Explaining Party Positions on Decentralization

Abstract

Debates about decentralization raise cultural questions of identity and economic questions of redistribution and efficiency. Therefore the preferences of statewide parties regarding decentralization are related to their positions on the economic and cultural ideological dimensions. A statistical analysis using data from thirty-one countries confirms this: parties on the economic right are more supportive of decentralization than parties on the economic left, while culturally liberal parties favour decentralization more than culturally conservative parties. However, country context – specifically the degree of regional self-rule, the extent of regional economic disparity and the ideology of regionalist parties – determines whether and how decentralization is linked to the two dimensions. These findings have implications for our understanding of the politics of decentralization by showing how ideology, rooted in a specific country context, shapes the ‘mindset’ of agents responsible for determining the territorial distribution of power.

In recent decades, established democracies have transferred political power away from the national level. While much attention has been paid to supranational (particularly European) integration, in many states the decentralization of governance is an equally important phenomenon: over the past fifty years, the migration of authority to the regional level has been an unmistakable trend across developed countries that has led to important processes of constitutional reform. ¹ As Marks et al. note, ‘Not every country has become regionalized, but where we see reform over time, it is in the direction of greater, not less, regional authority.’ ²

Structural approaches to decentralization that privilege social, economic and historical factors can explain pressures to establish ‘congruence’ between the state and society. But given their focus on deeply rooted factors, these approaches overlook the role of political agency and thus cannot account for the timing, rhythm and scope of territorial reforms that transfer authority to regions. ³ Structural perspectives are valuable insofar as they set the stage for examining the role of partisan actors who incarnate social and territorial cleavages and take the decision to decentralize power and whose preferences are therefore central to understanding the dynamics of decentralization. Decentralization has in fact become an important issue of political debate and electoral competition: while some statewide political parties have endorsed the transfer of authority to regional governments, others have opposed it. What lies behind parties’ positions on this issue?

This question has so far received scant attention in the literature, in contrast to party positioning on European integration, which has been studied extensively. ⁴ So far, a cross-national analysis exists only for the salience of decentralization in Britain, Italy and France ⁵ ; another study ⁶ considers Spain's Socialist and Conservative parties’ positions – and the importance they assign to the regionalist issue – during regional and state elections.

Understanding the positions of statewide parties on decentralization is important, since decentralization touches the core of politics: the power of the state to make and execute laws. However the study of multilevel governance has too often been a ‘party-free’ area of inquiry that has mainly studied the vertical and horizontal interactions of actors in the policy process. ⁷ Similarly, neoclassical approaches to the distribution of authority have treated decentralization as a depoliticized question, which is reducible to the establishment of an optimal number of territorial jurisdictions. ⁸ Yet the evidence from recent processes of constitutional reforms in different countries such as Belgium, Germany, Italy, Spain or the United Kingdom shows that the territorial distribution of authority is deeply contested by political parties.

In this article, we focus on statewide parties. Since they usually form the core of national governments, they strongly influence when and how changes to the territorial distribution of authority are implemented. ⁹ This means that the positions of statewide parties on jurisdictional questions are central to understanding political outcomes on decentralization. Thus as Marks and Hooghe argue, it is important to ‘place politics – contestation about the good society – at the centre of a theory of authority allocation’. ¹⁰

We ask what explains the positions of statewide parties on decentralization. Our central argument is that views on this issue are related to preferences on the two core ideological dimensions of party competition: the economy and culture. ¹¹ This is because decentralization is a jurisdictional question related to the nature of governance that, following Hooghe and Marks's post-functionalist theory of integration, raises questions of efficiency and redistribution on the one hand, and questions of identity on the other. ¹²

Specifically, we argue that parties on the economic right should favour decentralization because it can improve the efficient production of public goods, while parties on the economic left should oppose decentralization because it hinders efforts to redistribute wealth. In contrast, culturally liberal parties should support decentralization because it recognizes diversity and local decision-making, while culturally conservative parties should be against decentralization because it erodes national unity and territorial integrity. For some parties, the two dimensions are consistent in determining positions on decentralization; for others, they may lead to internal ideological tensions. To understand the ideological foundations of a party's position on decentralization, we thus need to know its views on both key dimensions of political conflict. In this, party positions on decentralization are similar to views on European integration. ¹³

We test this claim using data for thirty-one countries from the Benoit and Laver expert survey supplemented with party- and country-level data. ¹⁴ We find strong support for our argument that, in general, decentralization taps into the logics of efficiency and redistribution on the one hand and identity on the other: overall, positions on the issue are clearly related to both the economic and cultural dimensions.

However we also argue that this general association should differ across countries, because each country has different institutional, structural or strategic characteristics that influence whether decentralization is seen primarily through an economic or a cultural lens. While questions of territorial authority always have the potential to address the logics of efficiency, redistribution and identity, how these logics shape actual party positions should therefore depend on country context. We suggest that decentralization positions are more influenced by economic ideology when regional self-rule and economic disparity are high, and our analyses confirm this. We also find evidence that the nature of ideological competition from regionalist parties can affect whether (and which) statewide parties support or oppose decentralization. Finally, the relationship of decentralization to the cultural dimension appears to depend little on country context.

This article is structured as follows. We start by elaborating our theoretical framework and presenting our expectations concerning party positions on decentralization and the effects of ideological and contextual factors. We then present the data and the statistical model before describing our results. We conclude by drawing implications for future research on the determinants of party positions on decentralization and of territorial reform processes.

Decentralization and Ideological Dimensions

Our argument that the positions taken by statewide parties on the issue of decentralization are related to the core ideological dimensions of competition builds on the post-functionalist theory of integration developed by Hooghe and Marks, which transcends the notion that the allocation of authority is an efficiency-oriented outcome driven by functional pressures and posits instead that it is a deeply political choice conditioned by domestic (non-economic) conflicts. ¹⁵ They note that ‘[g]overnance has two entirely different purposes’: the first is to supply public goods, for example by increasing efficiency and redistributing resources; the second is to express a sense of identity with a specific territorial political community.

This distinction highlights the notion that decentralization is a multidimensional concept. ¹⁶ The first goal of governance relates to material values; it is expressed through the economic dimension of party competition and implies a distributional logic of contestation. There are three institutional components of territorial autonomy that touch on questions of efficiency and redistribution: (1) legislative autonomy on an exclusive, concurrent or shared basis; (2) control over administrative resources such as personnel and agencies and (3) control over financial resources, either in the form of central transfers or ‘own’ revenues. ¹⁷

The second goal of governance relates to ‘pre-material’ values; it is expressed through the cultural dimension and implies an identity-based logic of contestation. ¹⁸ In practice, this cultural component of decentralization may be expressed through normative debates surrounding, for example, the definition of the political community or the recognition of national pluralism through symbolic gestures or asymmetric territorial autonomy. Decentralization is therefore a multifaceted process that can address questions of efficiency and redistribution linked to economic preferences as well as questions of identity linked to cultural preferences. We elaborate on this below.

Decentralization and the Economic Dimension

The two economic questions related to decentralization are thus: (1) how does it affect the efficiency of governance and (2) how does it affect the distribution of wealth and welfare? Those on the economic right often welcome decentralized decision making, as it can promote a more efficient form of government. In one view, policy making should be decentralized as far as is necessary for the production of public goods to reflect local preferences. ¹⁹ A related argument states that decentralized forms of government are conducive to ‘market-preserving federalism’ ²⁰ since they create economic competition between jurisdictions and limit the central government's ability to encroach on the market. This line of reasoning maintains that such features have positive effects on economic growth. These arguments resonate with right-wing economic ideology, as they suggest that minimizing the writ of central government is necessary to ensure efficiency in policy making and economic prosperity. They reflect the thinking of parties on the economic right, such as the German Liberal party and the US Republican party, which have consistently sought to restore or increase the autonomy of the federal states.

Because decentralization removes power from the central government, it may also hinder the extent to which a country can undertake redistribution between its citizens. As a result, decentralization may prevent the implementation of key policy goals of the economic left: inter-territorial and individual equality. As Wildavsky puts it, ‘federalism means inequality’ ²¹ : decentralization can generate regional economic disparities in a variety of outcomes – such as economic growth, educational attainment or welfare state provision – because it limits the central government's ability to redistribute wealth and provide universal public services. ²² Just as the logic of efficiency means that the economic right may be in favour of decentralization, the logic of redistribution means that parties on the economic left, such as the Australian or British Labour parties, may oppose it. ²³ Specifically, our resulting hypothesis is:

Hypothesis 1: The more economically right wing a party is, the more it will support decentralization.

Decentralization and the Cultural Dimension

The main cultural question related to decentralization is: how does it reflect the distribution of territorial identities and political communities across a country? Culturally liberal parties, including for instance most ecological parties, are likely to favour decentralization for two reasons. First, culturally liberal parties are often committed to fostering multiculturalism and defending political minorities, and decentralization allows a country's institutions to coincide with the distinct identities and aspirations of its different communities. In this, cultural liberals may be driven by their opposition to the positions of culturally conservative parties. ²⁴ Secondly, the support of culturally liberal parties may also stem from post-materialist values; decentralization is seen as a way to improve the quality of democracy by enhancing civic participation, fostering political deliberation and augmenting the accountability of decision makers. ²⁵ As Marks et al. note, the process of regionalization in the 1970s coincided with the cultural shift toward post-materialism, which challenged conventional norms such as ‘centralized decision-making’. ²⁶

In contrast, culturally conservative parties, such as the Partido Popular in Spain or the Greater Romania Party, oppose the decentralization of power. Cultural conservatism implies a commitment to preserving the existing order on political, social and cultural issues. Parties with such views generally praise traditional values such as ‘the nation’, identify with a single national political community and regard the recognition of cultural diversity as a source of erosion of the integrity of the national community and democratic citizenship. They also emphasize the hierarchical nature of political authority and are thus suspicious of any mass involvement in (local) politics that may threaten political stability and elite-based decision making. In sum, our expectation is that culturally liberal parties will support decentralization, while culturally conservative parties will oppose it. More specifically, our hypothesis is:

Hypothesis 2: The more culturally liberal a party is, the more it will support decentralization.

Decentralization and Contextual Factors

The way in which decentralization raises questions of efficiency, redistribution and identity will vary across countries as a function of their social and territorial divisions, institutional setting and the structure of their political competition. Such factors will influence the sensitivity of individual parties to wealth creation and distribution as well as to identity claims, and thus shape how they interpret and position themselves on the issue of decentralization and determine how decentralization is linked to the economic or cultural dimension. Here we examine the effects of four potentially important differences between countries: the degree of self-rule, the degree of regional economic disparity, the presence of regionally based ethnic groups and the ideology of regionalist parties.

Degree of Regional Self-Rule

How the economic dimension determines party positions on decentralization may depend on the existing degree of regional self-rule. This is because the existing level of self-rule may affect whether the issue of decentralization is likely to raise questions of efficiency and redistribution. When a country has a high degree of regional self-rule, and when decentralization affects taxation powers and welfare functions, the issue may be seen largely through the lens of redistribution and efficiency, as was the case with Switzerland's 2001 fiscal equalization reforms or those of Germany in 2009. Economically right-wing parties endorse decentralization because they favour institutional changes that prevent the central government from imposing higher levels of taxation on richer regions and from interfering with regional governments’ economic policies. Economically left-wing parties oppose decentralization because they are critical of the effect of high levels of self-rule on the central government's ability to redistribute wealth and ensure uniform outcomes in social service provision. Thus our first contextual hypothesis is:

Hypothesis 3: The greater the level of regional self-rule, the stronger the association between the economic dimension and positions on decentralization.

Regional Economic Disparity

The influence of economic ideology on parties’ positioning may also depend on the degree of economic disparity between regions within a state. Economic production can be strongly regionalized, which can lead to substantial regional disparities in economic wealth. ²⁷ For example, the success of wealthy powerhouses such as Baden-Württemberg, Catalonia, Lombardy and Rhône-Alpes contrasts with poorer regions of each country. ²⁸ In countries with large economic disparities, decentralization raises questions of efficiency and redistribution. Economically strong regions may favour decentralization because it allows them to retain their wealth and their economic model, while weaker regions will plead for the redistribution of economic gains across the country.

We expect that in countries with such disparities, statewide parties will adopt positions on decentralization that reflect their attitudes toward wealth, redistribution and equality. Thus parties on the economic right are likely to welcome decentralization because it gives regions control over the resources necessary for encouraging growth – such as investment in human capital, communication and transport networks ²⁹ – and because it is conducive to creating an optimal link between fiscal policy and the provision of public goods that reflects the ideal preference of the regional median voter. ³⁰ Parties on the economic left will oppose decentralization, since they will support a tighter fiscal union that ensures redistribution from wealthy to poorer citizens and regions. For example, during Germany's reform of territorial financing in 2009, the Liberal and Christian Democratic parties supported decentralization to a greater extent than the sceptical Social Democratic party. Thus our second contextual hypothesis is:

Hypothesis 4: The greater the degree of regional economic disparity, the greater the association between the economic dimension and positions on decentralization.

Regionally Based Ethnic Groups

How decentralization links to the cultural dimension may depend on the presence of regionally based ethnic groups that display distinct cultural and political identities and articulate different policy preferences. ³¹ The presence of regionally based ethnic groups, such as the Scots in the UK, the Corsicans in France or the Basques in Spain, will enhance the appeal of the logic of identity. Thus decentralization may be an element of a broader ‘ethnicization’ of politics. ³² In such countries, decentralization deals mostly with the symbolic recognition of national pluralism and the establishment of structures of regional government that map onto national communities.

Statewide parties’ positions on decentralization depend on whether they support the aspirations of regionally based ethnic groups. Given our theoretical approach, we expect support for decentralization to be more closely linked to the cultural dimension. Culturally liberal parties will endorse decentralization because this may provide national minorities or stateless groups with group-based territorial rights, defuse conflict by ‘containing nationalism’ ³³ or decentralize political tensions. ³⁴ In contrast, culturally conservative parties will be even more strongly opposed to decentralization, as recognizing and empowering regional groups may weaken national identity and the national community. Thus, our third contextual hypothesis is:

Hypothesis 5: Cultural liberalism is more strongly associated with support for decentralization when there is a regionally based ethnic group.

Regionalist Party Ideology

A final contextual factor is the ideology of regionalist parties. While such parties usually concentrate on the need for decentralization, they also adopt positions on the economic and cultural dimensions. ³⁵ For example, the Scottish National Party (SNP) is traditionally on the economic left and culturally liberal, the Catalonian Convergencia i Unio (CiU) is relatively centrist in economic and cultural terms, and the Belgian Vlaams Belang is on the economic right and culturally conservative. Where they exist, regionalist parties may use their blackmail or coalition potential to persuade statewide parties to acknowledge and consider their demands for territorial autonomy, self-determination or independence. ³⁶ Which parties are threatened by regionalist competitors will depend on each statewide party's economic and cultural policies and the area of the policy space it occupies. ³⁷

Statewide parties may respond to this threat by strategically adopting positions on decentralization in order to maximize their share of the vote and to attain office, rather than out of ideological considerations. ³⁸ A statewide party can respond to a threat from a regionalist party by adopting either an accommodative or adversarial strategy. ³⁹ If it is directly threatened, a statewide party may adopt an accommodating strategy and take a pro-decentralization stance in order to challenge a regionalist party's ownership of the decentralization issue, recoup electoral losses and avoid losing future votes. If it is not directly threatened, but its statewide competitors are, it may adopt an adversarial strategy and take an anti-decentralization stance in order to distinguish itself from its statewide rival and raise the salience of decentralization, in the hope that regionalist parties continue to sap votes from its statewide rival. For example, if a regionalist party has a centre-left economic position, such as the Scottish and Welsh nationalists in the United Kingdom, then an economic left-wing statewide party would adopt an accommodative strategy, while an economically right-wing party would follow an adversarial strategy. ⁴⁰ The opposite prediction would hold if the regionalist party were on the economic right. In essence, we thus argue that electoral strategies can overcome the associations between decentralization and the core ideological dimensions. Our next hypothesis is therefore:

Hypothesis 6: The closer a regionalist party's ideological position is to that of a statewide party, the more that statewide party will support decentralization.

Data and Model

Ideological Scales

To test our hypotheses, we need information on party positions on decentralization as well as on the economic and cultural dimensions. Our source of data in this article is the expert survey of party positions carried out by Benoit and Laver in 2002 and 2003. ⁴¹ This survey contains assessments of party ideology on economic and cultural matters as well as on decentralization. After excluding those countries not listed as fully ‘free’ by Freedom House in 2003 ⁴² and four countries with missing values on key controls, ⁴³ thirty-one countries remain in this dataset. ⁴⁴ We present our results using this expert survey rather than the newer Hooghe et al. data, which also includes questions on these three issue areas, because the Hooghe et al. survey only covers EU countries. ⁴⁵ We also ran all analyses and robustness checks using the Hooghe et al. data, and the results are consistent across these two datasets (details available from the authors).

As noted above, decentralization is a multifaceted process that may have various forms and meanings across countries; the term may refer either to different types of institutional change or different levels of governance (regional or local). For example, in multinational states such as Spain, decentralization includes both institutional and cultural components, ⁴⁶ while in the United Kingdom the cultural component is less salient, at least for statewide parties. In homogeneous states such as Denmark, decentralization may simply mean the shift of administrative powers to regions or the transfer of expenditure powers to municipalities. This variation in the meaning of decentralization makes it difficult to obtain a single valid cross-national measure for the position of parties on decentralization. However, there is currently no better alternative measure of decentralization available. ⁴⁷ This article aims to uncover differing interpretations of decentralization across countries by capturing the effects of the economic and cultural dimensions and contextual factors on party positioning.

In the Benoit and Laver survey, experts were asked to assess party positions on decentralization. Low values on this scale indicate that the party ‘opposes any decentralization of all administration and decision-making ‘ and high values indicate that the party ‘promotes decentralization of all administration and decision-making’. ⁴⁸ A histogram and rug plot of party positions on decentralization are presented in Figure 1, which shows that non-regional parties differ substantially in their positions on this topic, though no such parties take particularly extreme views.

Fig. 1 Party positions on decentralization

Note: data from Benoit and Laver 2006. Countries included are listed in footnote 44; parties included are in Appendix 2.

To assess positions on the economic dimension, we use the public services versus taxes scale, in which low values mean that the party ‘promotes cutting public services to cut taxes’ and high values that it ‘promotes raising taxes to increase public services’.

Measuring the cultural dimension is less straightforward. In this article, we use party attitudes toward personal freedoms and traditional values. In the Benoit and Laver survey, this scale is called the ‘social policy’ dimension: low values signify that the party ‘opposes liberal policies on matters such as abortion, homosexuality and euthanasia’ and high values that it favours such policies. ⁴⁹ This question only covers part of the issues that comprise the cultural dimension, as topics such as nationalism and immigration are not included. ⁵⁰ To make sure that our results do not depend on how we measure the cultural dimension, we also ran all our analyses using the immigration or nationalism scale, where available, as well as with indices constructed by averaging the immigration/nationalism and social policy scales. Our results do not change when these alternative measures are used.

We model our dependent variable, decentralization, using linear regression. Since our data structure is best described as hierarchical (parties nested in countries), we run a multilevel model with country-level random intercepts. Crucially, setting up the model in this ways recognizes the potentially clustered nature of our observations while allowing us to include country-level controls and cross-level interaction effects. All models are run using the xtmixed command in Stata 11. ⁵¹

Contextual Factors

We argued above that the influence of the economic and cultural dimensions on decentralization positions may depend on a series of contextual factors. First, we measure the existing level of decentralization using the Marks et al. measure of self-rule, which is the extent to which sub-national units can run their own affairs independently of the central government. ⁵² We take the country-wide average for 2002–03, when the Benoit and Laver survey was carried out. Countries such as Belgium, Spain, Italy and Germany have high values on this variable, and countries such as the Baltic States, Bulgaria, Portugal and Slovenia have low values.

Secondly, we measure the regional disparity in economic prosperity using information on Nomenclature of Territorial Units for Statistics (NUTS)-2 regions for EU countries ⁵³ and on Organisation for Economic Co-operation and Development-defined regions for all other countries. ⁵⁴ For each country, we calculate the coefficient of variation (the standard deviation divided by the mean) of gross domestic product (GDP) per capita across all regions. ⁵⁵ This variable captures the extent to which regional mean prosperity varies within a country. ⁵⁶ Countries with notably high regional disparity include Belgium, Hungary, Romania, Slovakia and the United Kingdom; Australia, Japan and the Netherlands have low levels of inequality between regions.

Thirdly, we measure whether a country has a regionally based ethnic group using information provided in the Ethnic Power Relations dataset and its geographic supplement. ⁵⁷ We code ethnopolitically relevant groups as regionally concentrated if they are either only or partly regionally based. ⁵⁸ The resulting variable is 1 if the country has a regionally based ethnic group, and 0 if not.

Finally, we assess the extent to which regionalist parties take a stance on the statewide party's economic position. To measure this, we first classify parties as regionalist parties if they were given the equivalent party code in the Comparative Manifestos Project dataset or in the Hooghe et al. survey. ⁵⁹ We then consolidate these codes given our own country-specific knowledge in order to remove statewide nationalist parties from this categorization. We then code the average position of regionalist parties as their mean position on each of the two ideological scales, weighted by their share of the vote in general elections. ⁶⁰

Party-Level Controls

We include as controls three party-level variables that may strongly influence decentralization positions. First, parties that participate in government at the national level may be less likely to support decentralization: they have an incentive to maintain the status quo in the division of power between the different levels of governance in order to widen their room for manoeuvre. Devolving power to lower levels may also mean handing power to other parties that are sub-nationally strong. ⁶¹ We measure government participation at the time of the survey using information provided by Benoit and Laver directly. ⁶² We code parties as 1 if they were in government, 0 if not.

Secondly, parties may have a strategic incentive to take a position different to that of their competitors. In other words, the more the government (opposition) parties are in favour of decentralization, the more the opposition (government) parties will be against it. For example, this pattern was found in the rivalry between statewide parties in countries such as France and Greece, where Social Democratic parties favoured decentralization during their long spell in opposition to Conservative governing parties. We thus include a variable that measures the mean position of the government (for opposition parties) and of the opposition (for government parties), weighted by their vote share.

Finally, smaller parties may be more likely to support decentralization because they may want to ‘shake up the party system’ by taking a position in favour of change. ⁶³ There is evidence that small parties are more likely to be Eurosceptic; ⁶⁴ a similar pattern may exist for the decentralization issue. We measure party size using the most recent party vote share information provided in the Benoit and Laver dataset. The variable ranges from 0 to 1.

Country-Level Controls

We also include a series of country-level controls in all models. These include some of the contextual factors introduced above, namely the level of self-rule, the degree of regional economic disparity and the presence of regionally based ethnic groups. We also include as a control whether the party system includes a regionalist party, because the presence of such parties may polarize the positions of statewide parties, especially if the decentralization issue becomes very salient. Depending on the responses of statewide parties, we may see more positive or more negative stances on decentralization as a result. Though we are uncertain of the direction of a regionalist party's influence, this is nevertheless an important control variable.

Support for decentralization by statewide parties may also be greater in geographically and demographically large countries. A large country may have a greater heterogeneity of local conditions and voter preferences, and thus a greater need (and support) for locally tailored policies. ⁶⁵ We coded the geographic size of a country (in km²) using the United Nations Demographic Yearbook. ⁶⁶ The 2002 population of a country (in millions) is taken from Heston et al. ⁶⁷ For both area and population, we use the natural logarithm of the raw values in our models.

Results

The results of our model are presented in Table 1. The dependent variable, which ranges from 1 to 20, is coded so that positive values indicate support for decentralization.

Table 1 Results from Multilevel Linear Regression Model Predicting Decentralization Positions

Note: the outcome variable in all regressions is party position on decentralization, scaled from 1 to 20, with 20 the most positive stance; standard errors in parentheses. ***p < 0.001, **p < 0.01, *p < 0.05.

Decentralization and Ideological Dimensions

Model 1 presents the simple results, without controls, for the influence of the two ideological dimensions on party positions on decentralization. In Model 2, we add the controls for government participation, the mean position of the government/opposition, party size, the presence of a regionalist party, the presence of a regionally based ethnic group, the level of self-rule, and the logged geographic area and population size.

We find strong support for the claim that decentralization raises questions of identity on the one hand and efficiency and redistribution on the other, since the core ideological dimensions of political contestation are clearly associated with parties’ positions on decentralization. We can see in Models 1 and 2 that positions on the cultural dimension affect party positions on decentralization in the expected manner: the more culturally conservative a party, the more it opposes decentralization. For the economic dimension, our findings indicate that the further parties are to the right, the more they support decentralization. The magnitude of the effect of the two dimensions is similar, although the economic dimension appears to have a somewhat stronger influence. We therefore find strong support for Hypotheses 1 and 2.

The effect of these two variables is shown in Figure 2 by plotting the predicted values on the decentralization issue according to the party's position on the economic and cultural dimensions (using the results from Model 2). All other variables are held at their mean except for government participation and the presence of a regionalist party, which are both set to 0. The darker the shaded portion of the graph, the more in favour of decentralization a party is predicted to be. We can see that an economically left-wing, culturally conservative party would be clearly against decentralization, while its ideological opposite – a culturally liberal, economically right-wing party – would be expected to strongly favour decentralization. Parties in the diagonal from the bottom left to the top right are expected to take relatively centrist positions on decentralization.

Fig. 2 Predicted decentralization positions by ideological dimension

Note: predicted values for party positions on decentralization on a 1–20 scale based on Model 2. Higher values indicate greater support for decentralization. Presence of a regional party, presence of a regionally based and disadvantaged ethnic minority, and government participation set to 0; all other variables held at their mean.

Of the control variables, we find effects for two of our party-level variables (party size and government-opposition rivalry). There is thus evidence that smaller parties tend to favour decentralization. We also find that there is an effect of government-opposition rivalry: the more the government (opposition) is in favour of decentralization, the more the opposition (government) will oppose it, and vice versa. This confirms the notion that decentralization is used strategically as an element of competition between statewide parties, both in and out of power, that wish to distinguish themselves from their main rival. ⁶⁸ The presence of regionally based ethnic groups and regionalist parties also matters. If there is a regionally based ethnic group but no regionalist party, then statewide parties are, on average, slightly more favourable toward decentralization. If there is a regionalist party, then statewide parties are less supportive of decentralization. Interestingly, the only control variables close to significance are those that reflect statewide parties’ strategic considerations: their size and the positions and presence of competitors (in or out of office). Institutional and structural factors such as the level of self-rule, the presence of a regionally based ethnic group and regional economic disparity matter little. Decentralization is thus an issue that can be manipulated strategically for partisan advantage.

Contextual Factors

How does the influence of the two ideological dimensions vary based on contextual factors? Specifically, does it depend on the degree of self-rule, the degree of regional economic disparity or the presence of a regionally based ethnic group? We test our hypotheses by interacting our ideological scales with these country-level variables. Again, Table 1 presents the regression results.

To test Hypothesis 3, we interact the positions on the economic dimension with the level of self-rule. The results are presented in Model 3 in Table 1. The interaction effect is significant. Figure 3 plots the marginal effect of the economic dimension according to the level of self-rule. ⁶⁹ The histogram underlying the graph shows the distribution of sample values of self-rule. The result is clear: the greater the level of self-rule, the more economic ideology is associated with positions on decentralization. The economic right is predicted to be more in favour of decentralization than the economic left mainly when the level of self-rule is relatively high. The effect of economic policy positions on decentralization positions is around 0.5 in countries with a high level of self-rule (around 20) but close to 0 in countries with low levels of self-rule (around 0). This means that we have strong support for Hypothesis 3. ⁷⁰

Fig. 3 The effect of socio-economic position conditional on average self-rule

Note: based on Model 3. Graph shows predicted marginal effect of a one-unit increase (rightward shift) in a party's economic position conditional on a country's level of self-rule. Grey bars show the distribution of values of self-rule in the sample countries. ‘Pipes’ below histogram show individual country values of regional GDP disparity.

To test Hypothesis 4, we interact the positions on the economic dimension with the values for regional economic disparity. The results are presented in Model 4 in Table 1. Again, we plot the marginal effect of the economic dimension conditional on the degree of regional economic disparity (Figure 4). We find clear support for Hypothesis 4: the greater the level of economic disparity between a country's regions, the more economic positions are associated with decentralization positions. ⁷¹ At high levels of regional disparity (values of around 0.4) we expect a clear association between economic ideology and decentralization positions. In contrast, when regional disparity in GDP per capita is low, then we expect a weak association between views on the economic dimension and views on decentralization. ⁷²

Fig. 4 The effect of socio-economic position conditional on regional GDP disparity

Note: based on Model 4. Graph shows predicted marginal effect of a one-unit increase (rightward shift) in a party's economic position conditional on a country's level of regional economic disparity in GDP per capita. Grey bars show the distribution of values of economic disparity in the sample countries. ‘Pipes’ below histogram show individual country values of regional GDP disparity.

To test Hypothesis 5, we interact positioning on the cultural dimensions with the presence of a regionally based ethnic group. The results, presented in Model 5, do not support our hypothesis. Instead, the association between positions on the cultural dimension and decentralization are, if anything, a little weaker when there is a regionally based ethnic group. However, this difference in association is not significant. Overall, there is clearly no evidence that the impact of the cultural dimension is greater when there is a regionally based ethnic group in the country. ⁷³

Our final contextual effect (Hypothesis 6) concerned the effect of the ideology of regionalist parties and how this influences the strategic incentives of statewide parties. We test this hypothesis by adding two additional variables that measure the distance of the statewide party's position from the (weighted) mean position of regionalist parties on both the economic and cultural dimensions. Our sample is reduced to the twelve countries in which a regionalist party competes. Models 6 and 7 in Table 1 present the results; Model 6 re-runs the main model for the reduced sample and Model 7 adds the new distance terms.

Model 7 demonstrates that the ideological position of regionalist parties matters, but only on the economic dimension. We illustrate this in Figure 5, where we calculate predicted positions on decentralization as in Figure 1. In Figure 5, the statewide party's economic ideology is on the x-axis and the statewide party's distance from the weighted mean position of regionalist parties on economic ideology is on the y-axis. The area in the centre at the top is left blank, as these are arithmetically impossible values. For example, a party at 10.5 on the 1–20 scale can only ever be 9.5 units from other parties. When moving from left to right, the figure clearly reinforces the previous pattern: there is an association between right-wing economic views and support for decentralization. When moving from low to high values on the y-axis, we see that support for decentralization decreases as the distance from regionalist parties increases. In other words, when the average economic position of regionalist parties is close to that of the statewide party, the statewide party chooses a more accommodative strategy and supports decentralization more. When the regionalist parties are, on average, more distant ideologically, the statewide party chooses a more adversarial ideology, so supports decentralization less. This may help to explain patterns of party competition in the United Kingdom, where the Labour party, threatened by the left-wing SNP, endorsed devolution, a position that was rejected by the Conservative party. Similarly, the right-wing Flemish regionalist parties forced the Christian Democratic and Liberal parties to adopt more ardently decentralist stances that contrast with the resilient centralism of Belgium's Socialist party. Thus we find strong support for Hypothesis 6, but only for economic ideology.

Fig. 5 The effect of regionalist party positions

Note: predicted values for party positions on decentralization on a 0–10 scale based on Model 7. Higher values indicate greater support for decentralization. Presence of a regional party, presence of a regionally based and disadvantaged ethnic minority, and government participation set to 0; all other variables held at their mean.

Discussion and Conclusion

In this article, our aim has been to follow the encouragement of Marks and Hooghe to ‘bring politics into the study of institutional choice’ ⁷⁴ by focusing on how decentralization debates tap into questions of efficiency, redistribution and identity and by analysing how ideology shapes the views of statewide parties on the territorial allocation of authority. We have shown that the positions that statewide parties take on decentralization depend on their economic and cultural views. Parties on the economic right are more supportive of decentralization than parties on the economic left, while culturally liberal parties favour decentralization more than culturally conservative parties.

This finding is important because it highlights the need to go beyond the simple left-right dimension in order to understand party preferences on decentralization – an issue of jurisdictional architecture that, like European integration, taps into two separate logics. The two dimensions may reinforce each other: culturally liberal, economically right-wing parties are expected to be most in favour of decentralization, while their ideological opposites should be least supportive. Yet most parties, especially in Western Europe, do not combine the two dimensions in this way, as cultural liberalism tends to be associated with left-wing economic views and cultural conservatism with right-wing economic views. ⁷⁵ Marks et al. have noted that this pattern can cause tension concerning whether or not to support European integration: for example, many right-wing parties are ‘rifted between nationalism and market liberalism’. ⁷⁶ This characterization may also apply to decentralization, with culturally liberal parties on the economic left torn between redistribution and recognizing diversity, and culturally conservative parties on the economic right divided between economic efficiency and nationalism. Further research should investigate in detail the internal rifts that these contradictory ideological motivations may cause, and how they are managed by individual statewide parties.

In addition to these general patterns, we also hypothesized that the association between economic and cultural ideology and party views on decentralization depends on the individual country context. We found that this is true only for economic ideology, which is more associated with decentralization positions when economic disparity between regions or the level of self-rule is high. Statewide parties also react to the strategic incentives presented by regionalist parties’ economic positions. If regionalist parties take up a similar economic position to statewide parties, then the latter adopt an accommodative strategy to counter the threat; if regionalist parties take up a distant economic position, then statewide parties will pursue an adversarial strategy. Given the static nature of the analysis, we can only demonstrate a strong association; we cannot demonstrate the dynamic effect of competitive interactions on statewide party position over time. We found no evidence of contextual effects of the influence of the cultural dimension. These results show that the link between economic ideology and views on decentralization depends on the country context: the relationship of decentralization to the logics of efficiency and redistribution is moderated by individual country institutional, structural and strategic factors.

These findings have implications for our understanding of politics in systems of multilevel governance, and go some way toward building a causal theory of authority allocation by offering insight into how ideology, rooted in a specific country context, shapes the mindset of agents responsible for determining the territorial distribution of power. They also raise several questions that deserve further investigation.

First, this article has concentrated on party positions at a single point in time. Future work should introduce a dynamic element to the analysis to see if (and how) statewide parties change their positions on decentralization over time, and what lies behind these shifts in position. Further research should examine the effects of competitive interactions and the salience of decentralization on the changes in party positions over time and across countries: ⁷⁷ when do parties change their position, when does decentralization become politicized and how does this influence decisions to reform the territorial allocation of power? ⁷⁸

Secondly, research on the process of decentralization in different countries could benefit from an explicitly party-political approach to understanding the timing and tempo of territorial reforms. ⁷⁹ Future research should examine how party- and country-specific factors – such as party organization, party competition and institutional arrangements – condition the effect of ideology and structure the incentives of statewide parties to either endorse or oppose decentralization.

Finally, the EU's jurisdictional architecture, particularly in monetary and fiscal affairs, is currently contested based on different views on efficiency, redistribution and identity. ⁸⁰ Given the salience of such polarization, how do parties’ views on decentralization and European integration fit together? Are party positions on both topics driven by the same logics? Future work should investigate if (and how) party positions on the different levels of governance are connected.