Introduction
According to partisan theory (Hibbs, Reference Hibbs1992), variations in policy choices and outputs originate from the party composition of the government studied (Schmidt, Reference Schmidt1996). As described by Lijphart (Reference Lijphart2012: 75–77), several dimensions of party-based conflicts in democratic societies will have such an influence. The most salient of these has been the socioeconomic that revolves around the classical left–right divide (Lipset and Rokkan, Reference Lipset and Rokkan1967). Although several other lines of conflict have gained strength, this dimension maintains particular relevance in relation to both one of the central policy instruments as well as the core issue regarding the financing of welfare societies, namely, tax policy (Steinmo, Reference Steinmo1989). As partisan theory has evolved through decades, moving away from the overly simplified assumption of left-wing parties opting for redistribution and extensive welfare provision while right-wing parties represent the opposite, concern has instead been directed to the importance of contextual and institutional settings as embedding policy output (Häusermann et al., Reference Häusermann, Picot and Geering2013). Even if it still can be assumed that party composition of the government will pose effects on the output of policy, including tax levels (e.g. Benoit and Laver, Reference Benoit and Laver2006), a more fine-grained analysis of party coalitions and the effect of government changes that cover the vital institutional settings of the decision-making arena would provide new insights.
To examine whether parties matter for public policy, we take a novel approach by linking changes in taxes with government changes in the local ruling. This departs from the dynamic that changes in policy and party composition shape better possibilities than exclusive analysis of the status quo. Although we expect to find that tax increases are connected to the formation of left-wing governments while tax decreases are linked to right-wing governments, this assumption is not made in a void. First, we advance the simple left–right distinction by adding bloc-crossing coalitions, thereby accounting for a left-centre-right trichotomy that has been proven to be particularly valid when it comes to economic policy (Schmidt, Reference Schmidt1996). Second, we assume that elements of demographic and economic character will influence such policy-making. We do not conduct such assessments in isolation, and to add a baseline scenario, we also provide an overview based on current situations of political rule and tax levels. Hence, we raise the following research questions to apply to our study:
-
Does the composition of governments affect levels of taxation?
-
Do changes in government rule affect levels of taxation?
The questions raised are applicable to Swedish local governments, and we find several reasons why this is a particularly feasible strategy. Local governments’ share of collected taxes stands in most developed countries for a minor proportion that, over the years, have additionally diminished. To that end, the Scandinavian countries and their unique model of local government represent an anomaly. It is expressed through their long tradition of autonomous municipalities with the right to tax their citizens (Loughlin et al., Reference Loughlin, Hendriks and Lidström2012). Among them, Sweden is the country in which local taxation of income stands out as particularly substantial (Ladner et al., Reference Ladner, Keuffer and Baldersheim2019). In fact, the Organisation for Economic Co-operation and Development (OECD, 2016) lists Sweden as the one nation among developed countries in which local taxes make up the largest share of total tax revenue, reaching almost 40%, and thereby represent the main tool for municipalities to calibrate revenue.
At a more fine-grained analytical level, it is visible that such average figures bear additional nuance within them. In fact, all of Sweden’s 290 municipalities pursue their own unique taxation policy, which is calculated according to their citizens’ income (kommunalskatt). Local discretion within this area has created considerable variation across municipalities. In 2021, the municipality with the lowest local tax had a 40% lower tax level than the municipality representing the other end of the spectrum. Since revenues from this source represent the dominating proportion for local governments in general, local autonomy makes it possible for elected officials and, in the end, municipal citizens to decide on their ambition of collecting revenues for the local government. This is paired with local governments being embedded within the same municipal systems, making it possible to keep potential institutional confounders constant.
Even though there are rare examples when scholars have targeted how the left–right divide influences the variation of local taxes (Pettersson-Lidbom, Reference Pettersson-Lidbom2008; Lakomaa and Korpi, Reference Lakomaa and Korpi2014), more thorough analyses of the composition of local governments have been overseen. This includes the relevance of local government changes. They could enable agency and function as a window of opportunity for altering tax levels. We draw on a dataset that contains official Swedish statistics and does so over a longer period of time, ultimately covering the government changes of seven elections from 1994 to 2018. We estimate the variation in tax policy through a pooled, cross-sectional time-series model.
Our findings provide affirmative answers to both research questions. Yes, the composition of local governments affects levels of local taxation in that mixed and right-wing coalitions are intimately related to lower tax levels than their left-wing alternatives. Decomposing this finding by addressing the strength of the main competing parties in Sweden shows similar effects. This interacts with how economic conditions either restrict or enable cross-case variations in tax levels. Furthermore, local government changes appear to function as an opportunity for modifications in tax levels. Transitions where right-wing governments lose their hold on power are utilized as a window for enforcing tax hikes, while shifts to left-wing governments yield inverted effects. These results contribute to international studies in which partisan-driven policy is uncertain (Imbeau et al., Reference Imbeau, Pétry and Lamari2001; Iversen and Soskice, Reference Iversen and Soskice2006), not the least at the local level (Ferreira and Gyourko, Reference Ferreira and Gyourko2009; Gerber and Hopkins, Reference Gerber and Hopkins2011; Potrafke, Reference Potrafke2018), while our closer assessment of ruling coalitions, which keep institutional conditions at a constant, provide new insights.
Theoretical points of departure and literature review
The literature that focuses on the standpoint that ‘parties do matter’ has long since constituted a core element for social scientists in their endeavour to examine government choices of policy instruments and output (for an overview, see Häusermann et al., Reference Häusermann, Picot and Geering2013). Building on the schematic idea of parties presenting clear-cut alternatives for voters that are thereafter to be transformed into policy by the election’s winner (Schmidt, Reference Schmidt1996), such assumptions have faced criticism from different angles. One of these involves how this theoretical modelling has downplayed the complexity of institutional conditions, for example, by assuming a majoritarian system in which the contenders are clearly divided into ruling and opposition parties. However, such distinct outcomes are uncommon in many political systems. Also, the clear-cut distinction of parties into the left and right has attracted disapproval. The famous remark by Lipset and Rokkan (Reference Lipset and Rokkan1967) regarding several societal cleavages is a common objective, and more recently, additional categorization of parties among lines of a cultural dimension (Kriesi et al., Reference Kriesi, Grande and Lachat2006) have added to that. Finally, the extent to which parties deliver what they have promised differs across political systems (Thomson et al., Reference Thomson, Royed and Naurin2017).
When applying this literature to our specific case of public policy – tax policy – it is evident that both institutional and partisan explanations for societies’ variation in tax levels have attracted the interest of scholars. Steinmo (Reference Steinmo1989) represents the first perspective and, in a comparative study of taxation policy in the United States, Sweden and Britain, argues that the institutional structure is crucial for understanding how tax systems in these countries have been shaped. By doing so, he notices important contextual variations across the cases that provide answers to their different trajectories. For example, while the US situation reveals a fragmented institutional landscape, the Swedish case represents the opposite, with its corporatist system comprising business, labour and government elites. Hence, for the Swedish case, Steinmo notices a few characteristics founded in the institutional context that, in general, obstruct radical changes and instead build upon ideas of continuity. However, if changes are posed, ruling actors need to take into account the interests of all stakeholders in the corporatist setting. The importance of such institutional settings, including restraining parties’ discretion, has also been disclosed by others (Beramendi and Rueda, Reference Beramendi and Rueda2007).
At the other end, proponents of the partisan line of arguments state that political issues are decisive for both the scope and the role of welfare systems (Esping-Andersen, Reference Esping-Andersen1990) and thereby their financing through the tax system (Cameron, Reference Cameron1978). This has its background in the class structure between capitalists and labour that has laid the basis for party systems between the left and right. As Downs (Reference Downs1957) argued, parties to the left will opt for a higher degree of state intervention in the economy, meaning that one could assume that voters and parties located to the left side of the political spectrum strive for a more extensive welfare sector and thereby accept higher tax levels than their contenders positioned to the right. Embedded in such viewpoints are also ambitions of redistributions of assets in which the tax system maintains the role of collecting resources for further transfer (Lipset, Reference Lipset1960; Korpi, Reference Korpi1983; Benoit and Laver, Reference Benoit and Laver2006).
Leaving these foundational theoretical assumptions, we turn to the literature covering local governments and taxation. Whether a partisan linkage exists appears to be highly bounded by the examined context (for an overview, see Potrafke, Reference Potrafke2018) and also by the character of the specific tax instrument that can follow from different ideological motives (Genschel and Schwarz, Reference Genschel and Schwarz2011). Several studies have found no convincing effect that local taxes – measured as total municipal taxation or decomposed into different sources of taxation – are generally higher across US cities with a Democratic mayor compared to those with a Republican holding office (Ferreira and Gyourko, Reference Ferreira and Gyourko2009; Gerber and Hopkins, Reference Gerber and Hopkins2011). Reasons for this can be drawn from both an institutional and partisan line of argument. The first refers to the constitutional constraint of local tax policy, which highly restricts local politicians’ room to manoeuvre. In addition to the US context, other scholars have found similar limitations for local governments to pursue taxation policy elsewhere (Gouvêa and Girardi, Reference Gouvêa and Girardi2021). Furthermore, the relevance of the ideological logic vanishes in some subnational contexts. Oliver et al. (Reference Oliver, Callen and Ha2012) apply the term managerial democracies to characterize US municipalities that often lack ideological divisions and instead revolve around certain local issues and the performance of local governments. However, later research has questioned the null findings by showing that democratic mayors indeed increase taxes, though moderately (de Benedictis-Kessner and Warshaw, Reference de Benedictis-Kessner and Warshaw2016). Their expansion in public spending is chiefly financed through increased debt.
Among European local government, the results appear to be somewhat more consistent. Migueis (Reference Migueis2012) discloses that, in Portugal, left-wing-controlled municipalities increase the likelihood of the adoption of business taxes. Similar results were reported by Fiva et al. (Reference Fiva, Folke and Sørensen2018) in a study covering Norwegian municipalities’ standpoints on property taxes. Even if such taxes only make up a minor part of municipal revenue, it was confirmed that a right-wing shift in the local council is associated with lowering taxes. Similar results were also observed in the subnational arena of the Netherlands (Allers et al., Reference Allers, de Haan and Sterks2001). Although not a direct analysis of tax variations but still of relevance, Toubeau and Vampa (Reference Toubeau and Vampa2021) disclose the existence of a partisan effect in public spending in Italian and Spanish regional governments. The occurrence of left-wing parties in governments mitigates how budget deficits otherwise lead to cutbacks.
Examining the situation in Sweden, Pettersson-Lidbom (Reference Pettersson-Lidbom2008) found that left-wing governments have higher municipal taxation of income and higher spending and maintain higher public employment compared with right-wing governments. These results imply that left-wing governments tax 2–3% more than right-wing governments. Lakomaa and Korpi (Reference Lakomaa and Korpi2014) updated Pettersson-Lidbom’s (Reference Pettersson-Lidbom2008) study with new data and used another operationalization of Swedish municipal coalitions and confirmed several of Pettersson-Lidbom’s findings. However, surprisingly, they did not reveal any difference between left- and right-leaning coalitions concerning their tax level and government employees per capita. Although not covered by these studies, parties could pose an effect on policy results even if they are not part of the ruling coalition. The reason for this is that a state of ‘quasi-parliamentarism’ prevails in Swedish local governments. This entails opposition parties with insight into the executive democratic processes (Bergman, Reference Bergman1999) that is often combined with a widespread consensus-oriented approach (Lidström, Reference Lidström and Pierre2016). For example, it cannot be dismissed that the two dominant parties in Sweden, the Social Democratic Party and the Moderate Party, as the main contenders since the 1979 elections, could affect the outcome of policy, even if they are not included in the local rule.
Even if researchers quarrel over the importance of institutional and partisan explanations on local tax policy, there are obvious motives for how the latter would operate, and the rationale behind why parties and politicians take different standpoints in the taxation issue has been addressed more closely. In a survey directed towards all local government politicians in Sweden, Gilljam et al. (Reference Gilljam, Karlsson and Sundell2010) showed that the left–right dimension in Swedish politics is the dominant dimension and is predominantly amplified at the local level. More particularly, questions regarding local politicians’ perceptions of local taxation were also raised. Substantial differences were found across parties that followed the left–right dimension. The claim that increasing local taxation is a better alternative than lowering municipal services was a position that politicians from left-leaning parties (Left Party, Social Democratic Party and the Green Party) supported, while their centre-right counterparts (Centre Party, Liberal Party, Christian Democratic Party and Moderate Party) expressed the opposite standpoint. Representatives of the Sweden Democrats took a balanced position on this issue. Furthermore, Nilsson (Reference Nilsson2015) convincingly showed that, for the individual local politician in Sweden, sticking to the party principle, irrespective of one’s own beliefs, is particularly evident in relation to tax policy. Adding the fact that fulfilment of election pledges is among the highest in Sweden (Thomson et al., Reference Thomson, Royed and Naurin2017) makes it possible to assume that promised pledges or alterations of tax policy will have better chances to yield an actual impact in this context than in others. The composition of parties and the common ideological position they represent in Sweden will therefore be an appropriate point of departure.
Formulating hypotheses
In this study, we expect parties to matter for tax policy and draw on the left–right divide found within party systems to explain variations. Hence, we expect that the class structure that embodies this conflict still prevails for both theoretical (Lipset, Reference Lipset1960; Korpi, Reference Korpi1983) and empirical reasons found within the studied context (Gilljam et al., Reference Gilljam, Karlsson and Sundell2010), with caution taken regarding the specific arena in which local governments in Sweden operate. To that end, we have formulated three hypotheses to apply to the set of Swedish municipalities and what influences their taxation of income:
H1: Right-wing coalitions maintain lower local taxation than left-wing coalitions do.
H2: Municipalities where the Moderate Party is strong will maintain lower local taxation than municipalities where the Social Democratic Party is strong.
H3: Tax cuts occur when right-wing coalitions reach office, and tax hikes occur when left-wing coalitions reach office.
However, these hypotheses have not been reached with the naïve notion that institutional and contextual conditions can be disregarded (Häusermann et al., Reference Häusermann, Picot and Geering2013). Quite the contrary, we assess such aspects through different strategies. First, since our empirical arena is Swedish local governments with their unusual discretion over tax policy, we state that the institutional constraints and influencing factors that are frequent in most other countries will be less relevant, as analysis occurs within a specific setting. To complement this argument, we still assess remaining potential predictors by accounting for demographic and economic conditions, and we assume that, in addition to a direct effect, they could also influence local taxes in conjunction with partisan issues (Allers and Greef, Reference Allers and de Greef2018). Finally, to consider the multiparty setting of Sweden, we assume that the existence and entrance of ruling coalitions among the constellation through an accounting of a left-centre-right trichotomy will provide a more valid scope.
Swedish municipalities, local governments, and taxation
Sweden consists of 290 municipalities. Although they vary considerably in size, they all fall under the same regulations. Every fourth year, the municipal council (fullmäktige) is elected in proportional elections, which follow the same party-based representative government that characterizes the national arena. As Swedish politics have followed the classical division along the left–right axis, the very same dimension has been applied to the subnational arena. At the same time, a more scattered party system has been increasingly evident during the studied period, which has resulted in newer parties that have gained national success also achieving local representation, while a diminished strength can be witnessed among the traditionally major parties, particularly the Social Democratic Party (Lidström, Reference Lidström and Pierre2016). Local branches of the national parties are still the major competitors, but unique local parties exist as well.Footnote 1 There are several examples of local governments, ranging from the increasingly rarer one-party majority to local rule made up of coalitions by left-wing or right-wing parties. Mixed coalitions, stretching across the traditional left-right divide, are also relatively frequent. The executive committee (kommunstyrelse) coordinates and leads the political work of the municipality. It is chaired by a politician who represents the ruling party that governs the municipality solely or, normally, through a coalition. In the committee, parties are proportionally represented in their relation to the distribution of the municipal council (Lidström, Reference Lidström and Pierre2016). After the election, the municipal council takes office on the 15th of October, while the new executive committee and the new committees usually take office on the 1st of January.
Each year, the council sets the municipality’s tax rate and budget for the upcoming year. At large are municipalities completely free to outline their own taxation policy.Footnote 2 The legislation only imposes that municipalities should see that they have good financial management and that, if a deficit emerges during a certain year, it should be handled during the approaching three-year period (SFS, 2017). Taxation is the main revenue for the municipal sector in general. Municipal taxes are calculated on citizens’ income, meaning that the actual revenue differs in municipalities due to both the tax and income levels in municipalities. Hence, an equal tax rise in a municipality where citizens hold higher income levels will increase revenues more than in a municipality in which income levels are lower. Swedish municipalities have no authority to tax their citizens beyond the income tax. What can be referred to as administrative tax changes occurred during the examined time period. This comprises of welfare obligations that, due to administrative reforms, have shifted either from the regional authority to a municipal commitment or vice versa. The implementation of such reforms reflects a zero-sum game for taxpayers who will not have higher or lower taxes and that only involves a transfer of the tax-raising organizations.Footnote 3
In addition to local taxes, there are other sources of revenue for Swedish municipalities. These mainly comprise of municipal fees and government grants; some of the latter are general, while others are dedicated to certain commitments. The government directs considerable proportions of its support to municipalities through the municipal system of redistribution (kommunala utjämningssystemet). The redistribution of assets builds upon a Robin Hood-based principle and contains, in addition to models of government funding, some elements of the redistribution of wealth generated in local governments. This system consists of several elements in which a redistribution based on citizen income and municipal costs are the most dominant. Assets are distributed from municipalities with citizens having the highest levels of income to the other municipalities. To compensate for the varying costs of local governments’ welfare provisions, redistributions are also provided based on nine different areas of local commitments. However, the system is not designed to redistribute differences that result from local priorities in policy or the efficiency of the public administration. As the system will influence local revenue, they must be taken into account when assessing local taxes.
Data and estimation strategy
We estimated effects through a pooled, cross-sectional time-series model using ordinary least square regressions with panel-corrected standard errors.Footnote 4 The data that was exploited represents a panel of all Swedish municipalities from 1994 to 2018 and covers seven elections, making the term of office our unit of analysis. To ensure the robustness of our estimations, we added several robustness checks, which are presented in the empirical section and discussed in more detail in the online Appendix. This includes accounting for the lagged dependent variables (LDVs) to remedy problems with autocorrelation (Beck and Katz, Reference Beck and Katz1995). We also estimated the models with other specifications to ensure that our main estimator was well specified and tested for additional control variables. To expand our understanding of the relationships among particularly partisan and economic conditions, we estimated interaction effects between partisan variables and demographic and economic controls.
The data on actual coalitions (1994–2018) were collected from the Swedish Association of Local Authorities and Regions (SALAR). Additional data were collected from Statistics Sweden (SCB) and from the Kolada database. In general, the data coverage was good, but a few cases have been excluded.Footnote 5
In the online Appendix, we include descriptive statistics and data sources for all variables.
Dependent variables
To be able to address the research questions, we included two dependent variables: the municipal tax rate (T) and the change in municipal tax rate (ΔT). How they were measured is displayed in Table 1, where y denotes the year and i the municipality. As mentioned above, the length of the term of office in Swedish municipalities is four years, and we measured the municipal tax for the last year of each term of office.Footnote 6 For example, for the election held in 1994, the term of office was 1995–1998, and consequently, for that term of office, we measured the municipal tax level in 1998. The municipal tax level change between elections was our second dependent variable, and it was measured as the change in municipal tax level between elections. Referring to the same example, that was measured as the difference between 1998 and 1995.
Description of measuring points for dependent variables
We estimated the effects in both directions jointly (tax decreases and tax increases). However, since the tax level change between elections may be positive or negative (‘upturns’ or ‘downturns’), we also examined the effects in both directions to single out more detailed effects. We used a similar research strategy as Teorell (Reference Teorell2010), and to estimate the effect of tax upturns, we set all downturns to zero, and to estimate the effect of downturns, we set all upturns to zero.
Independent variables
In the first models (1–4), we used the ruling government in the municipalities as the independent variable to test H1 using T. In the period covered in this study, the left–right dimension was very prominent in Swedish politics, and traditionally, the main parties have been divided into two blocs: the left bloc and the right bloc. The Social Democrats, the Green Party and the Left Party are considered left-wing parties, while the Centre Party, the Liberals, the Christian Democrats and the Moderate Party are treated as right-wing parties. The far-right party, the Sweden Democrats, is considered bloc independent and has not been included in local government rule through to 2018. However, coalitions also exist across the left–right dimension (coalitions involving at least one left-wing party and one right-wing party). This type of mixed coalition exists in various constellations. As an example, the Social Democrats can govern together with the Centre Party and give rise to a more centre orientation. Also, local parties are part of the ruling coalition in several municipalities. When a local party governs together with left-wing parties, that government is coded as a left-wing government. Similarly, when a local party governs together with right-wing parties, that government is coded as a right-wing government. Governments are coded as left-wing coalitions, right-wing coalitions and mixed coalitions, where parties from both the right and left blocs govern together (a coalition across the left–right dimension).Footnote 7 Ruling governments were operationalized with three dummy variables, and left-wing governments were used as the reference category.Footnote 8
To estimate the effect of H2, we ran additional models (5–7) and included the strength of the largest right-wing party and the largest left-wing party (on the national level) as our independent variables and T as the dependent variable. The largest right-wing party is the Moderate Party, and the largest left-wing party is the Social Democratic Party. Each of the two party variables were operationalized as the party’s share of the total seats in the municipal council (in percentage).
In the final models, we examined the effect of H3 on Models 8–10, with government changes as the independent variable and ΔT as the dependent variable. With government changes, we refer to alterations in local rule that follow from the trichotomy previously displayed. Such were coded with seven dummy variables to capture all the alternatives that arise from the structure of three types of ruling coalitions.Footnote 9 The dummy for no government change was used as the reference category.
Control variables
Since decisions on local tax rates are not made in a vacuum, a battery of control variables needs to be included to make it possible to single out the potential effects of our assumptions. The control variables reflect economic and demographic conditions and are similar to those normally applied in the field.
Of economic relevance are circumstances that affect the municipal budget and its underlying components. As a measurement of the current internal economic conditions of the municipality, the financial solvency ratio was included. This variable indicates to what extent municipal assets are financed by the municipality’s own capital and is displayed in per cent. Since a higher proportion indicates a better economic situation for the municipality, we expected a negative relation with the dependent variables. This variable could function as a mediating condition that either restricts the local rule or expands their possibility to pursue their own ambitions. The average taxable income among citizens will be a crucial piece of information for local decision-makers when settling tax rates and their modifications. All other things being equal, higher average incomes in the municipality will result in tax hikes gaining more substantial revenue for the municipality (cf. Pettersson-Lidbom, Reference Pettersson-Lidbom2008). Hence, we expected a negative effect of this variable. The variable was measured as the average of taxable income (thousands of Swedish crowns) among all citizens in a municipality. Finally, as the municipal system of redistribution is a crucial source for municipal revenue, it needed to be included in the models (Blöchliger et al., Reference Blöchliger, Merk and Charbit2007). We controlled for how this system affects each municipality by measuring the net revenues or costs per citizen in thousands of Swedish crowns. As a higher and positive value of this variable has implicit greater dependency on the system of redistribution and, thereby, lower effects on tax rises, we assumed a negative relation with our dependent variables. Control of this character has not been included in the few Swedish studies but has shown relevance in other contexts (Allers et al., Reference Allers, de Haan and Sterks2001).Footnote 10 To ensure well-specified models, we also exploited the municipal tax rate as a control when examining H3 and assumed that already high tax levels would lower the inclination for further rises. Correspondingly, we included tax changes (between the local level and the region level) during the present term of office as a control when examining H1 and H2.
We included three demographic control variables. The first was a dummy variable for rural municipalities, where rural municipalities were coded 1 and urban municipalities were coded 0. Regarding the operationalization of the variable, we followed a suggestion made by SALAR and defined a rural municipality as a municipality that has less than seven inhabitants per square kilometre and less than 20,000 inhabitants in total (Glesbygdsverket, 2008). Consequently, both criteria must be fulfilled if a municipality is to be classified as a rural municipality. We expected that rural municipalities would have higher tax rates and higher tax increases than urban municipalities due to their more challenging demographic situation, and therefore, we expected that the coefficient would be positive.Footnote 11 The second demographic variable was population size, which refers to the total population in each municipality, and it was measured in thousands of people. We expected that municipalities with larger populations would have lower municipal taxes and lower tax increases due to the better preconditions that this implies (e.g. Gouvêa and Girardi, Reference Gouvêa and Girardi2021). Accordingly, we expected that the coefficient for population size would be negative. The population size variable has a highly skewed distribution, and therefore, we transformed it to a logarithmic scale. Finally, we included a control for the age structure of the population, making a third demographic control the average age in the municipalities. We expected that municipalities with a higher average age would have higher municipal taxes and higher tax increases due to additional welfare commitments. Thus, we expected a positive coefficient.
Caveats
One main caveat of our aim is that we have not controlled for explanations in tax variations that are due to different ambitions in local government policy. One could obviously assume that municipalities that undertake additional tax revenues also would have more far-reaching ambitions in the provision of welfare services. Similarly, it could also be assumed that municipalities with lower tax revenues deliver such services more efficiently than other municipalities. The municipal redistribution system has the ambition to correct for differences in municipal preconditions but not in these two potential effects. However, there are several reasons why we did not take differences in aspirations and efficiency into account. First, municipalities’ commitments range over several different sectors of welfare services. Estimating quality and efficiency in them is not easy. Second, attempts to examine differences of that character have noted inconsistent results (Andersson, Reference Andersson2011; Kommuninvest, 2020). Put differently, no solid evidence has been found to explain variations in local tax levels through either different levels of efficiency or quality in the deliverance of welfare services. Third, if such an effect still exists, we assume that we can partly grasp it through the focus on political ideology, assuming that parties or ruling coalitions constitute a proxy for ambitions of local welfare provisions. This would thereby lower the risk of underspecified models. Another caveat with the design is that we could not control for shifts in local government rules between elections, as such data were not available for the examined period. Our own impression is that such shifts in government have become somewhat more frequent in later years but were previously quite uncommon.
Analysis and findings
We present our findings in turn, starting with descriptions of the two dependent variables, and thereafter continuing with the multivariate analyses.
Descriptive statistics
The box-and-whisker plots in Figure 1 illustrate the distribution of the municipal tax level in the Swedish municipalities between 1998 and 2018. We remind the readers that the term of office is four years and that we have measured the municipal tax for the last year of each term of office. As revealed, Swedish municipalities showed an upward trend in local tax levels, which, on average, increased around 0.73 percentage units, while the median tax level has increased by 0.79 percentage units. In total, the municipal tax rate ranges between 14.90 and 23.90%, which indicates that there were considerable variations between municipalities. These upward tendencies counteract assumptions on how globalization may put pressure on societies’ labour taxes (Swank and Steinmo, Reference Swank and Steinmo2002). As Figure 1 also shows, several outliers had extraordinarily low municipal tax levels. Most of these municipalities are geographically found in the metropolitan Stockholm area and characterized by high average incomes.
Municipal tax levels in Swedish municipalities for 1998–2018.
The box-and-whisker plots in Figure 2 illustrate the municipal tax level changes in Swedish municipalities between different terms of office. We remind the readers that the comparisons of tax changes illustrated in Figure 2 are between two governments in a row. For example, the first comparison (1994–1998) is between the last year for the government in rule from 1992 to 1994 (the last period when the term of office was three years) and the government in rule from 1995 to 1998. Accordingly, this shows the difference in the local tax levels between 1994 (the last year in office for the ruling government of 1992–1994) and 1998 (the last year in office for the ruling government of 1995–1998). As mentioned previously, during the periods covered in this study, the municipalities took over activities from the regional level, and the municipalities raised the tax by the same level as the regions reduced the tax. We have taken these tax changes into account in Figure 2.
Municipal tax changes in Swedish municipalities.
The municipal tax changes varied between a decrease in the tax level of 1.32 units and an increase in the tax level of 2.55 units. As shown in Figure 2, tax increases (504 cases) were much more common than tax decreases (123 cases) among Swedish municipalities. The data also show that a considerable proportion of municipalities did not change their tax levels between terms of office.
Assessing local governments and party effects on municipal tax levels
In the first models, we tested H1 using T as the dependent variable and examined if there were any differences in the tax level depending on the ruling coalitions. The results are displayed in Table 2. In the Appendix, the standardized coefficients from all regression models are visualized.
Multivariate regressions on municipal tax level models 1–4
Note: Entries are for all forthcoming tables; the unstandardized regression coefficients with panel-corrected standard errors are in parentheses. ***significant at the .001 level, **significant at the .01 level, *significant at the .05 level. This information is also applicable also for Table 3 and Table 4.
In Table 2, the dummy variables for the ruling governments in the municipalities were included, and left-wing governments were used as the reference category. As shown in Model 1, the coefficients for right-wing governments and mixed governments were negative and statistically significant. The results indicate that municipalities that are governed by right-wing and mixed governments maintain lower levels of municipal taxes in comparison with municipalities that were governed by left-wing governments. Concerning the effect size, the results suggest that, on average, municipalities that were governed by right-wing governments had taxes that were 0.93 percentage units lower than municipalities that were ruled by left-wing governments. Correspondingly, local governments that were governed by mixed governments had taxes that were 0.26 percentage units lower than their left-wing counterparts.
In Model 2, the economic predictors were added to the dummy variables for the ruling governments. The results show that the dummies for the type of governments that were included in Model 1 continue to be significant, with expected signs. The size of the coefficient for the right-wing governments decreased (to −.594), while the coefficient for mixed governments increased (to −.328). The control of financial solvency was negative and statistically significant, while the redistribution variable was positive and statistically significant. The controls average taxable income and tax change were both negative and statistically insignificant.
In Model 3, the three demographic variables are included. The results for the dummies referring to the ruling governments and the economic controls are similar to those in Model 2. The dummy variable for rural municipalities was positive and statistically significant, and these results suggest that the municipal tax level was 0.20 percentage units higher in rural municipalities than in urban ones. The coefficient for population size was negative and statistically significant, and these results predict that population size has a negative effect on the tax level. The control average age was positive, as expected, and was statistically significant.
In Model 4 we included an interaction between the dummies for the ruling governments and financial solvency. A graph (based on Model 4) that illustrates the interaction effect is shown to the left in Figure 3. Our motivation for this was that a problematic internal economy of municipalities paired with local rule could provide more details on the function of the partisan effect. The findings indicate that right-wing governments’ effects on lower tax rates were much more distinct in municipalities with higher levels of financial solvency. In municipalities with low levels of financial solvency, the tax rate level differed only marginally depending on who governs. This shows both that under financial constraints, municipalities, irrespective of their partisanship, are restricted to maintain higher tax levels while right-wing government exploits a better local economy to sustain significantly lower taxes than their left-wing counterparts.
Interaction effect between ruling government and financial solvency (left) and interaction effect between the Moderate Party’s seat share and financial solvency (right).
The main findings from the first three models summarize that municipalities where right-wing parties govern had lower municipal tax levels compared with municipalities that were governed by left-wing parties. The results also show that municipalities that were governed by a mix of right-wing parties and left-wing parties maintained lower levels of municipal taxes compared with left-wing governed municipalities. Accordingly, the results are in line with our expectations, as expressed in H1, and add to a similar and previous pattern disclosed from the Swedish context (Pettersson-Lidbom, Reference Pettersson-Lidbom2008). However, as Model 4 showed, the financial context is important, as it appears to add knowledge regarding under which conditions tax levels vary.
Next, we examine whether the strength in terms of the number of seats in municipal councils of the largest right-wing party and the largest left-wing party on the national level has any effect on the municipal tax level, as reflected by H2. Table 3 presents the results.
Multivariate regressions on tax level models 5–7
In Model 5, we included the percentage of the total seats of the Moderate Party and the Social Democratic Party in the councils as the independent variables, together with the economic controls. The regression coefficient for the seat share of the Moderate Party had the predicted sign and was statistically significant, which suggests that a larger share of seats for the Moderate Party in the municipal councils had a negative effect on the municipal tax level. The coefficient for the Social Democratic Party was statistically significant and positive. Thus, the results show that an increase in the Social Democratic Party’s share of seats was associated with a tax hike. The economic controls financial solvency and the system of redistribution were statistically significant, yielding a negative effect for the first and a positive for the latter. The other control variables were statistically insignificant.
In Model 6, we added the three demographic variables. The results for our independent variables and the economic controls were similar to the previous one. Regarding the demographic controls, rural municipalities and average age were statistically significant, both with positive coefficients. The control population size was statistically insignificant. In general, a 10% increase in support for the Moderate Party was associated with a tax cut of about 0.5 percentage units, and a similar increase in support for the Social Democratic Party was associated with a tax increase of about 0.1 percentage units. In the Appendix, we illustrate the marginal effects disclosing the influence of the Moderate Party, and the Social Democratic Party (based on Model 6).
Finally, in Model 7 we included an interaction between the Moderate Party’s seat share and financial solvency. A graph (based on Model 7) illustrating the interaction effect is shown to the right in Figure 3. As shown, the Moderate Party´s effect on lower tax rates was much more substantial in municipalities with high level of financial solvency compared with municipalities with worse internal economies.
To sum up this section, in general, the findings are in line with our expectations of H2, showing that a larger seat share for the Moderate Party in the council is related to lower levels of municipal tax levels, while a larger share for the Social Democratic Party indicate the opposite pattern. It should also be noticed that the effect of the Moderate Party was substantially larger than that of its main contender. The analysis of interaction effects show that the combination of a stronger Moderate Party and a better municipal economy can be associated with lower tax levels.
The party and local government effect on municipal tax changes
In our final empirical section, we examine whether government changes affect the municipal tax rate, and Table 4 reports the results from Models 8–10. Hence, in these estimations, the accuracy of H3 is examined through regressing dummies on government change on ΔT. Here, we display both complete models (8) and specify them further by conducting separate estimations of upturns (9) and downturns (10).
Multivariate regressions on tax level changes models 8–10
In Model 8, we included the dummy variables for government changes while using no change of government as the baseline. Three of the independent variables were statistically significant. First, the results showed that a transition from a left-wing to a right-wing government decreased the tax level by around 0.14 percentage units compared with no change of government. Second, the opposite shift, from right-wing to left-wing government, increased the tax level by around 0.14 percentage units compared with no transition of government. Third, the results show that a change from a mixed government to a left-wing government increases the tax rate by 0.06 percentage units. Although the other independent variables were in the expected direction, they were statistically insignificant. These findings add a clearer pattern to what prior compilations have described as a scattered result (Kommuninvest, 2020). Considering the economic control variables, the system of redistribution and the tax rate yielded significance. The first can be interpreted as higher revenue from the system being correlated with lower levels of tax rises while the latter clarifies that those municipalities with already higher tax levels increase taxes to a greater extent than other municipalities. Among the demographic controls only population size showed relevance, with a negative effect. In an analogy with the tests of the first two hypotheses, we also estimated a model that included interactions between the government change dummy variables and financial solvency. However, contrary to previous convincing results, no coherent findings were reported.Footnote 12
From this, we turn to the more specified estimations in Model 9, which covered cases of tax hikes. Among the independent variables, the same three types of government changes reported similar statistical effects as in Model 8. Two of the controls were statistically significant; the system of redistribution and population size, both with a negative effect.
Finally, in Model 10, we examined tax decreases separately. To make it easier to interpret the results from Model 10, we reversed the scale for downturns so that higher positive numbers indicated higher decreases in taxes. The results should be taken with caution, as the number of cases were limited. One form of government change – a transition from a left-wing government to a right-wing government – yielded a statistical effect, and the results show that type of government change resulted in a tax decrease. Among the controls, an already higher tax was associated with a lower tax decrease, while the population size reported a positive effect on tax decreases.
Robustness controls
To ensure the trustworthiness of the analyses, a series of robustness controls were performed. Initially, we found it relevant to test an alternative operationalization of the first dependent variable (T). According to the literature on political budget cycles, tax cuts may be used to win votes prior to elections (see e.g. Alesina and Paradisi, Reference Alesina and Paradisi2017 for an overview). This implies that tax rates may change shortly before elections for opportunistic reasons and may not necessarily be related to partisan issues. As a robustness test, we applied the average tax rate for the term as an alternative dependent variable in Models 1–7. These models produced similar results as those presented in the original models. The models with the alternative operationalization are included in the Appendix. Furthermore to examine whether multilevel linkages affect policy-making, as recently suggested by Garritzmann et al. (Reference Garritzmann, Röth and Kleider2021), we included a dummy variable for the national government in office (left-wing or right-wing government). However, any effect of that character could not be witnessed in the study.
In estimations of our second dependent variable (ΔT), we have applied a different coding in Models 8–10. For example, we created three dummy variables (one for change to right governments, one for no change of government and, one for changes to left governments) and included them in the models instead of the original dummy variables. The results for the alternative dummies were similar to that of the original dummies and show that moving to a right-wing government leads to tax rate decreases and moving to a left-wing government leads to tax rate increases. The Models (8–10) of this alternative coding of the government change dummies are displayed in the Appendix.
Additional control variables have also been accounted for in the models, including the unemployment rate (data available only for the last three terms), close races (i.e. the margin of victory between the two largest parties) and accounting for the special circumstances of 1997–1999. Inclusion of these controls did not affect the results for the independent variables that we present.
All models were re-estimated with fixed and random effects, and our independent variables have, in general, shown similar effects. The same goes after adding an LDV to the original estimations (cf. Beck and Katz, Reference Beck and Katz1995). We report further details about the robustness controls in the Appendix, and richer presentations of interactions effects are also displayed there.
Conclusion
In this study, we have raised one of the core issues of political science, namely, whether parties matter in policy output. While our answer is convincingly affirmative, it draws on the empirical setting of Swedish municipalities in terms of not only the substantial financial control considering both the high level of financial self-reliance and the high level of fiscal autonomy (Ladner et al., Reference Ladner, Keuffer and Baldersheim2019) but also the great variations in local taxation.
We examine the partisan effect with the theoretical help of three hypotheses tested through two dependent variables. Building on the traditional divide between left–right positions in party systems that originate from conflicts between labour and capital (Downs, Reference Downs1957), we find that right-wing governments tax their citizens’ income at significantly lower rates than left-wing governments. This result nurtures the finding that parties located on the left side of the political spectrum strive for a more extensive welfare sector and thereby accept higher tax levels than their contenders positioned to the right (Lipset, Reference Lipset1960; Korpi, Reference Korpi1983; Benoit and Laver, Reference Benoit and Laver2006). Furthermore, we find a similar effect, though weaker in magnitude, when zooming closer into the support of the two dominant parties during the examined time period and thereby assessing the second hypothesis. A stronger position for the Moderate Party is associated with a negative effect on local taxation. Whether an opposite linkage is found in terms of the size for the Social Democratic Party, though, is not fully evident. The effect for this party is smaller, and since robustness checks report ambiguous results on the mere significance of the predictor, this linkage cannot be completely established. In general, the second hypothesis will obviously overlap with the first but could also imply that not just the composition of the local rule could be in play but also situations in which dominant parties could exert influence, irrespective of their inclusion in the local rule.
The final hypothesis has a different approach through the regressing of government changes on changes of taxes. Prior studies within this area have not reported any unified patterns (Ferreira and Gyourko, Reference Ferreira and Gyourko2009; de Benedictis-Kessner and Warshaw, Reference de Benedictis-Kessner and Warshaw2016). Nonetheless, our findings imply effects, most notably that municipalities turning from right-wing to left-wing rule are associated with tax hikes, while this effect is inverted with the opposite shift in local government. To further specify the direction of the effect, separate estimations of up- and downturns of tax changes were conducted. While tax increases reported similar results as the complete model, tax decreases were only associated with when left-wing rule altered to right-wing governments. These findings report that certain government changes at the local level are associated with changes in local tax policy. Although these findings contradict prior international experiences of a local arena being free from ideological battles (Oliver et al., Reference Oliver, Callen and Ha2012), it is more in line with results identified in European studies (Allers et al., Reference Allers, de Haan and Sterks2001; Pettersson-Lidbom, Reference Pettersson-Lidbom2008; Fiva et al., Reference Fiva, Folke and Sørensen2018). It verifies that the varying ideological opinions that are highly evident among local politicians in Sweden (Gilljam et al., Reference Gilljam, Karlsson and Sundell2010) also lead to policy changes when possibilities are given (Imbeau et al., Reference Imbeau, Pétry and Lamari2001).
Our conclusion that ‘parties matter’ for tax policy is not a finding reached in a void. We have taken careful measures to nuance the naïve stance that has disregarded contextual and institutional settings (Häusermann et al., Reference Häusermann, Picot and Geering2013). A first example of this includes accounting for the left-centre-right trichotomy that has been proven to be particularly valid when it comes to economic policy (Schmidt, Reference Schmidt1996). That bloc-crossing coalitions take an intermediate position has been empirically determined and verifies our expectations. Another example is how the internal municipal economy can function as a crucial condition for tax levels. Even if there are some ambiguities concerning this effect as reported in the Appendix, the general outcome of it is that the partisan effect is compressed if the local economy is challenging while amplified if it is affluent. That such contextual circumstances are decisive adds to Steinmo’s (Reference Steinmo1989) general remark that these policy choices are reached in a delicate setting of conditions and that, particularly in Sweden, the extensive welfare sector has created a certain political-economic logic.
To conclude, in this study, we have reported that parties appear to affect local taxation among Swedish municipalities based on their ideological positions. How voters cast their votes in local elections will ultimately influence how they are taxed. Bearing in mind the remarkably good setting for addressing variations in local tax policy that Sweden offers, we would still encourage future studies to examine whether similar results could be found elsewhere and to look closer at the mechanisms specifying the noticed pattern by further scrutinizing the circumstances that influence tax levels and how tax changes occur.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1755773922000388
Acknowledgement
The authors would like to thank the three anonymous reviewers and the editors for their helpful comments on earlier drafts, as well as colleagues at Mid Sweden University who provided valuable feedback. In addition, we are grateful for suggestions made by Emelie Värja at Kommuninvest.
John Högström wishes to acknowledge support from Riksbankens Jubileumsfond Grant SAB21-0051 for him to finish the work with this study.
Conflict of interest
The corresponding author states that there is no conflict of interest.
Data availability statement
The dataset generated during the current study are available from the corresponding author on reasonable request.
Open access
