Towards a scholarship on regional elections

ABSTRACT

This introduction article inaugurates the first annual special issue on regional elections which sets out to nurture a community of regional election scholars who seek to understand the factors driving regional voters, regional election outcomes, and regional electoral dynamics across the globe. This is a much-needed research effort given that the territorial scope and importance of regional elections have increased considerably over the past four decades yet most scholarship remains focused on national elections. Collectively the articles and reports presented in this first special issue enable us to distil three contributions to the scholarship on regional elections: a refinement on the scope conditions of the second-order election model; the introduction of the concept of barometer elections; and the insight that each regional (and national) election is regionalized or nationalized to a certain extent, either in outcome and/or in relation to the processes that sustain the outcome.

KEYWORDS:

• Regional elections
• methodological nationalism
• barometer election
• second-order election

Introduction

This is the first annual special issue on regional elections. The main objective of the annual special issue is to systematically and comparatively report on regional elections across the globe. Such an endeavour is much needed since regional elections are increasingly common but electoral scholarship is mainly focused on national elections. National elections are regularly surveyed through national election studies in most countries, they are systematically documented through initiatives such as the Political Data Yearbook produced by the European Journal of Political Research,¹ and they are thoroughly analysed in academic journals such as Electoral Studies, Journal of Elections, Public Opinion and Parties, Party Politics, and Representation. During the past years, election scholars have observed a rise of regional elections and they have surveyed regional voters, published election reports, and analysed regional elections in several journals, most notably in Regional and Federal Studies. However, these studies appear in a rather ad hoc manner and they tend to focus on the ‘usual suspects’ such as Catalonia and the Basque Country in Spain and Scotland and Wales in the United Kingdom. Although regional voters are regularly surveyed in the German Länder and Spanish comunidades autónomas, analyses of regional elections in these countries are often written in German and Spanish which limits their impact among a global (predominantly English-speaking) audience. Conversely, election surveys in English, as is often the case with Scottish and Welsh election studies, habitually, compare with election outcomes in other regions within the same country. An increasing number of studies look at subnational elections taking place in autocratic regimes or new democracies (e.g. Mexico, Indonesia, and Russia) but these often employ country-specific analytical frameworks which do not lend themselves for comparison and developing and testing theories. The Regional and Federal Studies annual special issue on regional elections arises from a desire to remedy these caveats and it seeks to nurture a community of regional election scholars who seek to understand the factors driving regional voters, regional election outcomes, and regional electoral dynamics across the globe.

All systematic research begins with good description (Landman 2000, 5). Descriptive studies serve as the raw data for comparative studies that aspire to higher levels of explanation. Once phenomena have been described, the comparativist can move on to search for those factors that may help explain what has been described (Landman 2000, 6). Regional and Federal Studies will publish a special issue on annual basis which contains articles and reports on regional elections. Both are country-specific and contain relatively much description on regional electoral institutions but also on regional government. This provides the context in which regional elections are conducted and which is often (partly) unknown to most election scholars. Authors of the reports and articles published in the annual special issue will also provide regional election results in a standardized excel format which will be published on the website of Arjan H. Schakel, one of the annual special issue editors.² In this way we strive to systematically and regularly collect regional election data over time and to make the data available to the whole scholarly community.

There are important differences between election reports and election articles. Election reports are shorter (about 4000 words) and remain largely restricted to descriptive analysis with an emphasis on what stands out in a particular election, region, or country in comparison to other elections, regions, or countries. Election articles are twice as long (up to 8500 words) and, in addition to describing the results of recent elections, they also seek to increase our understanding of these outcomes. Often this requires taking a (much) longer time period into consideration and the inclusion of regional elections from several decades into the analysis. Articles focus on regionalized and federal countries with powerful regions and with a longer tradition of a scholarship on regional elections that is regularly country-specific. Reports concentrate on elections held in regions in unitary countries which often have fixed election cycles and which are generally less powerful than their peers in regionalized and federal countries. Finally, and perhaps most importantly, election articles go through an anonymized peer review process whereas election reports are reviewed by the editors of the annual special issue.

This first special issue features four election articles and four election reports. The election articles cover elections held for 6 states and 2 territories in Australia, 9 states (kililoch) in Ethiopia, 16 German Länder, and 15 ordinary statute regions (regione a statuto ordinare), 4 special statute regions (regione a statuto speciale), and 2 special statute provinces (province a statuto speciale) in Italy. The election reports cover elections held in 13 regions (kraje) in the Czech Republic, 13 régions in France, 81 provinces (mga lalawigan) in the Philippines, and 41 counties (județe) in Romania. This selection of countries is not random. Considering that most regions and countries have four-year election cycles, we have set up four special issues whereby we have tried to find an optimal balance between a wide comparative scope and coverage soon after the elections have been held. A second annual special issue is in preparation and a call for election articles and election reports for the third annual special issue will be issued in September 2018.³ After the fourth special issue (to be published in 2020 or 2021) we will cover new rounds of elections held in countries included in this first annual special issue. In this way, we hope to ensure a systematic and regular coverage of regional elections as well as to nourish advancement in regional election scholarship since authors can take insights from previous special issues on board in their analyses.

One additional issue concerning the scope of the annual special issue on regional elections needs to be discussed. The focus is on elections held in regions and this requires us to define regional government. We follow the definition of Hooghe et al. (2016) who define regional government as an intermediate tier between the lowest, local, and national government. When countries have more than one such intermediate tier, we focus on the most authoritative tier. However, this does not mean that other types of election are not considered. On the contrary, national elections are often incorporated in election articles for analytical reasons. This requires authors to compare regional to national elections as is in the case of, for example, the second-order election model and the barometer election model (these two models are discussed in detail below). Election reports will (partly) cover other subnational and national elections especially when they are held simultaneously with regional elections, but authors may also decide to include other types of election to highlight electoral trends over time.

In the introduction to each annual special issue we draw comparative lessons and discuss the implications of the findings for our understanding of regional elections. The election articles and reports are country and region-specific but we need comparison to develop and advance theory on regional elections. Once regional elections have been described, the comparativist can move on to search for those factors that may help explain what has been described (Landman 2000, 6). Comparison allows for testing hypothesized empirical relationships among variables (Lijphart 1975). In turn, this enables predictions about election outcomes in other countries and regions based on generalizations from the initial comparison (Caramani 2008, 4–5; Landman 2000, 10). This first special issue starts the comparative exercise and in subsequent special issues we hope to accumulate and advance further knowledge on regional elections. Before we delve into the main findings of this first annual special issue we discuss in more depth the research rationale and need for developing a regional election scholarship.

Addressing methodological nationalism in electoral research

The territorial scope and importance of regional elections have increased tremendously over the past four decades. Both trends are systematically traced for three world regions – America, Asia, and Europe – by the regional authority index (RAI). The RAI tracks the authority (powers) of regional governments within 81 countries since 1950. Figure 1 displays the number of citizens (in 100 million) that are subject to regional government with directly elected assemblies or parliaments.⁴ Four lines are shown: one for America, one for Asia, one for Europe, and one for the total number of citizens subject to regional government.⁵ The difference between the sum of the lines for the three world regions and the total line indicates the number of citizens which are subject to regional government but without opportunity to directly elect their regional representatives. The lines take into account when direct elections for regional government are held in part of the country.

Figure 1. Territorial coverage of directly elected regional assemblies in 81 countries clustered in three world regions.

In 1950, the number of citizens subject to directly elected regional representatives was 750 hundred million in America, 427 hundred million in Asia, and 317 hundred million in Europe.⁶ Together, they constituted about 1.5 billion people out of a total of almost 2.0 billion people in countries with regional government. Hence, 78 per cent of the population was governed by directly elected regional assemblies. By 2010, this number had risen to 2.0 billion people out of a total of almost 2.3 billion people −812 million people in America, 622 million people in Asia, and 672 million people in Europe – or 92 per cent of the total population. Hence, nearly all citizens who can vote in national and local elections can also cast a vote in regional elections.

The territorial scope or coverage of regional elections has increased considerably over the past decades. In addition, regional assemblies also have become more powerful. Figure 2 displays average regional authority index scores for the three world regions on the basis of scores for directly elected assemblies alone. In other words, a regional authority index score is only included when the regional assembly or parliament is directly elected.⁷ Region scores are weighted by regional population size when deriving a country score and country scores are averaged for the three world regions. Between 1950 and 2000 regional authority has definitely increased in Asia and Europe and has remained almost the same for America. Scores increased from 7.7 to 11.2 in Asia, from 7.3 to 9.2 in Europe, and from 6.7 to 6.9 in America.

Figure 2. Average regional authority index score for directly elected regional assemblies in 81 countries clustered in three world regions.

The lines in Figure 2 display some significant declines between the 1960s and 1980s for America and Asia, and in the early 1990s and late 2000s for Europe. The main cause is the inclusion of countries because they gain independence or countries which turn towards authoritarianism. For example, the inclusion of Jamaica (1962), Trinidad and Tobago (1962), Barbados (1966), Guyana (1966), Bahamas (1963), and Suriname (1975) can explain the decrease in regional authority for America between 1962 and 1975. Similarly, the inclusion of Singapore (1965) and Brunei (1984) and the ‘authoritarian turns’ in Indonesia (1974–2000), the Philippines (1979 and 1984) and in South Korea (1961–1990) can account for the decline in regional authority between 1965 and 1987 in Asia. And the inclusion of 14 post-communist countries (1989–1991) and the ‘authoritarian turn’ in Russia (2005) and the break-up of Serbia and Montenegro (2006) underlie the declines in regional authority observed for Europe. When one takes this into account, then the increase in regional authority becomes more apparent. Average regional authority increased from 3.7 in 1980 to 6.9 in 2010 in America, from 6.6 in 1984 to 11.2 in 2010 in Asia, and from 7.8 in 1971 to 10.3 in 1989 or 10.4 in 2006 in Europe.

An increase of two to four points on the Regional Authority Index reflects considerable change. For example, regions may have increased their policy scope and gained competences in cultural-educational and welfare policy in addition to economic policy (+2 points) and witnessed an increase in their tax and/or borrowing autonomy. For example, instead of having the authority to set the rate of a minor tax, regions may set also the base of a minor tax (+1 point) or determine the rate of personal income, corporate, value added, or sales tax (+2 points). A region’s borrowing autonomy may also increase and a region, instead of borrowing under strict rules and with prior authorization by the central government, may borrow with posterior central government approval (+1 point) or may borrow without centrally imposed restrictions (+2 points).

Regional elections are clearly on the rise across the globe, both in territorial scope and in the level of authority exercised by directly elected regional representatives. However, these trends have not spurred a thriving scholarship on regional elections. Charlie Jeffery and Daniel Wincott (2010) offer an explanation in their discussion on ‘methodological nationalism’ in election research. Methodological nationalism refers to the tendency of social scientists to presume that social and political life is naturally encompassed by nation-states and takes the nation-state as the ‘natural’ unit of analysis.⁸ Jeffery and Wincott (2010, 179) note that social-scientific electoral analysis is mostly concerned with statewide elections and overlook ‘the possibility that elections to sub-state parliaments might be, or have become, an arena for patterns of voting behavior that diverge from “nationalized” patterns of statewide voting’. Theories on elections and voting behaviour do not require election scholars to test their hypotheses against regional election data, hence they do not look at regional elections considering that collecting data on regional elections is more cumbersome than for national elections.

The critique of methodological nationalism not only concerns empirical data, it also involves theory. Scholars who look at regional elections tend to use theoretical frameworks which incorporate the presumption of supremacy of national politics. This is clearly evident with regional election research which sets out to test the second-order election model (Schakel and Jeffery 2013). This model assumes that non-national, non-statewide elections are subordinate to national or statewide elections because there is less ‘at stake’ in second-order elections compared to first-order national elections (Reif and Schmitt 1980). The second-order election model assumes that voter behaviour is driven by evaluations of national politics, issues, and politicians. As a result, voter turnout in regional elections is low because these elections are conceived as not or less important than national elections. Voters who go to the ballot box tend to be dissatisfied and use regional elections as a means to signal their discontent with national politicians and parties. In sum, regional elections are either not ‘on the radar’ of election scholars because theory draws their attention to statewide elections or scholars who look at regional elections tend to employ theoretical perspectives which underline the importance of national politics.

The upshot is that there might be a ‘self-fulling prophesy’ at play: the second-order election model ‘imports the assumption that other electoral competitions are subordinate to statewide politics, [and thereby] pre-empts alternative possibilities’ (Jeffery and Wincott 2010, 179). Theory which presumes a supremacy of national politics tends to be tested against nationally aggregated data which subsequently drives theoretical innovation on electoral dynamics in statewide or national elections. As a result, scholarship on elections finds itself in a reinforcing cycle whereby potential ‘territorial effects’ are defined out of existence (Keating 2009). A similar ‘self-fulfilling prophesy’ can be observed in studies of nationalization which set out to trace the homogenization of electoral results across the state. For example, Caramani (2004, 291) concludes that there is no evidence for a regionalization of the vote despite a strong trend of institutional decentralization. However, his seminal study is based on statewide elections only and may have overlooked significant processes of regionalization in regional elections (Schakel 2013a, 2013b). The critique of methodological nationalism invites election scholars to address regional elections on their own terms and this ‘might generate a different or at least more nuanced picture’ (Jeffery and Wincott 2010, 179). But should regional election theory require more nuance or does it need to be different from national election theory? And if so, how would such a theory look like?

The literature offers two alternative theoretical approaches to regional elections (Jeffery and Hough 2009; Jeffery and Wincott 2010; Schakel and Dandoy 2013a; Schakel and Dandoy 2017; Swenden and Maddens 2009a). The first looks at the institutional setting and, in particular, at the functional responsibilities assigned to regional government. Voters differentiate between national and regional elections based on variations in regional policy autonomy. Decentralization of authority gives parties and voters incentives and opportunities to mobilize and respond to locally defined issues (Thorlakson 2007). Voters may understand that the regional level has independent policy-making capacity and may vote according to their evaluation of the performance of regional (instead of national) government. In turn, this creates an incentive for regional parties to adapt their policies to the regional electoral arena when it is politically opportune to deviate from statewide party policies (Maddens and Libbrecht 2009; Swenden and Maddens 2009b).

The second approach focuses on voting according to regional specific preferences which lead to a different vote choice in regional elections than for national elections. Specific regional interests can be expected when there are territorial cleavages based on, for example, history, language, and ethnicity. The basis of territorial cleavage theory lies in sociological approaches which explain dissimilarity between regional party systems by the extent to which territorial cleavages are politicized (Hearl, Budge, and Pearson 1996; Rokkan and Urwin 1983). Evidence from voter surveys held in Scotland and Wales clearly reveal a pattern of multilevel or dual voting based on territorial interests and spurred by regionalist parties such as Plaid Cymru and the Scottish National Party (Scully, Jones, and Trystan 2004; Trystan, Scully, and Wyn Jones 2003; Wyn Jones and Scully 2006). Territorial cleavages also drive regional voting in autocratic regimes or new democracies as, for example, is shown in the case of Ukraine (Henderson and Romanova 2016).

There is now a large and still accumulating body of evidence that increasing regional authority and strong regional identities decrease the nationalization of regional elections. Regional party systems become increasingly dissimilar from national party systems and the second-order nature of regional elections declines to the extent regional government becomes more powerful and/or regional cleavages are present and politically mobilized (Dandoy and Schakel 2013; Hough and Jeffery 2006; Romanova 2014; Schakel 2017). This body of literature thereby clearly depicts a ‘more nuanced picture’ of nationalization theory and the second-order election model. For example, the second-order election model expects that second-order election effects – lower turnout, vote share losses for parties in national government, and vote share gains for opposition, new, and small parties – relate to the timing of the regional election in the national election cycle whereby second-order election effects are largest around mid-term and tend to decline to the extent that a regional election is held closer to the preceding or forthcoming national election. The literature on regional elections reveals an additional important electoral cycle effect namely the timing of the regional election vis-à-vis other subnational elections. For example, horizontal simultaneity among regional elections tends to increase second-order election effects whereas vertical simultaneity with local elections tends to decrease second-order election effects. This example of electoral timing reveals that a comprehensive understanding of regional electoral dynamics may require the development of a ‘different picture’.

Remedying methodological nationalism in electoral research calls for conceptualizing and analysing party systems as multilevel electoral systems whereby spill-over between electoral arenas may go top-down from the national to regional and local levels but also bottom-up from the local to regional and national levels as well as horizontally between electoral arenas within a country but also transnationally between countries (Schakel 2018; Swenden and Maddens 2009a). Such a conceptual framework will increase our understanding of the causes as well as the effects of regional electoral dynamics. For example, analysing multilevel electoral dynamics within and across countries may draw our attention to the impact that regional elections may have on political regimes and democracy (see e.g. Romanova 2013; Ross 2011a, 2011b). With the annual special issue on regional elections we would like to identify the ‘building blocks’ which enable us to develop a ‘different picture’ on understanding the causes and effects of regional electoral dynamics and with this first issue we hope to make a useful start.

Comparative lessons of the first annual special issue on regional elections

In the remainder of this introduction, we draw some general observations from the election articles and election reports and we discuss the implications of these observations for our understanding of regional elections. In how far do the articles and reports provide evidence for a more ‘nuanced or different picture’ of regional elections? We identify three major analytical contributions: (1) a refined specification of the scope conditions of second-order elections, (2) the introduction of the concept of barometer elections, and (3) the insight that all regional elections are regionalized and nationalized to a certain extent, either in outcome and/or in relation to the processes that sustain the election outcomes.

The scope conditions of second-order elections

The articles by Jochen Müller and Emanuele Massetti on regional elections in respectively Germany and Italy enlarge our understanding of the scope conditions under which regional elections tend to be treated by voters as second-order elections. Müller teases out the impact of party identification of voters on second-order voting whereas Massetti’s longitudinal analysis draws our attention to (high) turnout, horizontal simultaneity and (low) regional authority as factors fostering a second-order status of regional elections.

Müller exploits a novel, innovative, and impressive dataset of 139 election surveys for Land elections held between 1978 and 2016 which includes more than 137,000 voters.⁹ This enables a much-needed exploration on the voter-level factors that drive second-order election voting and Müller presents two important results. First, the impact of party identification on casting a vote in Land elections follows a cyclical character. The likelihood that government party identifiers vote for a party in national government and that opposition party identifiers vote for the party in national opposition is lower around mid-term of the federal election cycle. In other words, ‘defection’ by voters from their first preference or sincere voting is more likely when a Land election is held around two years after the previously held federal election. Second, third-party identifiers – defined as voters that identify with a party that is not in federal government or opposition – are more likely to vote for their preferred party around mid-term. These two findings highlight the importance of party identification as a modifying variable for second-order voting.

Massetti finds that elections held in the 15 ordinary regions during the First Republic (1946–1992) can be conceived as ‘barometer elections’, i.e. ‘reliable opinion polls on national politics’ whereas during the Second Republic (since 1993) the barometer status of regional elections held in the ordinary status regions becomes less reliable. During the Second Republic, only the 2000 and 2005 rounds of elections appear to have a barometer status whereas second-order election effects are more visible in 1995, 2008–2010, and 2012–2015 election rounds. Massetti attributes this change mainly to lower levels of turnout when compared to previously held national elections and to regional elections held during the First Republic. But other factors may also have an impact because, since 2010, horizontal simultaneity has largely disappeared and decentralization reforms have significantly increased the powers of regional government during the 2000s. Interestingly, the establishment of a directly elected regional president and the majority bonus of seats for the coalition of parties that support the winning candidate for the regional presidency does not appear to have significantly increased the second-order status of regional elections. A relatively well-established finding in the literature on second-order elections is that second-order effects tend to be larger in bipolar party systems (Marsh 1998; Schakel and Dandoy 2013b). The upshot is that Massetti concludes that regional elections held in ordinary regions seem to ‘float’ in between a second-order and barometer status.

Barometer elections

The regional election reports written by Dragos Dragoman and Andreea Zamfira on județe elections in Romania and by Michal Pink and Otto Eibl on kraje elections in the Czech Republic also provide evidence that regional elections can, in some cases, be conceived as barometer elections.¹⁰ It might therefore be useful to discuss the characteristics and causes for a barometer status of regional elections. These elections do not exhibit second-order election effects because parties in national government do not systematically lose vote share and parties in national opposition and small and new parties do not consistently win vote share. Barometer elections can still be conceived as nationalized elections since regional election results do reflect the popularity of statewide parties based on their (perceived) performance in the national electoral arena. To our knowledge, the term barometer election was first coined by Anderson and Ward (1996, 447) who define this type of elections as ‘elections that reflect changes in citizens’ attitudes toward the government in response to changing political and economic conditions, absent the opportunity to install a new [national] executive.’ This definition resembles the characterization of second-order elections which also pertain to voting in regional elections based on the popularity of parties in national government despite the fact that national representatives will not be elected. However, Dragoman and Zamfira, Pink and Eibl, and Massetti, actually use the term barometer election (or litmus test) to categorize a type of regional election which is different from a second-order election. The main difference concerns the observed election outcomes, in second-order elections the outcome can be predicted: parties in national government lose vote share whereas opposition, small and new parties win vote share. In contrast, the outcome of barometer elections themselves is difficult to predict but these regional elections may function as reliable predictors of the outcomes of forthcoming national elections.

The introduction of the category of barometer elections is interesting because it points out that there are three forms of nationalization of regional elections (instead of two as defined by Schakel and Dandoy 2013b): nationalized, second-order and barometer elections. These types of elections share the feature that national politics spills-over into the regional electoral arena and (partly) drives regional election outcomes. They differ with regard to the election outcomes: nationalized regional elections can be recognized by highly similar regional and national party systems, second-order elections lead to dissimilar regional and national party systems according to the magnitude of second-order election effects, and barometer elections have similar outcomes as in the forthcoming national election but they may be dissimilar from previously held national (and regional) elections according to the extent of volatility observed for the statewide electorate.

The main factor that may be favourable for barometer elections is, according to Massetti, similar levels of turnout in regional and national elections as is often the case with compulsory voting (which was applied in Italy during the First Republic). Dragoman and Zamfira attribute the barometer status of Romanian județe elections mainly to the close timing of these regional elections to the forthcoming national elections six months later but also to similar electoral systems employed for regional and national elections whereby the constituencies are the same (the 41 județe) as well as the rules translating votes into seats (proportional rule). The case of kraje elections in the Czech Republic leads us to single out the timing of the regional election relative to the forthcoming national election as the probably most important factor inducing a barometer status. Pink and Eibl find that the 2016 round of regional elections, in contrast to previous rounds of kraje elections, can be conceived of ‘as a kind of litmus test’ which indicate ‘trends in electoral support nationwide’. What has also changed between the latest and previous rounds of kraje elections is that the former rounds were held around mid-term of the national election cycle whereas the 2016 round was held when the national government was in its third year in office. Apart from the previously mentioned factors we would also like to add two probable factors which are favourable for inducing a barometer status. These are similar levels of (limited) autonomy and horizontal simultaneity among regional elections; both factors are present for kraje elections in the Czech Republic, județe elections in Romania, and elections held in ordinary statute regions in Italy during the First Republic.

Regionalized or nationalized regional elections?

On the basis of a comparison of regional election outcomes in 13 West European countries, Schakel and Dandoy (2013b, 277–278) observe that regionalized regional elections are characterized by similar or higher levels of turnout compared to national elections, minimal second-order election effects, and electorally strong non-statewide parties with vote shares close to or above 30 per cent. This is the case for regional elections held in Belgium, Faroe Island and Greenland in Denmark, special statute regions in Italy, historic communities in Spain, and devolved elections in the United Kingdom. As a result, these regional party systems tend to be quite different from the statewide party system, but voters tend not to switch between their regional and national vote. In other words, party systems are incongruent whereas elections (outcomes) are congruent. This classification presumes that regionalized regional elections can be differentiated from nationalized regional elections by carefully comparing aggregate level outcomes across the types of election. The election article on Ethiopia and the election report on the Philippines written respectively by Ayele Zemelak and Julio Teehanke cast new light on how regional elections can be conceived as being regionalized and nationalized at the same time. Zemelak and Teehanke reveal that for the question whether regional elections are regionalized or nationalized, one should also take into account informal political processes that underlie aggregate level outcomes, at the very least in the case of less democratic countries.

Teehanke reports that in the Philippines since 1987, on average almost 34 per cent of all lower house representatives have switched party. About 60 per cent of the party switchers moved allegiance towards the party of President in office. The main reason for this extensive party switching is because the Philippine Congress (lower house of parliament) functions as a nexus between local and national elites and switching to the party of the winning presidential candidate increases the chances for Congressional members to draw perks towards their province. The consequence is that after a presidential election a significant nationalization trend can be observed because of extensive party switching. However, the underlying process sustaining the nationalization trend has a clear regional dimension. Regionalization becomes less visible in aggregate level outcomes because it moves from being an inter-party to an intra-party phenomenon. The election report on the Philippines also underlines the impact that (a lack of) party (system) institutionalization can have on the extent of regionalization or nationalization of the vote (Mainwaring and Scully 1995).¹¹ Zemelak describes how in Ethiopia – a federation that used to be a unitary state before 1995 – intra-party and inter-party processes at the national level also sustain a regionalization of elections. The Ethiopian Peoples’ Revolutionary Democratic Front (EPRDF) is a coalition of four ethnic parties of which each has an electoral stronghold in one of the four most populous regions which together comprise more than 70 per cent of the total Ethiopian population. The other five regions are represented by ethnic parties affiliated to the EPRDF and collectively these parties control all seats of all national and regional parliaments. Zemelak reports that this ruling coalition produces a de facto one-party system which manages to dominate all parliaments through a long ‘menu of institutional manipulations’. For both Ethiopia and the Philippines, one could argue that elections are regionalized and nationalized at the same time because regional electoral strongholds are sustained by processes taking place at the national level which reinforce the position of regional elites.

The insight that regional elections can be regionalized and nationalized at the same time may cast a different light on elections which are conceived to be highly nationalized. Bolwell and Eccleston report a comparatively unusual high congruence between state and federal elections held in Australia whereby more than 90 per cent of the vote share is similarly cast in the two types of election. However, that does not mean that regional politics is absent. First, Bolwell and Eccleston conclude that

elections in Australia are highly nationalised in the sense that voting patterns between electorates, elections and party shares are all similar to federal patterns. While conversely it might be observed that national elections are highly regionalised for the same reason, the distinction is pointless; elections at both tiers have similar characteristics – voters tend to vote the same way whether nationally or sub-nationally.¹²

Second, Bolwell and Eccleston also report that in September 2017, the Western Australian Liberal Party voted to examine if Western Australia could become financially independent. The vote followed after a heated debate on secession which was triggered by resentment with Western Australia’s declining share of revenue from the federal Goods and Services Tax. The secession of Western Australia has been on the political agenda several times over the past century – in 1900, 1933, and during the 1970s – and thereby a certain degree of regionalism has never been absent in Australian elections.

The case of Australia shows that all regional (and national) elections exhibit some degree of regionalization (and nationalization) of the vote even in the most (in)congruent multilevel party systems. The extent of regionalization and the balance between nationalization and regionalization differs across regions, countries, and type of election but, most importantly, a certain degree of regionalization is always present. Territoriality in the vote also plays a significant role in France. In his report on the French regional elections of 2015, James Shields explains why the Front National (FN) could be the winner of the elections in the first round but the loser in the second round. In the first round, the FN came first nationally with 27.7 per cent of the vote, ahead of the Union of the Right (UR) with 26.6 per cent and the Union of the Left (UL) with 23.1 per cent. The FN was ranked the largest party in 6 out of 12 regions (Corsica excluded), was the second largest party in 2 regions, and the third largest party in the remaining 4 regions. In the second round, the picture was completely different, the FN was the third largest party in 8 regions, the second largest party in 4 regions, and the FN did not manage to become the largest in any of the regions. The difference between the first and second round becomes even more surprising when one takes into account that the number of votes for the FN increased from 6.0 to 6.8 million and that the party polled similar vote shares nationally, 27.7 per cent in the first round and 27.1 in the second round. What happened in between the two rounds was that there was a collective effort by voters and statewide parties to prevent the FN from winning a majority of the vote that would grant it control of any regional council. Shields notes that turnout increased considerably between the two rounds – from 49.9 per cent in the first round to 58.4 per cent in the second round – most notably in regions where the FN had won the first round and looked strongly placed for the second round. Another important factor mentioned by Shields is concerted rather than divided opposition which even led to the withdrawal of the UL from the second round in two regions where the FN won more than 40 per cent of the vote. By withdrawing, the UL made it easier for the UR to win the second round against the FN.

The electoral system obviously plays an important role in an explanation of how the FN became a winner in the first round but turned into a loser in the second round.¹³ Territorial heterogeneity of the vote also plays a significant role. The FN, UR, and UL all tend to have their regional strongholds but the vote shares for UR and UL are more equally distributed than for the FN. The coefficient of variation is a measure of territorial heterogeneity of vote shares and it is derived by dividing the standard deviation of regional vote shares by the average vote share across regions. In the first round, the coefficients of variation were 0.27 for the FN, 0.15 for the UR, and 0.23 for the UL.¹⁴ In the second round, and not taking into account the two regions where the UL withdrew from the electoral competition, the coefficients of variation were 0.34 for the FN, 0.26 for the UR, and 0.25 for the UL. In other words, the territorial heterogeneity in vote shares was higher for the FN than for the UR or UL in both rounds of elections. The UR and UL have greater and more uniform territorial coverage which may put them at an advantage in mobilizing electoral support to keep the FN out of regional office. In contrast, the FN has particular regional strongholds which probably depend more on the perceived attractiveness of the head of the regional lists than on a strong and deeply entrenched party organization. Shields reports that the FN received more than 40 per cent of the votes in two regions where Marine Le Pen, the national leader of the FN, and Marion Maréchal-Le Pen, the niece of Marine Le Pen, headed the regional lists of the FN. It would be interesting to explore in further depth the hypothesis that party organization can explain when and where parties win elections. For example, it may explain the territorial heterogeneity in regional vote shares observed for the centre-left, centre-right, and Movimento Cinque Stelle (M5S) in Italy. Similar to the French case, the centre-left and centre-right managed to win all the regional presidencies despite the fact that M5S managed to become the most voted party in one region and the second most voted party in seven regions (Massetti, this issue). Also similar to the French case, all regions in Italy except for one employ majority bonuses for the coalition/party that supports the winning regional presidential candidate.

Conclusion

The annual special issue on regional elections aims to advance regional election scholarship by seeking to understand the factors driving regional voters, regional election outcomes, and regional electoral dynamics across the globe. This is much-needed effort given that the territorial scope and importance of regional elections have increased considerably over the past four decades. In 2010, 92 per cent of the population in 29 American countries, 11 Asian countries and 41 European countries with a total of about 2.3 billion people are ruled by directly elected regional representatives. Many regions can tax corporate and personal income, can borrow freely on the financial market, and have competencies in key policies such as education, health, culture, and welfare. Despite the increasing importance of regional elections, scholarship on explaining regional voting behaviour and regional election outcomes has relatively been lacking behind. This can be largely explained by the presence of methodological nationalism in election research, that is, the tendency of election scholars to focus on statewide elections both when they develop theory and when they test that theory. This leads them to overlook regional dynamics or regional effects. The annual special issue on regional elections seeks to remedy this caveat in election research and this first issue marks the beginning of this effort. The election articles and election reports of this special issue enable us to distil three contributions to the scholarship on regional elections: a refinement on the scope conditions of the second-order election model; the introduction of the concept of barometer elections; and the insight that each regional (and national) election is regionalized or nationalized to a certain extent, either in outcome and/or in relation to the processes that sustain the outcome. We envisage to further develop the scholarship on regional elections in subsequent annual special issues.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 http://www.politicaldatayearbook.com

2 http://www.arjanschakel.nl

3 The second annual special issue will -depending on the outcomes of the anonymized peer review process- include election articles on Canada, India, Kenya, Russia, Switzerland, and the United Kingdom and election reports on Croatia, Portugal, and Slovakia. The call for papers for the third annual special issue will include election articles on Austria, Bosnia and Herzegovina, Brazil, Malaysia, Sri Lanka, and the United States and election reports on Denmark, Greece, Poland, South Korea, and Sweden.

4 A regional government is included when the score on assembly is a 2 (‘the region has a directly elected assembly’) according to the Regional Authority Index (Hooghe et al. 2016, 77). Population data refers to census data for 2000–2012 (Statoids, http://statoids.com, consulted on 20 March 2018).

5 America (29 countries): Argentina, Bahamas, Barbados, Belize, Bolivia, Brazil, Canada, Chile, Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Suriname, Trinidad and Tobago, United States, Uruguay, and Venezuela.

Asia (11 countries): Australia, Brunei, East Timor, Indonesia, Japan, Malaysia, New Zealand, Philippines, Singapore, South Korea, and Thailand.

Europe (41 countries): Albania, Austria, Belgium, Bosnia and Herzegovina, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Kosovo, Latvia, Lithuania, Luxemburg, Macedonia, Malta, Montenegro, Netherlands, Norway, Poland, Portugal, Romania, Russia, Serbia, Serbia and Montenegro, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey, and the United Kingdom.

6 The reported population data are taken from census held in the 2000s (see end note 4) and therefore do not take into account the population growth since 1950.

7 Some countries have two regional tiers with directly elected assemblies, for example, départements and régions in France, Kreise and Länder (and Bezirk elections in Bavaria) in Germany, Italian and Spanish provinces and regions, and counties and states in the United States. The regional authority index scores of all these regional tiers are included when calculating a country average.

8 The main point of reference for Jeffery and Wincott (2010) concerns scholarship on well-established democracies in Europe, however, Hughes and Sasse (2001, 8) also criticize the literatures on democratic state-building and on transitions to democracy for taking the nation-state as the main unit of analysis.

9 Müller has updated a dataset originally published by Völkl et al. (2008) but, importantly, Müller is the first to introduce the dataset to a non-German academic readership (but see Völkl 2016).

10 Other instances of barometer elections are iller (provincial) elections in Turkey (Massetti and Aksit 2017).

11 Teehanke writes that the Philippines formally restored democratic rule in 1987 after an authoritarian regime that was in place since 1972.

12 A similar conclusion is derived by Dandoy (2013, 65) who analysed regional elections in Belgium where regional and federal elections are also highly congruent: ‘However, one can barely talk of a nationalization of elections. On the contrary, the high election congruence may equally indicate that national elections are regionalized.’

13 Shields explains that any list winning an outright majority of votes in the first round secures 25 per cent of the seats, effectively guaranteeing a majority in the council, while the remainder seats are distributed proportionally above a 5 per cent threshold. In case no list wins a majority in the first round, a second round is held with the lists winning 10 per cent or more of the vote which can fuse with smaller parties having won over 5 per cent of the votes.

14 The calculations are based on the vote shares displayed in Table 1 in Shields (this issue).