How does the welfare state affect the restrictiveness of immigration policies?

(1)

1

How does the welfare state affect the restrictiveness of

immigration policies?

The relationship between immigration and the welfare state has garnered increasing interest in recent years. Yet, scholars rarely address the question of how the presence of a large welfare state affects the restrictiveness of immigration policies. Based on existing studies and theories, one could both expect a positive and negative relationship as well as varying affects by welfare state programme. This thesis investigates the effect of welfare state effort/generosity on the restrictiveness of immigration policies in 17 OECD countries over the period 1870-2010. Analyses show that until the 1980s higher total welfare state effort and generosity were associated with stricter immigration policies. Since then the effect of spending has largely vanished, while that of generosity remains. It is argued that this could be indicative of a trade-off between immigrants’ social and admission rights. There is also evidence that welfare programmes which entail the zero-sum redistribution of resources but not others negatively influence on the openness of immigration policies.

Maximilian Kiecker Student Number:2075768 Supervisor: Alexandre Afonso

(2)

2

Introduction

In recent years, few issues have been as consistently at the top of the political agenda of advanced industrial democracies as immigration. Not only did immigration play a major role during the UK’s 2016 Brexit referendum campaign, but it has also become a prominent issue in the remaining European Union and the United States (Becker, Fetzer and Novy, 2017). Trump's flamboyant pledge to “build a great wall” and let “Mexico pay for it” is one case in point (Sides, Tesler and Vavreck, 2018 p. 140). Similarly, the aftermath of the 2015 refugee “crisis” has created new divisions among EU leaders and many member states have witnessed the rise and mainstreaming of right-wing populist parties following the great recession (Keskinen, Norocel and Jørgensen, 2016; Krastev, 2017). The increasing salience of immigration has also been confirmed by political scientists, who have demonstrated that party manifestos today pay unprecedented attention to the topic(Alonso and Claro, 2011).Some even argue that immigration increasingly structures party political competition in post-materialistic societies (Odmalm and Super, 2014). An issue which has regularly taken centre stage in these political debates is the relationship between immigration and the welfare state. Right-wing populist parties typically not only call for strict immigration policies but also embrace welfare chauvinism by portraying immigrants as attracted by and a threat to national welfare states (Keskinen, Norocel and Jørgensen, 2016; Röth, Afonso and Spies, 2018). Germany’s AfD has for instance regularly warned that Germany cannot become “the world’s welfare state” or that “the welfare state cannot be expanded to Cape Town”(Häusler and Roeser, 2016; Bundestag, 2018). Such views are however neither unique to the populist right nor are they novel sentiments. Conservative politicians across Europe have frequently displayed similar sentiments over the last decades ranging from Griffiths notoriously racist 1964 election campaign to the CDU’s pledges to stop migration into social systems (Die Welt, 2019; Yemm, 2019). On the other side of the political spectrum, some left-wing parties have likewise argued for strong limits to immigration and have worried that immigration could undermine social progress (Alonso and Claro, 2011; Hinnfors et al., 2012). The rising prominence of these topics has also stimulated the interest of political economists, which has led to a boom in the academic literature on the immigration-welfare state nexus since the late 1990s (Straeten, 2013; Afonso and Devitt, 2016).

This literature which is broadly concerned with the question of whether open immigration and generous welfare states are compatible is characterized by heterogeneous research agendas. Scholars are currently predominantly concerned with four distinct but interrelated questions. These are: firstly, whether immigrants are net contributors or beneficiaries of the welfare state and hence how they affect its fiscal sustainability. Secondly, whether the welfare state can act as an immigration magnet that causes problems of adverse selection. Thirdly how immigration affects micro-level preferences for redistribution and reversely how the presence of welfare states shapes individual attitudes towards immigration. The fourth research area builds on this and seeks to determine whether immigration affects welfare policies and generosity on a more aggregate level. Even though these issues have received ample attention, scholars continue to fiercely debate the prospects of an open and generous welfare state.

In this debate, the “generic” view has long been one of pessimism (Brady and Finnigan, 2014). A widely cited 1986 paper by Freeman, for instance, holds that politically “migration has been little short of a disaster” for the welfare state (Freeman, 1986, p.61) and scholars such as Razin Cohen and Sarka (2008) warn that immigration poses substantial fiscal challenges for policymakers. More recently, Alesina and Glaeser’s (2005) influential work has also predicted that continued immigration

(4)

4

will put pressure on European welfare states and will lead to an Americanization of welfare politics. According to these scholars, immigration creates a progressive’s dilemma in which pro-welfare state parties need to choose between embracing either immigration or redistribution (Brady and Finnigan, 2014; Spies, 2018). In stark contrast, others hold that the relationship is inverse and expansive welfare provisions and immigration are politically complementary. Römer (2017), for instance, holds that generous universal welfare states promote inclusionary norms that politically undermine the pursuit of exclusionary migration policies. Similar Spies (2018), by drawing on insights from Esping Andersen (1990) and Pierson (1996), sees little risk of an Americanization of European welfare states, even in times of large scale migration. Scholars such as Fenwick (2019), Burgoon (2012) and Magni Berton (2014) even argue that there are good reasons to believe that increased immigration should boost support for redistribution and overall welfare state effort.

These markedly different positions are underpinned by a considerable variety of theoretical approaches which can broadly be distinguished as either focusing on economic or sociological factors. Starting with the former, numerous economic channels that link immigration and the welfare state have been identified. The most intuitive of these being the fiscal channel through which net contributing and benefiting immigrants directly affect the financial capacity of the welfare state. Accordingly, the so-called efficiency hypothesis holds that states will adjust their welfare effort as immigration increases (Fenwick, 2019). Another relevant channel is the labour market, which may matter in several ways. Firstly, classic economic theory holds that a general increase in the labour supply should benefit capital and reduce wages for labour (Borjas, 2018). Secondly, if immigrants are predominantly lower or higher skilled, they affect the relative factor abundance ceteris paribus leading to higher wages of the scarcer sector (Gaston and Rajaguru, 2013; Borjas, 2018). Thirdly, immigration may create labour market insecurity, which can either lead to calls for migration restriction or calls for more expansive welfare measures as a means of insurance as the so called compensation hypothesis holds (Fenwick, 2019). It should be noted that these economic explanations are prima facie indeterminant and the impact of immigration on the welfare state depends on the relative skill and income level of immigrants (Cohen, Razin and Sadka, 2008). The lack of determinism does not apply to the main sociological theories put forward. The most influential one of these is the ethnic fractionalization hypothesis advocated for by Alesina and Glaeser (2005), which posits that a generous welfare state requires a feeling of social solidarity with members of a clearly defined community. It is argued that immigration introduces ethnic heterogeneity, which weakens the feeling of social solidarity and can readily be exploited by anti-welfare politicians, thereby, undermining support for policies that appear to benefit migrants if not the entire welfare state (Brady and Finnigan, 2014). A more optimistic view is that the presence of the welfare state itself can positively affect citizens’ opinions of immigrants by promoting universal norms that run counter to exclusionary policies (Boräng, 2012).

These theories have attracted a significant body of empirical literature in recent years which is reviewed in the next chapter. However, the results continues to be mixed (Schaeffer, 2013; Straeten, 2013).Despite disagreements, the current literature has considerably improved our knowledge of the relationship between migration and the welfare state. Yet it still leaves many questions unanswered (Afonso and Devitt, 2016) and possesses serious blind spots. One such blind spot is the literature’s treatment of immigration policy. Not only is it seldomly a core variable of interest but it is also rarely included as a control variable in quantitative studies. This is surprising because after all immigration policy is the key lever through which governments can influence, if not manage, the scale, ethnic makeup, and skill level of immigrants as well as their post-entry social and labour market rights. It should consequently be of high theoretical relevance, given that these are precisely the factors that

(5)

5

should determine welfare state outcomes according to current theories. Concretely, governments concerned with the fiscal costs of immigration or its distributive effects may attempt to mitigate these effects by pursuing policies that alter the flow or skill composition of immigrants (Boräng, 2012; Römer, 2017; Kolbe and Kayran, 2019). Similarly, individuals and interest groups may demand particular policies to limit or expand immigration flows (Facchini and Mayda, 2009b; Boräng, 2012). For instance, workers fearing labour market competition may not only call on their union to lobby for an expansion of social insurance but also directly demand measures to restrict “outsiders” from entering the labour market or to favour the recruitment of highly skilled individuals that share the cost of welfare expansion (Facchini and Mayda, 2009b; Afonso and Devitt, 2016). Inversely, employers or high skilled voters could either support low skilled migration to expand the labour supply and boost their factor scarcity or take a restrictive approach to reduce the risk of upward tax adjustments and wage competition in the sheltered sector (Facchini and Mayda, 2009b). As these examples illustrate, immigration policy may very well vary systematically depending on the size and type of national welfare states.

The limited number of studies that have investigated this question do indeed suggest the existence of such a systematic link. Looking at the skill selectivity of immigration policy in 20 countries between 2000 and 2010 Kolbe and Kayran (2019) find that decommodification reduces the skill selectivity of immigration policy, while it rises with increases in social spending. In related fashion, Römer (2017) finds that greater welfare state generosity is associated with greater post entry rights for immigrants. Furthermore, a recent study by Kalm and Lindvall (2019) investigated how early welfare state development relates to the introduction of immigration controls in Europe between 1880 and 1920, a much earlier period than the 1990s and 2000s, around which most political economy studies are clustered (Kalm and Lindvall, 2019). Their analysis shows that early adopters of welfare policies (e.g. Germany and Denmark) were also pioneers in immigration control and that the adoption of welfare policies in countries such as the UK was followed by a tightening of migration restrictions (Kalm and Lindvall, 2019). This is particularly interesting as it supports the notion that welfare state construction and immigration restrictions went hand in hand, which has long been a standard assumption of historians studying nation-state formation (Oltmer, 2005).

As is apparent from this short overview, the lack of immigration policy studies presents a clear gap in the literature which is not explained by theoretical considerations. In fact, as I will show in the next chapter, many of the existing theories on immigration and the welfare state also have significant implications for immigration policy. The lack of studies on this issue can instead be attributed to the lack of comparable immigration policy data. However, this practical hurdle has recently been overcome with the creation of various migration policy indices most notably the IMPIC dataset (Helbing et al., 2017) and Margaret Peters’ (2017) low skill immigration policy index, which not only covers many developed economies but also goes back to the mid-19th_{century. Drawing on this index,} I seek to close the current literature gap in this thesis by answering the research question: “How does the welfare state affect the restrictiveness of immigration policies?”. In answering this question, I, firstly, want to determine whether welfare state generosity is positively or negatively associated with immigration policy restrictiveness. Secondly, whether this relationship and its strength are consistent across time or if it was for instance particularly strong during the nation-state formation in the late 19th_{century, a possibility raised by Kalm and Lindvall (2019). Thirdly, I want to find out whether the} effect on immigration policy is driven equally by all social programmes or whether specific areas of welfare spending are more strongly associated with stricter/looser policy restrictions. This is an important question given that, based on the welfare chauvinism literature, one may, as I will explain in more detail later, expect programmes that are perceived to favour immigrants to have stronger

(6)

6

immigration policy implications. To answer these questions, I follow a three-step approach drawing on various time-series-cross-section (TSCS) regression analyses that use Peters’ (2017) immigration policy index as the dependent variable and welfare state spending and generosity as main explanatory variables. In the first step, I present a barebones model which analyses the relationship between the welfare state and immigration policy over the period from 1870-2010. Thereafter, I analyse the pre-war, inter-war and post 1980 periods separately to see how this relationship has varied across time. Thirdly, I zoom in on the period 1980-2010, for which the availability of a larger set of independent and control variables allows me to determine the effects of individual welfare programs. Before discussing these analyses further, it is important to take a closer look at the theoretical link between welfare and immigration policy to outline testable hypothesis. This is what the next section will do. Thereafter, I give a more detailed account of the data and methods used here, before I present and discuss the main results in my final chapter.

Literature Review and Theory

So far, no comprehensive and systematic review of the relationship between existing theories and immigration policy exists, largely because of the rarity of studies addressing the issue head-on. It is, therefore, both sensible and necessary to explain, from scratch, how immigration policy fits into existing theoretical frameworks. To do so, this chapter starts with a review of the main theories that link immigration and the welfare state. Thereafter, immigration policy is introduced to the fray by explaining its types, how it should affect immigration decisions and how it is politically determined. Building on this, the third part of this chapter demonstrates what current theories of the immigration-welfare nexus imply for immigration policy and formulates the concrete hypothesis tested in this thesis. It also discusses why we might expect different relationships across time.

Theories of Immigration and the Welfare State

The key theories of immigration and the welfare state can as aforementioned be broadly classified as sociological or economic in nature. The most relevant economy-focused theories are the efficiency hypothesis and the compensation hypothesis, which will be discussed in turn. Thereafter, I look at the sociological explanations, which are the ethnic fractionalization theory, welfare chauvinism hypothesis and the generosity hypothesis. It should be noted that the terminology applied to these different theories is somewhat inconsistent in the literature.

Efficiency Hypothesis

The basic argument of the efficiency hypothesis is extremely straightforward. It is based on the assumption that immigrants moving to advanced industrial societies are predominantly lower-skilled and rely more heavily on welfare as they are more prone to unemployment and have more children while contributing less in tax (Nannestad, 2007; Burgoon, Koster and van Egmond, 2012; Römer, 2017; Fenwick, 2019). One theoretical reason to expect this pattern comes from the welfare magnet literature, which holds that low skilled immigrants who are more likely to rely on welfare programs tend to settle in generous welfare states, while net contributing migrants prefer lower tax jurisdictions (Borjas, 1999; De Giorgi and Pellizzari, 2009; Razin and Sadka, 2014; Soroka et al., 2016). Additionally, immigrants may also be in a weaker labour market position than similarly skilled natives. This is because immigrants may experience racially motivated discrimination and be less

(7)

7

likely to engage in collective bargaining due to their more precarious legal status (Afonso and Devitt, 2016; Römer, 2017). Consequently, open immigration is thought to result in a net-fiscal burden for generous welfare states, which face the choice between raising taxes, adjusting benefits downwards or both if they seek to consolidate their budget (Facchini and Mayda, 2009a).

The evidence for the efficiency hypothesis is mixed. While the welfare magnet hypothesis has been affirmed by various scholars (Borjas, 1999; De Giorgi and Pellizzari, 2009), others argue that the observed clustering of immigrants in welfare states may simply be a result of network effects and that other factors are more likely to drive migration decisions (Afonso and Devitt, 2016). Similarly, numerous studies find that immigrants may indeed be a net burden on the welfare states in the short term (Razin and Sadka, 2000; Gaston and Rajaguru, 2013; OECD, 2013). However, these negative fiscal effects are estimated to be relatively small (OECD, 2013) and some scholars, such as Boeri (2010), even find that taking all economic channels into account, immigration results in a slight fiscal gain. Furthermore, macro-level studies generally only find weak to no evidence for the efficiency hypothesis (Gaston and Rajaguru, 2013; Brady and Finnigan, 2014; Soroka et al., 2016; Fenwick, 2019). There are however some reasons to treat these negative results with caution. Most macro studies typically use immigration flows or the share of foreign-born population as an independent variable when it may in fact matter more whether voters and policymakers fear or anticipate a negative fiscal effect (Kvist, 2004; Burgoon, Koster and van Egmond, 2012). Taking this approach, Kvist (2004) shows that the fear of increased immigration as a consequence of eastern EU enlargement resulted in strategic changes to welfare eligibility in various EU member states, the majority of which also temporarily banned labour migration. Alternatively, the negative results could be explained by the so-called tax adjust model, which holds that the fiscal pressure of immigration is absorbed by higher taxes rather than benefit cuts (Facchini and Mayda, 2009b). If this explanation is true, there are good reasons to believe that it should affect immigration policy as is explained below.

Compensation Hypothesis

The compensation hypothesis as it relates to immigration is heavily inspired by an equally named compensation hypothesis from the broader welfare state literature (Fenwick, 2019). This equally named hypothesis was originally formulated by Rodrik (1998) to explain why open economies appeared to have larger governments. Rodrik (1998) holds that unlike the sheltered sector the trade-exposed sector is subject to volatility in foreign markets, which induces significant labour market risks for trade-exposed workers. He argues that these workers, in turn, demand an expansion of welfare programmes as a means of insurance against their increased labour market risks. Consequent studies appear to confirm this theory as they show that the size of the trade-exposed sector and volatility is indeed related to welfare expansion (Burgoon, 2001). Furthermore, Walter (2010) has recently provided data in support of the micro-foundations of this theory by demonstrating that Swiss trade-exposed workers are more likely to hold pro-welfare state positions and to vote for pro-welfare state parties.

According to the commonly used Ricardo-Viner and Stolper Samuelson trade models, it should not matter whether people or goods move as both are subject to factor price equalization, meaning that prizes between exporting/sending and importing/receiving countries should converge (Burgoon, Koster and van Egmond, 2012, p. 291-292). Based on the observation that immigrants into Europe are typically lower-skilled than natives, one would consequently assume that largescale immigration should impose downward pressure on wages and increase labour market risks (Burgoon, Koster and

(8)

8

van Egmond, 2012). It thus appears reasonable to assume that the compensation effect for trade in goods should equally hold for the movement of people (Burgoon, Koster and van Egmond, 2012; Fenwick, 2019). In this case, those working in sectors in which immigrant labour is concentrated should be particularly exposed to wage and labour market risks and similar to the trade-exposed sector demand an expansion of social protection as a means of insurance (Burgoon, Koster and van Egmond, 2012; Fenwick, 2019). Overall, increasing immigration that entails labour market competition should thus ceteris paribus lead to larger welfare states.

The empirical evidence for a compensation effect in response to immigration is stronger than for the efficiency hypothesis but also not yet conclusive. Based on individual-level survey data and occupational classifications, Burgoon et al. (2012) have demonstrated that workers in sectors with a high share of foreign-born individuals do indeed take more favourable views of redistribution, which are furthermore conditional on and amplified by the fear of unemployment and personal income losses. Furthermore Magni Breton (2014) finds that when citizens perceive immigrants more as a labour market risk than a fiscal liability their demand for redistribution increases. Likewise, Brady and Finnigan (2014), using foreign-born population and net migration flows as an independent variable and individual support for specific welfare programmes as their dependent variable, conclude that their results are most closely aligned with the compensation hypothesis. Moreover, looking at the macro relationship between immigration and welfare generosity, measured as social spending Fenwick (2019) shows that increases in immigration are associated with an increase in welfare state effort. In contrast, Burgoon (2012), studying party manifestoes, suggests that unlike trade and capital openness, immigration is more negatively associated with support for welfare expansion, which indicates that parties may be less likely to insure workers against immigration risks.

Sociological Explanations:

In addition to these two economic explanations, three sociological theories are also relevant here. By distinguishing these from the theories above, I do not mean to imply that they reject economic rationales completely. Instead, they simply place a stronger emphasis on factors such as ethnic diversity and norms like solidarity and deservingness.

Ethnic Fragmentation and the welfare state

Perhaps the most discussed and longstanding sociological account of the relationship between immigration and the welfare state stems from the idea that ethnic heterogeneity undermines the support for the provisions of welfare and other public goods (Schaeffer, 2013). The most prominent representatives of this school of thought are arguably Alesina and Glaeser (2005). The assumption that underlies their argument is that generous welfare states are preconditioned on high levels of social trust and a mutual feeling of social solidarity among citizens (Alesina and Glaeser, 2005). It is argued that solidarity and trust are particularly strong in relatively homogenous societies because citizens are more likely to identify and be empathetic with others that share similar personal characteristics such as moral and religious believes, cultural affinities and language (Alesina and Glaeser, 2005; Afonso and Devitt, 2016). Accordingly, individuals in highly homogenous societies, such as the Nordic countries, are assumed to be unusually willing to share resources with their compatriots, thereby, providing fertile ground for the growth of national welfare states. In contrast, individuals in societies characterized by ethnic and cultural heterogeneity, such as the United States, are thought to have less social trust and be less likely to share resources with others that are

(9)

9

perceived to be different. From this perspective, immigration, in particular from different cultural backgrounds, is seen as an impediment to the development and maintenance of welfare states (Alesina, Glaeser and Sacaerdote, 2001; Alesina and Glaeser, 2005).

The question of whether diversity can undermine the provision of public goods more generally is also discussed by economists, political scientists and sociologists (Schaeffer, 2013; Straeten, 2013; Soroka et al., 2016). Results of these studies are roughly split down the middle with a slight majority of findings being confirmatory (Schaeffer, 2013). The evidence of studies directly addressing the welfare state, however, remains contested (Straeten, 2013). Replications of Alesina and Glaeser (2005) regression analysis of ethnic fragmentation and welfare by Burckhard and Mau (2009), who replace ethnic fractionalization with non-western immigrants, similarly find a significant negative effect on the public support for the welfare state. They, however, highlight that the effects are economically small and hold that the prospect that Europe could face an Americanisation of welfare is unlikely. Others, similarly, urge caution, reviewing Alesina and Glaeser’s (2005) book. Pontusson (2006) questions whether Europe is not also more heterogeneous than implied by Alesina and Glaeser(2005). Similarly, Spies (2018) has highlighted that the US case may not be transferable to Europe. Because immigrants and minority groups elsewhere did not start from the same disadvantaged socioeconomic position as for instance African Americans(Spies, 2018). Furthermore, unlike in the US, European welfare states predate largescale immigration and may thus be locked in through path dependency (Spies, 2018).

Welfare chauvinism

A second sociological account that is complementary to the position of Alesina and Glaeser(2005) is the welfare chauvinism hypothesis, which holds that citizens’ support for redistribution may only extend to their compatriots but exclude immigrants (Brady and Finnigan, 2014). This hypothesis, however, does not rely on the degree of overall societal fragmentation but instead revolves around the concept of deservingness (van Oorschot, 2008). For several reasons it is believed that citizens view immigrants as less deserving of welfare support (Soroka et al., 2016). This belief is thought to be especially strong when the public perceives immigrants to be more reliant on welfare services and to present a net drain on social insurance. Importantly, the perception of deservingness is also thought to depend on the nature of individual welfare programmes (Van Der Waal, De Koster and Van Oorschot, 2013; Soroka et al., 2016; Römer, 2017). Programmes that are seen as particularly vulnerable to chauvinism are those that include zero-sum redistribution of resources and offer immigrants a comparable standard of living to natives (Brady and Finnigan, 2014; Cavaille and Ferwerda, 2016). Concretely, this means that programmes such as healthcare, housing and family benefits should bear the brunt of welfare chauvinism (Cavaille and Ferwerda, 2016; Soroka et al., 2016; Ennser-Jedenastik, 2018). Chauvinistic attitudes have also been shown to vary by welfare state regime with liberal and conservative welfare states more at risk than social democratic ones (Van Der Waal, De Koster and Van Oorschot, 2013). This is thought to be the case because these regimes vary in the extent to which they have institutionalized deservingness as a criterion to access welfare benefits through means-testing (Van Der Waal, De Koster and Van Oorschot, 2013; Römer, 2017). Evidence in support of this view comes from Van Oorschot (2008), who finds that immigrants are indeed perceived as less deserving of welfare than other vulnerable groups, such as the elderly. Furthermore, it is well established that citizens take a less favourable view of redistribution when immigrants are perceived as numerous and reliant on welfare. Macro studies also provide support to

(10)

10

the welfare chauvinism hypothesis. For instance, Fenwick (2019) finds that generally, the compensation hypothesis holds while also observing a chauvinism effect. Furthermore, Soroka et al. (2016) show that a higher share of the foreign-born population significantly reduces spending on employment-related programmes and pensions, while not significantly affecting other programmes.

Welfare State Norms and Immigration

A final, more optimistic sociological theory, which arguably has particularly strong implications for immigration policy, is the idea that the generous welfare state promotes inclusive and solidaristic norms (Boräng, 2015; Römer, 2017; Kolbe and Kayran, 2019). The idea is that universal welfare states create conditions in which it is perceived normal that individuals have a substantial right to the resources necessary to lead a dignified life. Universality is further important as it is based on values of equality by granting everyone the same social rights irrespective of past contributions, ethnicity, religion etc. Individuals that are socialized with these values are viewed as less likely to “categorize people based on a hierarchy of deservingness and hence less prone to think in categories of deservingness”, which underpin welfare chauvinism (Römer, 2017, p.177). A secondary rationale which supports this theory is based on the concept of decommodification and closely related to the compensation hypothesis. It contends that a high degree of decommodification, associated with Nordic welfare state regimes, insulates citizens from market pressure, including the above-discussed labour market effects of immigration (Römer, 2017; Fenwick, 2019). By shielding individuals form downward wage pressure and increased labour market insecurity, it is argued that individuals are less likely to view immigrants as competitors and to develop anti-immigration sentiments (Römer, 2017).

Scholars investigating this theory mostly find evidence in its support. Answering the question of whether generous welfare states grant citizens the same rights to welfare as immigrants, Römer (2017) found that more generous welfare states are more likely to pursue inclusive policies. Similarly, Boräng (2012, 2015) finds that generous and universal Nordic welfare state pursue more liberal refugee policies and Kolbe and Kayan (2019) show that these states are also less likely to deploy skill-selective immigration policies.

Before we look at what these theories imply for immigration policy, it is necessary to take a step back and look at the different types of immigration policy, how they are thought to affect migration flows and how immigration policy is set politically.

Introducing immigration policy

As outlined in the introduction, immigration policy is the principal instrument through which governments can seek to steer the number and composition of immigrants (Kolbe and Kayran, 2019). However, it represents a diverse and often complex policy field, in which legislators have a considerable variety of measures at their disposal which can be combined in multiple ways. Broadly speaking, we can categorize these measures as either requirements that focus on the point of entry or those that define the rights and obligations of an immigrant once inside the country (Sainsbury, 2006b; Ruhs and Martin, 2008; Kalm and Lindvall, 2019). The former include for instance passport and visa controls, work permit issuance or straight out immigration fees. The latter include but are not limited to permanent income requirements to preserve one’s status, restrictions on access to the welfare state or ultimately the ease of acquiring citizenship (Sainsbury, 2006b). In practice, the

(11)

11

strictness of chosen policy packages varies drastically not just across countries but also time. For instance, during the mid-19th_{century, most European countries did not even require passports for} immigration and generally had low to no restrictions on inwards migration (Torpey, 1999). Furthermore, immigration policy can either universally treat all immigrants the same concerning entry and post-entry requirements or can discriminate based on geographical origin, citizenship, income or skills. As is well known, EU countries for instance grant each other citizens and their family members full freedom of movement and, with very narrow exceptions, the same rights as their nationals, while applying more restrictive policies to third-country nationals (Kvist, 2004). Similarly, skill selective regulations, which frequently grant easier entry and extended rights to highly skilled immigrants, appear to be on the rise in advanced industrial societies (Liebig and Sousa-poza, 2004; Kolbe and Kayran, 2019). In fact, some countries such as the Netherlands now offer a tax incentive to these immigrants, which is not even granted to their nationals (Liebig and Sousa-poza, 2004, 141). To understand how these policy measures are thought to alter immigration patterns, it is useful to look at them in the context of a simplified account of the migration decision models as they are used in the welfare magnet literature.

These models essentially hold that a rational individual who has decided to migrate will choose their destination by comparing the expected utility that they will gain from living in these respective countries (De Giorgi and Pellizzari, 2009). Utility may depend on various factors such as the prevalence of unemployment, economic conditions and cultural considerations, but expected income is held to be of dominant importance and the sole focus of the widely used Borjas model (Liebig and Sousa-poza, 2004; De Giorgi and Pellizzari, 2009). The expected income is thought to predominantly depend on one’s relative level of skills and education (Borjas, 1999). Welfare policies affect migration decisions by altering the expected income of an individual at a specific skill level by granting them net transfers or demanding taxes to finance the welfare state. The expected income is thus comprised of the market wage plus welfare benefits net of taxes. Freeman (1986, p.55) calls this “the social wage” (Borjas, 1999). If a more welfare-dependent low skilled immigrant thus faces the choice between similar countries with and without a redistributive welfare state, they are thought to prefer a country with generous welfare states. The reverse applies to highly skilled individuals whose net income would be lower if social security contributions are subtracted(Facchini and Mayda, 2009b). This adverse selection like process forms the basis of the welfare magnet argument. Whether a migration decision is made and to which country an individual immigrates however also depends on the fixed cost of resettling, which, if it is high, may offset the wage gains one would incur from choosing a faraway country over a neighbouring one or could even make migrating completely unprofitable (Borjas, 1999).

What then does this imply for immigration policy? Bans and quotas should simply restrict the flow of immigrants. However, immigration policies that fall short of direct bans and simply increase general immigration costs such as passport, visa and work permit requirements can be conceived as a fixed cost that equally applies to low and highly skilled individuals. Like fixed reallocation costs, these should reduce the incentives to immigrate to a country or redirect some immigration elsewhere. Changes to post-entry rights, such as reduced or no welfare access, and skill selective policies, such as tax incentives for highly skilled immigrants, can equally be understood in the context of migration decision models. Reducing immigrants’ social rights should make a country less attractive to low skilled individuals as they are no longer entitled to a country's social wage. Skill selective tax incentives, on the other hand, should make it a more attractive destination for skilled individuals(Liebig and Sousa-poza, 2004; Kolbe and Kayran, 2019). Both restricting social rights and increasing skill selectivity should, therefore, shift the skill composition of immigrants upwards. In summary, we can see from this discussion that pre- and post-entry immigration policies, whether skill selective or not, should shift the overall skill composition of immigrants upwards and reduce the

(12)

12

“pull” of the welfare magnet. Policymakers may thus be free to pursue different approaches and for instance substitute restrictions on immigrants’ social rights for generally more open policies(Ruhs and Martin, 2008; Kalm and Lindvall, 2019).

The politics of immigration policy.

Bearing these considerations in mind for the discussion below, let us now very briefly turn to the political determination of immigration policy. It is generally agreed that policy decisions are shaped by a variety of factors, including the preferences of individual voters, interest groups, political parties as well as countries’ institutional features (Boräng, 2012; Peters, 2017; Röth, Afonso and Spies, 2018). A good schematic overview to make sense of how these factors and actors interact to create the demand and supply of immigration policy comes from Facchini and Mayda (2009), who adapt Rodrik’s (1995) well-known model of protectionism to immigration policy. According to this model, the demand for immigration policy is firstly determined by the preferences of individual voters and firms, which are channelled through interest groups, parties and social movement. The supply, on the other hand, is thought to be determined by the interests of policymakers, which are subject to institutional constraints of government.

Figure copied from Facchini and Mayda (2009, p. 4).

Importantly, this implies that the preferences of voters do not fully shape policy decisions, something that has been empirically well documented (Facchini and Mayda, 2009). Instead, the capacity of

(13)

13

voters to translate preferences into political demands hinges on the degree of organisation of voters and the relative power of interest groups and political parties. Similarly, on the supply side policymakers may have certain preferences and incentives stemming for instance from spill-overs of other policy areas, such as foreign and security policy, which for instance have historically shaped Germany’s early migration regime (Oltmer, 2005). This may also include the fiscal constraints set out by the efficiency hypothesis. In pursuit of their preferences, policymakers are however equally constrained by institutions and must factor in the demand for policies stemming from voters and interest groups (Facchini and Mayda, 2009). What then do these discussions imply for the immigration policy-welfare state nexus?

The welfare state and immigration policy:

As I outline in this section, three main hypotheses on the relationship between the welfare state and immigration policies can be derived from the forgone theoretical discussion these are:

Hypothesis 1: States with larger welfare states pursue more restrictive low skill immigration policies.

Hypothesis 2: Universal and generous welfare states pursue less restrictive low skill immigration policies.

Hypothesis 3: Welfare programs that are vulnerable to welfare chauvinism lead to more restrictive low wage immigration policies.

In addition, two secondary hypotheses can be derived which hold that the relationship between immigration policy and the welfare state should change systematically over time. Before I, however, turn to these auxiliary theories, let me explain each of my main hypothesis in turn.

Why larger welfare states may pursue stricter immigration policies:

My first hypothesis which posits a negative relationship between welfare state effort and immigration openness is primarily derived from the efficiency hypothesis. We recall that if this hypothesis holds, we expect immigration to be a net drain on the welfare state that creates pressure for budgetary consolidation which can either be achieved through tax hikes or welfare cuts(Facchini and Mayda, 2009b, 2009a; Fenwick, 2019). These choices implied by the efficiency hypothesis are however known to be politically unpopular. On the one hand, retrenching the welfare state is believed to be difficult because, as originally argued by Pierson (1996), the expansion of welfare states has itself created organized and powerful welfare constituencies that office-seeking politicians find harder to ignore than the more diffuse interests of taxpayers. Accordingly, this new politics of the welfare state perspective views retrenchment as unlikely no matter the partisanship of government (Pierson, 1996). On the other hand, the efficiency hypothesis on the relationship between capital mobility and welfare spending, after which the immigration efficiency hypothesis is named, argues that welfare spending is more likely to decline than taxes being raised (Burgoon, 2001; Fenwick, 2019). The argument draws on the tax-competition literature and holds that the existence of footloose capital limits the ability of government to levy higher taxes(Burgoon, 2001; Swank and Steinmo, 2002). Governments seeking budgetary consolidation thus face a highly uncomfortable situation if the efficiency hypothesis holds.

Immigration policy, however, may offer a way out of this choice between a rock and a hard place (Boräng, 2012; Römer, 2017, Nannenstad, 2007). This is because it allows governments to overcome

(14)

14

the adverse selection-like problem of the welfare magnet and consequent fiscal pressures in two ways (Boräng, 2012). Firstly, by subjecting immigrants to stricter skill and income tests, governments can identify potentially bad risks from a social insurance point of view and either deny them entry completely or reduce their post-entry entitlements to welfare programmes. Secondly, governments can choose to substantially increase the one-off costs of immigrating to the point where they outweigh the potential benefits of the welfare state received by low skilled immigrants. No matter which option or mix thereof is chosen, immigration policy restrictiveness for low skilled immigrants should increase and the larger the welfare state the more restrictive policies need to be implemented to offset the pull factor of the magnet (Römer, 2017). This also implies that policy restrictiveness should increase with welfare expansion.

Opting for stricter immigration policies rather than general welfare cuts or tax rises should also face relatively little political opposition. Unlike welfare constituents and taxpayers, immigrants do not possess political rights and are not in a position to mount any political resistance (Magni-Berton, 2014; Römer, 2017). Business groups and highly skilled individuals, who are thought to profit from low skilled immigration through the labour market channel, may equally not come to the defence of open immigration policies (Facchini and Mayda, 2009b; Borjas, 2018)This is because, if the tax adjusts model holds and is large enough to offset labour market gains, these groups should equally favour more skill selective policies over tax hikes (Facchini and Mayda, 2009a 2009b). Furthermore, as is outlined in more detail below, business groups and right-wing parties may not be as unified in their support of open immigration as is traditionally believed. This is because employers in the service sector could inter alia fear that immigration allows competitors to engage in wage competition (Afonso, 2012; Afonso and Devitt, 2016). Similarly, right-wing parties may also be ideologically more inclined to follow restrictive immigration policies especially in an age of post-materialism (Odmalm and Super, 2014).

In addition to this explanation based on the assumption that office-seeking politicians may make a rational and strategic choice to limit immigration rather than adjusting taxes or welfare spending, there are also sociological reasons for H1. These can be derived from the ethnic fragmentation hypothesis of Alesina and Glaser (2005). If citizens’ level of social trust and willingness to engage in redistribution depends on ethnic homogeneity, this implies that citizens in these societies may be more suspicious of newcomers from diverse backgrounds and more prone to adopt chauvinistic attitudes. This is what the widely used contact theory holds, which argues that trust towards ethnic minorities increases with the level of exposure (Schaeffer, 2013). Rather than accepting immigrants with different ethnic and cultural backgrounds into a homogenous society, governments may impose strict entry requirements, preventing an increase in ethnic fractionalization in the first place (Boräng, 2012 p.44; Römer, 2017, Nannenstad, 2007). Furthermore, the path dependency of the welfare state may, as argued by Pierson (1996), counteract the potential pressure for a reduction in overall welfare, as theorized by Alesina and Glaeser (Spies, 2018). In the face of increasing ethnic fragmentation, policymakers may thus again prefer to impose stricter immigration limits rather than pursuing a reduction in welfare generosity.

Distinguishing between the potential drivers of H1, i.e. fiscal concerns on the one hand and lower trust in people with different ethnic, racial or cultural backgrounds on the other hand, is practically difficult in quantitative research design as the one chosen here. This is because I can only observe the general relationship holding constant various controls. While I, as explained in my methodology section, control for the stock of immigrants in the country, which should be a rough proxy for ethnic fractionalization, I can only reduce the risk but not conclusively rule out that a fear of ethnic

(15)

15

fractionalization could drive increases in immigration policy in more homogenous societies. This should be kept in mind as a potential limitation of this thesis.

Why larger welfare states may pursue more open immigration policies:

An opposite conclusion to H1 can be derived from the theory that generous universal welfare states promote inclusive norms and the compensation hypothesis. According to the generosity hypothesis, individuals in countries with such welfare states are thought to be socialized in an environment that promotes egalitarian and solidaristic norms (Römer, 2017). This socialization could not only be incompatible with the differences in treatment between immigrants and natives (Römer, 2017) and high and low skilled immigrants (Kolbe and Kayran, 2019) but also contain anti-immigration preferences of voters and policymakers alike (Boräng, 2012). This would lead to lower demand and supply of skill selective and restrictive immigration policies. The more generous a welfare state is, the more such norms should be based on societal consensus.

There is also a second reason to believe that H2 should hold. This comes from the labour market effect that underlies the compensation hypothesis. In the broader literature on globalization and the welfare state, the compensation hypothesis is accompanied and flanked by the so-called embedded liberalism theory (Hays, Ehrlich and Peinhardt, 2005). Like the compensation hypothesis, this view holds that globalization entails increased labour market risks which could result in a protectionist backlash if trade and capital restrictions are loosened. To overcome this political resistance, politicians expand the welfare state to ensure exposed workers to gain their consent for more open trade policies (Hays, Ehrlich and Peinhardt, 2005). Relying again on the idea that the movement of workers and goods should have a similar impact, the same dynamic could be at play when it comes to stopping immigration. Consequently, large welfare states should result in increased levels of decommodification and ease labour market risks for workers (Kolbe and Kayran, 2019). Workers with access to such a generous social safety net should, in turn, be less concerned about labour market competition from immigrants and thus also moderate calls for the exclusion of labour market outsiders through restrictive immigration policies (Boräng, 2012; Kolbe and Kayran, 2019). If true, this would mean that in the presence of a generous welfare state, policymakers have less of an incentive to pursue restrictive immigration policies than in its absence. To be clear, I am not claiming that policymakers make a conscious trade-off and generally prefer increasing immigration, but that more open policies could simply be a side effect of other causal factors that drive welfare expansion. Practically differentiating between these two causal mechanisms is again difficult and something that should be kept in mind.

Why some programmes could lead to stricter immigration policies but not others:

Aside from H1 and H2, which spell out a general relationship between total welfare state effort and generosity and immigration policy, there are also reasons to believe that some welfare programmes are more strongly related to immigration policy than others. This third hypothesis can be derived from the propositions of the welfare chauvinism literature. As outlined above welfare chauvinism is believed to be influenced by welfare state regime types and characteristics of individual welfare programmes. Programmes vulnerable to welfare chauvinism are those that entail a zero-sum redistribution, raise the standard of living of immigrants and are perceived to disproportionately benefit them (Brady and Finnigan, 2014; Cavaille and Ferwerda, 2016). If such vulnerable programmes, for instance, healthcare, are thus sizeable in a country that is an immigration

(16)

16

destination, it may lead to increased levels of chauvinism, that can be exploited by anti-welfare and anti-immigration politicians(Cavaille and Ferwerda, 2016; Ennser-Jedenastik, 2018; Kolbe and Kayran, 2019; Forkert, 2020). Conventionally, this is assumed to result in cutbacks in these types of welfare programmes(Brady and Finnigan, 2014; Soroka et al., 2016). However, as in the case of the efficiency hypothesis, such general cutbacks again seem a relatively crude method to accommodate the feeling of unjust redistribution among welfare chauvinistic voters. Given that such cutbacks do not target immigrants especially but all voters that rely on welfare benefits and social services, they appear like a high-risk strategy for office-seeking politicians, at least if one takes Pierson (1996) seriously. Rather than face organized opposition from pro-welfare constituents, politicians may again find it more convenient to ease chauvinistic concerns by tightening entry conditions for immigrants or limiting their social rights. Both should in effect have an interchangeable effect(Ruhs and Martin, 2008) and limit the number of immigrants that rely on these programmes and reduce the risk of potential exploitation by chauvinistic politicians. Importantly, such strategies also seem more politically expedient because welfare chauvinism is in itself not opposed to the welfare state per se but rather its allegedly excessive use by immigrants(Cavaille and Ferwerda, 2016; Ennser-Jedenastik, 2018). Following the logic of welfare chauvinism, general welfare cuts would be throwing out the baby with the bathwater.

Given these factors, it appears more plausible that welfare chauvinism will lead to reductions in immigrant’s welfare rights and restrictive immigration policies rather than programme specific welfare cuts stipulated by Soroka et. al. (2016,p.3-4). The hypothesis postulated here is also arguably more in line with the literature on radical right parties than the conventional chauvinism hypothesis. Studying the social policies of populist right-wing parties Afonso and Rennwald (2018) show that many of these parties have increasingly adopted pro-welfare state views over the last decades. Furthermore, Ennser-Jedanastik (2018), who examines populist right party manifestos, shows that populist radical right manifesto pledges classified as welfare chauvinistic typically focus on limiting immigrants’ access rights rather than cutbacks that apply to natives and immigrants alike. There is also good evidence that populist radical right parties oppose welfare retrenchment when in government for fear of alienating their constituents (Röth, Afonso and Spies, 2018), which further indicates that politicians are unlikely to prefer a retrenchment strategy to the alternative of stricter immigration controls.

To summarize, three hypotheses for the relationship between the welfare state and immigration policy can be derived from the current literature on immigration on the welfare state. H1 posits a positive relationship between welfare state size and restrictive immigration policies. Either because rational office seeking policymakers react to the potential fiscal costs of immigration by imposing stricter immigration policies or because they react to a voter’s unwillingness to share resources with immigrants from different backgrounds. H2, in contrast, holds that generous welfare states reduce the restrictiveness of immigration policies. Either because the solidaristic norms of the welfare state conflict with exclusionary immigration policies or because voters in these states are less concerned with labour market competition by immigrants. Lastly, H3 holds that welfare programmes that are prone to chauvinism, i.e. involve a zero-sum like distribution and are perceived to disproportionately benefit immigrants, lead to calls for stricter immigration measures. Before the methods used to test these hypotheses and the results of my thesis are discussed, it is warranted to look at reasons for why this relationship could change over time, which is important given the lengthy period under observation here.

(17)

17

Time and Immigration Policy:

One reason to believe that the relationship between the welfare state and immigration policies should vary over time comes from Kalm and Lindvall (2019), who, as aforementioned, find that the early European adopters of national welfare states developed restrictive immigration policies at the same time or shortly after the creation of welfare programs. They attribute this primarily to welfare chauvinistic attitudes among citizens and policymakers but conclude that these attitudes were probably especially strong in this period and may not be as relevant in later historical periods and today. Their explanation for this suspected variation across time is twofold. On the one hand, they point out that countries simply lacked experience with the effects of novel national welfare programmes and may thus have been more prone to limit access to these institutions. On the other hand, they argue that solidaristic and universalist norms were not yet strongly developed because citizens had not been socialized in solidaristic welfare states. They thus believe that as the welfare state expanded such norms became increasingly widespread and policymakers learned how to respond to and deal with welfare chauvinistic attitudes. Accordingly, they expect the relationship between welfare generosity and immigration policy restrictiveness to decline over time (Kalm and Lindvall, 2019). In my results chapter, I refer to this argument as auxiliary hypothesis 1 (AH1). There is however also an argument to believe the exact opposite, i.e. that the relationship between welfare and immigration policies should be stronger today than historically. Globalization and technical progress have substantially lowered transportation and information costs of migrating (Facchini and Mayda, 2009). At the same time differences in the standard of living still vary largely across countries. The reduction in fixed relocation costs induced by technical progress means that migrating should ceteris paribus be more attractive today than it was 50 or 100 years ago. Similarly, it means that the pull factor of higher wage differences between sending and receiving countries as well as the welfare magnet effect should be stronger today which makes the absence of larger immigration flows surprising(Facchini and Mayda, 2009b). If countries want to maintain a stable rate and composition of immigration this should mean that the costs imposed by migration policy must offset declining transport and information cost. Likewise, policy measures to offset the pull factor of welfare expansion should be stricter today than in the past to have the same effect. Hence, my auxiliary hypothesis 2 (AH 2) expects that the effect of welfare state expansion on immigration policy should become stronger over time, which should be a good proxy for declining transportation costs.

Data and Methods:

To test the outlined hypothesis, I follow a time-series-cross-sectional research design. This chapter proceeds by outlining the sampling procedure and briefly recapping the chosen research approach before discussing the data collection and operationalization of the dependent, independent and control variables. Thereafter, the statistical model and applied diagnostic tests are described and justified.

(18)

18

Data was collected for 15 advanced industrial economies over the period 1870-2010. This period was chosen because on the one hand immigration or social spending data for key countries is not available before then and on the other hand, as outlined by Kalm and Lindvall (2019), the creation of national welfare states started in the late 19th century with Germany’s 1883 health insurance being seen as the first modern welfare programme. The included countries are Austria, Australia, Belgium, Canada, Denmark, France, Germany, Japan, the Netherlands, New Zeeland, Norway, the United Kingdom, the United States of America, Sweden and Switzerland. Additionally, data on Ireland and Spain were collected for the period 1950-2010 as pre-1950 Immigration data was not available. The reason for selecting these countries is threefold. Firstly, they are all OECD members with historically developed welfare states, a precondition for testing whether welfare state efforts affect immigration policy. Secondly, due to their economic status, they can be considered prime immigration destinations and as high-income countries should be subject to downward wage pressure as a consequence of immigration if the welfare magnet hypothesis is to be believed (Fenwick, 2019). Thirdly, other OECD countries were excluded for pragmatic reasons given that immigration policy data was either fully unavailable or in the case of Finland only available for a very select period of 3 years.

As already briefly mentioned in the introduction, I follow a three-step research approach firstly to test Hypotheses 1 and 2, secondly to find out whether the effects change over time and thirdly to see whether immigration policy is especially sensitive to particular welfare state programmes as held by H3. To carry this out, a series of pooled time-series cross-section models are used. Due to the difficulty of obtaining data for all relevant control variables over the entire period, I run a barebones model that covers the entire period as well as a more complete model with a larger set of controls from 1960 onwards.

These models are further split down into different periods, the first one ranging from 1880-1950, the second one from 1950-1980 and the last period from 1980-2010. These periods were chosen because they differ in terms of the politics of the welfare state. The 1880-1950 period represents the pioneering of national welfare states and early development, throughout which all industrial societies covered here developed their first welfare programmes. The period 1950 to 1980 is generally considered to be the golden age of welfare state expansion. The period from 1980 onwards is characterized by Pierson(1996) as the era of the “new politics of the welfare state”. Due to the election of new right politicians, such as Reagan, Thatcher and Kohl, and accelerating globalization, the welfare state arguably faced increasing pressure for retrenchment rather than expansion during this period. Furthermore, if Pearson (1996) is to be believed, the effect of partisanship on the size of the welfare state has also vanished since 1980.

In the final step of this research design, I look at the effect of changes to individual welfare programmes rather than the welfare state as a whole. This is important to test hypothesis 3 because, as explained above, the characteristics of individual welfare programmes may very well aid or limit the extent of welfare chauvinisms influence on immigration policy.

Dependent variable:

The availability of reliable immigration policy data has long been a major obstacle for scholars of immigration and the welfare state (Helbing et al., 2017). With the recent creation of various indices, this has largely been overcome. Nonetheless, the only available index that covers the entire period in question continues to be Margaret Peters’ (2017) low skilled immigration policy index. Others, such as the IMPIC database, remain limited to a very specific period, in the case of IMPIC 1980-2010

(19)

19

(Helbing et al., 2017), or, like the DEMIG (2015) collection of immigration policy changes, do not easily lend themselves to the type of quantitative analysis carried out here.

Luckily, the index drawn up by Peters (2014, 2017) offers several advantages in the context of this study. It is explicitly designed to measure the openness/restrictiveness of low skilled immigration policy, which should be more sensitive to changes in welfare state effort than policies that target highly skilled immigrants. It furthermore follows a transparent codebook (Peters, 2017 p. 244-294), which has allowed Kalm and Lindvall (2019) as well as Shin (2017, 2019) to substantially expand the number of covered countries. Thirdly, it is also designed to take into account the complexity of the immigration policy field and to capture the large variety of measures available to governments through its 12 sub-categories.

These 12 sub-categories, which are summarized in table A1 of the appendix, measure various low skilled immigration policy indicators, using an ordinal 5 point scale, which ranges from 1 for very restrictive to 5 for very open (Peters, 2017 p.244-294). To give an illustrative example from Peters’ codebook, the skill selectivity indicator ranges from 5 for no skill restrictions for any visas and no excludable classes to 1 only highly educated, high-income earners allowed and many excludable classes. In this case, a 3 indicates a preference for high skill workers, but many opportunities for low skill workers and some excludable classes (Peters, 2017,p.244-294). A full overview of the coding scheme including a list of the underlying policy measures is available in the appendix of Peters’ (2017,p.244-294) trading barriers. A disadvantage of this dataset is that because governments can seek to achieve the objective to tighten and restrict immigration access to the welfare state in multiple ways, it may not capture all strategies perfectly. This is especially true if governments for instance heavily restrict immigrants’ social rights but do otherwise pursue open immigration policies(Ruhs and Martin, 2008). In such a case the aggregate index would suggest a more liberal immigration policy, even though in terms of welfare state access, it is rather restrictive. Because the individual subcategories rely on an ordinal 5 point scale and change very infrequently a disaggregate TSCS analysis of individual categories is also not straightforward. This should be kept in mind as a potential limitation of this thesis.

Data on the 17 countries included here was collected from three databases. The data from Klam and Lindvall (2019) covers Austria, Belgium Sweden, Norway and Denmark from 1870-1920, with data from 1920 to 2010 stemming from Shin (2017, 2019). Data on Ireland and Spain (1950-2010) equally comes from Shins’ (2017, 2019) dataset with the remainder being drawn directly from Peters’ (2014, 2017) original database. Having merged these datasets, I followed Peters’ (2017) instructions and combined the 12 policy indicators into an immigration openness indicator using a component factor analysis. The factor loadings are presented in table A2 of the appendix. The key dependent variable here, immigration openness, ranges from -2.6 to 1.6 with a mean of -2.26e-09 and a standard deviation of 1 with higher values indicating a more open immigration regime and lower ones indicating a more restrictive one. The development of immigration policy in the 17 countries used in this sample matches the one used by Peters (2017) and Shin and are presented in figure 2. As we can see immigration policy in almost all countries has become increasingly strict overtime and was considerably more open before WW2.

(20)

20

Figure 2:

Independent variable:

How to quantitatively capture the expansion and retrenchment of the welfare state is a widely discussed issue in the welfare state literature, where it is known as the “dependent variable problem” (Green-Pedersen, 2004). While the size of the welfare state serves as the main explanatory variable rather than as the dependent variable in this study, the theoretical concerns of this debate are nonetheless relevant here. Traditionally, welfare state effort is measured as spending either as a percentage of GDP or per capita (Green-Pedersen, 2004; Fenwick, 2019). The former measure, on which I primarily rely on in this study, has several advantages. Practically, it extends back much further than per capita measurements as historical comparable population data is hard to come by. More importantly, however, the denominator, real GDP, captures the real economic output and allows for a comparison that is adjusted for inflation and exchange rate changes, for which a per capita measure would have to additionally be adjusted (Fenwick, 2019). For a more recent period spending as a percentage of GDP is also readily available for individual programmes, such as unemployment and pensions, which allows for the testing of hypothesis three.

Measuring the welfare state in terms of spending, however, also has some drawbacks. As Esping Andersen famously remarked “no one struggled for spending per se” as aggregate spending does not necessarily capture important dimensions of the welfare state(Green-Pedersen, 2004, p.5). Firstly, it does not account for differences in the ease of access of the welfare state, such as eligibility requirements or whether or not means testing is deployed. It also does not capture the actual

How does the welfare state affect the restrictiveness of immigration policies?