While there is a wealth of political economy research on support for redistribution, we know much less about support for inclusiveness. This is surprising, because political struggles have been fought not only about the generosity of the welfare state, but also about its inclusiveness. Who should have access to the welfare state has once again become a salient issue due to rising levels of immigration and the electoral success of the populist right across Western Europe and in the United States. Analyzing support for immigrants' social rights – that is, views on whether immigrants should have equal access to social goods and services, will therefore help us better understand these recent political developments.Footnote 1
Existing research in political economy assumes that voters want to exclude immigrants from the welfare state to protect their own access to valuable social goods and services. Since poor voters rely most heavily on social provisions, new demands from immigrants should lower their support for social rights for immigrants the most (Kitschelt Reference Kitschelt1995). Although it is intuitive that concerns about benefit competition influence support for immigrants' social rights, previous studies have struggled to provide strong empirical evidence for the claim that material self-interest and benefit competition affect these attitudes. This seems to be related to three main issues.
First, most studies have focused on aggregate voting behavior. While they show that immigration increases the vote shares of the populist right (see, for example, Cavaillé and Ferwerda Reference Cavaillé and Ferwerda2018; Rydgren and Ruth Reference Rydgren and Ruth2011), such aggregate analyses cannot shed light on how immigration affects individual policy preferences. Secondly, surveys measuring individual attitudes towards immigrants' social rights have long been scant, and the few existing studies rely on cross-sectional data (see Degen, Kuhn and van der Brug Reference Degen, Kuhn and van der Brug2019; Mewes and Mau Reference Mewes, Mau and Svallfors2012; Reeskens and van Oorschot Reference Reeskens and van Oorschot2012). While cross-sectional analyses provide some valuable insights, they struggle to exclude confounders at the individual or context level. Thirdly, prior research has often used the level of immigration to measure benefit competition, but this is a key problem for two reasons. On the one hand, immigration does not necessarily lead to competition for social goods and services. On the other hand, the residential patterns of immigrants across and within countries are not random. The self-selection of immigrants can lead to an underestimation of the effect of competition if immigrants choose to live areas with a welcoming native-born population.
To address these issues, this paper combines high-quality individual-level panel data on support for immigrants' social rights with a novel and exogenous way to measure benefit competition. This measure relies on administrative data from a mandatory refugee dispersal system that distributes new refugees across municipalities and allocates social housing to them.Footnote 2 By focusing on the local level, where most social goods and services are provided and consumed, and by tracking the attitudes of the same individuals over almost a decade, this article sheds light on the question of whether and when benefit competition lowers support for immigrants' social rights.
I focus on the Netherlands, which has the largest social housing market among the advanced democracies (OECD 2018). With more than one-third of the Dutch population living in social housing, the highly inelastic nature of the social housing supply and a mandatory refugee dispersal system that requires municipalities to provide permanent housing to new refugees, social housing has all the elements to become a site of conflict.
I find that benefit competition has a meaningful effect on individuals' support for immigrants' social rights. Using the novel measure, I find that lower-middle-income individuals become less supportive of immigrants' social rights when they are more exposed to social housing competition. By contrast, competition does not reduce support among the rich, who are less eligible for and less reliant on social housing, or the very poor, who are more shielded from competition due to the allocation rules for social housing in the Netherlands. Moving to the political consequences, I also find that benefit competition increases support for the populist right more among lower-middle-income voters than among the rich. These results provide compelling evidence that benefit competition can erode support for immigrants' social rights and influence electoral politics.
This study makes three contributions. First, it is the first study to examine individual support for immigrants' social rights over time and to model unobserved individual-level heterogeneity, thereby reducing concerns of omitted variable bias. Secondly, it leverages a mandatory refugee dispersal system to construct a novel and exogenous measure of social housing allocations to refugees. This measure circumvents problems of immigrant self-selection while capturing immigrants' benefits use more precisely because municipalities rely almost exclusively on the general social housing stock to meet their dispersal targets. Lastly, this study provides evidence of the consequences of benefit competition for policy and political preferences. The findings show that material self-interest contributes to our understanding of changes in voters' support for immigrants' social rights and their support for the populist right.
Explaining Support for Immigrants' Social Rights
To explain whether (and when) benefit competition affects changes in individuals' support for immigrants' social rights, this section first specifies the conditions for benefit competition and emphasizes that competition can only occur when the supply elasticity of welfare provisions is low. It also states that existing welfare institutions influence how benefit competition affects individuals' welfare attitudes. The more that benefit eligibility depends on income, the more the economic risk of benefit competition is concentrated among the poor, and the greater their demands for restricting immigrants' social rights.
Conditions for Competition
Previous research has demonstrated that four conditions must be met before immigration leads to benefit competition, defined as a situation in which potential gains for some result in potential losses for others. First, benefit competition presupposes that native-born citizens and immigrants want to obtain the same social goods and services (Dancygier Reference Dancygier2010; Money Reference Money1997). This seems plausible, because many individuals plan their lives around the expectation that they will have access to these provisions (King and Waldron Reference King and Waldron1988).
Second, competition assumes that immigrants have access to welfare provisions. If they cannot make legal claims, they cannot threaten the material interests of the native-born population. This is more realistic for some countries and immigrants than for others. For example, new immigrants in the United States have been ineligible for most benefits since the 1996 welfare reform, but such restrictions are mostly absent in Western Europe. Likewise, undocumented immigrants and asylum seekers have weaker social rights than permanent residents and refugees (Sainsbury Reference Sainsbury2012). Thirdly, competition assumes that immigrants are net beneficiaries of redistribution, and not net contributors. While immigrants have, on average, lower incomes than the native-born population in most advanced democracies, the degree of their over-representation among the poor varies across countries and immigrant groups.
A final and crucial precondition for competition is that the supply of welfare provisions cannot be expanded in the short term (Dancygier Reference Dancygier2010; Freeman Reference Freeman1986; Money Reference Money1997). If shortages can easily be alleviated, individuals would not have to compete for the same goods. Since the supply elasticity varies greatly across social programs, space and over time, new demands from immigrants should be a greater threat to the material interests of natives in some situations than in others. For example, if an immigrant receives social assistance, this rarely affects the amount of social assistance that others can receive. By contrast, greater demand for welfare provisions with an inelastic supply, such as social housing, education and health care, can quickly lead to longer waiting times and a lower quality of goods for others. Low elasticity thus seems to be related to strict budget constraints or the nature of certain goods (for example, it takes time to train new staff to deliver health care services or to build new homes).
Welfare Institutions, Competition and Voters' Preferences
Benefit competition threatens individuals' material self-interest by introducing the risk of not receiving a social good.Footnote 3 Since the poor are often more reliant on these goods, most studies expect competition to reduce support for immigrants' social rights more among the poor than the rich (Kitschelt Reference Kitschelt1995; Mewes and Mau Reference Mewes, Mau and Svallfors2012). While this is a useful starting point, it does not consider that policy preferences are formed in the context of existing welfare institutions. This can explain why income is a better predictor of these attitudes in some contexts than in others.
Studies on immigration preferences, for example, show that rich voters are more likely to oppose liberal immigration policies when they live in areas with high levels of social spending or progressive taxation because they fear higher taxes in the future (Facchini and Mayda Reference Facchini and Mayda2009; Hanson, Scheve and Slaughter Reference Hanson, Scheve and Slaughter2007). Other studies find that the negative effect of income on redistribution preferences depends on tax progressivity, labor market regulations, and on whether benefits are tied to employment (Beramendi and Rehm Reference Beramendi and Rehm2016; Gingrich and Ansell Reference Gingrich and Ansell2012). While these studies show that welfare institutions moderate the effect of material self-interest on immigration and redistribution preferences, they have not focused on risks related to losing access to social goods and services or on support for immigrants' social rights specifically.
I combine these insights to understand how benefit competition affects support for immigrants' social rights among rich and poor voters. I focus on benefit eligibility because it influences who is most at risk from benefit competition. When social programs are universal, the poor and the rich are both eligible for and reliant on the welfare state. As these institutions equalize economic risk, a person's income should be a weaker predictor of their attitudes, and benefit competition should increase demands for social protection across the board. For many social programs, however, eligibility and reliance depend on income.Footnote 4 Since only individuals within the target population(s) are eligible, the economic risk of benefit competition is concentrated among those with lower incomes who should respond by demanding more social protection.
I therefore expect that the more benefit eligibility depends on income, the more the economic risk of benefit competition is concentrated among those with lower incomes, and the more individuals' incomes predict demands for social protection.Footnote 5 Their demands can include more exclusionary eligibility rules, more social spending or both. These strategies are not mutually exclusive, but I focus on the former because it may be a more successful strategy in the short term during times of austerity.
In ethnically homogenous societies, the poor can often secure their access to scarce social goods by introducing or strengthening means testing in social programs. In ethnically heterogeneous societies, however, introducing an income threshold would be less effective if immigrants are over-represented among the low-income population. Voters can then successfully target scarce resources to themselves by restricting immigrants' access to the welfare state.
In the remainder of this article, the main analysis focuses on support for immigrants' social rights in the context of competition for social housing in the Netherlands, a social program that mostly low- and middle-income families rely upon and that is characterized by a highly inelastic supply. I expect that low- and middle-income voters become less supportive of immigrants' social rights when benefit competition with immigrants increases in their area. In addition, I explore the political consequences of benefit competition. If voters feel economically threatened by competition with immigrants, they may also be more likely to vote for right-wing populist parties, which are the strongest and most vocal opponents of immigrants' social rights (Schumacher and van Kersbergen Reference Schumacher and van Kersbergen2016). I therefore test whether benefit competition increases support for the populist right more among those with lower incomes.
Social Housing and Refugee Dispersals in the Netherlands
I analyze the effect of immigration-related social housing competition on individuals' policy and political preferences over almost a decade and across 388 municipalities in the Netherlands. In order to attribute any differences in attitudes to social housing competition, I must first demonstrate that social housing is a good desired by immigrants and natives alike, and that its supply is highly inelastic in the short term. In addition, I must provide evidence that the native-born population cannot easily avoid competing with immigrants, and that immigrants do not sort themselves into areas based on the native-born population's attitudes towards immigrants' social rights. In this section, I make this case.
Social Housing
The housing market in the Netherlands consists of three segments: the owner-occupied market (60 per cent), the social housing market (34 per cent) and the rental market with non-regulated rents (6 per cent) (Ministry of the Interior and Kingdom Relations 2016, 25, 28). Not-for-profit housing associations are the main providers of social housing: they own three-quarters of the social housing stock (WoonOnderzoek 2015).Footnote 6 These actors operate at the local level and have a public task to provide housing for disadvantaged groups (Haffner et al. Reference Haffner2009, 206). Construction subsidies were abolished in 1995, but the sector still relies heavily on public funding: housing associations receive loans from government-owned banks with below-market interest rates, and often buy land from local governments at below-market prices. In addition, over one-third of all social housing tenants receive means-tested housing benefits, amounting to 3.6 billion euros in 2015 (Ministry of Finance 2018).Footnote 7
Social housing is of high quality, spread across neighborhoods within cities, and dwelling satisfaction is high among renters (Aedes 2017; WoonOnderzoek 2015). Combined with low rent prices, this makes it a desirable good for many who do not want or cannot obtain homeownership. Households with a non-Western migrant background are over-represented among social housing tenants: they constitute 12 per cent of the population, but 18 per cent of all social housing tenants (Statistics Netherlands 2018).Footnote 8 Nevertheless, native households remain the largest consumers of social housing, accounting for over 70 per cent of all tenants.
Despite a growing demand for affordable housing caused by population growth, urbanization and a rise in single-person households in recent decades, the social housing stock has not kept pace. Between 2000 and 2016, it even decreased by 2.2 percentage points (pp) as the number of households grew by 13.5 pp (SN 2018). This has contributed to severe housing shortages in some municipalities. The limited construction of social housing has many causes, but one important reason is a social landlords levy, introduced in 2013 as an austerity measure, which significantly reduced housing associations’ ability to invest in new projects. Social housing is thus a desirable and scarce social good. To assess whether individuals can avoid competition with immigrants, I now turn to the social housing allocation mechanism.
Although all legal residents were eligible for social housing until 2011, low- and middle-income households were always over-represented among its tenants (Haffner et al. Reference Haffner2009). In 2011, social housing became formally means tested with the introduction of an income threshold for new allocations (see Appendix S.1). Since then, housing associations have been required to allocate at least 90 per cent of their vacancies to their main target group – households with gross annual incomes up to 34,000 euros, which corresponds to 40 per cent of the population. They may allocate up to 10 per cent freely. Many housing associations try to target social housing in this segment to middle-income households (Kromhout and Zeelenberg Reference Kromhout and Zeelenberg2014, 13).
The social housing allocation system is based on local waiting lists. After individuals register with a local housing association, they can actively apply to advertisements of their local association or, if the latter has formed a regional partnership with other housing associations, the advertisements of the partnership. The local character of this system limits individuals’ inter-regional mobility.Footnote 9 The housing association then ranks all applicants based on their registration time and offers the dwelling to the highest-ranked applicant, who can refuse the offer without penalty. Although waiting lists are the main allocation mechanism, municipalities can require housing associations to allocate vacant dwellings to priority groups who then bypass others on the waiting list.Footnote 10 These groups include victims of domestic abuse, caregivers and receivers and, until 2017, refugees (Article 12 of the Housing Law 2014). These needs-based allocations disproportionally affect the chances of middle-income households.
Refugee Dispersals
To rule out the possibility that immigrants move to areas based on the local native population's support for immigrants' social rights, I leverage a mandatory refugee dispersal system that distributes new refugees across Dutch municipalities and provides them with permanent housing. Refugees account for 13 per cent of the total migrant population in the Netherlands,Footnote 11 but due to the dispersal system they are over-represented among new recipients of social housing.
After asylum seekers register their asylum claim, they are dispersed to large-scale asylum seekers' centers spread across the country, where they remain until their application is approved. When that happens, they enter the refugee dispersal system, which aims to reduce the concentration of refugees in large cities and foster their integration (Arnoldus, Dukes and Musterd Reference Arnoldus, Dukes, Musterd, Robinson, Andersson and Musterd2003). Twice a year, the national government announces refugee dispersal targets for all municipalities based on their population size and the estimated number of refugees in need of permanent housing. Larger municipalities receive more refugees than smaller municipalities, and the availability of social housing does not play a role in the distribution formula.Footnote 12 The provinces monitor the targets and have instruments to induce compliance: from signaling and inquiring to finding accommodation for refugees on behalf of the municipality. Refugees are then matched to municipalities, which have twelve weeks to find permanent housing for them. Although municipalities do not own social housing, they can require housing associations to allocate a portion of vacant social housing to priority groups. Municipalities and housing associations often aim to spread refugees across neighborhoods to foster their integration (Smits van Waesberghe and Razenberg Reference Smits van Waesberghe and Razenberg2016, 8).
Several factors reduce the likelihood that new refugees can self-select into municipalities. First, they receive a single offer for a home on a take-it-or-leave-it basis. While they can reject it, they have few alternatives: they lose their right to stay in the asylum seekers' center and they will not receive priority for social housing in other municipalities. Secondly, they could join the waiting list in their desired municipality, but their chances would be very low as they would have to join the back of the queue. Moreover, recent refugees often lack the resources to rent privately due to their poor labor market integration. Even if they find their own accommodation, they count for that municipality's target. Although refugees are free to move after their initial placement, their mobility is limited in the short term: 84 per cent of all refugees dispersed between 2011 and 2013 still lived in their assigned municipality in 2014 (Wissink and Lijzenga Reference Wissink and Lijzenga2014).
Figure 1 displays the national targets and dispersals of new refugees between 2007 and 2016. In total, more than 160,000 refugees were dispersed across the country in this decade. In most years, the number of dispersed refugees is around 10,000, but there were two large peaks. The first resulted from a large-scale regularization of 27,500 asylum seekers in 2007 for which a separate dispersal system was set up.Footnote 13 The second peak occurs after 2013 when the inflow of refugees increased rapidly due to the war in Syria.Footnote 14

Figure 1. National targets and dispersals of new refugees (2007–2016)
Social Housing Allocations to Refugees
I use the refugee dispersal system to measure the share of vacant social housing allocated to new refugees. This captures competition for social housing when we can assume that native individuals desire social housing and that the demand exceeds the supply.Footnote 15 Under such conditions, allocating social housing to one household comes at the expense of other interested households.
Social housing allocations to refugees are assumed to be exogenous to the welfare attitudes of the native-born population and unrelated to other confounders. This seems plausible, as the dispersal system distributes new refugees evenly across municipalities and gives them limited control over their first placement. Native-born voters also have very little influence over these dispersals as all municipalities must participate and the provinces can impose sanctions on municipalities that do not. While voters could demand an expansion of the social housing stock, this process is lengthy under normal conditions and even more so in the aftermath of the 2008–09 economic crisis. Avoiding competition is also challenging given the local character of social housing waiting lists. This lends support to the assumption that immigration-related social housing competition is unrelated to the welfare attitudes of native-born voters.Footnote 16
Finally, I assume that voters are aware of the allocations of social housing to refugees in their area. Ideally, this assumption would be tested directly (see Newman et al. Reference Newman2015). Although the survey does not include items on voters' perceptions, the assumption seems plausible as most municipalities are small (the median population size 26,500), most refugees are visible ethnic minorities in the Netherlands, and many housing associations publish data on allocations to refugees on their website. Appendix S.4 also demonstrates that regional newspapers cover the issue of housing allocations to refugees, though less frequently than other important local issues.
Empirical Strategy
In the Netherlands, eligibility for social housing depends on income in two ways. Individuals are first divided into a primary target group, which includes those with household incomes below the threshold, and a secondary target group, which includes both lower middle-income and high-income individuals. While the chances of receiving social housing do not depend on income for those in the primary target group, they decrease with income for those in the secondary target group. The latter's chances are also more affected by needs-based allocations, including to refugees, as these take priority over allocations to the secondary target group. I therefore expect an individual's income to be a better predictor of support for immigrants' social rights in the secondary target group. To take this into account, I run models for the full sample, the primary target group of social housing (defined as those with a net monthly household income up to EUR 1,800) and the secondary target group (net monthly household income above EUR 1,800).Footnote 17 I also account for housing market tightness in the analysis.
To examine whether benefit competition reduces voters' support for immigrants' social rights, I use conditional fixed effects logit models. By focusing on changes within individuals, fixed effects models eliminate an important source of omitted variable bias resulting from unobserved and observed time-invariant differences between individuals. This is a major advantage over previous research based on cross-sectional data.
All models include year fixed effects to control for shocks that affect all individuals in a given year. Due to data availability on the social housing stock, this analysis relies on six waves between 2009 and 2015. I also explore the political implications of my argument by testing whether social housing competition increases support for the populist right more among lower-income voters than among higher-income voters.
Data
To analyze support for immigrants' social rights and support for the populist right, I use the Longitudinal Internet Studies for the Social sciences (LISS) from CentERdata. This is a high-quality household panel survey first administered in 2007 to 4,500 households and 7,000 individuals drawn from a true probability sample of Dutch households.Footnote 18 Refreshment samples were added in 2009, 2011 and 2013. I use the Politics and Values module, which includes items that measure political attitudes. I combine nine waves between 2007 and 2016, and limit the sample to native-born respondents aged 18 and over.Footnote 19 In total, this study analyzes the attitudes of 10,118 respondents, 8,194 of whom completed the questionnaire multiple times (1,727 individuals twice, 1,013 individuals thrice, 754 individuals four times and 4,700 individuals more than four times).
Attrition, or the dropout of respondents, occurs in all panel studies. It can lead to biased results if those exiting the panel differ systematically from those who remain. To examine the degree of attrition, I compare the characteristics of individuals as a function of the number of waves they participated in (see Appendix S.5). I find that the groups do not differ greatly in terms of observable individual- or municipal-level characteristics, such as social housing competition, the foreign-born population or the low-educated population.
I match respondents to municipal data from Statistics Netherlands and administrative sources. To deal with shifting municipal boundaries, I adjust all municipal-level variables to reflect the borders of 1 January 2017.Footnote 20 This leads to 388 municipalities with an average of 43,000 inhabitants.
Support for Immigrants' Social Rights
The main dependent variable, Support for Immigrants' Social Rights, is measured using an item that asks individuals if they strongly agree, agree, neither agree nor disagree, disagree or strongly disagree with the following statement: ‘Legally residing foreigners should be entitled to the same social security as the native-born population.’ The term ‘social security’ is used in a broad sense to refer to government policies that promote a decent standard of living for everyone. Appendix S.6 shows the overall distribution of responses for all respondent-years. In 58.5 per cent of all observations, respondents either strongly agreed or agreed with the statement that immigrants should receive equal social rights, while in 16.5 per cent of all observations respondents strongly disagreed or disagreed with the statement. Given the relatively high levels of support and the strong norm against overt discrimination in advanced democracies, I interpret the neutral response (neither agree nor disagree) as a lack of support for immigrants' social rights.
I recode the variable into a binary variable coded 1 for those who strongly agreed or agreed, and 0 otherwise.Footnote 21 This allows me to capture voters' changes in positions on immigrants' social rights. This measure is a difficult test of the benefit competition argument since it captures support for immigrants' social rights and not support for refugees' right to social housing. I therefore interpret the results as a lower bound of the effect of social housing competition on support for immigrants' social rights.
Figure 2 shows the average support for immigrants' social rights (with 95 per cent confidence intervals) over time. In 2007, almost two-thirds of all respondents favored equal social rights for immigrants. This declined in subsequent years, but more than half of all respondents continued to express support, even after the refugee crisis. Moreover, Table 1 displays the changes in support for immigrants' social rights compared to the previous year. It shows that in almost 75 per cent of all respondent-years, individuals were as supportive of immigrants' social rights as in the previous year. Nevertheless, a substantial number of respondents changed their attitudes: 13.4 per cent became less supportive of immigrants' social rights whereas 11.7 per cent became more supportive. The analysis examines these changes.

Figure 2. Proportion of support for immigrants' social rights (2007–2016)
Table 1. Changes in support for immigrants' social rights compared to the previous year

Income
Material self-interest would ideally be measured using an item that asks whether individuals want to move into social housing. Unfortunately, such items are not available. I therefore use income as a proxy for material self-interest because eligibility for and reliance on social housing are strongly correlated with household income. I expect lower-income individuals to be less supportive of immigrants' social rights than higher-income individuals because they are more likely to compete with immigrants for social goods. The continuous income variable sums the net monthly income of all household members. I adjust it for inflation using 2015 as the base year. After rescaling, one unit corresponds to 1,000 inflation-adjusted euros.
Social Housing Allocations
The main independent variable at the municipal level is the share of social housing allocations to refugees. I expect more allocations to reduce support for immigrants' social rights, holding all else equal. Since municipalities rely on social housing to meet their refugee dispersal targets and the supply of social housing is fixed in the short term, more houses allocated to refugees leads to fewer houses available for others, including native-born households. I measure social housing allocations as the share of vacant social housing allocated to new refugees in each municipality j in each year t:

I discuss the construction and sources of this measure in detail in Appendix S.2. The number of social dwellings allocated to refugees is estimated using administrative data on the inflow of newly dispersed refugees, and dividing these by the average refugee household size. I estimate the number of vacant social housing units by multiplying annual data on the social housing stock by the local annual turnover rate. I focus on vacant social housing instead of the total social housing stock because individuals are competing for the former.Footnote 22
Social housing allocations to refugees were less than 5 per cent until 2013 and rose to 6.1 per cent in 2014 and 10.6 per cent in 2015. Even though refugees make up a small portion of the total population, they receive a substantial share of vacant social housing. Figure 3 illustrates the regional and temporal variation in social housing allocations across 388 municipalities in 2009 (the earliest date for which I have data on social housing) and 2015. Darker areas indicate municipalities in which more vacant social housing is allocated to refugees. The regional differences are starkest in 2015, but even in 2009 the refugee dispersal system placed greater strain on some municipalities than on others. Social housing allocations to refugees are relatively lower in the four largest cities, which also have among the largest social housing markets. Among the remaining urban and rural municipalities, however, there is substantial variation in social housing allocations.Footnote 23

Figure 3. Share of vacant social housing allocated to new refugees across 388 municipalities
Note: Panel A: 2009; Panel B: 2015.
Source: Author's dataset.
Figure 4 shows the distribution of changes in social housing allocations for each respondent-year in the full sample. On average, allocations to refugees increased by 0.9 pp. While these within-changes are small, the variation is wide: one-tenth of the sample was exposed to increases of at least 6 pp while another tenth experienced decreases of 3 pp or more.

Figure 4. Distribution of changes in social housing allocations in the sample
Control Variables
I control for the following individual-level variables: age (in years), a dummy for having a university degree, household size and a dummy for being unemployed. To capture differences between municipalities, I include the percentage of foreign-born individuals, a dummy for the presence of an asylum seekers' center, the percentage of social housing and the percentage of low-educated individuals.
It is also important to control for the tightness of the local housing market because social housing allocations to refugees are more likely to induce competition in pressured markets for affordable housing. Since common measures, such as waiting lists for social housing and the ratio of buyers to sellers, are not available for the local level, I use two proxies for housing market tightness. First, municipalities with a long-term population decline should have looser housing markets;Footnote 24 93 of the 388 municipalities fall into this category. Secondly, I use the ratio of single-person households to rental housing units.Footnote 25 Since single-person households are the main drivers of increased housing demand, and this household type puts more pressure on the Dutch housing stock, which consists predominantly of family homes, a higher ratio captures tighter markets for affordable housing. Appendix S.6 describes the operationalization of these variables.
Results
Table 2 presents the results from conditional fixed effects logit analyses for the full sample, the primary target group of social housing and the secondary target group. I first run reduced-form models that include the main independent variables: a person's income, the share of social housing allocations to refugees in their municipality and their interaction. Model 1 reports the results for the full sample. I find that income has a positive effect on individuals' attitudes: as income increases, support for immigrants' social rights increases. As expected, I also find that an increase in social housing allocations to refugees is associated with a decrease in support for immigrants' social rights. The positive interaction indicates that this decrease will be larger for those with lower incomes. While the coefficients are in the expected direction, they are statistically insignificant. To shed more light on this, I move on to the subsample analyses.
Table 2. Social housing allocations and support for immigrants' social rights

Note: estimates from conditional fixed-effects logit models. Logit coefficients are odds ratios. **p < 0.01, *p < 0.05, +p < 0.1.
Model 2 shows that the effect of social housing allocations on support for immigrants' social rights does not depend on income in the primary target group. The coefficients for income, social housing allocations and their interaction are in the same direction as before but they are insignificant. These null findings suggest that the poorest members of society are more shielded from social housing competition with refugees because those types of allocations have a smaller effect on their chances of receiving a house.
By contrast, Model 3 shows that the effect of social housing allocations to refugees depends on income in the secondary target group, as expected. Income and social housing allocations have a negative main effect, while the interaction term is positive and significant. This means that in the secondary target group, these allocations reduce support for immigrants' social rights more among those with lower incomes than among those with higher incomes. This finding makes sense given that housing associations often consider income for their free allocations (Kromhout and Zeelenberg Reference Kromhout and Zeelenberg2014, 13). The results suggest that social housing allocations to refugees matter most to lower-middle-income individuals, who often earn too much to be fully eligible for social housing and not enough to enter the private rental or owner-occupied market.
The fourth model explores whether the differences between the lower-middle-income and higher-income individuals in the secondary target group hold after controlling for individual- and municipal-level characteristics. I find that age, education, household size and becoming unemployed cannot explain why social housing allocations reduce support for immigrants' social rights more among those with lower-middle incomes.
Regarding municipal-level characteristics, I find that the effects of the level of immigration and the presence of an asylum seekers' center are positive; the former is significant at p < 0.1. This could be interpreted as support for the contact hypothesis, which proposes that positive interactions between immigrants and native-born individuals lead to more positive attitudes towards immigrants (Allport Reference Allport1954). It could also reflect that individuals with more inclusive attitudes towards immigrants self-select into urban areas, the main destination of immigrants (Maxwell Reference Maxwell2019). The size of the social housing market and the percentage of low-educated population are both poor predictors of changes in support for immigrants' social rights.
The last models account for housing market tightness using the two proxy measures described earlier. Models 5 and 6 show that neither long-term population decline nor the ratio of single-person households to rental units has a direct effect on voters' support for immigrants' social rights. Importantly, the main findings hold when I control for housing market tightness.
The findings from Table 2 show that lower-middle-income voters reduce their support for immigrants' social rights when refugees receive more social housing. If this is about competition, as I argue, these allocations should have a stronger effect in tight housing markets where these voters have few alternatives. As a more stringent test, I therefore re-run the analysis for respondents in the secondary target group split by housing market tightness (see Table 3).
Table 3. Social housing allocations and support for immigrants' social rights by housing market tightness

Note: estimates from conditional fixed-effects logit models. Logit coefficients are odds ratios. All models include respondents from the secondary target group only. **p < 0.01, * p < 0.05, +p < 0.1.
Models 7 and 8 focus on long-term pressures on the housing market by dividing the sample into those living in municipalities with and without a long-term population decline. In Models 9 and 10, I subset by the tightness of the rental housing market and divide the sample into those living in a municipality with a ratio of single-person households to rental units below or above the median. I find that the main coefficients of interest are statistically insignificant for those in less pressured housing markets (Models 7 and 9). Individuals thus do not seem to lower their support for immigrants' social rights simply because they dislike out-group members receiving these goods. By contrast, allocating social housing to refugees has a negative and significant effect on lower-middle-income voters in more pressured housing markets (Models 8 and 10). Overall, this evidence supports the interpretation that social housing allocations to refugees trigger benefit competition, and that competition for scarce social goods can make some voters more exclusionary towards immigrants' social rights.
To illustrate the substantive effects, Figure 5 presents the average marginal effects of social housing competition conditional on the level of income based on Model 8 in Table 3.Footnote 26 As expected, Figure 5 shows that social housing competition significantly decreases support for immigrants' social rights among lower-middle-income voters. They demand a less inclusive welfare state when they face competition for scarce social goods and services in their area. As income increases, the effect of social housing competition on support for immigrants' social rights diminishes and the confidence intervals widen, particularly for net household income values over EUR 5,000 (95th percentile).

Figure 5. Social housing competition, income and support for immigrants' social rights
For the very rich (above the 96th percentile), the effect of competition even becomes positive: this suggests that the very rich become more supportive of immigrants' social rights when the level of social housing competition in their municipality increases. This curious finding probably reflects the low number of observations in the tail end of the income distribution. Alternatively, this group of voters may realize that providing permanent housing to refugees is cheaper than having them stay in an asylum seekers' center. The latter is not only more expensive in the short term (for example, in 2017, the Dutch government estimated the costs of sheltering at EUR 32,300 per asylum seeker per year). It also hampers refugees' socioeconomic integration as they generally start their language and integration courses after their placement in a municipality.
I also calculate the predicted probabilities for lower-middle-income and high-income voters, defined here as having a net monthly income of 1,900 (30th percentile of the total income distribution) and 4,400 (90th percentile) inflation-adjusted euros, respectively, while holding the control variables at their means. The probability of supporting immigrants' social rights decreases significantly from 0.63 to 0.61 for lower-middle-income voters if we move the level of social housing competition from zero to the mean, and it drops to 0.60 if we move the level of competition one standard deviation above the mean. These differences are significant. By contrast, the effect of social housing competition is positive for high-income voters, but statistically insignificant. Given the relative stability of policy preferences (see, for example, Maxwell Reference Maxwell2019; O'Grady Reference O'Grady2019), these modest effects are meaningful as politics often happens at the margins and one-tenth of the sample witnessed rises in social housing competition of 6 pp or more during this decade. Since social housing competition with refugees is only one channel through which immigration can lead to benefit competition, the effects of immigration-related benefit competition may thus reasonably be larger.
Robustness Tests
Although one should always be careful when causally interpreting observational data, the within-subject design combined with an exogenous measure of social housing allocations allows me to exclude a wide range of alternative explanations. To further increase confidence in my results, in this section I discuss the main findings from a series of robustness tests (for more, see Appendix S.7).
First, low-income individuals may become less supportive of immigrants' social rights because they are more prejudiced, have authoritarian personalities or harbor more cultural fears (see van der Waal et al. Reference van der Waal2010; on general immigration attitudes, Sides and Citrin Reference Sides and Citrin2007). This explanation has been influential in the literature. To the extent that these predispositions are time invariant, the individual fixed effects should capture them in the longitudinal analysis. In addition to controlling for education, a strong predictor of cultural fears, I show that the core findings also hold when I include social class and a dummy for authoritarianism.
Secondly, previous studies show that the urban–rural divide is important for understanding support for the populist right and immigration attitudes (see, for example, Maxwell Reference Maxwell2019). To examine whether urbanization drives these findings, I control for population size, population density and a dummy for living in an urban area. This does not change the results. I also find that the relationship between income and social housing competition does not differ across highly urbanized and less urbanized municipalities.
Another concern could be that refugee dispersals trigger fairness concerns, as refugees often receive housing before others on the waiting list. While this explanation is not incompatible with my argument, it cannot explain why allocating social housing to refugees would matter more to lower-middle-income voters. Individuals also seem more than willing to violate norms of fairness, as data from the 2008 Cultural Changes survey suggests: while 66.5 per cent of Dutch respondents would not discriminate between the native-born population and immigrants in times of housing shortages, a large minority of 31.2 per cent would give priority to native-born households over immigrant households.
Moreover, some may argue that individuals should not care about competition if they have already secured social housing. However, this neglects the fact that Dutch persons move, on average, eight times during their lifespan. Current social housing tenants may therefore reasonably worry about their future chances of moving into a new social dwelling. More importantly, housing surveys show that more than one-third of Dutch households wants to move homes, and more than half of them wants a rental dwelling or has no preference between rental and owner-occupied dwellings (WoonOnderzoek 2015). The group of potentially affected individuals is thus large.
Finally, the main finding that social housing competition matters more to those with lower incomes is robust to controlling for political ideology, homeownership, being out of the labor force, local unemployment rates, the percentage of industrial jobs, the percentage of benefit recipients, average housing values and the size of the private rental market. I also find similar results when I use a gross income measure, restrict the analysis to individuals in ethnically homogenous municipalities or use an index of pro-immigration attitudes. Appendix S.7 describes these and other tests.
Competition and Support for the Populist Right
Having demonstrated that competition for social housing reduces support for immigrants' social rights among lower-middle-income voters, this section explores whether it also increases support for the populist right more among these voters. I estimate two-way fixed effects linear models with respondents' evaluations of the main populist right party in the Netherlands, the Party for Freedom, as the dependent variable.Footnote 27 This variable ranges from 0 (very unsympathetic) to 10 (very sympathetic).Footnote 28 Support for the Party for Freedom is, on average, 2.95 (s.d. = 2.4). In order to save space, I focus here on the results for the secondary target group by housing market tightness.Footnote 29
Table 4 summarizes the results. I find that allocating a higher share of social housing to refugees increases support for the populist right and that this effect is stronger for lower-middle-income voters than for high-income voters. As before, I find that these effects are strongest for voters in tighter housing markets (see Models 12 and 14). This finding suggests that these social housing allocations trigger feelings of competition and can push lower-middle-income voters to populist parties that propose to protect them against competition. Moving on to the control variables, I confirm previous findings that education and becoming unemployed are important predictors of support for the populist right (see, for example, Rydgren Reference Rydgren2007). By contrast, the level of immigration does not predict support for the populist right in the Netherlands.
Table 4. Social housing allocations and support for the populist right by housing market tightness

Note: estimates from fixed-effects linear models with robust standard errors. All models include respondents from the secondary target group only. **p < 0.01, *p < 0.05, +p < 0.1.
For voters with a net monthly income of 1,900 inflation-adjusted euros (30th percentile of the total income distribution), support for the populist right increases from 2.81 to 2.87 (out of 10) if we move from zero to the mean of social housing competition, and to 2.91 if we move one standard deviation above the mean of social housing competition. This mirrors the earlier findings on support for immigrants' social rights, and shows that the populist right can attract voters on economic as well as cultural issues. By contrast, social housing competition seems to push rich voters away from the populist right. For voters with a monthly income of 4,400 inflation-adjusted euros (90th percentile), support decreases from 3 to 2.95 when competition increases from zero to the mean and decreases further to 2.92 when competition increases to one standard deviation above the mean.
Conclusion
This article tested whether immigration-related benefit competition lowers support for immigrants' social rights more among poor voters than among rich voters. It focused on competition for social housing in the Netherlands, a social program in which benefit eligibility and reliance depends on income. Using individual-level panel data and a novel measure of social housing allocations to refugees based on the mandatory refugee dispersal system, I found that lower-middle-income individuals became less supportive of immigrants' social rights when social housing competition intensified in their municipality. By contrast, competition did not influence support among the rich, who are less eligible for and less reliant on social housing, or among the very poor, who are more shielded from social housing competition. Using the same panel data, I also showed that social housing competition increased support for the populist right more among lower-income voters than among higher-income voters. Benefit competition can thus have political implications.
Although recent findings suggest that immigration has a stronger negative effect on the redistribution preferences of the rich than the poor (see Alt and Iversen Reference Alt and Iversen2017; Finseraas Reference Finseraas2012; Rueda Reference Rueda2018), this article demonstrates that immigration also matters to lower-income individuals because it can reduce their support for an inclusive welfare state. This confirms the importance of examining welfare attitudes other than general support for redistribution (see also Cavaillé and Trump Reference Cavaillé and Trump2015). These findings also challenge the interpretation that lower-income voters are lured away from the left by the anti-immigrant rhetoric of right-wing populist parties (Alesina and Glaeser Reference Alesina and Glaeser2004). Instead, some of these voters seem to align themselves with a party that promises to protect their material self-interest by restricting immigrants' access to valuable social goods. To maintain broad support for generous and inclusive welfare states, it is therefore crucial for governments to balance the demands of the native-born population and immigrants when welfare provisions are scarce.
This article moved beyond the labor market to illustrate how immigration impacts welfare attitudes when it puts pressure on social goods and services in local communities. I focused on social housing because it is a tangible and important good for many. My argument, however, can be applied to other welfare provisions with low supply elasticity, such as education and health care. Large segments of the population rely on these social programs throughout their lives, and governments spend a substantial portion of their budgets on their provision. This would be an interesting avenue for future research. The Netherlands was a useful case because a large portion of its population depends on social housing and its mandatory refugee dispersal system creates exogenous variation in competition for social housing. These findings can also increase our understanding of welfare attitudes in other countries with extensive social housing markets and immigration, such as Germany, France or the Nordic countries. Given the more universal character of its public housing system, Sweden would be particularly interesting for further study once individual-level data on support for immigrants' social rights become available. The findings may travel less well to countries such as the United States and the United Kingdom, where social housing is more residual and stigmatized, and many immigrants are already ineligible due to legal restrictions. In these countries, immigration may increase competition in the private rental market where other factors play a role.
The analysis also has some limitations. First, I focused on social housing competition across municipalities, but I did not address ethnic residential segregation within municipalities. If the native-born population eschews diverse neighborhoods and immigrants self-select into them, there may not be competition between groups. While this is a valid concern, Dutch municipalities and housing associations aim to spread refugees evenly across neighborhoods; thus the results also hold for voters living in ethnically homogenous municipalities. Secondly, I cannot rule out the possibility that social housing competition only affected individuals' willingness to express their attitudes towards immigrants' social rights instead of their position on this issue, as my argument holds. Yet even if this were the case, these findings matter because political actors may follow up on voters' signals and translate them into more restrictive social policies.
Supplementary material
Data replication sets are available in Harvard Dataverse at: https://doi.org/10.7910/DVN/9TSJTY and online appendices at: https://doi.org/10.1017/S0007123420000150. Some of the datasets are restricted access, but instructions to obtain the data are provided in the Dataverse.
Acknowledgements
I thank David Rueda, Desmond King, Jane Gingrich, Pablo Beramendi, Silja Häusermann, Jeremy Ferwerda, Helen Milner, Inyoung Cho, Raluca Pahontu, Isabel Perera, Verena Fetscher and Sadia Zafar for helpful comments. I also thank participants in research seminars at the University of Oxford, the WZB Berlin Social Science Center, as well as the 2017 and 2019 International Conferences of Europeanists. I am grateful to Joris Mulder (CentERdata), Quinten Snijders (Aedes), and Susanne van Opijnen (COA) for helping me find the data used in the analysis. The article has greatly benefited from the comments and suggestions of the Associate Editor and three anonymous referees. All faults are my own.