Various faces of localised voting in Sweden

ABSTRACT

This article investigates the extent of and characteristics of localised voting in a Swedish context. Localised voting is defined as an independent act of choice in relation to local elections. Contrary to most previous research, this article suggests that localised voting should also include those who are well informed about each election but vote for the same party in the different elections and not only those that split their tickets. A citizen survey conducted in the four northernmost counties indicates that two-thirds of the voters are local; 38 percent are informed same-party voters and 28 percent split their tickets. To a large extent, the two types of localised voting are explained by different factors. The informed same-party voters tend be older and are critical of the state of democracy in their municipality whereas the split-ticket voters have weak party allegiance and generally support a small party.

KEYWORDS:

• Localised voting
• Sweden
• local democracy
• local elections
• first order elections

The literature on voting in local elections suggests that these generally have a second-order status. Those who take part are expected to make less of independent choices, assuming that less is at stake, and instead model their choice of party and candidates largely on their preferences in the national elections. Originally developed in relation to European Union elections, a study by Reif and Schmitt suggested that ‘many voters cast their votes in these elections not only as a result of conditions obtaining within the specific context of the second-order arena, but also on the basis of factors in the main political arena of the nation’ (Reif and Schmitt 1980, 9). Second-order elections are also characterised by lower turnout and better prospects for smaller and newer parties. Many studies have confirmed that this also seems relevant at the local level (see, for example, Miller 1988), although others identify a considerable amount of ‘first-order thinking’ (Cabeza 2018) or ‘localized voting’ (Kjær and Steyvers 2017, 2019), or even characterise local elections as being ‘one and three-quarter elections’ (Heath et al. 1999).

Localised voting is here understood as an independent act of choice in relation to local elections. Extensive localised voting is crucial for the quality and legitimacy of local democracy and for a functioning chain of representation between the local electorate and their representatives. One expression of localised voting is the extent to which people split their votes between different parties. However, here it is argued that localised voting is more than just split-ticket voting as it should also include informed same-party voters, i.e. those who make conscious and informed choices for each election but decide to support the same party in the local and national elections. This has not been taken into account in previous research which risks generating a skewed picture of the extent of and characteristics of localised voting and of the democratic quality and legitimacy of local elections.

The aim of this article is to investigate the extent of localised voting in a Swedish context and to capture the characteristics of the voters who make an independent choice in local elections. Particularly relevant is the question whether informed same-party voting is explained by the same factors as split-ticket voting.

Sweden is chosen as a least likely case for localised voting. This is the only European country that consistently holds local and national elections on the same day.¹ It is also a country where local government plays a clear national role by providing extensive national welfare services for citizens and where 95 percent of the local councillors belong to any of the parties in the national parliament. Although turnout is almost as high in the local as in the national elections,² this may include voters who would have abstained from voting if there was a separate local election day. For these reasons, local elections are particularly at risk of being dominated by national issues. Indeed, this potential problem has been a subject of longstanding debate since Sweden introduced the joint election day in 1970, and there have been recurring proposals to return to separate days for local and national elections (Kjellgren 2001; Mörk, Erlingsson, and Persson 2019). For this reason, Sweden is a good critical case for testing assumptions about local elections as second-order elections and the extent to which there is localised voting. If local elections are subject to individual choice by Swedish voters, they stand a good chance of being so in other countries as well.

Previous research in Sweden has only concerned split-ticket voting (for example, Johansson 2010; Erlingsson and Oscarsson 2015; Berg, Erlingsson, and Oscarsson 2019), but we do not yet know the extent of informed same-party voting. This article is a result of the first research in Sweden to take into account the full picture of localised voting and, thereby, provides a more comprehensive analysis of the phenomenon.

In relation to previous studies, the joint election day has several analytical advantages. It makes it possible to ensure that the time lag between local and parliamentary elections has no effect on the analyses. The misconception that voters use the local elections to express a mid-term verdict on the popularity of the national government in office can be avoided, although there is no guarantee that Swedish voters use three votes as ways of expressing general support for their favourite parties. Indeed, one of Reif and Schmitt (1980) original assumptions was that second-order characteristics would be strongest in the mid-point between two national elections. Hence, concurrent elections should also make localised voting less likely, which is also a strong argument for Sweden as a less likely case. Voting on the same day also means that split-ticket voting, i.e. voting for different parties in different elections, is a simultaneous activity and does not need to be constructed. Many other studies have had to rely on respondents’ stated voting intentions, assuming that there was a national election on the same day (e.g. Marien, Dassonneville, and Hooghe 2015; Kjær and Steyvers 2017).

Expected variations in localised voting

Localised voting may vary among individuals as well as among local contexts. In this section, we provide a summary of assumptions in the literature on the mechanisms that enhance localised voting and the type of variables that will be used in the subsequent empirical analysis when testing their relevance in the Swedish setting. Localised voting can be seen as another type of public participation, as an expression of connection to the local territory, as a particularly sophisticated type of electoral choice or as driven by the political preferences and party identification of the individual. In addition, we will explore the role of the municipal context and, in particular, whether population size and density and the supply of political parties enhance localised voting. The potential explanations are regarded as complementary rather than conflicting. For example, localised voting can be both strongly connected to the local territory and be enhanced by non-socialist political preferences.

Individual resources, engagement and networks

One way of explaining localised voting is to understand the local focus as an act of public participation. Those who typically participate in public life would not only be more likely to vote generally but also expected to make individual choices in local elections. The assumption is that the factors that enhance participation generally should also make citizens more able to discriminate between different types of elections. Therefore, the drivers for participation should also be relevant as predictors for localised voting. One particularly influential participation theory, the Civic Voluntarism Model, emphasises the individual’s resources, engagement and network membership as the explanation for why he or she takes part. Verba, Schlozman, and Brady (1995) argued that previous explanations in terms of socioeconomic conditions were too limited. They claimed that inactivity in political life was attributable to a lack of resources (citizens being unable), to issues of engagement (citizens being uninterested) or a failure in recruitment (citizens not being asked to get involved). This theory has been supported by numerous empirical studies (for example, Oliver 2001; Barkan 2004). The importance of individual resources for localised voting has been confirmed in previous research (Johansson 2010; Berg, Erlingsson, and Oscarsson 2019; Cabeza 2018; Oliver, Ha, and Callen 2012; Elklit and Kjær 2013).

In this analysis, we will investigate the role of gender, hence treating it as a resource variable. Previous studies have shown that women in Sweden are slightly more likely to split their votes than men are (Berg, Erlingsson, and Oscarsson 2019). Having a Swedish background would also facilitate access to the political system in a way that has turned out to be more difficult for people with an immigrant background. Further, the individual’s level of education is commonly seen as a resource that facilitates participation, and there is clear evidence that this enhances split-ticket voting in Sweden (Johansson 2010; Berg, Erlingsson, and Oscarsson 2019), participation in local elections in the Netherlands (Lefevre and van Aelst 2014) and first-order thinking in local elections in Spain (Cabeza 2018). Age can also be regarded as a resource, but previous research results vary with regard to its direction. Young people seem more likely to split their tickets, while older people may have more experience and insights, which may result in their being more likely to make informed choices (Erlingsson and Oscarsson 2015; Cabeza 2018).

The two additional components suggested by the Civic Voluntarism Model, i.e. to be engaged in politics and to be involved in networks, are also connected to the idea that citizens’ general civic orientation matters for their political engagement (Almond and Verba 1963) and with social capital theory (Putnam, Leonardi, and Nanetti 1993). Hence, we would assume that those who are engaged and involved in networks are also more likely to be localised voters. These people, more than others, would be better able to distinguish between the different levels of government as well as interested in having an influence at each level. The indicators used here will be interest in politics and activism in voluntary associations.

Localism and access to local information

Another way of reasoning is to expect localised voting to be a function of links to the local community. In the literature on urban governance, the distinction between localism and city regionalism is an analytical dimension that helps explain differences in views on metropolitan integration and reform (Briffault 2000; Strebel 2018). The localist position prefers as much decentralisation as possible (Evans, Marsh, and Stoker 2013) and far-reaching municipal autonomy. Regionalists, on the other hand, would be more in favour of city-regional institutions and territorial redistribution (Barron and Frug 2005; Owens 2010). The assumption here would be that those with the tightest links to the municipality would also be those who would make informed and conscious choices with regard to local elections.

Localist connections to the municipality may be physical or cognitive. The duration of residence in the municipality is an obvious physical connection. It has been shown that, in the US, long-time residents are more likely to vote in local elections (Oliver, Ha, and Callen 2012). Physical connection may also involve ownership. Indeed, according to the same study, home ownership is the main explanatory factor for localised voting in the US (Oliver, Ha, and Callen 2012). But localism may also be cognitive, for example, as the subjective proximity that citizens feel in relation to the municipal territory (Marks 1999; Laczko 2005; Lewicka 2011). We would expect those with strong identification with the municipality to be more localised in their voting. We will also assume that people who regard the politics of their municipality to be legitimate (Wheatherford 1992) are more likely to be localised voters. We will use citizens’ satisfaction with democracy in the municipality as an indicator of this attitude. Another type of cognitive local connection is knowledge about the local area. Voters who know more about the local community and the stakes involved in the local elections are expected to be more localised in their voting (Lefevre and van Aelst 2014). The indicators capturing how well-informed voters are on local matters are the voters’ self-assessed knowledge about party proposals and knowledge about candidates in the municipal elections, and the extent to which they read a daily local newspaper.

Voting sophistication

Making a separate and independent choice in the local elections in a predominantly second-order context may require a higher degree of sophistication than a routine vote (Downs 1957; Sniderman, Brody, and Tetlock 1993). Compared to the media-exposed national elections, where citizens are informed by intensive campaigning, it could be expected that citizens are required to think more strategically in relation to local elections (Karp et al. 2002; Marien, Dassonneville, and Hooghe 2015). Suggested by theories of representative democracy, this may involve basing the choice on either an evaluation of previous records or a decision on promises for the future set out in a party’s manifesto (Prezworski, Stokes, and Manin 1999).

Hence, citizens who exert either evaluative or programmatic voting are assumed to be more localised in their voting than others. We will also assume that the time for the voting decision will matter and will suggest that the longer one waits, the more information one can collect, and the more likely it is that the voter will make a localised choice. Hence, localised voters would be likely to make up their minds during the last month or even in the last week before the election. At least in Denmark, the split-ticket voters make a late decision (Elklit and Kjær 2013).

Another type of sophistication could involve making the effort to identify not only which party to vote for but also to select a specific candidate. Knowledge of local candidates has been shown to enhance localised voting in Belgium (Marien, Dassonneville, and Hooghe 2015). Although the Swedish election system is essentially a choice between parties, voters can also express their preference for one candidate on the list. If the candidate is endorsed by at least 5 percent of the voters for the party in the municipality, he/she is moved to the top of the list. Personalised voting will be captured by two variables: whether the voter indicated a specific candidate and whether he/she generally emphasises person rather than party as a basis for electoral choice.

Political preferences

Localised voting may also have a party dimension. Previous studies have found connections between moderate ideological positions and localised voting (Marien, Dassonneville, and Hooghe 2015) and between weak party identification and split-ticket voting (Karp et al. 2002). In Sweden, Social Democratic voters have turned out to be less willing to split their tickets, and it has been suggested that this reflects a collectivist tradition and a stronger sense of party attachment among its supporters. Voters for non-socialist parties, on the other hand, are more likely to split their votes, and their voters also tend to identify with a particular party to a lesser extent (Oscarsson and Holmberg 2016). For this reason, we will assume that support for a non-socialist party distinguishes local from non-local voters and that convinced party supporters are less localised in their voting.

The local context

Finally, we would expect that localised voting is determined not only by personal characteristics but also by the type of community where the voter resides (compare Kjær and Steyvers 2019). A number of specific contextual factors could be relevant, such as the popularity of the political regime and local scandals. As data on such events are difficult to obtain, we will focus on general contextual conditions. First, we would assume that small municipal size and sparse population would facilitate independent choices in local elections. In such a community, it is more likely that citizens know candidates personally (Dahl and Tufte 1973; Denters et al. 2014). In Norway, local matters are more important reasons for party choice than national matters in the smallest municipalities (Bjørklund 2017).

Second, the supply of political parties in a municipality would also enhance informed local choice (Johansson 2010). The political parties that are represented in the national parliament have candidates in all municipal elections, but local or regional party candidates run for office only in some municipalities. The presence of local parties in the municipality is expected to make localised voting more common and should, in particular, enhance split-ticket voting.

Extent of localised (and non-localised) voting

Methodology and data material

This article adds to a field of research in which individual-level data are rare but much desired (Clark and Krebs 2012; Kjær and Steyvers 2019). The analysis is based on a citizen survey conducted in the four northernmost counties, consisting of 44 municipalities, immediately after the general elections on the 19 September 2010. A total of 4000 respondents were randomly selected and posted a survey. The survey was distributed on the 20^th of September and the final answers were received by the end of January 2011. The response rate was 63.5 percent, which is very good for this type of survey. The survey was conducted as part of an analysis of a proposed regionalisation reform (which was not carried out) and included a number of questions on the different elections, which makes it useful for the analysis in this article as well. Support for the different parties in the data set corresponds closely to the real election results, although Social Democratic and Left Party voters are slightly overrepresented (Blomgren 2012). Only respondents who had reached the voting age of 18 were included in the analysis.

There is no other dataset in Sweden that allows for analyses of localised voting beyond split-ticket voting. However, the four northernmost counties differ from the rest of Sweden in that they are more sparsely populated and do not include any large metropolitan areas. They also diverge in a political sense as citizens are more supportive of parties to the left. In the country as a whole, there are no significant differences in split-ticket voting between cities and the countryside (Berg, Erlingsson, and Oscarsson 2019), but as the analysis below will show, localised voting in the broader sense is partly related to municipal population size and density, but these differences disappear after controlling for individual characteristics. Hence, it is highly likely that the results would be similar in other parts of the country as well. For this reason, the findings of this study can be regarded, at least preliminarily, as relevant for the whole of Sweden. Future research in Sweden with questions that allow for a more comprehensive understanding of localised voting will be able to corroborate or refute the results of this study.

Measuring localised and non-localised voting

When defining localised voting as an independent act of choice in relation to local elections, the ideal would be to measure this by asking people to what extent local matters are relevant when they decide how to vote. Unfortunately, the survey did not include any such straightforward question.³ However, it should still be possible to identify this group on the basis of two sets of information from the survey. The first is actual split-ticket voting between local and national elections. The second is the response to a question about whether one makes separate decisions in the different elections.

Although some who choose different parties in the local and national elections may do so as a way of spreading their votes among several preferred parties without considering the level of government that the vote concerns, it seems reasonable to assume that most voters who split their tickets are aware that this involves different levels of government. Table 1 summarises straight and split ticket voting in the parliamentary and municipal elections. Notably, the county council elections are not included in these figures, and as this study is limited to the municipal level, they will also be omitted in the subsequent analyses.

Table 1. Straight and split-ticket voting in the 2010 parliamentary and municipal elections

	N	%
Straight voting (for the same party)	1396	69.7
Split-ticket voting	508	25.4
Voted in only one election	99	4.9
All	2003	100

Notes: The table summarises straight and split-ticket voting in the parliamentary and municipal elections in the four northernmost Swedish counties in 2010. Omitted from the table are blank voters, non-voters and those who did not answer the question. A blank voter is one who submits an empty ballot paper, i.e. without any party name or name of a candidate. Number of cases included in the analysis: 1192 (1076 are missing in a total population of 2268).

A quarter of the electorate split their tickets which is fairly close to the approximately 27–28 percent reported in the national surveys that were conducted in relation to the 2010 elections (Berg, Erlingsson, and Oscarsson 2019). However, as suggested previously, those choosing the same party in the parliamentary and municipal elections but who reached this conclusion after separate decisions in each election will also be regarded as localised voters. As shown in Table 2, this turns out to be very common as 60 percent fully or partly agree with such a statement. We do not know exactly at which elections these voters made separate decisions – it may be only two of the three. However, split-ticket voting between municipal and parliamentary elections is the most common type of split-ticket voting (Oscarsson and Holmberg 2018), and we will, therefore, assume that those who fully or partly agree with this statement are localised voters in this sense.⁴ There is considerable consistency between split-ticket voting and responses to the question of separate decisions. Almost 90 percent of those splitting their ticket between parliamentary and municipal elections also claim to make separate decisions. There are 50 respondents who split their vote but disagree with the statement, and these may use the different tickets as a way of spreading their support among preferred parties without considering that it concerns elections at different levels of government. These respondents have been omitted from the analysis.

Table 2. Separate decisions for the different elections

	N	%
Fully agree	658	32.8
Partly agree	546	27.2
Partly disagree	174	8.7
Fully disagree	380	19.0
No opinion	247	12.3
All	2005	100

Notes: The general wording of a set of questions was ‘How did you reason when making up your mind in the elections?’ Respondents were asked to respond to the statement ‘I made separate decisions for each election (parliament, county council, municipality).’ It is acknowledged that the wording is not ideal as informed voting may be a matter of degree. Also, the question may suffer from a social desirability bias.

One point of criticism could be that this design identifies politically conscious voters in general rather than only localised voters. However, an understanding in second-order theory is that national elections are the norm, and any independent position of elections at other levels of government requires consciousness about that particular level. Therefore, a voter who makes conscious choices for different levels of government is treated as a localised voter by definition.

We can now identify the full category of localised voters. These are considered to be those who vote for different parties in the parliamentary and municipal elections, those who vote for the same party but make separate decisions in the different elections, and those who vote only in the municipal elections. The remaining voters are considered to be non-localised voters. These voters are either voting for the same party in the parliamentary and municipal elections, without making separate decisions for each election, or are voting only in the parliamentary elections. As shown in Table 3, with this definition, localised voters would amount to 67 percent of all voters. Apart from the 28 percent who are local split-ticket voters, we would also add the 38 percent who vote for the same party in the two elections but make separate decisions and a small group who voted only in the municipal elections.

Table 3. Localised and non-localised voters

	N	%	All
Localised voters:
Split-ticket voting between parliamentary and municipal elections	508	27.5
Voting for the same party but making separate decisions	696	37.6
Voted only in the municipal elections	37	2.0	67.1
Non-localised voters:
Voting for the same party and not making separate decisions	546	29.5
Voted only in the parliamentary elections	62	3.4	32.9
Total	1849	100

Two-thirds of the electorate is a surprisingly large share, in particular when considering that the joint election day has been criticised for downplaying the scope of local politics. The localised voters are actually outnumbering the non-localised voters.⁵ This also provides strong evidence for Swedish local elections being independent elections and not just elections of a second-order determined by national political issues. Recalling the initial discussion about the state of local democracy, this would suggest an unexpectedly healthy state of local democracy in Sweden. The level of localised voting is only slightly higher than in Denmark, where 62 percent of the electorate were localised voters in the 2013 local elections.⁶ In Norway, 43 percent claimed that local matters were the most important reason for their party choice in the 2015 local elections, and this was an increase from 29 percent in 1995 (Bjørklund 2017).⁷

The results also suggest that previous studies of localised voting that take only split-ticket voting into consideration risk being highly misleading. Those who make informed decisions but choose the same party in the different elections may have different characteristics than the split-ticket voters. Whether this is the case will now be explored.

Who is the localised voter?

Localised voting has been operationalised as consisting of two components – split-ticket voting and informed same-party voting.⁸ Remaining voters are non-informed same-party voters. With binary dependent variables and with independent variables at both individual and municipal levels, this suggests a multi-level design using logistic regression analysis. Such analyses, conducted with all three dependent variables, indicate that there is significant variation at the municipal level only with split-ticket voting as a dependent variable. Such voting tends to be more common in smaller and more sparsely populated municipalities and in municipalities where local parties are represented in the council. However, the second-level effects (at the municipal level) lose their significance when first-level variables are entered. This suggests that differences between municipalities are due to the population composition of the areas. For this reason, all analyses will be performed using ordinary logistic regression techniques. As the three components are mutually exclusive and not ordered, we will use multinomial logistic regression analysis.

Although presented as separate sets of assumptions, a few of them are empirically interrelated. For example, people who are middle-aged and older tend to read local newspapers (rxy =.345 and .326 respectively), and strongly convinced party supporters are also more interested in politics (rxy = .216). However, there are no signs of multicollinearity, as no independent variable generate a VIF-value above .4. Perhaps surprisingly, those with stronger individual resources, such as more education, do not score high on any of the indicators of access to local information or voting sophistication, which underlines the empirical relevance of treating these as separate categories of assumptions.

The results of the logistic regression analyses are presented in Table 4. The table presents the coefficients and standard errors for two of the three components in the dependent variable – split-ticket voting and informed same-party voting. Non-localised voting is the reference category. The overall relevance of the model is relatively good, with a Cox & Snell R square of .31 and a Nagelkerke R square of .35. There is support for all four sets of assumptions about how variation in localised voting can be explained, but individual resources and civic orientation does not explain variation in split-ticket voting and party preference does not discriminate among the informed same-party voters.

Table 4. Explaining variation in localised voting. Multinominal logistic regression analysis

	Spilt-ticket voting		Informed same-party voting
	B	SE	B	SE
Individual resources and civic orientation
Male (gender)	0,150	0,185	0,115	0,158
Swedish background	0,412	0,416	−0,692*	0,327
Education (ref: Basic)
Upper secondary (1)	0,117	0,332	−0,532*	0,256
Unfinished higher (2)	0,502	0,354	−0,368	0,282
University degree (3)	0,144	0,349	−0,333	0,275
Age (ref: 18–29)
30-49 (1)	−0,123	0,314	0,179	0,290
50-64 (2)	0,154	0,347	0,361	0,316
65-85 (3)	0,187	0,413	0,988**	0,356
Interest in politics	0,346	0,192	0,441**	0,165
Active in voluntary association	−0,053	0,192	0,095	0,166
Localism and access to local information
Duration of municipal residency	0,148	0,190	0,044	0,162
Home ownership	−0,070	0,228	−0,028	0,195
Municipal identification	0,177	0,203	0,250	0,175
Satisfied with local democracy	−0,066	0,229	−0,535**	0,193
Knowledge: Party proposals	0,931***	0,236	0,604**	0,202
Knowledge: Candidates	0,072	0,243	0,269	0,208
Local newspaper reader	−0,139	0,254	0,157	0,235
Voting sophistication
Evaluative voting	−0,139	0,272	0,249	0,265
Programmatic voting	−0,079	0,249	0,691**	0,233
Time for decision (ref: Before spring)
Spring/summer (1)	0,954***	0,271	0,370	0,229
Final months (2)	1,137***	0,268	0,608**	0,231
Final week (3)	1,462***	0,267	0,508*	0,245
Indicate a specific candidate	0,239	0,208	−0,035	0,173
Emphasise person rather than party	0,590***	0,083	0,274***	0,077
Party preferences
Party (ref: Social dem)
Moderate Party (1)	−0,049	0,272	0,046	0,213
Centre Party (2)	0,904**	0,295	−0,238	0,293
Liberal Party (3)	0,804*	0,332	−0,503	0,333
Christian Democrats (4)	−0,056	0,513	−0,306	0,426
Left Party (5)	0,579	0,320	−0,201	0,294
Green Party (6)	0,791*	0,354	−0,016	0,330
Sweden Democrats (7)	−0,481	0,899	0,251	0,567
Convinced supporter (ref: No)
Fairly convinced (1)	−1,355***	0,329	−0,278	0,243
Strongly convinced (2)	−0,586**	0,208	−0,175	0,191
Intercept	−5,265***	1,331	−2,694*	1,159
Pseudo R-Square:
Cox and Snell: 0,307
Nagelkerke: 0,347
McFadden: 0,170

Notes: The coefficients are B-values from the multinomial logistic regressions and the standard errors. The third value of the dependent variable, non-localised voting, is the reference category. Levels of significance: * 0.05; ** 0.01; *** 0.001. For the wording of survey questions and the construction of the variables, see Appendix. Voters with preferences for local parties have been excluded from the party variable as these by definition would be split ticket voters, thereby coincidingly too closely with the dependent variable.

Many features are common for both types of localised voters. Hence, independent of split-ticket and informed same-party voting, the localised voters are better informed about the party proposals in the local elections and tend to make up their minds late during the election campaign. They also emphasise person rather than party in their choice. However, other than that, split-ticket voters and informed same-party voters clearly belong to different groups of citizens. Indeed, the two components are strongly negatively correlated (rxy = −.478).

Split-ticket voters are more common among the supporters of the Centre and the Liberal parties and also, surprisingly, of the Green Party. The common denominator is that all these parties are among the smallest in the Swedish political system. Those with weak party allegiances are also more likely to split their votes. To some extent, these patterns confirm findings of previous research in Sweden on split-ticket voting (compare Berg, Erlingsson, and Oscarsson 2019). Those who make informed same-party voting decisions, on the other hand, have a basic level of education, are older and are interested in politics. They also tend to be programmatic voters and, contrary to expectations, they have an immigrant background and are also dissatisfied with the state of democracy in their municipality.

The differences between the two types of localised voters suggest that studies that consider split-ticket voting only will not provide the full picture. Omitting the part of the electorate that makes an informed choice but ends up voting for the same party in both local and national elections underestimates the extent of localised voting. It also fails to recognise that this vote may often be a protest vote. For this reason, the operationalisation of who the localised voter is has significant consequences for conclusions about the status of local elections.

Several of the initial assumptions have turned out to be irrelevant for any form of localised voting. There are no evident differences between men and women, and those with the highest level of education are not more likely to be localised voters than others. The prediction based on participation theory that activism in associations would enhance localised voting was not relevant. Perhaps most surprisingly, many measures of localist orientation were not supported by evidence, including duration of residence, home ownership and municipal identification. A strong attachment to the home municipality does not enhance localised voting in Sweden.

Summary and conclusions

The aim of this article has been to investigate the extent of and characteristics of localised voting in a Swedish context. Its main contribution in relation to previous research on the first-order characteristics of local elections is to include those who make an informed choice but vote for the same party in the local and national elections in the concept of localised voting. Localised voting is more than just split-ticket voting.

The picture that has emerged is, in several important respects, different from what the research literature has found previously. Although split-ticket voting largely follows the same pattern that has been identified elsewhere, in Sweden as well as in other countries, those who make an informed choice but select the same party represent a different group. A major difference is that the split-ticket voters have a specific party-political profile. In the municipal elections, they choose of the smaller parties and have weak party allegiances. In contrast, those who are well-informed and vote for the same party tend to have different personal characteristics. They are overrepresented among people with an immigrant background, older citizens and those who are interested in politics, make a programmatic vote, and consider the parties’ proposals for the future. Surprisingly, those who are critical of the state of democracy in their municipality are also more likely to be informed same-party voters. Hence, with this more comprehensive definition of the localised voter, a broader set of electors, with different characteristics, is included.

The overall extent of localised voting amounts to two-thirds of all voters. Of these, 28 percent are local split-ticket voters and 38 percent vote for the same party in the two elections but based on informed and separate decisions. Even with a more conservative estimation, counting only those who fully agree with the statement that they make separate decisions in the different elections, almost half the electorate would be localised voters. This is a surprisingly large proportion, in particular when considering that the joint election day has been criticised for downplaying the scope of local politics. Despite the joint election day, a nationalised system of local government and the dominance of national parties in the local councils, our data suggest that local elections in Sweden are at least as much first as second order. Although the dataset is limited to a part of the country, there is good reason to believe that similar patterns would emerge if the country as a whole were to be investigated as all significant predictors are at the individual level and do not concern differences between municipalities. In this study, Sweden was regarded as a critical case of the test of the first-order character of local elections. Since Swedish local elections turned out to be subject to individual choice to a considerable extent, perhaps by as much as two-thirds of the electorate, local elections stand a good chance of being independent elsewhere as well. This is an important contribution to the literature on how the democratic quality and legitimacy of local elections should be assessed.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Anders Lidström

Anders Lidström is Professor of Political Science at Umeå University, Sweden, with a research focus on local, regional and urban politics and governance. His empirical studies concern both Swedish conditions and comparisons between countries. He is the author, with Jefferey Sellers and Yooil Bae, of Multilevel Democracy. How Local Institutions and Civil Society Shape the Modern State (Cambridge University Press, 2020).

Notes

1. Elections in Sweden are to the national parliament, county councils and municipal councils. Between 1970 and 1994, they were held every third year, but from 1994 onwards, every fourth year. A proportional, list-based system is applied on all levels. Political parties provide ballot papers with the party name and a list of their candidates. A party label is compulsory, i.e. independent candidates are not allowed; however, groups of citizens are free to form new and local parties. Each list gets seats according to the share of the votes that the list receives. Candidates are normally elected in the order in which they are listed, but as it is possible to indicate support for one particular candidate, this candidate supersedes the top candidate as listed by the party if he/she receives at least 5 percent of the share for the party.

2. In the 2018 elections, 86 percent voted in the national and 84 percent in the local elections.

3. Both the Norwegian and Danish local-election studies have questions on the extent to which local matters are important for voting choices in local elections (for the most recent surveys, see Kjær and Steyvers 2017; Saglie and Christensen 2017).

4. In the subsequent empirical analyses, whether the results would be different if we chose a narrower operationalisation by including only those responding that they fully agree with the statement has been tested. This has only a marginal impact on the patterns that are identified, suggesting that there are no major differences between the respondents replies of fully and partly agree.

5. The share of localised voters would still be considerable if we were to apply, instead, a more conservative definition of ‘Voting for the same party but making separate decisions’ by including only those fully agreeing with this statement. Almost a fifth, or 19.3%, of the respondents would fall into this category, which would mean that the total percentage of localised voters would be 48.8%.

6. This is the sum of the split-ticket voters (31%) and those voting for the same party but claiming to have a local focus in their choice (31%) (Kjær and Steyvers 2017). As Denmark holds national and local elections on different days, the local election survey asked about party preferences if there had been a national election on the same day. We assume that those with a local focus correspond to those in the Swedish survey claiming to make separate decisions in the different elections.

7. However, this question is constructed differently, which makes direct comparisons difficult. For example, in the Norwegian election studies, there is also a response alternative ‘general trust for the party’ that was selected by 38 percent in the 2015 elections.

8. In the subsequent multivariate analysis, informed same-party voters will also include those only voting in the municipal elections, see Table 3.

Appendix

Variables in the logistic regressions

Table

Variable	Variable construction/question (question wording in italics)	Variable values
Indicators of the dependent variable
Localised voting		1. Split-ticket voting 2. Informed same-party voting 3. Non-localised voting
Indicators of the independent variables
Male (gender)	Gender of respondent	1. Female 2. Male
Swedish background	Foreign background is defined as having immigrated oneself, having lived most of one’s childhood abroad, or having a parent who lived most of his/her childhood abroad. All others are defined as having Swedish background.	1. Foreign background 2. Swedish background
Level of education	What is your highest level of education? Seven response alternatives were reduced to four in the analysis.	0. Basic, compulsory school, etc. 1. Upper-secondary school, etc. 2. Attended university or college 3. Earned a university or college degree
Age		0. 18–29 years 1. 30–49 years 2. 50–64 years 3. 65 ≠ 85 years
Interest in politics	How interested are you in politics generally?	1. Not particularly/not at all interested 2. Fairly/very interested
Active in voluntary association	How many voluntary associations are you active in?	1. None 2. At least one
Duration of municipal residency	Duration of residency in the home municipality	1. Moved to the municipality 2. Have lived in the municipality almost since birth
Home ownership	Are you owning or renting your home?	1. Renting 2. Owning
Municipal identification	To what extent do you feel at home in the municipality where you live?	1. Not fully feeling at home 2. Feel fully at home
Satisfied with local democracy	To what extent are you satisfied with the state of democracy in your home municipality?	1. Not particularly/not at all satisfied 2. Fairly/very satisfied
Knowledge: Party proposals	How would you assess your knowledge about the parties’ proposals in the 2010 local elections?	1. Very/fairly good 2. Very/fairly bad
Knowledge: Candidates	How would you assess your knowledge about the parties’ candidates in the 2010 local elections?	1. Very/fairly good 2. Very/fairly bad
Local-newspaper reader	How many daily local newspapers do you read?	1. None 2. At least one
Evaluative voting	To what extent do you agree with the following statement? ‘I based my voting decision on what the parties had done previously’.	1. Fully/partly disagree 2. Fully/partly agree
Programmatic voting	To what extent do you agree with the following statement? ‘I based my voting decision on the parties’ statements for the future.’	1. Fully/partly disagree 2. Fully/partly agree
Voting decision	When did you make the decision about how to vote in the 2010 elections (held on 19 September 2010)?	0. Before the spring 1. In the spring or summer 2. During the final month but not the final week 3. During the final week
Indicates specific candidate	Did you indicate a specific candidate in the 2010 municipal elections?	1. No 2. Yes
Emphasise person rather than party	What was more important for you when voting in the municipal elections – person or party?	1. Party was crucial 2. Party was more important 3. Of equal importance 4. Person was more important 5. Person was crucial
Party preferences	Which party did you vote for in the 2010 local elections?	0. Social Democratic Party 1. Moderate Party 2. Centre Party 3. Liberal Party 4. Christian Democrats 5. Left Party 6. Green Party 7. Sweden Democrats
Convinced supporter	Are you a convinced supporter of a particular party?	0. No 1. Fairly convinced 2. Strongly convinced