SEARCH

SEARCH BY CITATION

Keywords:

  • nationalisation;
  • methodological nationalism;
  • multilevel party systems;
  • decentralisation;
  • regional elections

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

The concept of ‘nationalisation’ is vigorously discussed in the literature and three dimensions have been proposed. A first dimension considers the extent to which a party's vote in territorial units varies across time and this is labeled ‘dynamic nationalisation’. ‘Distributional nationalisation’ focuses on the degree to which there is an equal distribution of party votes across territorial units. Finally, ‘party-linkage nationalisation’ concerns the extent to which candidates link together under common party labels. In addition to a conceptual debate there has been a simultaneous debate on the measurement of the various forms of nationalisation. This article contributes to both debates and argues that most of the literature on nationalisation suffers from a methodological nationalism bias – that is, the tendency of many scholars to choose the statewide level and national election as the natural unit of analysis. This claim is supported by a conceptual and empirical analysis regarding the effects of decentralisation on nationalisation. The conceptual analysis shows that the non-robust findings of many studies concerning the effects of decentralisation on nationalisation can be related to the methodological nationalism bias. An effect of decentralisation is found once nationalisation is conceptualised with regard to its multilevel dimension and the measurements of nationalisation are differentiated according to parties, regions and type of elections (national or regional). An empirical analysis on the nationalisation of party systems, parties and regions in 18 countries for national and regional elections held between 1945 and 2009 shows that regional authority has a significant and robust effect on regions and regional elections but not on parties, party systems and national vote shares.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

Nationalisation is a concept in political science which has drawn the attention of various scholars over the past decades. The term refers to the territorial homogenisation of voting behaviour. Nationalisation processes ‘represent a broad historical evolution toward the formation of national electorates and party systems [whereby] peripheral and regional specificities disappear’ (Caramani 2004: 1). One of the first scholars who addressed the phenomenon of nationalisation was Donald Stokes (1967), who considered the degree to which a party's vote in territorial units varies across time. In this perspective, nationalisation refers to the extent to which voters respond similarly to national stimuli and this is labeled ‘dynamic nationalisation’ by Morgenstern et al. (2009: 1322). After Stokes, a process of ‘concept-stretching’ took place and a variety of definitions of ‘nationalisation’ followed suit (Caramani 2004; Chhibber & Kollman 2004; Clagett et al. 1984). A second dimension, identified as ‘distributional nationalisation’ by Morgenstern et al. (2009: 1322), focuses on the degree to which there is an equal distribution of party votes across territorial units. In this view, party nationalisation occurs when a party obtains similar vote shares across the territory. Chhibber and Kollman (2004) and Cox (1997) add a third dimension which they conceive of as the extent to which candidates link together under common party labels (‘party-linkage nationalisation’) whereby nationalisation refers to the process of party aggregation.

Next to a conceptual debate, there is an intensive debate on the measurement of the various dimensions of nationalisation. The main debate on the measurement of dynamic nationalisation centres on the measurement of the ‘uniform swing’. This occurs when vote share changes between two elections are similar across the districts (Stokes 1967). In this perspective, dynamic nationalisation refers ‘to the degree to which a shift in the distribution of the vote in one district is identical to the shift in the vote in each other district’, but this may underestimate nationalisation since the term can also be extended to cover ‘the degree to which different areas respond to the same electoral forces, whether or not the effects of these forces are numerically identical across space’ (Katz 1973: 817).

Similarly, there are different measurements of distributional nationalization and a good overview is provided in Bochsler (2010). Scholars most commonly consider the distribution of vote shares across the territory. For example, Rose and Urwin (1975) adopt a measurement which focuses on the dispersion of regional vote shares around the national mean. However, this measurement is sensitive to the size of the party because those with larger vote shares tend to have larger deviations from the national mean. These considerations lead Caramani (2004) to use an index adjusted for party size and the number of regions. This measurement is also based on the differences between a percentage a party obtains in each constituency and the national result, but the ‘core expression’ is divided by the number of regions and average party vote share. The latest development in the measurement of distributional nationalisation relies on the Gini coefficient, which takes into account the different number of voters or votes across the districts (Bochsler 2010; Jones & Mainwairing 2003).

A final dimension of nationalisation considers the extent to which candidates present themselves under one common party label and this may be called ‘party-linkage nationalisation’. A simple measure would be to calculate the percentage of regions in which a party competes. A party is nationalised when it presents candidates in all districts across the territory and it is denationalised or regionalised when it presents candidates in one district only. The main objection of this indicator is that it is highly sensitive to the number of districts in a country. Therefore, Chhibber and Kollman (2004) and Cox (1997) use an index which subtracts the average of the effective number of parties at the district level from the same number at the national level. Nationalisation occurs when the effective number of parties is the same at the district and national level, but a country denationalises when the number of parties at the national level increases because different parties compete across the districts.

Despite the considerable progress in the scholarship on nationalisation I argue that there is a fundamental problem in the conceptualisation and measurement of it. Most nationalisation dimensions and measurements are subject to a ‘methodological nationalism bias’ (Jeffery & Wincott 2010) – that is, the tendency to choose the statewide level as the natural unit of analysis. As a result, most nationalisation studies focus on party systems and on national elections (Gibson & Suarez-Cao 2010; Swenden & Maddens 2009). The argument put forward in this article is that the concept and measurement of nationalisation should acknowledge the multilevel dimension of party systems which entails that measurement should be differentiated according to regions and parties and that regional elections should be included.

Multilevel nationalisation ‘stretches’ the concept further, but there are good reasons for doing so. Most studies on nationalisation look at party systems and national elections and some scholars do find evidence for a relationship between the degree of decentralisation of authority and the nationalisation of party systems (Chhibber & Kollman 2004; Harbers 2010) whereas others do not (Caramani 2004; Deschouwer 2009; Lago-Peñas & Lago-Peñas 2010). In this article I hypothesise that regional authority may affect individual regions and regional elections but decentralisation does not necessarily need to have an impact on (all) parties, party systems and national elections. An empirical analysis on national and regional election vote shares in 18 countries for elections held between 1945 and 2009 shows that decentralisation has a statistically significant and robust effect on the nationalisation of regional elections and regions but not on the nationalisation of parties, party systems and national elections.

The structure of the article is as follows. The next section discusses the literature on the relationship between decentralisation of authority and nationalisation of party systems. The section that follows introduces the hypothesis that regional authority may have an effect on particular parties or regions rather than on (all) parties and party systems. The fourth section shows that commonly used measurements of nationalisation suffer from a ‘methodological nationalism bias’. A multilevel perspective of party system nationalisation is then proposed, which leads to the operationalisation of various measurements of distributional and dynamic nationalisation. An empirical analysis follows, which corroborates the argument that regional authority impacts on regions and regional elections but not on (all) parties, party systems or national elections. The final section concludes.

Decentralisation and denationalisation of party systems

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

The relative power of each level of government in a country is often considered as a key institutional variable capable of influencing subnational electorates and subnational party strategies (Hough & Jeffery 2006; Pallarés & Keating 2003; Swenden 2006). However, only a few authors have explicitly addressed the causal relationship between decentralisation and its effects on party systems. Two examples are Chhibber and Kollman (2004) and Thorlakson (2007, 2009).

Chhibber and Kollman (2004: 19) start with the premise that ‘politicians have always seen it in their collective and individual interest to establish linkages across district lines, to aggregate their votes across districts to create regional or national parties that can influence policy or run the government’. The degree to which politicians will aggregate themselves in regional or national parties depends on the authority of the national government vis-à-vis the subnational government. If local governments make the important decisions, politicians may find it important to communicate local party labels to the voters because the candidate wants to make clear with which group he or she will work with, once voted in office. The opposite situation arises when national governments make the important decisions. Then it becomes important for candidates to communicate national party labels in order to position themselves relative to national government policies and to make clear that the candidate could become part of the national government. Chhibber and Kollman (2004: 20) add that the amplitude of political centralisation and, by extension, the incentives of candidates to adopt party labels can vary by matter of degree.

Thorlakson (2007) argues that decentralisation gives both parties and voters incentive and opportunity to mobilise and respond to locally defined issues which may lead to the development of ‘unique’ party systems at the subnational level. With decentralisation the subnational level becomes a more important site of competition. Voters may recognise the importance of the subnational level in its own right and vote according to their assessment of the performance by subnational government. Parties in the subnational arena have an incentive to deviate their policies from the statewide party when adhering to the statewide party policies involves electoral risks in the subnational arena (Hough & Jeffery 2006; Maddens & Libbrecht 2009).

Decentralisation also creates multiple arenas of competition which leads to the potential for issues to be mobilised differently across the arenas resulting in variation in dimensions of conflict and variation in voter and party alignments (Thorlakson 2009). A shift of authority in fiscal matters and policy making from the national to the subnational level intensifies these processes. Subnational governments can use policy to ‘build’ distinctive state identities, mobilising issues related to industrial, agricultural or economic policies, where differences in the economic base can be more easily translated into policy demands among state voters that differ from those of voters in the country as a whole (Thorlakson 2000).

There are strong theoretical expectations that decentralisation of authority should affect politicians, voters and parties to such an extent that party systems become territorially distinct. However, the empirical results in nationalisation studies are mixed at best. Some do not find evidence for a relationship between the degree of decentralisation of authority and the nationalisation of party systems (Caramani 2004; Deschouwer 2009; Lago-Peñas & Lago-Peñas 2010) whereas others do (Chhibber & Kollman 2004; Harbers 2010; Thorlakson 2007). How can we account for these conflicting findings? Surely, these studies differ with respect to the operationalisation of the dependent variable, sample of cases and whether the focus is on cross-national or cross-time variance – factors which may explain the difference in findings. However, I argue that a more fundamental problem underlies these studies – namely, most analyses do not specify in detail under which conditions one should expect to find an effect of decentralisation on the territoriality of the vote. The next section discusses these conditions.

Decentralisation and denationalisation of regions and parties

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

Caramani (2004: 291–292) offers an interesting hypothesis for the non-impact of regional authority on nationalisation: ‘[R]ather than being a cause of territorialization of voting behavior, federal structures reduce the expression of regional protest in the party system by opening up institutional channels of voice.’ In this view, one does not expect decentralisation to have an effect on homogenisation of vote shares in national elections but one does for regional elections.

Caramani's hypothesis forces one to break away from the statewide party system perspective to analyse the effects of decentralisation on the constituent units of party systems which are parties and regions. A similar conclusion has been reached by Swenden and Maddens (2009: 4–5) who point out that nationalisation studies are all based on results for national elections alone (see also Gibson & Suarez-Cao 2010). As a consequence, regional elections are disregarded which leaves out an increasingly important component of party system nationalisation since ‘regional governments frequently control important aspects of capital-intensive policies such as education, industrial development, transport, infrastructure, health, or social assistance’ (Swenden & Maddens 2009: 4–5). One of the few nationalisation studies that include federal and state elections is the one by Lori Thorlakson (2007), and she finds clear evidence that party systems are least congruent in decentralised federations. There is, therefore, a strong case to broaden the study of party system nationalisation to regions, parties and regional elections (see also Deschouwer 2003; Hopkin 2009).

Studies on regional elections have been dominated by a second-order election perspective. The concept of second-order elections was introduced by Reif and Schmitt (1980) who studied the first election to the European Parliament (EP). They observed a seeming subordination of EP election results to ‘first-order’ national-level politics. Reif and Schmitt argued that there is less ‘at stake’ in second-order elections than in national elections, prompting voters to use them as an opportunity to vent their spleen about national-level politics – through protest votes for fringe and opposition parties or by not bothering to turn out – before returning to their ‘real’ electoral choice at the next national election (Reif & Schmitt 1980: 8–9).

The relation between the second-order election concept and the literature on nationalisation is clear. When regional elections are second-order, as Reif and Schmitt (1980: 8) argue, then regional elections are ‘repeated national elections’. Several authors applied the second-order election theory to regional elections (e.g., Bromley 2006; Cutler 2008; Floridia 2010; Pallarés & Keating 2003; Tronconi & Roux 2009). These studies reveal that ‘some places are less second-order than others’ (Jeffery & Hough 2009: 231), that ‘the second-order model has limited explanatory power in explaining regional election outcomes’ (Schakel & Jeffery 2012) and that ‘only some of the regions can be considered clearly second-order’ (Dupoirier 2004: 590; my translation). These studies also find that the degree to which regions display second-order election effects ‘depends on how much is at stake’ (e.g., institutional authority) and ‘depends on the depth of territorial cleavages’ (Jeffery & Hough 2009: 223).

The evidence on regional election results is in line with studies on the regional voter. Recent work on American (Ebeid & Rodden 2006), Canadian (Cutler 2008), German (Völkl et al. 2008) and British (Johns et al. 2010) regional elections suggests that many voters are able to compartmentalise the regional electoral arena from the national arena and make different kinds of judgment in each of them. In addition, the evidence is also congruent with the studies on decentralisation and regional party strength which shows that regional authority leads to stronger regional parties (Brancati 2008; Massetti & Schakel 2011). In sum, most of the evidence shows that regional authority has a significant effect on the nationalisation of (particular) regions and regional elections.

In contrast to regions, there are several reasons to expect that not all (statewide) parties are necessarily confronted with diverging vote shares across the territory after regional reform. First, decentralisation may lead to the establishment and rise of regional parties. These parties compete with statewide parties on traditional, economic left–right issues as well as on decentralisation issues (Meguid 2010). Regional parties can be placed on the left side as well as on the right side of the ideological spectrum and, hence, statewide parties on the left or the right may lose vote share (De Winter et al. 2006; Massetti 2009). Another strategy by statewide parties is to ally with the regional party. The parties compete under separate labels in regional elections or the statewide party refrains from participating in regional elections as part of the deal (De Winter et al. 2006). However, the parties compete in national elections under the same statewide party label. The regional party may want to participate in this deal in exchange for concessions of the statewide party on the decentralisation issue. The benefit for the statewide party is regaining or preventing a possible vote share loss.

An important lesson from this overview is that there is no a priori reason to assume that all parties are equally affected by decentralisation of authority. Regional authority does not necessarily have an effect on all parties, but we may find significant effects of decentralisation of authority for regions and regional elections. Nationalisation scholars have been primarily concerned with explaining party system nationalisation at the statewide level for national elections and measurements have been developed accordingly. It comes as no surprise that these studies have not found a strong, robust and significant effect of subnational authority on party system nationalisation. The next section will show that commonly used measurements suffer from a ‘methodological nationalism bias’.

‘Methodological nationalism bias’ in measurements of party system nationalisation

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

An example may illuminate how measurements of distributional and party-linkage nationalisation, when applied to the party system level, may miss important effects at the region and party levels. In Figure 1 one may find a hypothetical voter and vote share distribution across four regions (W through Z) and four parties (A through D) for two countries (I and II). Figure 1 also displays nationalisation scores for regions, parties and party systems.

image

Figure 1. Dynamics in nationalisation measures for parties, regions and party systems: Comparison across countries

Notes: Shown are scores on nationalisation measures for parties, regions and party systems for two countries (I and II) and for four parties (A, B, C and D) and four regions (W, X, Y and Z). There are two differences between country I and country II. First, the voter distribution is equal across the regions for country I but unequal for country II: there are 1,000 voters in regions W and X in country I but there are 250 voters in region W and 1,750 voters in region X in country II. The second difference is the vote share distribution: in region W in country I, 25 per cent of the voters vote for party A and 75 per cent vote for party D whereas 100 per cent of the voters in region W in country II vote for party D. The various measurements are discussed in the section ‘Methodological nationalism bias in measurements of party system nationalisation’.

Download figure to PowerPoint

Nationalisation of parties is measured by the standardised party nationalisation scores (PNSS) developed by Bochsler (2010) which measures the heterogeneity in vote shares across the territory according to a Gini coefficient. The difference between the PNSS and the measurements by, for example, Caramani (2004) and Jones and Mainwairing (2003), is that the PNSS weights for the sizes of territorial units and considers the different number of such units in different countries. The measurement indicates a homogeneous distribution when ‘every territorial unit will cast a number of votes for this political party which is approximately proportional to the unit's size, or when a party will win a similar vote share in every territorial unit. In the case of heterogeneous vote distributions, however, most of the votes are concentrated in a few territorial units’ (Bochsler 2010: 161). The index runs from 0 to 1, with 0 indicating that the party obtains all its vote share in one region (denationalised), and a score of 1 reflects a perfectly equal vote share distribution across the regions (nationalised).

Nationalisation of regions is measured by a dissimilarity index (Floridia 2010; Jeffery & Hough 2009; Johnston 1980; Pallarés & Keating 2003; Tronconi & Roux 2009) which is calculated by subtracting a party vote share obtained in a national election from a vote share obtained in a regional election. The absolute values of party scores are taken, which are subsequently summed and divided by two to avoid double counting (one party's loss is another party's gain). The mathematical formula for calculating dissimilarity (DIS) is the following:

  • image(1)

whereby iN stands for the party vote share obtained by party i in the national election and iR stands for the party vote share obtained by party i in the regional election. Scores run from 0 per cent where the party system in the region is completely the same as that of the statewide party system (nationalised) through 100 per cent where the party system at the regional level is completely dissimilar (denationalised).

Party system nationalisation is measured by four indicators which are commonly used in the literature. The first, PNSSw, is a cumulative and weighted (by party size) index of PNSS party scores and the second, DISw, does the same for DIS region scores (weighted for region size). The Index adjusted for Party size and the number of Regions (IPRw) developed by Caramani (2004) has the same purpose as the PNSS – that is, to measure the extent to which vote shares are equal across the territory. The index is weighted for party size and aggregated at the party system level. Scores run from 0 per cent (nationalized) through to 100 per cent (denationalized). The fourth measure is the index of party aggregation introduced by Chhibber and Kollman (2004)1 and this measurement is only available at the party system level. This index subtracts the average of the effective number of parties at the district level from the same number at the national level. The idea is that the national party system will become larger when different parties compete across the regions. There is no (theoretical) upper limit, but denationalisation is indicated with higher scores.

There are two differences between country I and II. The first concerns the voter distribution across the regions. The regions in country I each contain 1,000 voters; in country II the voter distribution is unequal. Region W has 250 and region X has 1,750 voters, whereas regions Y and Z both have 1,000 voters. The second difference concerns the vote share distribution. In region W, in country I, 25 per cent of the voters vote for party A and 75 per cent vote for party D, whereas 100 per cent of the voters in region W, in country II, vote for party D.

The example leads to two insights in the behaviour of measurements of nationalisation. First, party system nationalisation may conceal significant variation in the nationalisation of parties and regions. The PNSS changes with magnitudes varying from –0.12 to +0.15, which, on a scale from 0 to 1, is equal to a 12–15 per cent change. Similarly, DIS changes from –8 per cent through +33 per cent, which are equivalent to a 8–33 per cent change on a scale running from 0 to 100. In contrast, party system nationalisation scores remain remarkably stable. The PNSSw changes by −0.001 which represents a 0.1 per cent decline. The DISw and IPRw change, respectively, by –4.2 and +3.1 per cent.

A second and related insight is that some parties and regions nationalise whereas others denationalise. Party B and region Y nationalizes, but parties A, C and D and regions W and Z denationalise. The net effect at the party system level is that countries I and II have practically the same scores for party system nationalisation.

The insights which the example in Figure 1 provides are important for the study on the effects of regional authority on nationalisation. Suppose that there is a third difference between countries I and II– for example, regional authority for region W in country II, which is a relatively small region, is higher than for the other regions. This is not a far-fetched scenario; take for example the special regions in Italy, the historic nations in Spain, Åland in Finland, and Scotland and Wales in the United Kingdom. These are all relatively small regions and they have a regional government which is more authoritative than that of the other regions (Hooghe et al. 2010). If we were to adopt a party system level perspective we would conclude that regional authority does not matter for nationalisation. However, if we were to adopt a region or party level perspective we would conclude that regional authority leads to nationalisation or denationalisation depending on the particular region or party on which we focus.

The party system nationalisation indicators have corrections for party and region size because they are designed to consider nationalisation of large parties and regions as more important than smaller ones.2 However, given that voter distributions across the territory are highly dissimilar across countries and that regional authority might affect particular parties and regions but not others, the effects of decentralisation may be unnoticed when one focuses at the party system level. Since most nationalisation studies are based on overall aggregate national patterns, ‘they cannot fully account for the potential importance of limited or aberrant “nationalization” in strong regions, such as Catalonia, Scotland, the Basque Country, Wales or even Bavaria, that represent relatively small sections of the population but play a key role in triggering the decentralization of the state’ (Swenden & Maddens 2009: 5).

The example in Figure 1 makes clear that party system nationalisation may be something different than nationalisation of parties and regions. More importantly, measurements of party system nationalisation may not be appropriate to study the effect of regional authority to the extent that decentralisation may affect some regions and parties but not others. The discussion of the literature in the previous section provides strong evidence that decentralisation indeed has an impact on particular regions and particular parties, but probably not on the political system as a whole. We have therefore strong arguments for broadening the analysis of the nationalisation of party systems to the nationalisation of regions, parties and regional elections. This implies that we should conceptualise party systems as ‘multilevel’ (Suarez-Cao & Freidenberg 2010; Swenden & Maddens 2009) or as ‘federalised’ (Gibson & Suarez-Cao 2010).

According to Swenden and Maddens (2009: 6), ‘the multilevel party system brings together a statewide party system which emerges from statewide elections and a set of regional party systems reflecting the outcome for regional elections’. A full understanding of party competition in federalised party systems ‘requires consideration of these separate party subsystems, as well as the interactions between them’ (Gibson & Suarez-Cao 2010: 37). Similarly, Swenden and Maddens (2009: 6) argue that nationalisation of multilevel party systems involves looking at the integration of the multilevel party system, which is indicated by the extent to which the same parties are represented at the various levels of the system and how evenly balanced their support is across the levels and regions of the state:

A multilevel party system is fully integrated if the same parties are represented at the statewide and regional levels in about equal strength. A multilevel party system is loosely integrated if the support for statewide parties varies considerably across the regions and between the statewide and regional levels. (Swenden & Maddens 2009: 6)

Measurement of nationalisation of parties and regions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

Figure 2 provides an example of how PNSS party scores and DIS region scores may be still of use in the measurement of nationalisation. In Figure 2 we display the same example as depicted in Figure 1, but now the number of voters across all the regions is the same. The example in Figure 2 allows us to see how the nationalisation measures behave when we focus on change between elections within one country. This comparison concerns dynamic nationalisation and involves the analysis of vote swings between elections. For example, one could compare an election at time t with an election at time t+1dynamic nationalisation ‘proper’– or one could compare a national election to a regional election, which can be seen to indicate the extent of party system integration. Again, the voters who vote for party A in election I (25 per cent of the total votes) vote for party D in election II and, as a result, this party sees its vote share rising from 75 to 100 per cent.

image

Figure 2. Dynamics in nationalisation measures for parties, regions and party systems: Comparison across elections

Notes: Shown are scores on nationalisation measures for parties, regions and party systems for two elections (I and II) and for four parties (A, B, C and D) and four regions (W, X, Y and Z). There is one difference in the vote share distribution between election I and election II. In region W in election I, 25 per cent of the voters vote for party A and 75 per cent vote for party D whereas 100 per cent of the voters in region W vote for party D in election II. The various measurements are discussed in the sections ‘Methodological nationalism bias in measurements of party system nationalisation’ and ‘Measurement of nationalisation of parties and regions’.

Download figure to PowerPoint

The first thing to notice is that the PNSSw and DISw indicate properly that election II is less nationalised than election I. Although the IPRw and Chi/Ko also indicate that election II is more regionalised, the amount of difference they depict is by far less than the changes we observe at the region and party levels.

At the party level (PNSS scores) we find that two parties (A and D) are affected and both parties denationalise. This is in line with the change in vote share distribution in region W and the PNSS seems to be an appropriate measurement to indicate party nationalisation. However, turning back to the example in Figure 1 reminds us that in a cross-sectional perspective, in which unequal voter distributions are likely, the PNSS measurement might be less useful. Suppose that regional authority is higher for region W in country II than for any other region. In that case, we cannot make use of party PNSS scores in a cross-sectional comparison because the differences in scores may just reflect variance in relative voter distributions across the territorial units. In other words, we do not know whether the differences in party PNSS scores are caused by the difference in regional authority or by the difference in voter distributions across the regions.3 In sum, the PNSS is a valid instrument of nationalisation of parties, but rather useless for studying the effect of regional authority in a comparative framework unless we have an indication (additional hypothesis) which particular parties should be affected.

In contrast to parties, all regions denationalise according to the DIS scores. This is a bit awkward since only the vote distribution within region W has changed. A closer look at how the DIS measurement is constructed reveals why this is the case. The DIS measure compares party vote shares at the regional level with those at the national level. In Figure 2 the DIS measures are annotated to reveal how the comparison is being made. The input in the dissimilarity formula may vary according to the type of election or vote share (i.e., national elections (N) or regional elections (R)) and with respect to the territorial unit of analysis (i.e., national level (N) or regional level (R)). Hence, the vote shares that can be plugged into the dissimilarity index are the following:

  • • 
    National electoral results aggregated at the national level (NN)
  • • 
    The vote share in a particular region for the national election (NR)
  • • 
    Regional electoral results aggregated at the national level (RN)
  • • 
    The vote share in a particular region for the regional election (RR)

Assume that election I is a national election. In that case, we compare vote shares for national elections (N) disaggregated at the regional level (R) with vote shares for national elections (N) aggregated at the national level (N). Hence the comparison is NN with NR whereby the first letter indicates the type of election and the second letter denotes the level of aggregation.

Suppose that election II is a regional election. In that case, we compare vote shares for regional elections (R) disaggregated at the regional level (R) with vote shares for regional elections (R) aggregated at the national level (N), hence RN versus RR. Since the change in the vote shares for parties A and D in regional elections leads to different regional election vote shares aggregated at the national level, all regions increase their score on the DIS measure. Suppose that regional authority for region W increases between election I and election II, but that the other regions are not subject to regional reform. In that case, we would conclude that decentralisation would not matter because regions X, Y and Z increased their DIS scores without having experienced regional reform. However, if one constructs a dissimilarity index which compares vote shares obtained in regional elections (R) disaggregated at the regional level (R) with vote shares for national elections (N) aggregated at the national level (N), hence NN versus RR, one would arrive at the correct conclusions. With this measure, regions X, Y and Z do not change on their DIS scores whereas region W increases its score in exact concordance with the change in vote share distribution between parties A and D. This holds also when the voter distribution across the regions is unequal because the region is treated by the measurement as one unit no matter how many voters it contains.

To sum up, the examples in Figures 1 and 2 have made clear that when decentralisation has a significant effect on parties and regions it could remain undetected by commonly used measurements of nationalisation. To fully capture the dynamics in multilevel party systems, nationalisation should be operationalised according to party systems, parties and regions. In addition, the measures of nationalisation should be applied to national election vote shares, regional election vote shares and to the difference between them. This amounts to 12 different operationalisations of nationalisation.

  • Parties

    • 1
      National elections: PNSS party scores based on national election vote shares
    • 2
      Regional elections: PNSS party scores based on regional election vote shares
    • 3
      Difference elections: (1) minus (2)
  • Regions

    • 4
      National elections: Dissimilarity between national election results aggregated at the national level and the national election results for a particular region (NN–NR)
    • 5
      Regional elections: Dissimilarity between regional election results aggregated at the national level and the regional election results for a particular region (RN–RR)
    • 6
      Difference elections: Dissimilarity between the national and the regional party system (NN–RR)
  • Party systems

    • 7
      National elections: The weighted (by party size) sum of PNSS scores obtained in national elections
    • 8
      Regional elections: The weighted (by party size) sum of PNSS scores obtained in regional elections
    • 9
      Difference elections: (7) minus (8)
    • 10
      National elections: The weighted (by region size) sum of DISNN–NR scores
    • 11
      Regional elections: The weighted (by region size) sum of DISRN–RR scores
    • 12
      Difference elections: The weighted (by region size) sum of DISNN–RR scores

The nationalisation measures applied to national and regional elections concern the measurement of distributional and party-linkage nationalization, whereas the integration of party systems is captured by dynamic nationalisation which is indicated by the difference measures. The remainder of the article will systematically test the effects of decentralisation on the nationalisation of multilevel party systems according to the various conceptualisations and measurements of nationalisation.

Empirical analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

This article analyses regional and national election vote shares for regions and parties in 18 countries (see Table 1).4 The sample of 18 countries provides a good mix whereby the extent of regional authority varies widely between them but also within them. In addition, some countries have held regional elections since 1945 whereas other countries introduced regional elections at a later date.

Table 1. Countries, elections, regions and parties
CountryRegionsNumber of regionsElection periodNumber of electionsNumber of parties
NationalRegionalNationalRegional
  1. Notes: The number of regions varies over time because of newly created regions: Australian Capital Territory (1989) and Northern Territory (1974) in Australia; Yukon (1978) in Canada; replacement of the 15 Amter with five Regions in 2005 in Denmark; Baden Würtemberg (1953), Saarland (1957) and the addition of six Eastern German Länder in 1990 in Germany; Okinawa (1972) in Japan; Flevoland (1987) in the Netherlands; Ceuta and Mellila (1995) in Spain; the merger of five län into two in Sweden in 1998 (Älvsborg, Gothenburg and Bohus, and Skaraborg into Västra Götaland; Kristianstad and Malmöhus into Skåne); and Jura (1978) in Switzerland. The number of French regions (Corsica), Swiss cantons and American states vary over time because of data availability.

  2. Source: Schakel (2010).

AustraliaStates, Territories6-81946–2007251471325
Austria Länder 91945–2008201281014
Belgium Gemeenschappen, Gewesten 41995–20074161418
CanadaProvinces, Territories10–111945–2008211941519
Denmark Amter/Regions15/51973–200712/2120/101113
France Régions 21–221986–20076871425
Germany Länder 8–161949–2009172021835
Greece Nomoi 501993–2009620077
Italy Regioni 201948–2008161864373
Japan Todofuken 46–471967–200915516219
Netherlands Provincie 11–121946–2006191822127
Norway Fylker 191973–2009101711111
Portugal Distritos 201975–2009131886
Spain Comunidades autonomas 17–191979–200891333250
Sweden Län 21–241973–200611255810
SwitzerlandCantons17–261947–2007164052232
United KingdomRegions, Northern Ireland, Scotland and Wales111997–20053958
United StatesStates, Washington, DC48–511968–2002188461811

The main independent variable of interest, regional authority, is measured by the Regional Authority Index developed by Hooghe et al. (2010). The authors provide region scores on a yearly basis which allows matching authority scores to regional level vote shares. In the case of nationalisation of parties and party systems, regional authority scores are aggregated in a similar way as dissimilarity scores – that is, they are weighed by region size (total number of regional valid votes divided by the total number of valid votes at the statewide level) before they are summed. The RAI is composed of two indicators: self-rule (authority exercised by a regional government over those who live in the region) and shared rule (authority exercised by a regional government or its representatives in the country as a whole). The scores vary between 0 and a maximum of 24 points.

Nationalisation is not solely associated with the vertical allocation of authority. Nationalisation studies point out that territorial cleavages, electoral systems, electoral cycles, size of the party and region, and presidentialism are also factors which should be considered (Bochsler 2010; Harbers 2010; Lago-Peñas & Lago-Peñas 2010; Morgenstern & Swindle 2005). Table 2 details the measurement and hypothesised effects of the control variables.5

Table 2. Control variables
VariableNationalisation ofScoresHypothesised effect (higher scores lead to)
  1. Sources: Ethnolinguistic fragmentation: Alesina et al. (2003); regional history and language index: Fitjar (2010); all other variables: Schakel (2011).

Ethnolinguistic fragmentationParty systems, partiesFrom 0 (homogeneous) to 1 (heterogeneous)Denationalisation
Regional history indexRegionsFrom 0 (not distinct) to 3 (very distinct)Denationalisation
Regional language indexRegionsFrom 0 (not distinct) to 3 (very distinct)Denationalisation
Electoral systemParty systems, parties, regionsMajoritarian/plurality systems score 0, mixed systems 0.5 and proportional systems 1Denationalisation or nationalisation
Vertical simultaneityParty systems, partiesMajority of regions hold elections simultaneously with national election (= 1)Nationalisation
RegionsThe number of days between a regional and national election divided by 100Denationalisation
Horizontal simultaneityParty systems, partiesMajority of regions hold elections simultaneously with other regional elections (= 1)Nationalisation
RegionsThe number of regions which hold their elections simultaneously divided by the total number of regions in the countryNationalisation
Presidential systemParty systems, parties, regionsA score of 1 for France (since 1958) and the United StatesNationalisation
SizePartiesParty vote share (%) at the national levelNationalisation
RegionsThe number of regional valid votes relative to the total statewide number of valid votesNationalisation

The units of analysis are nationalisation scores obtained in elections which are clustered within countries. This constitutes a typical cross-section cross-time dataset which is analysed within a Beck and Katz (1995) framework with panel corrected standard errors and a correction for autocorrelation. In the case of region and party nationalisation scores, we are confronted with an additional layer: election scores are clustered in parties or regions which are subsequently clustered within countries. In order to account for this extra tier, this article analyses multilevel mixed-effects linear models with controls for autocorrelation.6

Table 3 presents the results for six models which analyse nationalisation scores for party systems. To ease interpretation, the number of the model matches the number of the particular operationalisations of the nationalisation measurement as discussed above. The models analyse party system nationalisation for distributional (models 7–8 and 10–11) and dynamic nationalisation (models 9 and 12). Two types of party system nationalisation scores are considered. First is PNSSw, which aggregates over parties (7–9) and second is DISw, which aggregates over regions (10–12). Distributional nationalisation is indicated by higher PNSSw scores but lower DISw scores. The extent of party system integration is captured by dynamic nationalisation measures which are operationalised by the difference between regional and national elections. Here, higher scores mean that regional elections are more denationalised than national elections.

Table 3. Nationalisation of party systems
 Parties (weighted and aggregated)Regions (weighted and aggregated)
(7)(8)(9)(10)(11)(12)
National electionsRegional electionsDifference electionsNational electionsRegional electionsDifference elections
  1. Notes: Shown are the results of a regression model with panel corrected standard errors (between brackets) and a correction for autocorrelation (Beck & Katz 1995). For an overview of included national and regional elections, see Table 1.The measurement based on parties is a cumulative score (weighted by party size) of standardised Party Nationalisation Scores for parties. The measurement based on regions is a dissimilarity index whereby scores are weighted by region size. * p < 0.05; ** p < 0.01 (two-tailed).

Regional authority−0.003** (0.001)−0.006* (0.003)−0.001 (0.002)0.144 (0.095)0.219 (0.165)0.411 (0.214)
Ethnolinguistic fragmentation−0.289** (0.081)−0.335** (0.087)0.098* (0.049)15.788** (4.972)28.158** (4.371)14.538** (4.361)
Presidential system0.134** (0.042)0.279** (0.054)−0.057* (0.027)−7.352** (2.360)−13.162** (3.783)−12.579** (2.842)
Electoral system−0.055** (0.029)0.039 (0.048)−0.136** (0.049)3.495 (1.954)3.267 (1.920)2.842 (6.146)
Vertical simultaneity0.031 (0.026)−0.037** (0.014)−1.513 (1.127)−1.696 (1.956)
Horizontal simultaneity−0.059 (0.035)−0.010 (0.017)0.947 (1.806)3.797 (2.427)
Constant0.951**0.824**0.086**4.621*6.922*11.986**
Rho0.8060.8490.6700.8020.8960.753
Wald chi224**52**52**21**59**25**
R20.790.810.200.360.410.27
Number of elections244222243244222242
Number of countries181818181818

The results for the regional authority variable are disappointing. In models 7 and 8 it pops up as statistically significant, but this result is not robust. The beta coefficient for national elections loses statistical significance once the analysis is extended to those national elections which have been held before regional elections were introduced. In addition, the beta coefficient for regional elections loses statistical significance once Belgium, Italy or the United Kingdom are excluded. The finding that regional authority does not seem to matter for party system nationalisation is in line with the argument of this article and corroborates earlier findings by Caramani (2004) and Lago-Peñas and Lago-Peñas (2010).

As argued above, it is more likely that one finds significant effects for the constituent elements of party systems – that is, parties and regions. Table 4 presents six models which analyse distributional and dynamic nationalisation scores for, respectively, parties (models 1–3) and regions (models 4–6).

Table 4. Nationalisation of parties and regions
 PartiesRegions
(1)(2)(3)(4)(5)(6)
National electionsRegional electionsDifference electionsNational electionsRegional electionsDifference elections
  1. Notes: Shown are the results of a multilevel mixed-effects linear model with a correction for autocorrelation (Rho). For an overview of included national and regional elections, see Table 1. The measurement for parties is the standardised Party Nationalisation Score. The measurement for regions is a dissimilarity index. * p < 0.05; ** p < 0.01 (two-tailed).

Regional authority−0.001 (0.001)0.002 (0.002)0.003 (0.002)−0.138 (0.105)0.427** (0.133)1.210** (0.154)
Ethnolinguistic fragmentation−0.692** (0.169)−0.581** (0.143)−0.057 (0.073)
Regional history index1.069 (0.734)1.748** (0.653)1.214 (0.648)
Regional language index1.886** (0.728)1.393* (0.654)1.245 (0.647)
Presidential system0.042 (0.120)0.024 (0.098)−0.016 (0.054)−10.530 (7.870)−9.212 (8.816)−4.806 (7.090)
Electoral system−0.030 (0.025)0.005 (0.056)−0.047 (0.034)−5.032** (1.057)−0.179 (0.206)7.419** (1.072)
Vertical simultaneity0.024 (0.013)−0.028* (0.013)0.019 (0.026)0.264** (0.028)
Horizontal simultaneity0.057 (0.036)−0.003 (0.029)−6.775** (1.038)−1.819 (1.113)
Size0.008** (0.0004)0.017** (0.001)−0.004** (0.001)−31.528** (6.950)−29.493** (5.678)−17.852** (5.761)
       
Constant0.679**0.333**0.153**23.139**21.831**5.804
Rho0.8070.7420.6710.5990.7280.619
Wald chi2464**1008**104**59**105**258**
Log likelihood−175015321272−13933−12349−14327
Number of observations161618091264444638884392
Number of parties281391201
Number of regions362361361
Number of countries181818181818

The first three models show the results for party nationalisation scores (PNSS). The evidence is clear: there is no effect of regional authority on the territorial distribution of party vote shares. The finding that regional authority does not seem to have an impact on the nationalisation of parties should not come as a surprise. As elucidated above, parties may adapt differently in response to decentralisation or may not be affected at all.

The next set of three models analyses nationalisation scores for regions. Here regional authority does have an effect for regional elections and for the difference between regional and national elections, but there is no effect of decentralisation for national elections. The results for regional elections (distributional nationalisation) are not robust when Italy is excluded, but the results for the difference between regional and national elections (dynamic nationalisation or party system integration) survives all robustness checks. One point on the regional authority index equals a 0.4 per cent increase in dissimilarity between regions for regional elections (RN-RR) and a 1.2 per cent increase in the difference between the statewide party system and a particular regional party system (NN-RR). Given that an average regional authority is about 14 points with a 5-point standard deviation, this means that dissimilarity may increase from 4 per cent, respectively, to 11 per cent for regions which score 9 on the RAI, to from 8 per cent, respectively, to 23 per cent for regions which score 19 on the index.

The models on the region scores also produce interesting results with respect to vertical and horizontal simultaneity. Vertical non-simultaneity has an impact on dynamic nationalisation or party system integration: the difference between regional and national vote shares increase by 0.26 per cent per 100 days difference between the dates of regional and national elections. Horizontal simultaneity seems to significantly affect distributional nationalisation. Dissimilarity scores between regions decline by 6.8 per cent when regional elections are held on the same date. These results do not come as a surprise given the particular vote shares which are analysed, but this means that the measurements of nationalisation should be adapted to the particular hypothesis one wants to test. The main finding of the empirical analysis is that decentralisation of authority leads to increasing territorial heterogeneity of the vote for regions, but not for parties. In addition, decentralisation has limited or no effect on national election vote shares, but a significant effect on regional election vote shares.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

An important insight from this article is that to find the effects of decentralisation, the party system has to be conceptualised as a multilevel party system which entails that one looks at the nationalisation of parties and regions for regional and national election vote shares, as well as at the extent of party system integration, as indicated by the differences between regional and national election vote shares. The implication of this insight goes beyond the study of the denationalising effects of regional authority. For example, horizontal simultaneity of regional electoral cycles significantly pops up in the analysis on distributional nationalisation, and vertical simultaneity between national and regional elections seems to matter for dynamic nationalisation or party system integration (Table 4). This suggests that scholars of elections in multilevel systems should specify hypotheses according to the conditions under which one expects to find effects of particular independent variables, and that they should also carefully develop measures accordingly.

This claim finds further support in the non-significant findings regarding the effects of decentralisation on the nationalisation of parties. It might well be the case that the nationalisation of one party may offset the denationalisation of another party or that only a couple of parties are affected by decentralisation. Unless we have further clues about the particular parties that should be affected by regional authority we will most likely not be able to find any effect of decentralisation. The findings in this article, therefore, also point out that the literature on elections and parties in multilevel settings is currently lacking theoretical refinement (Meguid (2010) may be considered an important exception) and, as a result, significant effects may remain undetected.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References

I am grateful to two anonymous reviewers and the editors of EJPR who contributed with relevant and helpful comments on previous versions of this article. I would also like to thank the British Academy for financial support through a Newton International Fellowship for 2009–2011 which enabled me to collect election data during a postdoctoral fellowship at the University of Edinburgh.

Notes
  • 1

    What follows equally applies to the inflation index used by Cox (1997).

  • 2

    E.g., the party nationalisation scores differ across the countries because the PNSS compensates for the number of voters within a region – i.e., the measurement is designed so that it reflects not only the vote share distribution, but also the amount of voters across the regions.

  • 3

    Even if we had empirical evidence that the differences in party PNSS scores can be linked to regional authority we would be confronted with three denationalising parties and one nationalising party.

  • 4

    PNSS scores are calculated with an Excel application that is provided on Daniel Bochsler's website: http://www.bochsler.eu/

  • 5

    An online appendix (http://www.arjanschakel.nl) provides further details of the measurement of regional authority and control variables, method, robustness of the results, and discusses hypotheses and the results for the control variables.

  • 6

    To test the robustness of the results, nationalisation scores have also been analysed by using fixed effects models with lagged dependent variables. The results for these models corroborate the results for the cross-sectional and multilevel models (results not shown).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Decentralisation and denationalisation of party systems
  5. Decentralisation and denationalisation of regions and parties
  6. ‘Methodological nationalism bias’ in measurements of party system nationalisation
  7. Measurement of nationalisation of parties and regions
  8. Empirical analysis
  9. Conclusion
  10. Acknowledgements
  11. References
  • Alesina, A. et al. (2003). Fractionalization. Working Paper 9411. Cambridge, MA: National Bureau of Economic Research.
  • Beck, N. & Katz, J. (1995). What to do (and not to do) with time-series cross-section data? American Political Science Review 89(3): 634647.
  • Bochsler, D. (2010). Territory and electoral rules in post-communist democracies. Houndmills: Palgrave Macmillan.
  • Brancati, D. (2008) The origins and strengths of regional parties. British Journal of Political Science 38(1): 135159.
  • Bromley, C. (2006). Devolution and electoral politics in Scotland. In D. Hough & C. Jeffery (eds), Devolution and electoral politics. Manchester: Manchester University Press.
  • Caramani, D. (2004). The nationalization of politics: The formation of national electorates and party systems in Western Europe. Cambridge: Cambridge University Press.
  • Chhibber, P.K. & Kollman, K. (2004). The formation of national party systems: Federalism and party competition in Canada, Great Britain, India and the United States. Princeton, NJ: Princeton University Press.
  • Clagett, W., Flanigan, W. & Zingale, N. (1984). Nationalization of the American electorate. American Political Science Review 78(1): 7791.
  • Cox, G. (1997). Making votes count. Cambridge: Cambridge University Press.
  • Cutler, F. (2008). One voter, two first-order elections? Electoral Studies 27(3): 492504.
  • Deschouwer, K. (2003). Political parties in multi-layered systems. European Urban and Regional Studies 10(3): 213226.
  • Deschouwer, K. (2009). Towards regionalization of statewide electoral trends in decentralized states? The cases of Belgium and Spain. In W. Swenden & B. Maddens (eds), Territorial party politics in Western Europe. Houndmills: Palgrave Macmillan.
  • De Winter, L., Gomez-Reino, M. & Lynch, P. (2006). Autonomist parties in Europe: Identity politics and the revival of the territorial cleavage. Barcelona: Institut de Ciéncies Politiques i Socials.
  • Dupoirier, E. (2004). La régionalisation des élections régionales? Un modèle d'interprétation des élections régionales en France. Revue Française de Science Politique 54(4): 571594.
  • Ebeid, M. & Rodden, J. (2006). Economic geography and economic voting: Evidence from the US states. British Journal of Political Science 36(3): 527547.
  • Fitjar, R.D. (2010). The rise of regionalism: Causes of regional mobilization in Western Europe. London: Routledge.
  • Floridia, A. (2010). The regional elections of 2010: Much ado about nothing? Bulletin of Italian Politics 2(1): 137145.
  • Gibson, E.L. & Suarez-Cao, J. (2010). Federalized party systems and subnational party competition: Theory and an empirical application to Argentina. Comparative Politics 24(1): 2139.
  • Harbers, I. (2010). Decentralization and the development of nationalized party systems in new democracies: Evidence from Latin America. Comparative Political Studies 43(5): 606627.
  • Hooghe, L., Marks, G. & Schakel, A.H. (2010). The rise of regional authority: A comparative study of 42 democracies. London: Routledge.
  • Hopkin, J. (2009). Party matters: Devolution and party politics in Britain and Spain. Party Politics 15(2): 179198.
  • Hough, D. & Jeffery, C. (eds) (2006). Devolution and electoral politics. Manchester: Manchester University Press.
  • Jeffery, C. & Hough, D. (2009). Understanding post-devolution elections in Scotland and Wales in comparative perspective. Party Politics 15(2): 219240.
  • Jeffery, C. & Wincott, D. (2010). The challenge of territorial politics: Beyond methodological nationalism. In C. Hay (ed.), New directions in political science. Basingstoke: Palgrave Macmillan.
  • Johns, R. et al. (2010). The Holyrood elections 2007: Explaining the result. Available online at: http://www.scottishelectionstudy.org.uk/paperspubs.htm
  • Johnston, R. (1980). Federal and provincial voting. In D.J. Elkins & R. Simeon (eds), Small worlds: Provinces and parties in Canadian political life. Toronto: Methuen.
  • Jones, M.P. & Mainwairing, S. (2003). The nationalization of parties and party systems: An empirical measure and an application to the Americas. Party Politics 9(2): 139166.
  • Katz, R.S. (1973). The attribution of variance in electoral returns: An alternative measurement technique. American Political Science Review 67(3): 817828.
  • Lago-Peñas, I. & Lago-Peñas, S. (2010). Decentralization and the nationalization of party systems. Environment and Planning C: Government and Policy 29(2): 244263.
  • Maddens, B. & Libbrecht, L. (2009). How statewide parties cope with the regionalist issue: The case of Spain – a directional approach. In W. Swenden & B. Maddens (eds), Territorial party politics in Western Europe. Houndmills: Palgrave Macmillan.
  • Massetti, E. (2009). Regionalist party positioning in the multi-dimensional ideological space: A framework for analysis. Regional and Federal Studies 19(4–5): 501531.
  • Massetti, E. & Schakel, A.H. (2011). Decentralization and Regionalist Parties' Electoral Fortunes: Unpacking a Complex Relationship. Paper presented at the sixth general conference of the European Consortium of Political Research, Reykjavik, 2527 August.
  • Meguid, B. (2010) Party competition between unequals: Strategies and electoral fortunes in Western Europe. Cambridge: Cambridge University Press.
  • Morgenstern, S. & Swindle, S.M. (2005). Are politics local? An analysis on voting patterns in 23 democracies. Comparative Political Studies 38(2): 143170.
  • Morgenstern, S., Swindle, S.M. & Castagnola, A. (2009). Party nationalization and institutions. Journal of Politics 71(4): 13221341.
  • Pallarés, F. & Keating, M. (2003). Multi-level electoral competition: Regional elections and party systems in Spain. European Urban and Regional Studies 10(3): 239255.
  • Reif, K, & Schmitt, H. (1980). Nine second-order national elections: A conceptual framework for the analysis of European election results. European Journal of Political Research 8(1): 344.
  • Rose, R. & Urwin, D.W. (1975). Persistence and change in Western party systems since 1945. Political Studies 18(3): 287319.
  • Schakel, A.H. (2010). Regional election database, University of Edinburgh. Available online at: http://www.arjanschakel.nl
  • Schakel, A.H. (2011). Congruence between regional and national elections. Comparative Political Studies published online 18 October 2011, doi: 10.1177/0010414011424112.
  • Schakel, A.H. & Jeffery, C. (2012). Are regional elections really second-order elections? Regional Studies, published online 15 June 2012, doi:10.1080/00343404.2012.690069.
  • Stokes, D. (1967). Parties and the nationalization of electoral forces. In W.N. Chambers & W.D. Brunham (eds), American party systems: Stages of political development. New York: Oxford University Press.
  • Suarez-Cao, J. & Freidenberg, F. (2010). Multilevel Party Systems and Democracy: A New Typology of Parties and Party Systems in Latin America. Paper presented at the annual meeting of the American Political Science Association, Washington, DC.
  • Swenden, W. (2006). Federalism and regionalism in Western Europe: A comparative and thematic analysis. Houndmills: Palgrave MacMillan.
  • Swenden, W. & Maddens, B. (eds) (2009). Territorial party politics in Western Europe. Houndmills: Palgrave Macmillan.
  • Thorlakson, L. (2000). Government building and political development in federations: Applying Canadian theory to the German case. Regional & Federal Studies 10(3): 129148.
  • Thorlakson, L. (2007). An institutional explanation of party system congruence: Evidence from six federations. European Journal of Political Research 46(1): 6995.
  • Thorlakson, L. (2009). Patterns of party integration, influence and autonomy in seven federations. Party Politics 15(2): 157177.
  • Tronconi, F. & Roux, C. (2009). The political systems of Italian regions between statewide logics and increasing differentiation. Modern Italy 14(2): 151166.
  • Völkl, K. et al. (2008). Wähler und Landtagswahlen in der Bundesrepublik Deutschland. Baden-baden: Nomos.