Voting Behavior of Immigrants and Their Children in Sweden

Abstract

We still know remarkably little about the voting behavior of immigrant populations, and in particular, the children of immigrants – who grow up in the same society as their contemporaries, but may be subject to different patterns of socialization. This article uses verified voting behavior in Swedish municipal elections to offer at least two new perspectives on these questions. First, we are able to separate out the impacts of family socialization, general societal socialization, and citizenship acquisition on electoral participation. Second, we are also able to add to our knowledge of the differences in political participation levels between different groups of foreign-background voters.

Keywords:

• turnout
• foreign-background voters
• Sweden
• socialization
• second-generation

Introduction

The starting point of democracy is that it represents ‘rule of the people’. But who exactly are ‘the people’? Diverse patterns of migration, citizenship, and residence rules on enfranchisement blur the boundaries of the demos. Countries may have significant proportions of residents who are not citizens; citizens who attained that status long after becoming residents; and citizens who are not residents. We still know remarkably little about the voting behavior of these groups, and – in particular – about the long-term effects of foreign background on a person’s political activity in their country of residence.

Available international evidence suggests that the electoral turnout of immigrant populations (where they are enfranchised and eligible to stand) is generally lower than for native citizens (Hutcheson & Russo, 2021). But there is controversy as to why, and to what extent levels of political participation can change after early socialization patterns are established. This is particularly relevant in the case of immigrant populations, who may be socialized in one country but live for long periods of time in another. Of particular interest is the effect that socialization has on the children of immigrants, who grow up in the same society as their contemporaries who do not have a foreign background, but may be subject to different patterns of socialization in their home and public spheres that may affect their political behavior into adulthood. It is with this group that this article particularly concerns itself.

Hitherto, our knowledge of this has been limited by the availability of data. Studies of the electoral participation of immigrants are often based on detailed case studies of particular geographical areas (e.g. Collard, 2010; Groenendijk, 2008; Ruedin, 2018; Togeby, 1999) or of particular ethnic groups but not immigrant populations in general (e.g. Fennema & Tillie, 1999; Jacobs & Tillie, 2004; Maxwell, 2010). Alternatively, extrapolations from national representative surveys mean that the political views of ‘immigrants’ in general can be described, but there is often a lack of specific detail about particular ethnic groups due to small sample sizes (e.g. Bass & Casper, 2001; Richard, 1999). Only a few studies (e.g., Wass et al., 2015) have looked at turnout among migrant voters in a systematic way within individual countries. Even then, there has been very little study of the so-called ‘second generation’ – those born and brought up in a country, but whose parents originated elsewhere. Even if there is a limited literature on the political attitudes of the children of immigrants in general (e.g. Moschion & Tabasso, 2014; Togeby, 2004), their electoral participation remains only tentatively researched (Bevelander & Hutcheson, 2017; Humphries et al., 2013).

Filling this gap is of vital importance if we are to fully understand the complete spectrum of political activity in a country. This article examines this under-researched aspect of political participation. As an empirical basis, it uses verified voting behavior in Swedish municipal elections from 2002 to 2014 (with a particularly detailed focus on the last of these). We offer at least two new perspectives. First, with the focus on electoral participation across multiple generations of people with foreign background, we are able to separate out the impacts of family socialization, general societal socialization, and citizenship acquisition as factors determining electoral participation. We do this by investigating the levels of turnout among citizens and non-citizens who are foreign-born voters; Swedish-born voters with a foreign background;¹ and Swedes with no foreign background. Second, we are also able to add considerably to our knowledge of the differences in levels of political participation between different groups of foreign-background voters.

The article proceeds as follows. First, using theoretical perspectives on migrant political participation, we develop a number of hypotheses about the participation of foreign-born and foreign-background voters. Next, we contextualize the Swedish case and show why it is a particularly good laboratory for testing these hypotheses. Finally, the hypotheses are tested, and the significance of the findings are analyzed.

Electoral turnout and foreign background: theoretical expectations

Turnout is often used as an indicator for political engagement. In representative democracies, higher turnout can confer greater agency on elected representatives to represent societal preferences. At the same time, if certain groups are excluded from participation by electoral rules, or if turnout levels amongst different societal groups of enfranchised voters differ, the result may be a ‘representation bias’ – in which the views of certain people are systematically over- or underrepresented in the polity (White & McAllister, 2007).

Before proceeding further, it is worth noting two points. First, there a strong interaction between the right to vote, and the citizenship regime of the country. How easy, or difficult, it is for non-citizens to become citizens is subject to strong variance from country to country (Orgad, 2017; Vink & Bauböck, 2013). (Resident) citizens above the age of majority in most countries can generally vote in almost all types of elections, but hardly any countries enfranchise non-citizens in national elections. In some cases, the franchise is available to non-citizens in local elections, most notably to mobile European Union citizens in other EU states, but (as in Sweden) sometimes going beyond that.

Second, the focus of our article is specifically on turnout in municipal and regional elections, rather than national ones. There is a stronger democratic case for granting voting rights to non-citizens in local elections than in national ones (Bauböck & Arrighi, 2017). Citizenship is a feature of self-governing states in an international state system, but local authorities are effectively ‘origin-blind’ to inter-municipal migration by national citizens. Thus the delivery of local services do not need, normatively, to be connected to place of origin.

In Sweden, municipal and regional governments are responsible for a significant amount of public services, as well having the right to levy taxes. They are also the only levels in which non-citizens of Sweden are enfranchised alongside national citizens. This allows a four-way comparison between citizens and non-citizens, and Swedish-born and foreign-born voters.

Turnout amongst foreign-born voters: theoretical expectations

As mentioned above, a there have been a number of studies of turnout and political activism of immigrant populations, albeit often based on a narrow range of cases.

Even where people are allowed to participate, there is also a disparity in activity levels between different ethnic groups and regions of origin (Bevelander et al., 2021; de Rooy, 2012; Jacobs & Tillie, 2004; Togeby, 2004). There is a multitude of possible explanations for immigrants’ lower participation: lower civic engagement and trust in institutions (Fennema & Tillie, 1999); ethnic residential concentration, where neighbors behave similarly (Togeby, 1999); the varying mobilizing effects of ‘bonding’ within communities and ‘bridging’ between them (Ager & Strang, 2008; Ocampo et al., 2018); differences in civic engagement between ethnic minority groups (Jacobs & Tillie, 2004); and levels of organizational membership among migrant populations (Berger et al., 2004). It has also been suggested that immigrant-specific theories of turnout may be unnecessary, and that ‘standard’ models of turnout still explain the majority of variation in participation between immigrant groups (Spies et al., 2020).

Two factors that are thought to play overriding roles are the incorporation or socialization and citizenship regimes. Various scholars (e.g. Adman & Strömblad, 2000; Bevelander & Pendakur, 2011; Maxwell, 2010; Messina, 2006; Oskarsson, 2003; Ruedin, 2018; Togeby, 1999; Voicu & Comşa, 2014) have noted an integration and naturalization effect on political participation. In general, foreign-born non-citizens vote less than those who have become citizens; and long-standing residents are more likely to vote than recent arrivals.

From this, we can extrapolate certain hypotheses about the re-socialization effects of living in a different society from the one in which one grew up. There is controversy as to how this mechanism works. Whereas early studies suggested that political predispositions develop early in pre-adult life and are relatively immutable thereafter (Easton & Dennis, 1969; Hess & Torney, 1967; Hyman, 1959; Merelman, 1986; Sears & Funk, 1999), such a view has become unfashionable. Instead, it has been suggested that exposure to alternative political environments is likely to modify voters’ behavior (Ramakrishnan & Espenshade, 2001; Wong, 2000). From this, we could hypothesize that integration in a new political setting would lead to a gradual convergence between the participation rates of foreign-born voters and native-born ones. A modified version of this theory suggests that early socialization experiences determine initial political behavior, but that the speed of resocialization varies according to how big a difference there is between earlier and later political experiences (Armony et al., 2004; Black et al., 1987; Finifter & Finifter, 1989; White et al., 2008). Thus we could expect participation rates to change over time, with the change most pronounced amongst people from systems least like the ones that they have entered. (For such people, the acquisition of new political knowledge requires a greater degree of relearning, and hence alters behavior more).

Citizenship acquisition would be expected to interact with and strengthen that effect (Bevelander & Pendakur, 2011). Hainmueller et al. (2015, 2017) found a causal effect of citizenship acquisition on political participation. Many of the same processes of relearning may be at work in the attainment of citizenship. The process of citizenship acquisition may involve a degree of civic learning about the political system of the new country. Moreover, citizenship gives an immigrant a long-term stake in the country that a non-citizen does not necessarily have (Bauböck, 2006; Dronkers & Vink, 2012; Howard, 2009; Janoski, 2010; Koopmans et al., 2012; Vink & de Groot, 2010).

Deriving from these observations, we can develop three general hypotheses about the turnout rates of foreign-born citizens:

H1: Being born outside a country makes an individual less likely to participate in a country’s elections than a native-born person, regardless of citizenship or other factors.

H2: The longer an individual has been living in the country, the more likely he or she is to participate.

H3: Immigrants who have acquired citizenship are more likely to participate in voting than non-citizens who are eligible to vote.

Children of migrant populations: socialization and turnout

The hypotheses above relate mainly to foreign-born voters (i.e. those who have themselves immigrated to a country – in this case, Sweden). What of the next generation: those who were themselves born in the country of study, but had one or more parent who was foreign-born? Foreign-background individuals have been seen in various studies to acquire certain elements of identity and behavioral traits deriving from their parents’ country (-ies) of origins, even if they have grown up in the country of their birth. There are often also differences in, for example, education attainment and economic status between children of immigrants and those whose parents were born in the country (Borgna, 2016; Dustmann et al., 2012; Hammarstedt & Ekberg, 2004). There has only been very tentative investigation into whether this transfers into the electoral sphere.

From broader studies of electoral behavior amongst the population as whole, it seems that there is some degree of intergenerational transfer in voting behavior (Smets & van Ham, 2013). This is down to a mixture of parental education and household mobilization (Gidengil et al., 2016; Kudrnáč & Lyons, 2017). But what of the grown-up children of migrants, more narrowly? People who are socialized from the outset in the social and political institutions of the state in which they are growing up (as well as by their parents) may be thought more likely to vote than ‘first generation’ immigrants, but limited studies among Latino voters in the United States and among two-generation Turkish households in Europe has suggested that voting turnout differences remain between foreign-background voters and those born to ‘native’ parents, even in this second generation (Spierings, 2016; Terriquez & Kwon, 2015).

Several factors may contribute to this and may not all be directly related to ethnic background. First, children of immigrants are not distributed homogenously; urban segregation may mean that they grow up in areas that contain a large concentration of other foreign-background people, leading to several intersectional identities and self-exclusion from mainstream civic engagement (De Wenden, 2014). It is also possible that there is an environmental effect, independent of individual factors. As Persson (2013) has shown, a low-education individual living in a district with people of generally higher education tends to be more likely to participate in elections than others of the same education level as him- or herself elsewhere. Studies from the United States (Terriquez & Kwon, 2015) have also indicated that barriers to ‘first-generation’ foreign-born voters’ participation often inhibit the participation of their children (though in some cases, the second-generation may in turn re-activate their parents).

On the other hand, people of foreign background may be more engaged with public society than their parents’ foreign-born generation. Moschion and Tabasso (2014) have shown that, in certain behavior traits and attitudes, the second generation has a hybrid position between that of natives and their first-generation immigrant parents. Notwithstanding urban and cultural segregation, growing up in a country will bring people with foreign-born parents into substantial contact with mainstream society through the education and childcare systems and the labor market (Hammarstedt & Palme, 2012; Humphries et al., 2013). Later in the article we will test whether this leads foreign-background individuals to be more politically active later in life than their foreign-born parent(s). Alternatively, the cultural influence of home may temper that of societal structures, leading to lower participation levels than amongst the same generation without a foreign background.

The above analysis leads to two hypotheses about this potentially hybrid identity.

H4: Foreign-background persons will have higher participation rates than foreign-born immigrants, all other things being equal.

H5: Foreign-background persons will generally have lower participation rates than people without any foreign-born parents.

Building on this, we can investigate whether the sense of cultural and behavioral difference varies according to whether one or both parents are foreign-born. People with one native-born parent (where the other is foreign) may not have very different socialization experiences from other citizens. We would expect that there would be a stronger non-native cultural influence where both parents are originally from other countries, as there will be less Swedish influence in their home lives than amongst those who have a Swedish parent.

This leads to the final hypothesis:

H6: Foreign-background persons with only one non-native parent are more likely to vote than those with two non-Swedish-born parents.

The Swedish case: electoral rights and the foreign-background population

Sweden is a particularly good country to test the differences of socialization amongst voters of foreign background and foreign birth. By 2020, 19.7 per cent of the population (2.05 million were foreign-born, and 25.9 per cent of the population were classified as ‘foreign background’ (foreign-born, or with two foreign-born parents). However, only 8.7 per cent of the Swedish population were not citizens of Sweden (Statistics Sweden, 2021), mainly due to the relatively liberal naturalization process (Bevelander et al., 2015, p. 6).²

Postwar immigration to Sweden divides into several periods. Until the early 1970s, most immigration was labor-driven, mainly from Finland and Mediterranean countries such as Greece, Yugoslavia and Turkey. By the mid-1960s, under pressure from trade unions, government policy began to restrict this. Thereafter, immigration came largely from family reunification and refugees, mainly from Chile, Poland and Turkey (1970s), Ethiopia, Iran and other Middle Eastern countries (1980s); Iraq and the former Yugoslavia (1990s); and these areas plus Somalia in the early 2000s. The entry the entry of Sweden into the European Union (EU) in 1995 led to a new stream of intra-EU labor mobility (Bevelander, 2010), and most recently there have been large numbers of refugees from Syria.

Citizenship in Sweden is based on the jus sanguinis principle. People whose parents are Swedish citizens are automatically granted citizenship. The children of non-Swedish citizens who are born in Sweden are not automatically entitled to Swedish citizenship, but Swedish legislation on naturalization is one of the most liberal in Europe (Citizenship Act, 2001, Art. 7-8; Solano & Huddleston, 2020). Citizenship acquisition is very high among immigrants from Africa, Asia and Eastern Europe. After 15 years in the country, about 90 per cent have acquired Swedish citizenship. Lower levels are found for European immigrants (around 35 per cent after 15 years), and Nordic immigrants have the lowest levels of naturalization (25 per cent after 15 years) (Helgertz & Bevelander, 2017).

Moreover, Sweden is one of the most open European countries for enfranchising (and allowing the candidacy of) non-Swedish citizens in municipal and regional elections (Globalcit, 2019). Since 1976, foreign citizens who have been registered for three years have had the right to vote in subnational electoral contests.³ After 1998, the three-year waiting period was waived for EU citizens and citizens of Iceland and Norway (Local Government Act, 2017, Chapter 1, Section 7).⁴ In principle, the electorate for sub-national elections comprises everybody over the age of 18 resident in Sweden, except for non-EU/Nordic citizens who have been resident for less than three years.

As Table 1 shows, non-Swedish citizens accounted for between 4 and 6 per cent of the electorate in the municipal elections between 2002 and 2014, primarily from other Nordic or European states. However, the other 94 per cent of the electorate were not a homogenous group, and contained a considerably number of people of foreign background. Only 73.9 per cent of voters in 2014 were Swedish-born citizens with Swedish-born parents.

Table 1. Composition of electorate by citizenship, municipal elections 2002–14.

Citizenship	2002	2006	2010	2014
Sweden, of which	95.3	95.3	94.7	94.0
Swedish-born, 2 Swedish-born parents	80.3	78.5	76.3	73.9
Swedish-born, 1 foreign-born parent	4.9	5.3	5.7	6.1
Swedish-born, 2 foreign-born parents	1.7	2.0	2.3	2.8
Foreign born	8.4	9.5	10.3	11.3
The Nordic countries excluding Sweden	1.9	1.9	1.8	1.6
Europe excluding the Nordic countries	1.5	1.7	2.0	2.5
Asia	0.7	0.6	0.8	1.0
Africa	0.2	0.2	0.3	0.5
North America	0.1	0.1	0.1	0.1
South America	0.2	0.1	0.1	0.1
Other	0.1	0.1	0.1	0.1
Total	100.0	100.0	100.0	100.0

Source: Statistics Sweden (2018).

Given this diversity, municipal elections in Sweden provide an ideal laboratory in which to examine electoral participation rates of different groups of voters with foreign backgrounds. For example, we can examine the effects of residence, citizenship status, duration of stay and region of origin on voting participation levels.

This is the main focus of the rest of this article. We place a particular focus on the so-called ‘second generation’ – i.e. people (whether Swedish citizens or not) who were born in Sweden, with at least one parent born outside it. Such people have largely been neglected in the literature. Members of this group are distinguished by the fact that they have been born, raised and socialized in Sweden – just like Swedish-born people with Swedish parents – but may also have been subjected to other cultural influences through their family backgrounds. We are interested in whether this makes their political behavior more like their parents’, or more like others who grew up in Sweden without a foreign background – and what this says about the effects of socialization on electoral participation.

Data and method

Statistics Sweden has consistently collected information about the electoral participation of the population since 1909 and has published the electoral participation rates of immigrants since 1988. While immigrant citizens vote on average about 8 per cent less than natives, the decline in voter participation over time by immigrants has been lower than for native-born citizens. Tracking the voting probabilities of non-citizens is more difficult, partly because a substantial number of non-citizens leave the country without informing anyone (Öhrvall, 2006). Nonetheless, it seems that participation rates have decreased substantially since first measured at around 60 per cent in 1976 (Hammar, 1979).

Until now, however, the participation rates of the children of immigrants have not been studied in detail. To address this, we draw on data that allow us a uniquely detailed picture of electoral turnout in Sweden. First, a series of datasets on electoral participation in the 2002, 2006, 2008, 2010 and 2014 elections contain information about individual electoral participation for eligible to vote in national, provincial and municipal elections. Since 1964, all the electoral surveys in Sweden have been based on the Labor Force Survey sample. Statistics Sweden matches the actual participation information provided by the provincial authorities in Sweden to this sample. This information is matched to the second source: personal registry data from Statistics Sweden, which contains demographic and citizenship information on every Swedish resident. Based on these two datasets, a database is created that contains individual level data on actual voting and also allows us to control for a number of individual characteristics. Such surveys have allowed us to track participation rates – including large samples of non-citizens – in each election from 2002 onwards. Since Bevelander (2015) focused on the 2002, 2006 and 2010 elections using a cohort analysis, to study immigrant voting integration, this study instead focuses on the 2014 election in more depth. The focus is on the turnout of both first- generation immigrants and their children born in Sweden.

The electoral significance of non-Swedish citizens varies geographically. They comprised less than 5 per cent of the electorate in two-thirds of municipalities, but there were over 450,000 non-citizen voters in total. Sweden’s three biggest cities (Stockholm, Gothenburg and Malmö) accounted for the largest individual counts, but municipalities at the edge of the country often had the largest concentrations of non-citizen voters in proportional terms (e.g., Haparanda, on the Finnish border, at 27.9 per cent). The general patterns were similar in the most recent election of 2018, with by then over half a million non-Swedish voters (Valmyndigheten, 2014, 2018).

Bearing in mind the differential concentrations of citizens and non-citizens in different municipalities, Table 2 shows turnout rates across Sweden as a whole, divided by region of voters’ citizenship, among those eligible to vote in Sweden at municipal elections since 2002. Turnout rates amongst Swedish citizens (including naturalized immigrants) have been consistently between 80 and 85 per cent, whereas participation amongst citizens of other countries has averaged less than half that. It has generally been higher among women than men in almost all groups. There has been considerable variation by region of origin.⁵ Of particular note is the fluctuation: turnout amongst non-citizens of African origin, for instance, was much higher in 2010 and 2014 than in earlier elections, while the opposite was the case amongst South Americans. To some extent this may be down to the fact that these are not constant groups. It is estimated that about a quarter of the non-citizen voters from each election over the last 20 years had naturalized by the time of the next election, with new mobile EU citizens and immigrants taking their place (Statistics Sweden, 2019, p. 35).

Table 2. Voting in municipal elections. Turnout by citizenship (per cent).

	2002		2006		2010		2014
	Women	Men	Women	Men	Women	Men	Women	Men
Sweden	80.5	79.6	82.1	80.9	84.9	83.5	86.7	85.1
Nordic countries	40.7	30.1	40.5	34.5	44.1	36.4	46.8	37.9
Europe (excl. Nordic)	35.0	33.3	38.5	35.4	32.2	30.0	29.5	28.4
Asia	35.0	23.8	35.9	30.2	39.4	29.3	34.3	24.1
Africa	25.1	25.8	35.9	30.8	46.8	40.6	49.5	43.3
North America	41.3	35.8	42.6	37.1	44.1	44.1	41.2	39.5
South America	50.4	38.6	42.5	40.9	41.7	37.7	40.9	36.3
Other countries	31.5	22.7	32.8	29	29.5	27.9	25.1	22.8
All	78.5	77.3	80.1	78.7	82.6	80.7	84.0	81.7

Voters are classified as ‘Swedish citizens’ if they hold Swedish citizenship, even if they previously held or currently hold another citizenship as well. Thus, the non-Swedish regional classifications refer to groups of people who are citizens of another country but not (also) of Sweden.

Source: Statistics Sweden, valdeltagandeundersökningar, 2002, 2006, 2010 och 2014.

To test the reasons for these differential turnout rates, and in particular to examine the patterns of turnout amongst ‘second generation’ Swedish-born individuals, we focus for the rest of the article on the 2014 municipal election in detail. The sample that is used in this study contains data on approximately 74,113 individuals, of whom 19,794 are foreign born (including 13,377 non-citizens) for the year 2014.

Earlier studies clearly have shown that both demographic- and socio-economic variables are important indicators for explaining voting or not voting. Thus, the models included in this analysis have the following demographic variables: sex, marital status, age; and important socio-economic variables: income quintile and educational level. The variables ‘sex’ and ‘marital status’ are dummy variables where males, and never being married, are respectively the reference categories. Age is used here as a linear control variable, whereas both the educational level and income variables are categorical variables. Education is classified by highest attained level, into the groups of primary schooling, secondary schooling, and low and higher university degrees. The income variable is created by classifying people into quintiles of income deriving from of paid work, self-employment, or transfers connected to earlier work. The reference categories are those who have no income from work and are either on study loans or social benefits.

As noted above, most previous studies analyzing voting behavior of immigrants show generally lower participation rates of immigrants versus natives. Since we not only want to analyze the immigrant-native voting gap, but also the gap between those socialized from birth in the Swedish political environment compared with those who have obtained citizenship subsequent to moving to Sweden, migration-specific variables including are included in the analysis: area of origin, years since migration, generation, and citizenship.

Whereas Table 2 showed overall turnout levels across the last four elections by region of origin, Table 3 narrows the focus to the 2014 election and distinguishes not only by region of origin, but also citizenship status and generation.⁶ Amongst the citizens, Swedes without a foreign background had a voting participation rate of almost 90 per cent in the 2014 elections. Amongst Swedish citizens born abroad, turnout was highest among those from Nordic countries, Latin and North America, and Oceania (over 80 per cent – slightly lower than among Swedish-born citizens, but still respectable by international standards) and lowest among individuals born in Africa and Asia.

Table 3. Voting participation (per cent) in 2014 Swedish municipal elections by area of origin (birth), citizenship and generation.

	Foreign citizen	Swedish citizen
Swedish-born	–	88.2
Nordic-born	43.2	80.1
EU28-born	29.2	71.8
Other European-born	27.6	70.7
North America and Oceania-born	40.0	82.0
Africa-born	33.8	68.3
Asia-born	29.2	68.9
Latin America-born	38.6	82.4
Foreign-born	34.3	72.4
Swedish-born (2 foreign-born parents)	45.0	77.4
Swedish-born (1 foreign-born parent)	52.0	85.1
Total	–	76.8

Source: Electoral Survey 2014, Statistics Sweden.

There are also apparent differences depending on the parental background of Swedish-born individuals. Those with only one foreign-born parent participated at a rate higher than that of any foreign-born group (at 85.1 per cent). By contrast, turnout was somewhat lower amongst Swedish-born individuals with two foreign-born parents. This provides some aggregate confirmation of hypotheses 6, but this cannot be confirmed until multivariate individual-level analysis is conducted systematically in section 4.

Voting rates for non-citizens varied by region of origin – but the most significant point is that turnout was much lower across the board among non-citizens than Swedish citizens from the same regions.

In the next section, we engage in a multivariate analysis of the electoral participation of immigrants in the 2014 municipal elections, using logistic regression. The use of logistic regression analysis is common in social sciences when the dependent variable is binary, in this case if people voted or not. For robustness checks we used fixed effect models which produced similar results. All regressions were conducted including individual covariates. We start with an analysis of immigrant voting patterns overall (testing hypotheses 1 to 3) and subsequently infer different groups of variables to determine their influence on voting behavior of immigrants and their children (hypotheses 4 to 6). In all regressions, we control for society-wide aspects by including a variable measuring municipality of residence. The results are presented in coefficients and odds ratios, relative to the reference categories.

Analysis of the immigrant electorate

Initially, the analysis focuses on a base-line specification of the native-immigrant voting gap, and on predicting the probability of voting, including all individual demographic and socio-economic covariates. Table 4 shows this baseline in model, A1.

Table 4. Logistic regression. Voting participation in 2014 Swedish municipal election.

	Model A1 (all voters)	Model A2 (foreign only)	Model B	Model C	Model D
Age	1.001	1.007***	0.995***	0.999	0.997**
Females	1.380***	1.341***	1.326***	1.281***	1.269***
Married	1.430***	1.185***	1.313***	1.309***	1.338***
Secondary education	1.387***	1.394***	1.295***	1.167***	1.134***
Lower university degree	2.227***	1.747***	1.823***	1.703***	1.638***
Higher university degree	2.351***	1.851***	2.063***	2.014***	1.858***
Income (quintile1)	1.685***	1.853***	1.667***	1.625***	1.620***
Income (quintile2)	2.125***	2.291***	1.913***	1.759***	1.695***
Income (quintile3)	2.878***	2.955***	2.399***	2.117***	2.045***
Income (quintile4)	3.816***	3.743***	2.936***	2.550***	2.469***
Income (quintile5)	5.221***	5.181***	4.056***	3.609***	3.477***
Born in Nordic country (except Sweden)	0.185***	0.273***	0.383***	0.781***	–
Born in E28 (except Nordic countries)	0.107***	0.178***	0.314***	0.508***	–
Born in Europe (except EU countries)	0.112***	0.181***	0.303***	0.467***	–
Born in North America or Oceania	0.092***	0.164***	0.297***	0.691***	–
Born in Africa	0.232***	0.385***	0.758***	1.301**	–
Born in Asia	0.158***	0.269***	0.466***	0.586***	–
Born in Latin America	0.142***	0.231***	0.390***	0.831*	–
0–10 years in Sweden			0.370***	0.656***	0.660***
11–20 years in Sweden			0.705***	0.763***	0.724***
21–30 years in Sweden			0.917	0.900	0.920
31–40 years in Sweden			1.033	0.973	1.024
Over 40 years in Sweden			1.293*	1.034	1.096
Swedish citizen				4.197***	3.743***
Swedish born 2 immigrant parents					1.292***
Swedish born 1 immigrant parent					1.857***
Constant	1.060	0.545***	1.086***	0.307***	0.248***
Nagelkerke R^2	0,329	0.219	0,245	0,306	0,326
N=	77 113	29 282	29 282	29 282	29 282

Reference categories: Males, Never married, Primary education, No income, Born in Sweden, Foreign citizen. In model 5 the reference category is Foreign born.

Significance: ***p < 0.01, **p < 0.05, *p < 0.10.

Voting probabilities amongst all foreign background voters

The estimates show that foreign-born individuals are less likely to vote than those born in Sweden, as noted already. This appears to confirm hypothesis 1. Interestingly, controlling for demographic and socio-economic factors, immigrants from primarily economic developing geographical parts of the world vote more relative to immigrants from those originating in European/North-American states (except the Nordic countries and Sweden).

When it comes to the covariates – sex and marital status – the analysis of the 2014 electorate confirms what earlier studies for Sweden and elsewhere have shown (Bäck & Soininen, 1994; Bevelander, 2015; Bevelander & Pendakur, 2011; Öhrvall, 2006; Togeby, 2004). Females are more likely to vote than males; and those who are (or have been) married are more likely to vote than those who have never been. However, once other variables are controlled for, age does not seem to have had much of an influence on voting participation in the 2014 Swedish municipal election, contrary to the expectations that would arise from mainstream electoral studies. (Among most populations, older voters generally have a greater propensity to vote (Blais, 2000; Wolfinger & Rosenstone, 1980)). However, the socio-economic covariates predict what was expected: higher education as well as higher income increase voting propensity.

The remaining models in Table 4 focus on the foreign-born and foreign-background part of the electorate. The baseline model A2 replicates model A1, but without those born in Sweden to two Swedish parents. In general, the demographic and socio-economic covariates show the same direction of influence as before. The reference category in this model comprises Swedish-born individuals with at least one foreign-born parent. As noted above, they had higher voting participation then all foreign-born groups. This result confirms the earlier descriptive results and holds when controlling for the variation in other individual characteristics among the studied population.

Hypothesis 2 suggested that the longer an immigrant had lived in Sweden, the more likely he or she would be to vote. Model B tests this, by adding a variable accounting for length of residence: ‘years since migration’. As predicted, foreign-born individuals are more likely to vote with increased time in the country. There appears to be a ‘catching-up’ process over time – in line with the expectations of the models of socialization discussed earlier. Recent arrivals have lower levels of participation that display a statistically significant difference from the reference category. There is no significant difference between immigrants of 20-40 years’ standing and those born in the country. Those who have lived more than 40 years in Sweden actually are more likely to vote than those born there, a finding that is statistically significant. Again, all other demographic and socio-economic covariates in the model show similar predictive levels and the same signs compared as the reference model A2. Once the length of time spent in the country is controlled for, however, the odds ratios connected to each region of birth also change substantially between models A2 and B, providing further evidence that the number of years spent in Sweden plays an important role in the likelihood of an individual voting.

Next, we turn to the third hypothesis – that the acquisition of citizenship would positively affect political participation. We test this in model C. Both the demographic and socio-economic variables show relatively stable odds ratios between models B and C, except again for the ratios relating to regions, which change substantially once citizenship acquisition is controlled for. However, the most important finding, which is in line with earlier studies for Sweden (Bevelander et al., 2015; Bevelander & Pendakur, 2011; Hainmueller et al., 2015, 2017) is that those who have obtained Swedish citizenship – in both generations – are far more likely to vote than foreign citizens.

‘Second generation’ voting probabilities

Hypotheses 4 to 6 related to the relative participation rates of foreign-born immigrants compared with the generation below. Hypothesis 4 suggested that Swedish-born people with at least one foreign-born parent would have higher participation rates than foreign-born immigrants (in other words, that the ‘second generation’ would have higher turnout rates than the ‘first generation’). For this purpose, the foreign-born group is used as the reference category, and we split the second-generation by the number of immigrant parents they have, at the same time controlling for all earlier included covariates. Model D shows that there is a significant difference in voting participation between (a) immigrants who are foreign-born; (b) Swedish-born people with two foreign-born parents; and (c) Swedish-born people with one foreign-born parent.

The findings lend credence to the relationship predicted in hypothesis 4, that Swedish-born voters of foreign background are more likely to vote than their parents’ generation of foreign-born immigrants. By itself, this is a relatively blunt discovery, and requires further investigation depending on the region of the world with which the foreign-background voters are connected. As Just and Anderson (2012) have pointed out, the difference in political culture between an individual’s place of birth and the country in which they settle in can make a difference as to how fast their political behavior changes. Similarly, we should examine whether the political culture into which parents were born may – though parental socialization – make a difference to how their children vote.

Table 5 contains several models that test this theory, looking at whether there are still differences in voting behavior amongst the foreign-background voters, depending on their parents’ regions of birth, and controlling for citizenship uptake.

Table 5. Logistic regression. Voting participation in 2014 Swedish municipal election (foreign background).

	Model E	Model F	Model G	Model H	Model I
Age	0.994***	1.000	1.002	1.003	1.003
Females	1.377***	1.238***	1.263	1.263***	1.273***
Married	1.756***	1.181**	1.182*	1.178*	1.181*
Secondary education	1.437***	1.514***	1.421***	1.417***	1.429***
Lower university degree	3.288***	2.883***	2.672***	2.664***	2.641***
Higher university degree	4.305***	4.330***	4.141***	4.128***	4.120***
Income (quintile1)	1.218***	1.152	1.120	1.120	1.129
Income (quintile2)	1.639***	1.566***	1.591***	1.589***	1.588***
Income (quintile3)	2.129***	1.582***	1.605***	1.605***	1.643***
Income (quintile4)	2.915***	2.078***	2.042***	2.042***	2.031***
Income (quintile5)	3.903***	3.423***	3.536***	3.526***	3.599***
Swedish citizen			4.807***	4.865***	4.807***
Swedish born one immigrant parent	0.670***	1.612***	1.353***	1.253***	1.447***
Swedish born two immigrant parents	0.418***
Swedish born immigrant mother				0.862*
Swedish born immigrant father					1.149
Constant	2.144	0.927	0.245***	0.279	0.208***
Nagelkerke R2	0.136	0.134	0.182	0.183	0.182
Population	54 319	6 843	6 843	6 843	6 843

Reference categories: Males, Never married, Primary education, No income, Born in Sweden (natives in model 6), foreign citizen, Swedish born two immigrant parents, Swedish born Swedish mother, Swedish born Swedish father.

In all regressions, we control for society-wide aspects by including a variable measuring municipality of residence.

Significance: ***p < 0.01, **p < 0.05, *p < 0.10.

Models E and F are our new baseline analyses. Model E aims to test hypothesis 5, that foreign-background persons will generally have lower participation rates than people without any foreign-born parents. It shows the difference in probabilities between Swedish-born voters with and without parents of foreign origin, controlling for demographic and human capital characteristics. Again, the results for demographic and socio-economic covariates are in line with earlier results. Females vote more than males, as do married people compared with unmarried ones. Voting participation increases with both educational attainment and economic affluence. But the most significant finding from model E that voters born in Sweden to foreign-born parents do indeed generally have lower participation rates than citizens without any foreign-born parents.

The remaining models in Table 5 focus only on such foreign-background (‘second generation’) individuals. Model F measures the difference in voting participation between those who have one foreign-born and one Swedish-born parent, and those have two foreign-born parents. It confirms hypothesis 6: those with one immigrant parent vote to a larger degree than those with two immigrant parents.

Having confirmed the six hypotheses, in the final part of this analysis we move beyond them in models G, H and I. Model G includes a covariate for those that have Swedish citizenship. This shows the expected high positive correlation – indicating that citizenship makes a difference even if an individual has grown up in Sweden.

In models H and I we include lastly whether household composition, in this case if either the father or the mother is Swedish- or foreign-born, affects the voting behavior of the individual. The results show that having a Swedish mother has a positive correlation with voting, whereas no significant effect can be measured of having a Swedish father. It is beyond the scope of this article to explain this in detail, but it is worth noting that the differential effects of maternal and paternal socialization have been well-documented in other spheres (Cassano & Zeman, 2010; Chaplin et al., 2005; Shortt et al., 2016; Wong et al., 2009).

Conclusions and discussion

Based on an investigation of the 2014 Swedish municipal elections, set against the backdrop of previous elections, this article has sought to move forward our knowledge about participation of voters of foreign background in the electoral process. Using unprecedentedly rich data that cross-references actual recorded turnout with the demographic and citizenship details of voters, it confirms earlier findings that the participation rates of foreign-origin voters are generally lower than that of those born in the country. The study has also confirmed that propensity to vote is correlated with higher levels of education and income, and that length of residence since immigration plays a role in making people more likely to vote. Even when controlling for all of these factors, the well-established ‘citizenship premium’ to political participation is confirmed once again.

This study has gone beyond previous studies, in two respects. First, it has established that the region of origin makes a difference. Foreign-born voters are generally less likely to vote than their Swedish-born counterparts, but after controlling for demographic and socio-economic factors, this effect varies by region. Future research would ideally investigate this further by disaggregating by country, but unfortunately this was not possible in this case, given the classifications applied by Statistics Sweden

Second, we have investigated not only the voting propensity of foreign-born voters, but also that of second-generation Swedish-born people whose parent or parents originated outside the country. This has allowed us to establish in much more detail the continuing socialization impact of a foreign background. Our findings are mixed. By comparison with the first generation, the Swedish-born voters of the next generation are more likely to vote. We understand that this is partly due to the socialization process of growing up in the country, rather than having to adapt to a new environment. At the same time, we can see the persistence of certain traits from the foreign-born generation. Swedish-born voters who are also Swedish citizens are much more likely to vote than those without citizenship. Those who have one Swedish parent – particularly a Swedish mother – are also more likely to vote than those who have two. In a survey study (Bevelander et al., 2021) focusing on children of immigrants, for some groups no statistical difference in voting behavior is measured when controlling for demographic, educational and social capital characteristics, in line with what Bevelander and Pendakur (2009) found for Canada.

This has important implications for our understanding of socialization processes, as well as more substantively for democracy. First, it appears to indicate that there is a continued political behavior legacy deriving from a foreign background, even if people are born in a country and socialized through its educational and societal structures. As noted earlier, this may be related to the fact that birth in Sweden alone does not mean that people are identically socialized. It is still likely that those with foreign background will come into contact with others of foreign background in early life through social and urban segregation and parental friendships. In turn, this means that we should revisit the traditional models of socialization. The ‘resistance’ model of political socialization, suggesting that political behavior is conditioned by early experiences alone, appears to be too rigid for application to the process of first- and second-generation political integration.

Moreover, there are important implications for the representativeness and legitimacy of the political system. As noted earlier, Sweden has a substantial proportion of people who were either born abroad or had at least one parent who was. Whilst many foreign-born immigrants have naturalized, the analysis above has shown that that there remain persistent voting differences (generally lower participation) even amongst such naturalized citizens and their children. Moreover, although Sweden has a relatively liberal qualification for the franchise in municipal elections, there are persistently lower rates of participation amongst non-citizens of the country, even though they have the right to vote. This means that they are less well-represented in the political system, potentially leading to the sort of ‘participatory distortion’ highlighted by Verba et al. (1995, p. 708) and others (e.g. Piven & Cloward, 2000).

On these points – the models of socialization and the ‘representation gap’ – much research remains to be conducted, but the present study provides a significant advance on our existing knowledge of generational differences in voting participation amongst voters of foreign background.

Acknowledgements

Preliminary work on this article benefited from support and comments from the Migration Studies Delegation (DELMI) of Sweden. Work on the article was also co-funded by the European Union’s Rights, Equality and Citizenship Programme (2014–2020) and the GOVCIT project (2017–2020) by The Norwegian Research Council. The content represents the views of the authors only and is their sole responsibility. The European Commission does not accept any responsibility for use that may be made of the information it contains.

Notes

1 In official Swedish statistics, ‘foreign background’ refers to individuals who were either themselves born outside Sweden, or had two parents born outside Sweden. Given that our explicit focus in this article is to examine the effects of different socialization factors (including the number of foreign-born parents a Swedish-born person has), we generally utilise a wider definition that encompasses anybody with at least one non-Swedish-born parent. When we refer to the narrower (official) definition, it is denoted with quotation marks.

2 In 2014, the year that is the main focus of the empirical part of the article, the numbers were slightly lower (16.5 per cent foreign-born, 21.5 per cent foreign background, and 7.6 per cent non-Swedish citizens), but still comparatively high by European standards.

3 For the full discussion around the introduction of voting rights for non-Swedish citizens, see SOU (Swedish Official Publications) (1975). Kommunal Rösträtt för invandrare (Stockholm: Kommundepartementet) No. 1975.05. Available at: https://data.kb.se/datasets/2015/02/sou/1975/1975_15%28librisid_14680805%29.pdf (accessed 17 August 2020).

4 The Local Government Act from 1991 was replaced in 2017. Apart from a minor change of wording, it enfranchised the same groups.

5 Although it would be ideal to track by country rather than region of origin – given the disparities between countries in the same regions – this is not possible due to the classifications applied by Statistics Sweden in the original dataset.

6 The geographical regions used in table 3 and the rest of this article differ slightly from those in table 2. This is due slightly different classifications in the dataset in 2014 compared with earlier years, and the need to standardize for comparability across elections.