nd no e¤ect of the variables of interest on the wages of the stayers and the same is true for the labour market conditions in the municipality. Low educated, young, female and married immigrants receive lower wages. Asians and Latin American immigrants in the sample get lower wages than East Europeans, while Africans receive the lowest wages on average. So far we have found that the variables of interest and the labour market conditions are very important in the migration decision, but they do not a¤ect refugee wages. Wages seem to be a¤ected only by the individual characteristics of the refugees. This would imply that a refugee placed in the capital Stockholm would receive the same wage if he or she had been placed in the poorest municipality in Sweden. More analysis is needed before we reach such a strong conclusion. We proceed then to consider simultaneously the e¤ect of the variables of interest on wages and amenities which implies making use of the information on wages for all refugees (not just the stayers) and extracting additional information from the migration decision of the refugees. 6.2 Simultaneous estimation of the e¤ect on wages and amenities In tables IV to X, the results are presented in …ve columns. The …rst two columns correspond to the coe¢ cients of the wage and amenity functions when we estimate equation (11). The last three columns correspond to the coe¢ cients of the wage functions of stayers (at the placement municipality) and movers (at the target municipality) and the amenity function when we lift the constraint forcing the coe¢ cients on the wage equation to be equal for stayers and movers. Table IV contains the results for the whole group of refugees in our sample, assuming that the residuals at the placement and target municipalities are independent of each other. We show the coe¢ cients of all explanatory variables except the number of asylum seekers coming from the same country of origin, the …xed e¤ects and the year e¤ects. In tables V to X we only display the coe¢ cients corresponding to the three …rst rows (the variables of interest), but the same covariates and controls are included in the regressions in all tables. When we restrict the coe¢ cients to be identical for movers and stayers, we 17 …nd that negative attitudes a¤ect both wages and amenities of the refugees in our sample. The measure we use is the natural logarithm of wages, so the coe¢ cients tell us the percentage increase in wages due to a small increase in the explanatory variables. Reducing negative attitudes from the average level of 0.5 to zero would allow these immigrants to have 5% higher wages. We do not …nd in this regression much evidence of networking, the coe¢ cient for the share of NDC immigrants is positive but not signi…cantly di¤erent from zero. The coe¢ cient for the interaction between negative attitudes and share of NDC immigrants is positive but small and not signi…cantly di¤erent from zero. The migration choice of the immigrants gives us an indication of the di¤erence in quality of life in the placement and target municipality and the cost of moving. A reduction in negative attitudes from 0.5 to 0 would increase the quality of life of refugees by an equivalent of 10% of their wages, while an increase of the share of NDC immigrants from zero to its average level (10%) increases their quality of life by an equivalent of 19% of their wages. Lower open unemployment, higher level of education in the population and higher share of income originating in the private sector increase the refugees’ wages and amenities. A higher share of small …rms in the municipality and lower municipal tax rates increase their welfare mainly through quality of life. Women get lower wages than men. Wages are higher for well educated, older immigrants and their cost of moving seems to be lower. Asian and Latin American refugees have lower wages than Eastern Europeans, while the African refugees receive the lowest wages. Married immigrants have a higher cost of moving. Asians and Latin Americans have a higher cost of moving and Africans a lower cost of moving than East Europeans. In the three last columns, we allow the coe¢ cients to di¤er for the wages at the placement municipalities (wages received by stayers) and at the target municipality (received by movers). We can examine which of these groups is more a¤ected by the variables of interest, but at the cost of accepting broader …xed e¤ects to increase the degrees of freedom. We …nd that negative attitudes a¤ect mainly the wages of the movers. If negative attitudes decreased from 0.5 to 0 movers would receive 7% higher wages and all refugees would enjoy an increase in their quality of life equivalent to 11% of their wages. Most covariates have similar e¤ects on the wages of stayers and movers and similar e¤ects on amenities as in the regression with identical coe¢ cients. Until now we have assumed that the residuals of the wages and amenities equations are independent of the placement and target municipalities. We explore now the consequences a correlation of residuals would have on our results. 18 6.2.1 Correlated residuals If, for example, a low educated immigrant has very high ability, the worker is likely to get a high wage that cannot be explained by the variables in our regression. If he or she moves to another municipality, his or her wage is likely to be high there as well. This means that the correlation in residuals is likely to be positive. If we could observe the wages of many immigrants before and after moving, then it would be possible to estimate this correlation. But most movers actually did move soon after the placement, long before we observe them in our sample. We will then simply introduce a wide range of correlations of residuals in our main regression and study how the coe¢ cients change with this introduction. In table V we present just the coe¢ cients corresponding to the variables of interest, but all regressions include the same controls as in table IV. We introduce three alternative correlation values (0.25, 0.50 and 0.75) with very small e¤ect on the coe¢ cients. Both the e¤ect of negative attitudes and of the share of NDC immigrants on the quality of life seems to be smaller the higher the correlation of residuals. The larger the correlation of residuals, the stronger the e¤ect of negative attitudes on the wages of stayers and the weaker the e¤ect on the wages of movers, but the changes are very small. Also the e¤ect of the variables of interest on the quality of life becomes smaller as correlation rises. Our interpretation is that assuming a positive correlation is a more realistic assumption that explains better the migration decision and therefore leaves less to be explained as quality of life and costs of moving in our regressions. In most of the following tables the results are displaced including both independent errors and a correlation of 0.50. In some cases we restrict to the results with correlated errors (where our results are weaker) for the sake of space.16 The immigrants are not an homogeneous group. We now study how they are a¤ected by the variables of interest depending on their education level. 6.2.2 Results for immigrants with di¤erent education levels In table VI it is investigated whether attitudes have di¤erent e¤ects on the immigrants’welfare depending on their education level. Low educated immigrants have completed high school education at the most. We have more detailed information on education (a seven level scale), but we prefer to divide into just two groups in order to minimize concerns about the di¤erences in the quality of education across countries of origin. One common criticism to discrimination studies is that di¤erences in wages between natives and immigrants to a large extent re‡ect di¤erences in the quality of education even when two individuals have formally reached the same level. As 16 We can provide results with independent errors and alternative correlations if requested. 19 we compare the situation of similar refugees across municipalities with di¤erent levels of attitudes, we are not a¤ected by this criticism. We would only be a¤ected if the employers in municipalities with more negative attitudes had better information about the low quality of education in the countries of origin than employers living in municipalities with less negative attitudes. This does not seem plausible. It does seem plausible that employers in municipalities with more negative attitudes perceive the quality of education as lower, but we interpret that as one form of discrimination. All the coe¢ cients for negative attitudes in the wage equations of the well educated refugees are negative and larger than the coe¢ cients we found in table V, but standard errors are large implying that most coe¢ cients are not sig- ni…cantly di¤erent than zero. When we assume independent errors it is only the coe¢ cient for the movers’ wages that is signi…cantly di¤erent from zero. The movers’wages would increase by 15% if negative attitudes decreased from the average level to zero. When we assume correlated errors instead, the e¤ect is found on the stayers’wages. But a positive coe¢ cient for the interaction between negative attitudes and the share of NDC immigrants means that the e¤ect of attitudes is only negative if the share of NDC immigrants in the economy is relatively low. Already in table V we saw that the larger the correlation of residuals, the stronger the e¤ect of negative attitudes on the wages of stayers and the weaker the e¤ect on the wages of movers. A rise in the share of NDC immigrants increases the quality of life of all well educated immigrants, but it reduces the wages of both well educated stayers and movers if errors are correlated. An increase in negative attitudes a¤ects only the quality of life of low educated refugees. A reduction of negative attitudes from the average level to zero increases the amenities of low educated refugees by 8 - 10%. An increase in the share of NDC immigrants increases both the quality of life and wages of low educated immigrants, indicating that networking is important for low educated immigrants only. 6.2.3 Results for immigrants with di¤erent continent of origin Table VII presents the same regression performed in four subgroups depending on the continent the immigrant came from. All regressions assume an error correlation of 0.5. Note that the number observations is quite small in some cases and these regressions are very demanding. Negative attitudes reduce the quality of life of the refugees from Eastern Europe and Asia, but they seem to a¤ect only the wages of the immigrants from Asia (particularly the stayers). An increase in the share of immigrants coming from non developed countries increases the quality of life of all refugees, independently of the continent they come from. This last e¤ect is stronger for Africans and Asians. 20 6.2.4 Results for immigrants with di¤erent gender and age Table VIII presents our results for di¤erent gender and age groups, assuming again an error correlation of 0.5. The quality of life of both female and male refugees are negatively a¤ected by an increase in negative attitudes towards immigrants and positively in‡uenced by a rise in the share of NDC immigrants. Both e¤ects are stronger for males than females. But only the females’wages are a¤ected by negative attitudes towards immigrants, particularly the female stayers. Reducing negative attitudes from the average level to zero would increase their wages by 7% (and the wages of all women by 6%). It would also increase the amenities of all women by 8% and the quality of life of males by 9-10%. An increase in the share of NDC immigrants increases the quality of life of males by 17%, while it increases the quality of life of women by 12-14%. The amenities of both refugees over and under 40 years old are negatively a¤ected by an increase in negative attitudes towards immigrants and positively in‡uenced by a rise in the share of NDC immigrants. The e¤ects are of similar magnitude for both groups. Negative attitudes seem to a¤ect wages of older immigrants (those over 40 years old) rather than younger ones. A reduction of negative attitudes from 0.5 to 0 would increase the wages of these immigrants by 8%. The e¤ect would be stronger for older movers, they would get 10% higher wages. The same reduction of negative attitudes would increase the quality of life of both older and younger refugees by approximately 9%. 6.3 Results for immigrants from Developed Countries We present in table IX the same regressions for a group of immigrants, who are not refugees and have never been placed. We study them as a placebo group. If some other factor that we have not considered in our regressions determines both an increase in the level of negative attitudes and a reduction in the wages and quality of life in the region,then we should estimate the same e¤ect on this group. We show the results for all refugees from developed countries (DC) assuming …rst independent errors and then an error correlation of 0.5. Then we present the regression results for females and for older immigrants assuming correlated residuals, as these are the groups for which we found stronger results in the previous analysis. Basically we can observe that negative attitudes have no e¤ect on the wages of DC immigrants and, when they a¤ect their quality of life (for older DC immigrants) it is actually in the opposite direction. We observe that the share of immigrants from developed countries increases the quality of life and in many cases even the wages of DC immigrants. This seems to indicate that developed 21 immigrants bene…t more from networking in the labour market. 6.4 Interpretation of the results These results may be evidence of discrimination of immigrants from non developed countries. The strongest e¤ects are found via the migration decisions of refugees, which can indicate that discrimination is a more serious problem in other areas than the labour market. Some potential examples are discrimination in schools, housing, hospitals, etc. We also …nd weak evidence of discrimination in the labour market. We interpret the fact that wages of well educated are more a¤ected than those of low educated as an indication that some well educated refugees may be performing jobs below their skill levels (for example, driving a taxi) if they live in a municipality with very negative attitudes. It is not necessarily the case that they get paid less for the same job, but it could be the case that they do not get access to jobs that correspond to their quali…cations. The fact that the wages of women are more a¤ected by discrimination than those of men may re‡ect the fact that they are less mobile as shown in table III. If men decide where their family lives, then it may be the case that a woman that su¤ers discrimination cannot move to a less discriminatory area unless her husband is also a¤ected by discrimination. Individual characteristics such as the kind of education or occupation may turn a woman more vulnerable to discrimination than her husband. The wages and the quality of life of immigrants from developed countries, our placebo group, are not a¤ected (or are a¤ected in the opposite way) by negative attitudes towards immigrants. This is an indication that we are not capturing the e¤ect of omitted variables that have a positive e¤ect on negative attitudes and a negative e¤ect on wages or amenities for all workers in a region. We provide two examples that may give a more concrete illustration of to what extent attitudes are of importance. The …rst example is Lund, a municipality in Skåne County, southern Sweden. The city of Lund has more than 76,000 inhabitants and is believed to have been founded around the year 990, when the Scanian lands belonged to Denmark. It soon became the Christian centre of Northern Europe with an archbishop and the towering Lund Cathedral. Lund University, established in 1666, is Sweden’s largest university. Lund is an island of immigrants’acceptance in a county where attitudes are very negative. Out of 91 refugees placed in Lund in our sample, 67 (74%) chose to stay there. Furthermore, 87 immigrants that had been placed in other municipalities chose to move there. The immigrants that chose to stay are on average younger (39 years old) and less educated (3.95 on a scale up to 7) than those who moved into Lund (41 years old and 4.36) and those who moved out (45 years old and 4.62). 51% of the immigrants placed in Lund came from Eastern Europe, mainly Poland. 42% of the immigrants that moved out went to 22 municipalities with large cities, mainly Stockholm. Only 20% of the immigrants that moved to Lund came from such municipalities, mainly Malmö. The second example is Dals-Ed municipality, in western Sweden, on the border to Norway. Its seat is located in the town of Ed, 366 km Northwest from Lund. In Dals-Ed there are about 400 lakes, a national park and several nature reservations. The northern-most oak tree forest of the province grows there. The municipality is also rich in ancient remains, around 60 grave mounds, stone formations and a stone circle from the late Iron Age are still preserved. It is the scarcest populated municipality in Västra Götaland County, with 6.7 inhabitants per square kilometer. Dals-Ed is one of the municipalities with most negative attitudes:In our sample we observe 67 immigrants that were placed in Dals-Ed during the period 1985 - 1994, but only one of them is still there by 2003. The average age of the immigrants placed in Dals-Ed was 43 and their average education was 3. 69% of the immigrants came from Asia, mainly Iran. Almost half of the movers went to municipalities with large cities, mainly Göteborg. The third example Härryda municipality, also situated in Västra Götaland County, is one of the municipalities with lowest negative attitudes and probably a more fair comparison to Dals-Ed. Its seat is located in the town of Mölnlycke, with about 15,000 inhabitants. Forests cover about half the municipality area and lakes about one twelfth. Out of 31 refugees placed in Härryda in our sample, only 12 (39%) chose to stay there. But, at the same time, as many as 49 immigrants that had been placed in other municipalities chose to move to Härryda. The average age and education of the immigrants placed in Härryda was 36.9 and 2.9, respectively. The same characteristics for the immigrants moving into Härryda were 35.6 and 2.8. We do not claim that attitudes is the only reason explaining why Lund and Härryda are more attractive for immigrants than Dals-Ed, but the numbers suggest it is one important reason. 7 Robustness Tests 7.1 Income instead of wages We have performed our study on the wages received by refugees, where part-time wages had been recalculated (by Statistics Sweden) as corresponding full-time wages. This is the best measure available to compare the situation of refugees in di¤erent locations, as it is not a¤ected by temporary unemployment spells while the refugee seeks for a new job after she or he has moved. As a robustness test, we will repeat the analysis looking at the e¤ect of the variables of attitudes on the refugees’labour income. This is the income result- 23 Table I Immigrants to Sweden from Selected Countries On average per year during the period 1990 –1994 Residence permits obtained by reason: Sample Country Refugees 1) Family reunion Labour market Total Ex-Yugoslavia 13860 3080 12 16952 1036 Poland 47 1356 19 1422 75 Romania 201 427 0 627 41 Russia 172 489 6 667 18 Ethiopia 623 575 0 1197 53 Somalia 1323 819 0 2142 118 Uganda 91 0 91 14 Cuba 88 0 88 4 Chile 135 436 0 571 36 Afghanistan 116 0 116 16 Bangladesh 86 1 87 13 Iraq 2663 955 0 3617 239 Iran 1542 1532 3 3077 171 China 78 220 19 317 21 Lebanon 596 650 0 1246 132 Sri Lanka 83 157 0 240 17 Syria 416 429 0 846 75 Turkey 401 1517 0 1918 94 Stateless/ Unknown/Oth.2) 892 1230 6 2128 65 Total coming from countries in our sample 23410 13872 64 37346 2230 Other countries 3) 69 8101 154 8254 Total coming from all countries 23479 21972 218 45600 1) Granted residence permits according to the Genève convention,de facto refugees, persons in need of protection,humanitarian reasons,special refugee quota 2) Stateless and unknown (mainly Palestinians), former Soviet Union and Peru, where many asylum seekers came from in the period. 3) For family reunions: mainly UK, Germany, USA,Philippines,Thailand, Vietnam. For labour permits: mainly UK,Germany,Netherlands, Greece, Canada, USA,Brazil,Japan, South Korea. Source: Migrationsverket 30 Table II Summary Statistics Obs Mean Std Dev. Min Max Individual Characteristics ln Wages 26425 9.74 0.25 7.60 12.33 Education 26168 3.52 1.37 0 7 Age 26425 38.04 9.30 18 64 Woman 26425 0.55 0.50 0 1 Married/Cohabitant 26425 0.68 0.47 0 1 Municipal characteristics where the Stayers live Negative Attitudes 11960 0.509 0.083 0.169 1.000 Share immig NDC 11963 9.3% 4.8% 0.6% 24.7% Unemployment 11963 4.3% 1.8% 0.4% 13.7% % well educated 11963 20.9% 7.7% 7.0% 42.9% Market support 11963 52.4% 6.9% 23.4% 69.0% Share small firms 11963 27.4% 4.8% 11.5% 40.0% Municipal tax rate 11963 30.8% 1.3% 26.5% 34.0% Min temp winter 11963 -4.6 2.3 -18.7 -1.3 Latitude 11963 58.52 1.67 55.37 67.17 Municipal characteristics where the Movers live Negative Attitudes 14414 0.508 0.084 0.169 0.886 Share immig NDC 14448 9.7% 4.8% 0.7% 24.7% Unemployment 14456 4.1% 1.7% 0.4% 13.4% % well educated 14462 21.2% 7.4% 6.8% 42.9% Market support 14462 53.2% 6.6% 24.7% 69.0% Share small firms 14456 27.2% 4.5% 5.3% 43.8% Municipal tax rate 14456 30.7% 1.3% 26.5% 34.1% Min temp winter 14448 -4.36 2.12 -20.0 -1.3 Latitude 14448 58.29 1.63 55.37 67.85 Municipal characteristics where the Movers were placed Negative Attitudes 14455 0.522 0.119 0.152 1.000 Share immig NDC 14462 5.3% 4.0% 0.5% 24.7% Unemployment 14462 4.3% 1.7% 0.9% 13.4% % well educated 14462 15.4% 6.6% 6.5% 42.9% Market support 14462 48.2% 7.1% 23.4% 69.0% Share small firms 14462 24.0% 5.1% 11.1% 43.7% Municipal tax rate 14462 31.5% 1.3% 26.5% 34.4% Min temp winter 14462 -6.30 3.54 -20.0 -1.3 Latitude 14462 59.31 2.45 55.37 67.85 Sample composed of citizens of the countries listed in table I that immigrated to Sweden in the years 85- 94,with the following proportions: Eastern Europe (46%), Asia (37%),South America (11%) and Africa (6%). The sample corresponds to the years 1996 –2003. Negative Attitudes is negative attitudes towards immigrants.Share immig NDC is the share of immigrants from non-developed countries.% well educated is the share of immigrants with more than high school education (education≥4). Market support is the share of income originated in the private sector.Share small firms is the share of firms with less than 50 employees. Min temp winter is the average minimum temperature in winter. 31 Table III Marginal effects in the probit estimation of the probability of staying where placed and Estimation of the effect on the wages of stayers correcting for the selection bias All workers Endogenous variables Prob(stayer) Wage of stayer Negative attitudes towards immigration -0.346 *** -0.11 (0.05) (0.12) Share of immigrants from non-developed countries (NDC) 1.743 *** -0.492 (0.16) (0.66) Negative Attitudes towards immigrants * Share of immigrants from NCD 0.773 (1.12) Well educated -0.116 *** 0.100 *** (0.01) (0.02) Age -0.020 *** 0.021 *** (0.01) (0.00) Age 2 0.0001 *** -0.0003 *** (0.00) (0.00) Woman 0.017 ** -0.101 *** (0.01) (0.01) Married / cohabitant 0.048 *** -0.017 (0.01) (0.01) Africa -0.159 *** -0.05 * (0.01) (0.03) Latin America 0,079 *** -0.06 *** (0.01) (0.02) Asia -0.019 ** -0.02 ** (0.01) (0.01) Latitude -0.017 (0.02) Average minimum temperature in winter 0.009 (0.01) Open unemployment 0.779 0.040 (0.56) (0.60) Share of well-educated inhabitants in the population 1.251 *** -0.059 (0.09) (0.15) Market support 0.409 *** 0.076 (0.11) (0.15) Share of small firms 1.277 *** -0.073 (0.15) (0.28) Municipal tax rate 2.397 *** -0.756 (0.67) (0.85) Fixed and year effects yes yes Observations 26488 11351 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. Regional fixed effects at the labour market area level.We further control for the number of asylum seekers from the same country of origin in the corresponding period. Standard errors clustered at the individual level displayed under the coefficients. 32 Table IV Simultaneous estimation –All immigrants –Independent errors Same coefficients Diff. coeff. stayers & movers Endog var Wages ΔAmenities 1) W Stayers W Movers ΔAmenities 1) Negative Attitudes -0.091 * -0.198 *** -0.049 -0.143 * -0.226 *** (0.05) (0.05) (0.07) (0.07) (0.05) % immigrants from NDC 0.368 1.890 *** -0.026 0.433 1.965 *** (0.37) (0.17) (0.54) (0.50) (0.17) Neg Attitudes *% immig NCD 0.040 0.654 -0.062 (0.63) (0.92) (0.82) Well educated 0.149 *** -0.034 *** 0.096 *** 0.190 *** -0.078 *** (0.01) (0.01) (0.01) (0.01) (0.01) Age 0.012 *** -0.006 ** 0.012 *** 0.012 *** -0.006 * (0.00) (0.00) (0.00) (0.00) (0.00) Age 2 -0.0001 *** 0.0001 *** -0.0001 *** -0.0001 *** 0.0001 ** (0.00) (0.00) (0.00) (0.00) (0.00) Woman -0.097 *** 0.009 -0.088 *** -0.106 *** 0.021 * (0.00) (0.01) (0.01) (0.01) (0.01) Married / cohabitant 0.005 0.026 *** 0.001 0.004 0.021 * (0.01) (0.01) (0.01) (0.01) (0.01) Africa -0.049 *** -0.036 ** -0.061 *** -0.042 *** -0.046 ** (0.01) (0.02) (0.01) (0.01) (0.02) America -0.024 *** 0.051 *** -0.002 -0.048 *** 0.074 *** (0.01) (0.02) (0.01) (0.01) (0.02) Asia -0.023 *** 0.035 *** -0.028 *** -0.027 *** 0.039 *** (0.01) (0.01) (0.01) (0.01) (0.01) Latitude -0.026 *** -0.026 *** (0.01) (0.01) Avge min temp winter -0.008 -0.009 * (0.01) (0.01) Unemployment -2.088 *** -6.757 *** -1.889 *** -2.546 *** -7.009 *** (0.30) (0.48) (0.39) (0.37) (0.44) Share well educated 0.209 *** 0.847 *** 0.136 * 0.263 *** 0.869 *** (0.05) (0.11) (0.07) (0.07) (0.10) Market support 0.219 *** 0.368 *** 0.403 *** 0.186 ** 0.292 *** (0.07) (0.12) (0.09) (0.09) (0.11) Share small Firms 0.095 0.408 *** 0.196 * -0.000 0.434 *** (0.09) (0.16) (0.11) (0.00) (0.14) Tax rate -0.322 -1.783 *** -0.384 -0.002 -2.035 *** (0.29) (0.56) (0.46) (0.01) (0.54) Fixed/year effect yes yes yes Yes yes Observations 25967 26013 1) ΔAmenities is the difference in amenities at placement and target municipality plus the cost of moving. * significant at 10% ; ** at 5% and *** at the 1% level. We control for the number of asylum seekers from the same country of origin.. The explanatory variables are defined as the differences in values between placement and target municipality.Standard errors clustered at the individual level displayed under the coefficients. 33 Table V Simultaneous estimation –All immigrants –Correlated errors Same coefficients Diff. coeff. stayers & movers Wages ΔAmenities 1) W stayers W movers ΔAmenities 1) Independent errors Negative Attitudes -0.091 * -0.198 *** -0.049 -0.143 * -0.226 *** (0.05) (0.05) (0.07) (0.07) (0.05) Share immig NDC 0.368 1.890 *** -0.026 0.433 1.965 *** (0.37) (0.17) (0.54) (0.50) (0.17) Att * Share immig 0.040 0.654 -0.062 (0.63) (0.92) (0.82) Correlated errors -Correlation: 0.25 Negative Attitudes -0.096 * -0.193 *** -0.062 -0.130 * -0.218 *** (0.05) (0.05) (0.06) (0.07) (0.04) Share immig NDC 0.241 1.666 *** 0.080 0.307 1.729 *** (0.38) (0.15) (0.50) (0.47) (0.15) Att * Share immig 0.101 0.491 0.033 (0.64) (0.86) (0.79) Correlated errors -Correlation: 0.50 Negative Attitudes -0.098 * -0.170 *** -0.069 -0.113 * -0.190 *** (0.05) (0.04) (0.06) (0.07) (0.04) Share immig NDC 0.116 1.376 *** 0.150 0.188 1.400 *** (0.38) (0.13) (0.47) (0.45) (0.13) Att * Share immig 0.143 0.304 0.097 (0.64) (0.80) (0.75) Correlated errors -Correlation: 0.75 Negative Attitudes -0.096 * -0.127 *** -0.069 -0.088 -0.138 *** (0.05) (0.03) (0.06) (0.06) (0.03) Share immig NDC -0.022 0.979 *** 0.136 0.073 0.967 *** (0.38) (0.09) (0.43) (0.42) (0.09) Att * Share immig 0.172 0.127 0.111 (0.64) (0.73) (0.71) Observations 25967 26737 1) ΔAmenities is the difference in amenities at placement and target municipality plus the cost of moving. The explanatory variables are defined as the differences in values between placement and target municipality. * significant at 10% ; ** significant at 5% and *** significant at the 1% level. Same covariates and controls as in Tables III and IV. Regional fixed effects at the labour market area when the coefficients are assumed to be identical for stayers and movers and at the county level otherwise.We further control for the number of asylum seekers from the same country of origin in the corresponding period. Standard errors clustered at the individual level displayed under the coefficients. 34 Table VI Simultaneous estimation –Heterogeneity by education level Same coefficients Diff. coeff. stayers & movers Wages ΔAmenities 1) W stayers W movers ΔAmenities 1) Well educated immigrants –Independent errors Negative Attitudes -0.166 -0.056 -0.173 -0.310 * -0.168 (0.13) (0.13) (0.15) (0.19) (0.11) Share immig NDC -0.662 1.451 *** -1.765 -1.190 1.489 *** (0.87) (0.37) (1.43) (1.12) (0.35) Att * Share immig 1.413 3.538 1.932 (1.46) (2.42) (1.88) Observations 8519 8537 Well educated immigrants Correlated errors Negative Attitudes -0.201 -0.042 -0.284 ** -0.222 -0.134 (0.13) (0.10) (0.15) (0.18) (0.09) Share immig NDC -0.033 0.995 *** -2.154 * -1.119 * 1.006 *** (0.90) (0.28) (1.26) (1.04) (0.27) Att * Share immig 1.700 3.991 * 1.624 (1.51) (2.14) (1.76) Observations 8519 8537 Low educated immigrants –Independent errors Negative Attitudes -0.033 -0.202 *** -0.001 -0.047 -0.202 *** (0.05) (0.05) (0.07) (0.07) (0.04) Share immig NDC 0.765 ** 1.813 *** 0.588 1.005 * 1.943 *** (0.35) (0.17) (0.52) (0.44) (0.17) Att * Share immig -0.622 -0.453 -0.857 (0.59) (0.89) (0.75) Observations 17448 17476 Low educated immigrants Correlated errors Negative Attitudes -0.028 -0.168 *** 0.014 -0.044 -0.169 *** (0.05) (0.04) (0.06) (0.06) (0.03) Share immig NDC 0.590 * 1.307 *** 0.903 ** 0.652 * 1.364 *** (0.34) (0.12) (0.43) (0.40) (0.12) Att * Share immig -0.612 -0.999 -0.534 (0.58) (0.73) (0.67) Observations 17448 17476 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. Same covariates and controls as in Tables III and IV.Error correlation: 0.50. 1) ΔAmenities is the difference in amenities at placement and target municipality plus the cost of moving. The explanatory variables are defined as the differences in values between placement and target municipality. 35 Table VII Simultaneous estimation –Heterogeneity by origin Same coefficients Diff. coeff. stayers & movers Wages ΔAmenities 1) W stayers W movers ΔAmenities 1) Eastern Europe –Correlated errors Negative Attitudes -0.057 -0.166 *** -0.021 -0.078 -0.188 *** (0.07) (0.06) (0.08) (0.08) (0.05) Share immig NDC 0.504 1.032 *** 0.410 0.879 1.247 *** (0.55) (0.21) (0.72) (0.63) (0.21) Neg Att * % immig NDC -0.449 -0.257 -0.789 (0.88) (1.17) (1.00) Observations 11656 11675 Asia –Correlated errors Negative Attitudes -0.220 ** -0.246 *** -0.220 * -0.144 -0.206 *** (0.11) (0.08) (0.11) (0.14) (0.07) Share immig NDC -0.419 1.699 *** -0.402 -0.469 1.472 *** (0.64) (0.21) (0.73) (0.82) (0.20) Neg Att * % immig NDC 0.931 1.240 0.980 (1.09) (1.25) (1.42) Observations 9734 9761 Latin America –Correlated errors Negative Attitudes 0.194 -0.061 0.200 0.286 -0.148 (0.17) (0.17) (0.17) (0.20) (0.13) Share immig NDC -0.357 0.954 *** -0.067 0.395 0.961 *** (1.25) (0.34) (1.34) (1.42) (0.33) Neg Att * % immig NDC 0.650 0.669 -0.908 (2.18) (2.39) (2.46) Observations 2871 2871 Africa –Correlated errors Negative Attitudes 0.314 -0.179 -0.056 0.060 -0.124 (0.28) (0.16) (0.45) (0.27) (0.13) Share immig NDC 2.034 1.806 *** 1.545 0.334 1.576 *** (1.48) (0.34) (2.30) (1.49) (0.35) Neg Att * % immig NDC -3.011 -1.243 -0.203 (2.63) (4.10) (2.59) Observations 1706 1706 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. See footnote in table V. Error correlation: 0.50. 36 Table VIII Simultaneous estimation –Heterogeneity by gender and age Same coefficients Diff. coeff. stayers & movers Wages ΔAmenities 1) W stayers W movers ΔAmenities 1) Females -Correlated errors Negative Attitudes -0.128 ** -0.163 *** -0.137 ** -0.112 -0.170 *** (0.06) (0.05) (0.07) (0.08) (0.05) Share immig NDC 0.199 1.142 *** -0.355 0.698 1.227 *** (0.44) (0.16) (0.50) (0.55) (0.16) Neg Att * % immig NDC 0.180 1.086 -0.460 (0.74) (0.84) (0.93) Observations 14126 14147 Males -Correlated errors Negative Attitudes -0.081 -0.180 *** 0.011 -0.128 -0.215 *** (0.10) (0.06) (0.11) (0.11) (0.06) Share immig NDC -0.146 1.711 *** 0.665 -0.541 1.673 *** (0.66) (0.20) (0.87) (0.71) (0.19) Neg Att * % immig NDC 0.364 -0.535 0.965 (1.10) (1.49) (1.18) Observations 11841 12866 Over 40 years old -Correlated errors Negative Attitudes -0.171 * -0.169 ** -0.105 -0.216 * -0.178 *** (0.09) (0.07) (0.10) (0.12) (0.06) Share immig NDC -0.689 1.762 *** -0.662 -0.946 1.732 *** (0.65) (0.24) (0.78) (0.82) (0.23) Neg Att * % immig NDC 1.339 1.571 1.893 (1.08) (1.33) (1.35) Observations 10749 10767 40 years old and younger -Correlated errors Negative Attitudes -0.033 -0.172 *** -0.044 -0.030 -0.196 *** (0.06) (0.05) (0.07) (0.07) (0.04) Share immig NDC 0.650 1.177 *** 0.620 0.918 1.215 *** (0.42) (0.14) (0.52) (0.47) (0.14) Neg Att * % immig NDC -0.767 -0.489 -1.142 (0.71) (0.89) (0.81) Observations 15218 15246 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. See footnote in table V. Error correlation: 0.50. 37 Table IX Simultaneous estimation –Immigrants from Developed countries Same coefficients Diff. coeff. stayers & movers Wages ΔAmenities 1) W stayers W movers ΔAmenities 1) All immigrants DC -Independent errors Negative Attitudes 0.053 0.233 -0.128 0.206 0.207 (0.12) (0.20) (0.12) (0.18) (0.17) Share immig DC 0.228 * 0.494 *** -0.010 0.488 ** 0.493 *** (0.13) (0.13) (0.09) (0.24) (0.12) Neg Att * % immig DC -0.044 0.136 -0.200 (0.22) (0.16) (0.45) Observations 6450 6450 All immigrants DC -Correlated errors Negative Attitudes 0.045 0.174 0.001 0.077 0.155 (0.12) (0.14) (0.12) (0.15) (0.13) Share immig DC 0.200 * 0.381 *** 0.077 0.507 *** 0.444 *** (0.12) (0.09) (0.09) (0.19) (0.09) Neg Att * % immig DC -0.042 -0.031 -0.293 (0.21) (0.17) (0.35) Observations 6450 6450 Females -Correlated errors Negative Attitudes -0.083 0.180 -0.110 0.009 0.142 (0.12) (0.16) (0.12) (0.16) (0.14) Share immig DC 0.085 0.344 *** 0.018 0.419 0.376 *** (0.11) (0.10) (0.09) (0.18) (0.09) Neg Att * % immig DC 0.107 0.077 -0.287 (0.18) (0.16) (0.30) Observations 3729 3729 Over 40 years old -Correlated errors Negative Attitudes -0.078 0.799 *** -0.069 0.037 0.541 ** (0.19) (0.27) (0.18) (0.27) (0.23) Share immig DC -0.051 0.183 -0.053 0.588 0.404 ** (0.24) (0.17) (0.19) (0.39) (0.17) Neg Att * % immig DC 0.487 0.276 -0.123 (0.38) (0.29) (0.69) Observations 2699 2699 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. Same covariates and controls as in Table III,except Share immig DC (share of immigrants from developed countries). Error correlation: 0.50. 38 Table X Robustness Tests -Effects on the Immigrants’ Labour Income Same coefficients Diff. coeff. stayers & movers Income ΔAmenities 1) Istayers Imovers ΔAmenities 1) All immigrants NDC -Independent errors Negative Attitudes -0.358 * -1.353 *** -0.440 -0.946 *** -1.625 *** (0.20) (0.26) (0.28) (0.25) (0.22) Share immig NDC 4.407 *** 7.392 *** -0.593 2.938 * 8.290 *** (1.32) (0.74) (2.09) (1.68) (0.75) Neg Att * % immig NDC -1.221 4.790 0.948 (2.22) (3.53) (2.81) Observations 82527 82708 All immigrants NDC -Correlated errors Negative Attitudes -0.325 * -0.944 *** -0.305 -0.836 *** -1.182 *** (0.18) (0.17) (0.23) (0.23) (0.15) Share immig NDC 2.899 ** 4.576 *** 2.467 0.997 5.241 *** (1.23) (0.49) (1.74) (1.47) (0.53) Neg Att * % immig NDC -0.792 1.908 2.557 (2.07) (2.92) (2.47) Observations 82527 82708 Males - Correlated errors Negative Attitudes -2.222 -0.800 *** -0.080 -0.822 *** -1.132 *** (1.89) (0.32) (0.36) (0.32) (0.22) Share immig NDC 1.392 5.068 *** 5.736 ** 0.269 5.700 *** (1.62) (0.75) (2.68) (2.06) (0.76) Neg Att * % immig NDC 7.169 -3.580 3.2304 (22.1) (4.50) (3.47) Observations 40566 40660 Middle-age -Correlated errors Negative Attitudes -1.279 -0.827 *** -0.397 -0.830 *** -1.131 *** (1.56) (0.26) (0.27) (0.27) (0.18) Share immig NDC 2.456 * 3.877 *** 1.211 -0.249 4.611 *** (1.35) (0.60) (2.03) (1.74) (0.61) Neg Att * % immig NDC -10.65 2.731 3.778 ** (18.4) (3.41) (2.94) Observations 573595 57485 * significant at 10% ; ** significant at 5% and *** significant at the 1% level. Same covariates and controls as in Table III.Error correlation: 0.50. 39