A comparative study about restricting immigration policies across 22 European countries

. Elaborating a popular assumption about the effects of immigration policies on the integration of migrants, we argue in this article that more restrictive immigration policies lead to the selection of immigrants with greater integration potential, and that this selection should foster migrant integration. To test this argument, we combine country-level data from the Immigration Policies in Comparison database with individual-level data on economic, political and social integration from multiple rounds of the European Social Survey across 22 European countries. As a matter of fact, first of all, more restrictive immigration policies do not increase the likelihood of more educated migrants to be admitted, but they do make it more likely for migrants from European OECD countries to be admitted, while making it less likely for migrants outside the OECD. Secondly, we find that immigration policies affect some forms of economic, political and social integration outcomes, but mostly for immigrants from non-OECD countries. In conclusion, the immigration policies do affect integration outcomes but that these effects are small and limited to specific integration outcomes and migrants from specific regions. Our study, therefore, relativises the underlying popular assumption that immigration restrictions foster migrant integration, bearing important implications for the currently salient debates on immigration policy-making.


Introduction
In the context of increasing migration inflows, debates over how to best integrate immi-grants have become very important in Western societies over the last decades. These debates not only revolve around the effectiveness of integration policies but also around the role of immigration policies. It is often assumed that a more targeted selection of migrants according to criteria that are aimed at helping them integrate (which we shall call integration potential) foster migrant integration. This was for example an important issue in the 2017 German elections when both left and right liberal parties alluded to the idea: The Free Democratic Party (FDP) wanted to reform the Blue Card that allows high skilled migrants to migrate to Germany in order to reach better social and economic integration outcomes (FDP 2017, 69). In addition, they proposed to introduce a point system that considers age, language capacities and skills, as in their view integration courses are not enough to build modern immigration societies. The Social Democratic Party (SPD) also promoted a point system to improve what they explicitly called the 'migrants' capacity to integrate' (SPD 2017, 77). However, despite the practical relevance of this idea, systematic research on this topic is rare, and we do not know whether restrictive immigration policies indeed lead to better integration outcomes.
Migrants are fully integrated if they participate on an equal basis with natives in major institutions of the host country such as the labour market or the political system and if they feel recognised as a part of the national community (Alba and Foner 2015, 5). Accordingly, we focus in this paper on migrants' employment and socio-economic occupational status, their political trust and participation as well as their social trust and perceived discrimination. To study the degree of migrant integration in these different domains most studies so far have focused on the policy effects of integration and citizenship policies (e.g. Fleischmann  There are only very few studies that have looked at the effects of immigration policies, which regulate who is admitted and allowed to remain in a country. Waldinger (2003, 263) and Massey et al. (1998) pointed to this selection effect already two decades ago. More recently, several studies have focused on very specific aspects only, be it single policy changes and/or effects in a small number of countries (Rinne 2012, 12-13). Söhn (2013) is the only one who looks at the effects of both immigration and integration policies and shows that for Aussiedler in Germany language skills and classes had positive impacts on structural integration. Chiswick and Miller (2004) found that the way visa categories were organised in the 1990s had an effect on immigrants' language skills in Australia. Tani (2019) showed that the introduction of more restrictive measures for labour migrants in Australia in the late 1990s influenced the human capital of the affected group. There was, however, no detectable impact on indicators measuring immigrants' skill utilisation. The same policy change was exploited by Cobb-Clark (2003), who found that the more restrictive selection criteria lead to better labour market integration. Constant and Zimmermann (2005) as well as Cangiano (2014) investigate the role entry channels play in Germany and Denmark in the early 2003, respectively for several European countries in 2008. They show that, compared to labour migrants, humanitarian and family migrants have lower salaries and are less likely to be employed.
Against the back ground of the limitations of existing studies, we attempt for the first time to test in a systematic and comprehensive way the role immigration policies play in integrating migrants economically, politically and socially. We argue that if policies are more restrictive, more criteria need to be fulfilled to enter a country, which in turn increases the selectivity of the regulations. In other words, by increasing the integration potential of admitted migrants, restrictive immigration regulations should lead to better integration outcomes. We also argue that selection effects are more likely to emerge for migrant groups with a relatively low integration potential or relatively high social distance, by which we mean immigrants from poorer countries whose citizens have, on average, a lower level of education.
To test our arguments, we combine rounds 1-8 from the European Social Survey (ESS) to measure economic, political and social dimensions of integration outcomes with data from the newly built Immigration Policies in Comparison (IMPIC) dataset to measure immigration policy restrictiveness (Helbling et al. 2017) across a large number of Euro-pean countries. We mostly focus on how migrants are selected based on their education and regions of origin to measure their skills and social distance. We show that, first, more restrictive immigration policies do not increase the likelihood of more educated migrants to be admitted, but the likelihood of migrants from European OECD countries compared to migrants from other regions. Second, we find that immigration policies affect some forms of economic, political and social integration outcomes, but mostly for immigrants from non-OECD countries.

Immigration policies and integration potential
Migration selection processes can serve various purposes. In some cases, the very aim of these policies was to keep migrants in the country for only a short period of time and to prevent them from integrating as it was the case with many guest worker programmes (Ellermann 2013). After the oil crisis of the 1970s more and more policies have been implemented whose aim it is to attract highskilled migrants while deterring low-skilled migrants (Cerna 2014;Doomernik et al. 2009, ix). Selection based on skill does not only constitute a means to meet the demand for high-skilled professionals (Constant and Zimmermann 2005). High-skilled immigrants are also increasingly viewed to be easier to integrate economically, socially and politically, as they purportedly possess the necessary language skills and level of education (Doomernik et al. 2009, 10).
Cultural characteristics constitute another group of important selection criteria. Since the 2000s several Western European states have introduced pre-arrival integration tests (mostly focused on language capacities and country knowledge) whose aim it is to select migrants that are easier to integrate (Goodman 2012(Goodman , 2014 In a nutshell, among other criteria current immigration regulations include numerous conditions regarding economic and cultural requirements (Bjerre et al. 2016, 9). Increasing restrictiveness regarding these criteria aims at selectively restricting the group of migrants to those who are financially more independent, better qualified and culturally closer to the host society. As Cangiano (2014, 423) and Bonjour (2014) have already noted, the establishment of such conditions constitute selection mechanisms that might lead to better migrant integration as migrants have better 'starting positions' as Söhn (2013, 320) puts it. We refer to this idea as integration potential. Various studies have shown that integration gaps can indeed be explained by the integration potential in terms of migrants' skill levels and social distance to the host society (Granato and Kalter 2001;Heath and Martin 2013;Ebner and Helbling 2016).
We argue that this relationship holds for economic, social and political integration to the same extent as high education can be seen as a resource that helps immigrants access the job market, social networks and become politically active. Such migrants are better qualified for (high-skilled) jobs (economic integration) and share similar values and thus have fewer difficulties to connect to the host society and to be accepted by them (social integration). We also know from existing research that higher education leads to more political participation (Aleksynska 2008;Maxwell 2010) and economic and social capital to increased political trust and political participation (Jacobs and Tillie 2004). It can thus be argued that more restrictive immigration policies lead to better economic, political and social integration outcomes as migrants are selected based on their integration potential (H1).
This selection process can occur right at the border when it is decided who can enter the country and who cannot. It can also happen through self-selection when potential migrants with a low integration potential are deterred by restrictive measures and decide, for example, against applying for a visa or prefer to move to a more liberal country (Docquier, Peri, and Ruyssen 2014). Such decisions might be affected by initiatives through which states try to attract high-skilled migrants and thereby convey clear signals about the kind of migrants they want (Shachar 2006). Migrants might also improve their integration potential through step-wise migration (Paul 2011). They first move to countries that are easier to access to gain certain skills that allow them to migrate to more restrictive countries.

Diverging selection effects across country groups
It has been heavily debated for some time whether immigration policies have any effects on migration flows. If there are no policy effects on flows, there is no selection of migrants with preferred characteristics. Castles (2004) and Sassen (1996, 63-105) argue that state immigration policies have failed and sovereignty in this field eroded, among others because international human rights treaties and principles of freedom of circulation have prevented nation-states from closing national boundaries. Meanwhile, Messina (2007, 244) concludes that 'the declining sovereignty thesis is largely exaggerated and unsubstantiated by the facts'. Nowadays, nation-states have better means to control their borders (Freeman 1994) and often externalise controls to non-state actors or control migration through venue shopping (Guiraudon and Lahav 2000;Zolberg 2003).
Several studies have already shown that there are indeed policy effects on immigration flows (Hatton 2004; Ortega and Peri 2013; Fitzgerald, Leblang, and Teets 2014; Helbling and Leblang 2018). It can be assumed that these effects depend on the demand-side of migration that has not been considered by studies investigating the immigration policy effects. This has also consequences for immigration policy effects on migrant integration outcomes. If it is people with a high integration potential from OECD countries that primarily want to move, policy effects disappear as migrants fulfil the eligibility criteria in both generous and restrictive immigration countries. Selection and consequently integration effects can thus only be observed if there is variation in the demand for migration among both groups of migrants with high and low integration potential. This variation depends to a large extent on the sending countries' characteristics. As Söhn (2013, 304) puts it: While 'poorly' educated immigrants from more developed countries usually have at least basic literacy skills, immigrants coming from lower social strata of less developed states are less likely to be literate and will have a hard time finding qualified jobs.
As Migali and Scipioni (2018) show, the socio-economic status of persons who intend to migrate varies considerably across low-and high-income countries. The share of potential migrants in lowincome countries that have a tertiary degree is much lower than in high-income countries. Given the fact that there is a high correlation between potential and actual migrants (Tjaden, Auer, and Laczko 2019), it can be assumed that immigrants from such countries have on average a much lower integration potential than immigrants from high-income countries that consist mostly of immigrants with a high integration potential. It can therefore be argued that the selection mechanisms only have an effect for groups that consists of migrants from low-income countries with an average integration potential that is relatively low (H2).

Reverse causality arguments
The argument that immigration regulations affect integration outcomes is based on the assumptions that regulations affect migrant integration and not vice versa. It is plausible to assume that a country's immigration regulations become more restrictive because migrant integration is poor. However, even if this is the case it is not clear whether new regulations lead to better integration outcomes. It is still of interest whether policies are effective, irrespective of the reasons of their introduction. There would only be endogeneity if restrictive policies are introduced because integration outcomes are very good. But there is little reason to expect such a relationship.
More importantly, reverse causality could only become a problem if migrant integration outcomes were aggregated to the national level and regressed on immigration policies at that same level. In this paper, we hold instead that to assess the potential association between immigration policies and migrant integration outcomes, individual migrants must be matched with the immigration policy they were exposed to. Hence, in our models immigration policies are temporally prior to migrant integration outcomes and, therefore, these outcomes cannot determine the specific immigration policies the individual migrants were exposed to.
Integration outcomes might also be affected by return migration if migrants with high and low integration potentials have different likelihoods to return. However, it is far from clear whether such a bias exists. It can be assumed that people with a low integration potential have a higher likelihood to return exactly because they did not integrate, did not find a job and/or did not feel at home (Dustmann 1996;Dustmann and Weiss 2007). However, these people also have a higher likelihood to come from less attractive countries, which makes a return less probable and naturalisation more probable (Dronkers and Maarten 2012). Those with a higher integration potential might also have a higher likelihood to return (or move to a third country) as they are generally more mobile regarding financial resources and social capital.

Immigration policies
To measure the restrictiveness of immigration policies we draw on the Immigration Policies in Comparison (IMPIC) database (Helbling et al. 2017). The IMPIC measures policy outputs (i.e. actual laws and regulations rather than their implementation or their related policy outcomes) for 33 countries of the Organization for Economic Cooperation and Development (OECD) from 1980 until BCP Business & Management

EMEHSS 2022
Volume 25 (2022) 855 2010. It covers regulations regarding labour immigration, family reunification, refugee and asylum policies, and policies targeting coethnics. 1 For each of these fields the comprehensive IMPIC index captures the restrictiveness of entry conditions and eligibility criteria that define how difficult it is to establish legal residence in a country. Moreover, the rights and the security of status associated with a respective entry permit are included that stipulate for how long immigrants can stay on the territory and to what extent they are granted certain rights such as access to the labour market. 2 In addition, the IMPIC measures immigration control mechanisms that cross-cut these policy fields, indicating how strictly the regulations are enforced, and how undocumented migrants are treated. Using categorical principal component analysis (CATPCA), it has been shown that the regulations in the three policy fields of labour migration, family reunification and asylum can be reduced to a single and consistent empirical dimension (Schmid and Helbling 2016). The special and usually marginal category of co-ethnics as well as the variables capturing control mechanisms form separate dimensions.
Building on the latter result, in the following analyses we will thus use one comprehensive index that combines the regulations in the three policy fields of labour migration, family reunification and asylum (see also Helbling and Leblang 2018). These are the fields that nowadays constitute the main legal channels of migration into advanced industrialised societies (Messina 2007, 20-46;Schain 2008, ch.1). Regulations targeting coethnic immigrants are excluded, because they exist only in a few countries and, as already mentioned, concern a very special category of immigrants. Finally, control mechanisms do not concern selection and therefore are irrelevant for testing the argument.
In Table A1 in the Appendix we list all regulations that we used to build an additive index that measures the restrictiveness of immigration regulations and that varies between 0 (liberal) and 1 (restrictive) (for codebook and technical details see Bjerre et al. 2016). Figure A1 in the Appendix illustrates the distribution of the restrictiveness of immigration policies comprehensively for all three policy fields in our twenty-two host countries for the years 1980-2010. It shows that immigration policy restrictiveness in these fields varies over time within countries, providing within-variation for our statistical analyses, which include fixed-effects for the receiving countries. Furthermore, it reveals that immigration policies in most countries became more liberal over time (see also Schmid and Helbling 2016).

Migrant integration
To measure immigrant integration outcomes on the individual level for a sufficient number of immigrants, we pool eight waves (2002-2016) of the European Social Survey (ESS). ESS is the only survey that allows us to investigate migrant integration across a large number of countries and years. For the purposes of our study we only include first-generation immigrants, which are defined as persons who have been born outside the country of destination, and if at least one of their parents also fulfils this criterion. 3 From this group, we exclude all those who have entered as EU citizens under the EU free movement regime, as national immigration restrictions do not apply to this group. EU citizens that entered before the free movement regime was in place are included. Moreover, we only include persons who have been resident for 20 years or less in the host country as we assume that the selection effects become negligible after such a long time. As IMPIC covers the years 1980-2010, we only include persons that entered the host country in this period. Furthermore, we do not know whether persons below the age of eighteen came with their parents and therefore were not themselves selected by immigration policies. Therefore, we only included people that were at least 18 years of age when they entered the country.
Using the ESS for research on immigrant integration may have certain drawbacks. The data is produced in lengthy face-to-face interviews which are conducted only in (one of) the host country's language(s). This has two potential implications. First, questioning immigrants about their integration in such a personal setting may aggravate problems of social desirability, making the respondents likely to report more 'positive' outcomes. Second, the fact that language proficiency of (one of) the host country's language(s) is required for participation in the ESS may create not only a general under-representation of immigrants in the sample but could also bias the sample towards immigrants who are more integrated. Nevertheless, while the first problem constitutes a generally unresolved issue in survey research, the second problem is partly mitigated by the inclusion of several individuallevel control variables (in some of our models) that are also associated with varying aspects and degrees of integration.
Following Alba and Foner (2015, 5) we define integration as a process that allows migrants to participate in the important domains of their host societies and to feel accepted as part of that society. These domains include the educational and political system as well as the labour and housing markets. The questions included in the ESS do not allow us to measure all but some of the most important forms of integration, namely the integration into the labour market and the political system as well as feelings of acceptance. While there certainly exist further items to measure relevant forms of integration the questions included in the different ESS waves allow us to analyse some important aspects of economic, political and social integration (see Table A2 in Appendix for question wordings). 4 For economic integration, our measures cover two aspects of immigrants' labour market outcomes: employment and socio-economic occupational status. The former is a dummy measuring whether in the last five years the respondent has not had a longer period of unemployment that is three months or more and has also been employed by Fleischmann and Dronkers (2010) to study economic integration (see also van Tubergen, Maas, and Flap 2004). For socio-economic occupational status we use the ISEI scale (International Socio-Economic Index; Ganzeboom, de Graaf, and Treiman 1992). The ISEI is derived from the International Standard Classification of Occupations (ISCO) combined with information on education and income and constitutes a measure of the status of specific occupations. In order to make sure to only include individuals that are economically active, we limited the economic outcome indicators to respondents that are not older than 65.
To capture immigrant political integration, we use four indicators that have been employed by Helbling et al. (2016) to measure political integration. We measure, first of all, political interest and political trust. While the former is a single four-point variable, the latter is a composite of trust in political parties, in the national parliament, and in politicians, which are all 11-point scale items. Secondly, we measure conventional and unconventional political participation. The first is a fouritem additive scale of being member of a political party, working in a party or action group, contacting a politician or government official, or working in another organisation or association. 5 The second is a four-item additive scale of having worn campaign badges and stickers, having signed a petition, having taken part in a lawful public demonstration, or having boycotted a product. We acknowledge that it might not be fully clear whether unconventional forms of participation, particularly demonstration activities, can be seen unambiguously as instances of political integration. Yet we are interested in this aspect of political participation, because especially if immigrants may feel unwelcome in the receiving country and politically inefficient, they may fall back on such unconventional forms of political participation.
Finally, to measure social integration we use two indicators: social trust and not having perceived discrimination. We define social trust as generalised trust, which denotes trust in people that are personally unknown (Stolle 2002). Our measure combines three 11-point variables available in the ESS and which have been shown to form a reliable scale in previous research (cf. Rosenberg 1956) and that have been used in the study by Dinesen and Hooghe (2010) to measure migrants' integration. Our measure for not having perceived discrimination is a dummy variable that measures whether respondents have not perceived discrimination in at least one of three domains: colour or race, nationality and ethnic group membership. We consider this aspect of integration to be an indicator for social integration in contrast to, for instance, labour market integration, because the absence of discrimination perceptions as measured with this variable is not restricted to a specific social situation or organisational domain (such as the labour and housing markets or the education system).
To facilitate visualisation and comparison, we standardise all dependent variables from 0 to 100, using empirical minima and maxima. Higher values indicate higher levels of integration. In addition, the key independent immigration policy variable is standardised with empirical minima and maxima BCP Business & Management

EMEHSS 2022
Volume 25 (2022) 857 from 0 (most liberal) to 1 (most restrictive), so that the total effect size can be gauged directly (summary statistics are shown in Table 1).

Sample
Our sample covers a total of 22 European countries with a maximum of eight ESS rounds that were fielded from 2002 to 2016. About 6 ′ 500 first-generation immigrants that have resided in the country for no longer than 20 years are included (see Table 2).  We differentiate between three broad country of origin categories that can be distinguished according to their social distance to European destination countries. This allows us to differentiate between migrant groups with generally high and low integration potential. We first use the distinction between OECD and non-OECD countries of origin and thus migrant groups with different degrees of education (Migali and Scipioni 2018), and, accordingly, with a higher and a lower integration potential. Second, within the OECD, we distinguish between European and non-European countries. We therefore assume that there are three main types of migrants with varying integration potentials. Table 3 illustrates the distribution of migrants with high, medium and low integration potentials in European countries. It also shows the distribution of these three migrant types overall, as well as distinguished by a dichotomisation of the immigration policy that was in place in the year of entering the country of each individual migrant (liberal for values above the average, restrictive for values below the average). Across both categories (all) it appears that most migrants in European countries have a low integration potential, followed by migrants with a high integration potential and lastly those with a medium integration potential. Moreover, we see that the distribution of these three migrant types differs by the type of immigration policy. Countries with restrictive policies have a higher share of migrants with a high integration potential and a lower share of migrants with a medium and a low integration potential, which reflects our initial assumptions about the effects of relatively restrictive immigration policies on migrant composition. We observe the opposite if relatively liberal immigration policies were in place. This overall unequal but systematic distribution of the types of migrants is also reflected in Table A3 in the Appendix, which shows the distribution of the three main migrant types across countries and ESS waves. Note: The cutoff-point for liberal/restrictive immigration policy values is defined as the arithmetic mean (=0.33) of the matched immigration policy indicator (see Table 1; see chapter on the analytical strategy for details).

Analytical strategy
To prepare the analysis we match the year of entry of individuals with country-year observations of the IMPIC indicator (see Table A4 for a description of the immigration policy indicator). Hence, in the context of this study, the IMPIC variable is located on the individual and not the country-level. In other words, it mainly varies across individual migrants according to their year of immigration rather than across countries. The ESS asks migrants how long ago they came first to live in the country (rounds [1][2][3][4] or what year they came first to live in the country (rounds 5-8). As the ESS data does not allow accounting for the exact entry category of immigrants, we cannot match individuals with BCP Business & Management

EMEHSS 2022
Volume 25 (2022) 859 the specific immigration policy field that applied to them. However, the fact that the regulations in different immigration policy fields can be reduced to the same empirical dimension (Schmid and Helbling 2016) increases the validity of our matching approach. As the resulting matched immigration policy variable is highly skewed, we transformed it 6 to better capture the relevant variation towards the lower end of the scale (see Appendix Figure A2).
To fully ascertain whether and how immigration policies affect migrants' integration outcomes, our analysis proceeds in three steps. First, we explore whether more restrictive immigration regulations increase the likelihood of a migrant to have a higher education and come from a country that is socially closer to the host society. This is a test of the assumption that more restrictive policies lead to better integration outcomes by affecting the composition of the migrant population. Second, we examine whether more restrictive immigration policies increase migrant integration outcomes and whether skills and social distance function as selection criteria in addition to age and gender. Third, we interact the restrictiveness of immigration policies with our country of origin categorisation to test whether the policy effects are conditional on the social distance between host and destination countries.
To test our arguments, we need to proceed differently from the studies that investigate the effects of integration policies. These studies control for all relevant individual factors in order to investigate the effect of integration policies on comparable immigrants in different countries. In our case including many individual-level variables would distort our estimate of how integration is affected by the composition of the migrant population. One of the aims of immigration policies is to affect this composition of the migrant population regarding their integration potential, which in turn affects the likelihood of successful integration. Therefore, for each dependent variable we estimate four models in which we successively introduce individual controls that are potentially relevant for a successful integration in order find out how they alter the immigration policy effect. This procedure allows us to see which immigrant characteristics are most affected by selection processes. For instance, if the introduction of the education variable into a model leads to a reduction or disappearance of the policy effect, this shows us that immigrants integrate better if they have been selected based on their level of education. Models 1 have no controls. To test the most important selection effects, as discussed in the theory section, we proceed as follows: in Models 2 we include age and sex as basic individual controls, in Models 3 we include age, sex and education (to measure the effect of skills), and in Models 4 we include age, sex and country of origin categories (to measure the effect of social distance). However, note that in all models we include receiving country and survey rounds fixed-effects as well as further time-varying country-level factors to control for cross-national differences as well as differences over time in different survey rounds (see below).
We use linear regressions for all dependent variables. This is because, on the one hand, our ordinal variables sufficiently approximate linearity. On the other hand, we favour linear probability models to analyse our binary outcomes, since they yield very similar marginal effects as logistic regression models and are more straightforward to interpret. The application of linear probability models for binary outcomes is increasingly viewed as a viable strategy in sociology and economics (Mood 2010; Angrist and Pischke 2009). Given our scaling, the coefficients of these regressions indicate the increase in the probability to observe a positive outcome for a change in the independent variable from most liberal to most restrictive.
To control for the time-invariant host country context in these individual-level models, we add country and survey rounds fixed-effects. We also control for various relevant time-varying factors at the host country level. First, we include measures for mean xenophobic attitudes and the percentage of the foreign-born population (see Table A5 in Appendix for sources). 7 The former captures the degree of tolerance or hostility towards immigrants more broadly and thus reflects the reception context that might improve or worsen integration outcomes. The size of the immigrant population is important as a general context and might also tell us something about the restrictiveness of immigration policies. Second, we control for GDP per capita 8 and the unemployment rate, which both reflect the availability of jobs on the labour market and thus may affect immigrants' economic integration outcomes and contentment with the political arena as well as the social environment. To control for general assimilation effects, we include the native level of the respective dependent variable (see Tables A6 to A9 in the Appendix for summary statistics of all macro-and individuallevel independent variables). 9 Lastly, we introduce integration policies as a further control variable. We use the MIPEX scores (Huddleston et al. 2015), as these constitute the most comprehensive cross-national indicators to measure integration policies. The MIPEX captures political, social and cultural rights of immigrant groups, which have been shown to affect immigrant political integration outcomes (Helbling et

Results
We do not find that more restrictive immigration policies increase the probability of migrants having higher skill-levels, as measured with their years of education. Although the sign of the coefficient is positive, Table 4 shows that the association between policy restrictiveness and education is neither statistically nor substantively significant. 11 As appears in Table 5, however, the restrictiveness of immigration policies affects the composition of migrants by countries of origin. 12 The table shows the probability of a migrant in the host country to belong to one of the three migrant groups depending on the degree of immigration policy restrictiveness at the time of entering the country. Relative to very liberal policies, very restrictive policies increase the probability of a migrant being of European OECD origin by 25 percentage points. At the same time, greater immigration policy restrictiveness is associated with a decrease in the probability of a migrant being of non-OECD origin by 22 percentage points. Meanwhile, immigration policies do not have a discernable effect on the probability of a migrant being of non-European OECD origin. Thus, more restrictive policies privilege migrants from regions with a lower social distance, namely those of European OECD origin, and discriminate those from regions with a higher social distance, namely those of non-OECD origin. Robust standard errors in parentheses.***p <0.001, **p< 0.01, *p< 0.05 Note: Linear regressions include country and survey fixed effects as well as cluster corrected standard errors at the country-level. Sex and age were included as individual-level controls. Dependent variable scaled from 0 to 38, and independent variable scaled from 0 to 1.
After having seen that restrictive immigration policies select migrants based on their countries of origin but not based on their level of education, we now turn to the question whether restrictive immigration policies affect migrants' integration outcomes. Table 6 shows the means of the dependent and key independent variables by migrants' origin region and restrictiveness of host country immigration policies. We first see that across all countries of origin migrants economic and political integration outcomes differ by the type of immigration policy. Compared to countries with liberal immigration policies, political integration is a little lower in countries with restrictive immigration policies and economic integration is higher. The other integration outcomes are almost the same in both groups. Regarding countries of origin, positive effects of restrictive immigration policies on economic integration outcomes are most apparent for non-OECD migrants (i.e. low integration potential) and to a smaller extent also for non-EU OECD migrants (i.e. medium integration potential). For all migrant groups the means on some of the political integration outcome indicators are a little lower in countries with restrictive policies as Standard errors in parentheses.***p <0.001, **p< 0.01, *p< 0.05. Note: Predicted marginal effects estimated using multinomial logistic regression. Models include country and survey fixed effects as well as cluster corrected standard errors at the country level. Sex and age were included as individual-level controls; independent variable scaled from 0 to 1. non-EU OECD and non-OECD countries (H2). However, political integration outcomes for these two groups are lower in countries with restrictive immigration policies. Figure 1 shows the effects of immigration policies on various outcome variables, each of which is analyzed by four model specifications, as explained above (see Tables A12a to A12d in Appendix for  full regression outputs). We observe only one statistically significant regression coefficient: A shift from very liberal to very restrictive immigration policies is associated with a higher probability of not having been unemployed in the past five years. The effect size is between 10 and 13 percentage points, depending on the model specification. While accounting for skills does not change anything, controlling for region of origin reduces the estimated effect size to some extent. This suggests that social distance rather than the skill-level acts as a selection criterion mediating the effect of immigration policy restrictiveness on unemployment. Otherwise, we observe no statistically and substantively significant effects of more restrictive immigration policies on integration outcomes. We therefore reject our first hypothesis.
In the last step of our analyses we like to know whether the policy effects vary across migrants from different regions. As explained in the theory section, we expect larger Note: Entries are coefficients from linear regression models; dependent variables standardised from 0 to 100; independent variables standardised from 0 to 1; solid bars represent 95 per cent confidence intervals; models includecountry and survey fixed-effects and cluster-corrected standard errors at the country-level; DV= dependent variable.
Selection effects for migrants from non-OECD countries as compared to migrants from OECD countries, and especially compared to those from European OECD countries. In Figure 2 we display graphs with marginal effects for statistically significant interaction terms between the origin region variable and the immigration policy predictor (see Table A13 in the Appendix for full regression outputs).
First, non-OECD migrants are statistically different from European OECD migrants regarding the association between immigration policy restrictiveness and socio-economic occupational status. For migrants from outside the OECD, the socio-economic occupational status increases by almost 7 points when we they entered a country under a very restrictive instead of a very liberal immigration regime, though the net effect is still below 10 points on the 100 points scale.
Second, non-OECD migrants differ to a statistically significant extent from European OECD migrants regarding the association between restrictive immigration policies and employment. For migrants coming from outside the OECD, the chances of being employed versus having recently been unemployed increase by almost 14 percentage points when they entered a country under a very restrictive instead of a very liberal immigration regime. Non-European OECD migrants are not statistically different from non-OECD migrants but substantively, their employment chances increase with greater immigration policy restrictiveness, though in a slightly less pronounced way and starting from a higher absolute level of employment. Hence, the general immigration policy effect we observe for employment outcomes is mainly driven by specific selection effects for non-OECD migrants.
Third, we also observe diverging effects on immigration policies on political trust among non-OECD and OECD migrants. In this case, however, it appears that for migrants from the OECD area, greater selection reduces political trust while it remains stable across different immigration regimes for non-OECD migrants. The net effect is also below 5 points on the 100 points scale. We speculate that non-OECD migrants with a higher social distance may experience the political system more positively when they have been granted access under a restrictive immigration regime, perceiving the possibility for settlement as a reward. For OECD migrants with a lower social distance, highly restrictive immigration policies may symbolise a more hostile political reception context. Fourth, we see diverging policy effects on social trust. While for non-OECD migrants the association is non-existent, for non-European OECD migrants it is positive and slightly below 10 points on a 100 points scale. It might be that selection is perceived as a higher reward and therefore leads to more social trust. As social trust can also be inter-preted as a very general integration indicator, this result may point to the specific effectiveness of selection when it comes to migrants that are different yet proximate enough in social terms. Overall, we see that our second hypothesis can be confirmed for all three forms of integration investigated in this paper (but not all measurements of these forms of integration).

Conclusion
In this study we have shown that, first, more restrictive immigration policies do not increase the number of more educated migrants but the number of migrants from Euro-pean OECD countries compared to migrants from other countries. This finding adds important evidence to the literature that investigates how immigrants are selected on cultural characteristics (FitzGerald et al. 2017). Second, we find that immigration policies affect some forms of economic, political and social integration outcomes, but only for immigrants from mostly non-OECD countries. The effect on employment is limited to non-OECD and non-European OECD migrants. While for non-OECD migrants political trust remains stable, for OECD migrants it decreases when immigration policies are more restrictive. And while for non-European OECD migrants social trust increases with greater selection, for non-OECD migrants and European OECD migrants the restrictiveness of immigration policies only has a very small effect. It thus appears that the demand-side of immigration plays an important role in understanding selection effects. If mostly migrants with a relatively high integration potential intend to migrate, which is mainly the case for European OECD migrants, selection mechanisms become irrelevant.
We conclude that immigration policies do affect integration outcomes (Hypothesis 1) but that these effects are limited to economic integration or migrants that come from countries outside the OECD (Hypothesis 2). Therefore, when viewed from the angle of the broader literature, both integration and immigration policies appear to play a minor role when it comes to migrant integration. It can be noted however that our analyses confirm for a larger sample of countries earlier economic studies that focused on individual countries and showed that immigration policies affect migrants' economic integration (Cobb-Clark 2003; Constant and Zimmermann 2005;Cangiano 2014). This stands in contrast to the literature on integration policies that did not detect any policy effects on the economic integration of migrants (Fleischmann and Dronkers 2010). These findings put in perspective the widespread assumption that greater immigration policy restrictiveness has a general and substantial positive effect on various migrant integration outcomes.
While this study helps us better understand how immigration regulations affect migrant integration it also faces certain limitations. It needs to be kept in mind that migrant participants in the ESS surveys are most likely relatively well integrated. Also, a common limitation of studies that investigate policy effects concerns the fact that the role of relatively highly aggregated measures are studied. Finding overall null effects does therefore not mean that individual regulations do not have very concrete effects.
Besides tackling these measurement problems future research also needs to investigate why exactly policies are relatively ineffective when it comes to migrant integration. As recent research has shown, immigration regulations do affect migrant inflows (Hatton 2004; Ortega and Peri 2013; Fitzgerald, Leblang, and Teets 2014; Helbling and Leblang 2018). As we have shown in this study, however, this does not necessarily imply that the selected immigrants integrate better. This might be due to the fact that restrictive policies can lead to unintended effects when the number of irregular migration increases and return migration is discouraged (Brekke, Roed, and Schone 2016; Czaika and de Haas 2016; Czaika and Hobolth 2016). We also need to take a closer look at how difficult it is even for high-skilled migrants to integrate when it comes to the transfer of skills from one country to another, when they lack the required social and cultural capital, or when they encounter discrimination (see Steinmann 2019). This would also be a reason to investigate the relationship between immigration and integration policies as well as their interaction in producing integration outcomes. Hence, while we hope our analysis has broken important ground in the study of the potential effects of immigration policies on migrant integration, much remains to be done. What seems certain is that the currently salient and heated debates on immigration policy-making need more systematic scientific evidence.