Vertical linkages between regional and national electoral arenas and their impact on multilevel democracy

ABSTRACT

By adopting a multilevel system perspective, we derive three insights from the analyses on regional elections presented in this third annual review of regional elections. First, we argue that Poland, South Korea and the USA can be conceived as aggregated bipolar party systems where competition at the nationwide level revolves around two parties whereas single parties dominate in the regions. Second, drawing on the cases of Bosnia and Herzegovina, Brazil, Kenya, and Nigeria, we claim that the integration of multilevel party systems (also) depends on the extent of collaboration between parties from the national and regional electoral arenas. Third, significant vote-switching revealed by election surveys held in Denmark and Sweden poses a significant challenge for theories on regional voting. We conclude by considering the wider implications of these three insights for multilevel democracy.

KEYWORDS:

• Multilevel party system
• multilevel election system
• aggregated bipolar system
• vote switching
• multilevel democracy

Introduction

This third annual review of regional elections includes five election articles and four election reports. The election articles cover elections held in two entities and 10 cantons (kantoni/županije) in Bosnia and Herzegovina, in 26 states (unidades federativas) in Brazil, in 47 counties (kaunti) in Kenya, in 36 states (jiha/ȯra/ìpínlẹ`) in Nigeria, and in 50 states in the USA. The election reports discuss elections held in five regions in Denmark, in 16 voivodeships (województwo) in Poland, in nine provinces (gwangyeok-si) and eight metropolitan cities (do) in South Korea, and in 21 regions (regioner) in Sweden. The main objective of the annual review is to systematically and comparatively report on regional elections across the globe. This plus the previous two annual reviews collectively cover elections held in a total of 595 regions, counties, provinces, and states in 24 countries with a total population of more than 3 billion people (Schakel and Romanova 2018, 2019).¹

Through the accumulation of annual reviews, we seek to increase our understanding of the factors that drive regional voters, regional election outcomes, and regional electoral dynamics (Schakel and Romanova 2018, 233–236). Articles and reports analyse the outcomes of recently held regional elections by comparing them with previously held regional elections or with other types of elections such as presidential, local, or national assembly elections. In each introduction, we set out to draw comparative lessons and discuss the implications of the findings for our understanding of regional elections. More specifically, we adopt a multilevel system perspective while considering the collection of articles and reports featured in this third annual review of regional elections.

A burgeoning literature on elections, parties and voting in multilevel systems has produced several concepts that are often used interchangeably. We propose to differentiate between multilevel party systems, multilevel election systems, and multilevel electoral systems. The concepts help by clarifying what the prime unit of analysis is, i.e. parties, elections, or electoral rules. A multilevel party system brings together the statewide and a set of regional party systems and considers the territorial pervasiveness of parties (i.e. participation in elections across the territory) and the type of elections in which parties participate (i.e. participation in nationwide and/or regional elections) as well as the strength of the organizational linkages between the statewide and regional party branches (Swenden and Maddens 2009, 9).² A multilevel election system perspective brings together nationwide and a set of regional elections and considers the vertical and horizontal interactions between and the integration of national and regional electoral arenas (Schakel 2018, 691). A multilevel electoral system is the set of rules employed in the regions and at the national level to conduct elections and to translate the number of votes into a number of seats in an assembly or parliament (Schakel and Romanova 2019, 298).³ In this introduction we identify three insights that surface when one adopts a multilevel system perspective, two insights relate to the study of multilevel party systems and a third insight concerns research on multilevel election systems.

First, we conceptually distinguish between aggregated and replicated polarized biparty systems. Competition at the national level in both types of party systems revolves around two parties that are split across an ideological divide whereby both win equally sized vote shares often close to 50 per cent. A polarized biparty system at the national level can be the result of either similar party systems at the subnational level (replicated) or of predominant party systems in the regions whereby one party dominates in one half whereas another party dominates in the other half of the regions (aggregated). In the first section, we discuss in more depth the aggregated polarized biparty systems in Poland, South Korea, and the USA and we discuss possible factors that sustain this kind of multilevel party systems.

We then proceed with the second insight which concerns the role that parties play in the horizontal and vertical integration of multilevel party systems. An emerging literature investigates the impact of internal party organizations on the integration of the multilevel party system. The cases of Bosnia and Herzegovina, Brazil, Kenya, and Nigeria reveal how the nature and extent of collaboration between parties from the national and regional electoral arenas should also be considered when one evaluates the level of integration of multilevel party systems.

Our third insight is based on election survey data which reveal significant vote switching (up to 30 per cent of voters) between national, regional, and local elections held in Denmark and Sweden. These are unitary countries where regions have not many competences (with the exception of health policy and hospitals) and where regional elections are held simultaneously with local elections (and also with national elections in Sweden). Both conditions should be favourable for uniform vote share swings, yet voters clearly make distinct choices. The presented election survey data poses a significant challenge for the assumptions regarding regional vote motivations upheld in the second-order election and nationalization of elections literatures.

In the discussion, we consider the wider implications of these three insights for electoral democracy. We evaluate how a multilevel system perspective sheds new light on how we may assess the role and contribution of regional elections in multilevel democracies. More specifically, the reports on Poland and South Korea and the article on the USA uncover that a highly competitive national electoral arena can be complemented by highly non-competitive subnational electoral arenas. In addition, the articles on Bosnia and Herzegovina, Brazil, Kenya, and Nigeria reveal that multilevel party collaboration has a significant impact on the consolidation of (electoral) democracy. Finally, the reports on Denmark and Sweden expose that regional elections may play a larger role in multilevel democracy than previously thought.

Aggregated polarized biparty systems

The election article on the USA and the election reports on Poland and South Korea describe the outcomes of nationalized regional elections held in polarized biparty systems. Following Sartori (2005, 73–75) we conceive of a polarized biparty system as the competition between two core parties (or two party alliances) that both win equally sized vote shares and which are split across an ideological divide. The competition between the Democratic and Republican parties in the USA is a classic example of a polarized biparty system. When one adopts a multilevel party system perspective, one may identify two different types of polarized biparty systems. First, a nationwide polarized biparty system can result from the accumulation of polarized biparty systems in most or all regions. In other word, similar polarized biparty systems can be found at the national level and in the regions across the statewide territory. Therefore, we use the label ‘replicated polarized biparty system’ to describe this kind of polarized biparty system. Second, a polarized biparty system at the nationwide level can also come about when two parties receive vote shares well beyond 50 per cent but one party wins those vote shares in one half whereas the other party earns its vote shares in the other half of the regions. In such a multilevel party system, one finds a polarized biparty system only at the national level whereas in the regions one finds predominant party systems, i.e. regions dominated by a single party (Sartori 2005, 173–177). This kind of multilevel party system we label as an ‘aggregated polarized biparty system’.

Aggregated polarized biparty systems can be found in the USA, South Korea, and to a lesser extent in Poland. Melusky and Richman (2020) observe that after the 2018 state elections in the USA, the Republican and Democratic Parties held state legislative seat shares that were more than 40 and less than 60 per cent in only 13 of the 50 states. In other words, in 37 states either the Democratic Party or the Republican Party holds more than 60 per cent of the seats. A trifecta is a situation whereby one party delivers the state governor and obtains an absolute majority of the seats in both the state legislative upper and lower chambers. Melusky and Richman (2020) note that after 2018 state elections, the Republican Party controls 22 trifectas and the Democratic Party holds 14 trifectas. Predominant party systems can be found in a stunning 36 out a total of 50 states whereby one single party controls the state legislative and state executive branches.

The contribution on South Korea (Jung and Kang 2020) illustrates how an aggregate polarized biparty system can transform within a couple of years into a multilevel predominant party system where a single party controls the legislative and executive branches at all tiers of government across the nationwide territory. In 2014, the liberal New Politics Alliance for Democracy won 9 out of 17 (53 per cent) metropolitan and provincial governors, 349 out of 849 (41 per cent) metropolitan and provincial council seats, 80 out of 226 (35 per cent) municipal mayors, and 1157 out of 2897 (40 per cent) municipal council seats (Jung and Kang 2020, Table 7). The respective wins for the conservative Saenuri Party were 8 out of 17 (47 per cent), 416 out of 849 (49 per cent), 151 out of 226 (52 per cent), and 1,413 out of 2897 (49 per cent) (Jung and Kang 2020, Table 7). Both parties had changed their names for the 2018 elections and the liberal Democratic Party of Korea (successor of the New Politics Alliance for Democracy) won 14 out of 17 (82 per cent) metropolitan and provincial governors, 652 out of 824 (79 per cent) metropolitan and provincial council seats, 151 out of 226 (67 per cent) municipal mayors, and 1639 out of 2926 (56 per cent) municipal council seats (Jung and Kang 2020, Table 7). The conservative Liberty Korea Party (successor of the Saenuri Party) suffered a dramatic loss and its respective wins were 2 out of 17 (12 per cent), 137 out of 824 (17 per cent), 53 out of 226 (24 per cent), and 1,009 out of 2897 (35 per cent) (Jung and Kang 2020, Table 7).

What explains this dramatic shift in party allegiance across the legislative and executive branches at multiple tiers in such a short period of time? Jung and Kang (2020) ascribe the dramatic swing to the highly nationalized nature of the 2018 elections which were overshadowed by the (un)popularity of the former and current South Korean presidents. The former president Park Geun-Hye was impeached because of corruption in December 2016 and this process made her and her party, the Liberal Korea Party, highly unpopular for several years. The current president Moon Jae-In from the Democratic Party of Korea was highly popular because of successful summits between the leaders of North and South Korea held on 27 April and 26 May 2018 just before the elections which were held on 13 June 2018. A polarized and highly nationalized multilevel party system offers only one other suitable alternative for voters in case they want to switch parties. Voters shifted en masse their support from one unpopular party to the other and ideologically more distant party.

Jung and Kang’s explanation resonates with the explanation provided by Melusky and Richman (2020) for the highly nationalized state elections in the USA in 2018. President Trump’s strategy was clearly to turn the elections into a popularity contest for his presidency. However, in contrast to South Korea, where the party of the sitting president benefited from a highly unpopular former president, the Republican Party did not gain from President Trump’s involvement in the campaigns which had both a nationalizing and a polarizing effect. Voters with intense views about Trump were more likely to vote and to cast a 2018 state legislative vote consistent with the partisanship of the 2016 presidential election choice (Melusky and Richman 2020). Overall, the mobilizing and polarizing effects of President Trump’s involvement in the state electoral campaigns appear to have had a greater impact on Democratic Party supporters than on Republican Party sympathizers. The Democratic Party increased their trifecta total with a net gain of six trifectas and won legislative seats in almost all 50 states whereas the Republican Party lost four trifectas and incurred legislative seat losses in nearly all 50 states.

A nationalizing and polarizing trend is also observed by Gagatek and Tybuchowska-Hartlińska (2020) for regional elections in Poland. In both the 2014 and 2018 regional elections three parties accounted for more than 70 per cent of the vote. In 2014, the vote shares were equally distributed across the parties: PiS (Prawo i Sprawiedliwość, social-conservative) won 27 per cent of the vote, PO (Platforma Obywatelska, liberal) 26 per cent, and PSL (Polskie Stronnictwo Ludowe, agrarian) 24 per cent. In 2018, the aggregate vote moved towards a two-party system and PiS won 34 per cent of the vote, PO (in alliance with Nowoczesna) 27 per cent, and PSL 12 per cent (Gagatek and Tybuchowska-Hartlińska 2020, Table 2). The development towards a two-party system is more evident when one looks at seat shares in the regional councils. In 2018, PiS won more than 40 per cent of the seats in eight out of 16 regions and obtained more than an absolute majority of the seats in five regions. PO won more than 40 per cent of the seats in five regions and between 30 and 40 per cent of the seats in an additional six regions. As a result, PiS forms a single-party executive government in seven regions and forms a coalition with a junior party (Bezpartyjni Samorządowcy) in one region whereas PO governs the other eight regions together with PSL (plus SLD (Sojusz Lewicy Demokratycznej) in two regions and plus MN (Mniejszość Niemiecka) in one region) (Gagatek and Tybuchowska-Hartlińska 2020, Table 3).

Poland might be moving towards a fully blown aggregated polarized biparty system. In the 2019 European (26 May) and national (13 October) elections, PiS won respectively 45 per cent and 44 per cent of the vote and PO in alliance with several parties won respectively 38 and 27 per cent. In addition, both PiS and PO have clear electoral strongholds, PiS mainly in the South East and PO mainly in the North West of Poland (Gagatek and Tybuchowska-Hartlińska 2020, Figures 3–5). Thus, a biparty system at the national level arises out of (nearly) predominant party systems at the regional level.⁴ Similar to South Korea and the USA, regional elections in Poland were nationalized and polarized. Gagatek and Tybuchowska-Hartlińska (2020) observe that PiS seized control over the judiciary and public TV and radio while in national government and this induced PO, PSL and other opposition parties to collaborate against PiS. The 2018 regional election campaign subsequently revolved around two poles with a pro-PiS camp on one end and a contra-PiS camp (PO, PSL, and allies) on the other end. All parties focused on national government policy, the opposition claimed that the PiS government had been disastrous whereas the PiS emphasized its government successes.

Collectively, the cases of Poland, South Korea, and the USA reveal that also in highly nationalized and polarized elections one may observe a high degree of territorialization in the vote. One may even argue that aggregated bipolar party systems should be more common than replicated bipolar party systems because most parties tend to have electoral strongholds in different kinds of territories produced by, for example, urban-rural or centre-periphery cleavages. South Korea, and to a lesser extent Poland, also reveals how swiftly an aggregate bipolar party system can transform into a multilevel predominant (single) party system. The three aggregate polarized biparty systems also raise normative issues with regard to democracy (Vampa 2018). At the statewide level, there is genuine electoral contestation for executive office between two parties. However, predominant (single) party systems are omnipresent at the subnational level where electoral contestation is (far) less competitive and often effectively absent.

Parties make or break multilevel democracy

Parties are crucial for democracy because they perform four important functions: (1) they aggregate interests by offering a package of positions on a wide range of issues, (2) they structure the choice offered to voters by presenting a limited number of alternatives and by organizing elected representatives in parliament, (3) they recruit candidates for political office and socialize them into existing political routines, and (4) they form executive governments (Lelieveldt and Princen 2011, 172). A key question is how parties perform these functions in multilevel democratic polities (Detterbeck 2012; Lelieveldt and Princen 2011, 171–174).⁵ A multilevel system perspective highlights the pivotal role that parties perform regarding the horizontal and vertical integration of a multilevel party system.

Holding elections at the regional and national levels increases the possibilities for parties to mobilize voters on the basis of different policy position packages (function 1) and voters may be confronted with a set of parties at the regional level that is different from the nationwide level (function 2). Regional authority is one of the key variables that heavily impacts the incentives for regional party branches to differentiate their policy positions from those of their central party at the statewide level (Chhibber and Kollman 2004; Thorlakson 2007). When regional governments have competences over important policy areas such as education, health, regional economic policy, and social welfare, voters will care more about what parties have to offer in regional elections. Regional party branches are more likely to tailor their policy bundles towards the interests of regional voters especially when regional voter preferences are different from other voters in the statewide territory. Powerful regional governments and specific regional interests also enable political entrepreneurs to establish new and region-specific parties. In addition, it is often easier to win a seat in a regional than in a national parliament simply because it requires less votes, i.e. the (absolute number of) votes to seat ratio is smaller for regional parliaments. Region-specific or regional parties are more often found in regions with specific voter preferences and in regions where voters have strong regional identities (Brancati 2005; Massetti and Schakel 2017). Voters with region-specific interests are not only concerned with being able to choose between a set of policy packages that is different from the statewide offer but they also care about which parties are governing over them and whether parties can be trusted to decide and implement policy in the region’s best interest (Massetti and Schakel 2015).

As a consequence, the party constellation of regional parliaments and executive governments also become increasingly different across regions and between regional and national electoral arenas (functions 3 and 4). The establishment of regional electoral arenas increases the differentiation between party systems across regions (horizontal disintegration) and between the national level and the regions (vertical disintegration). The pressure towards a disintegration of a multilevel electoral system is higher when regional authority increases and when region-specific identities and preferences become stronger. However, statewide parties can affect the integration of the multilevel party system to the extent that they compete in elections and participate in governments across the statewide territory at all levels of government. The extent to which statewide parties can monopolize electoral competition depends first and foremost on their internal organization.

In well institutionalized party systems, regional branches of statewide parties are often bounded by the central party organization and are often not allowed to deviate regional positions too much from the national policy position. Statewide parties prefer to voice clear policy positions that serve as cues for voters to assess how policy packages differ between parties and that enable voters to determine which policy package is closest to their preferences. Voters may get confused when regional branches of statewide parties start to offer different bundles of policies. The extent to which regional branches of statewide parties can tailor their policy package towards region-specific interests depends on how much leeway they get from the central party organization. Similarly, regional voters may prefer candidates from their own region to govern over them and regional branches of statewide parties can better cater towards those preferences when candidate selection procedures are decentralized. It is therefore not surprising that the development towards multilevel party systems has stimulated research on internal party organizations (for example, Astudillo and Martínez-Cantó 2019; Fabre 2008, 2011; Ribeiro and Fabre 2019; Saglie 2019; Thorlakson 2009, 2011).

Interactions between parties can also contribute to the integrating of multilevel party systems. Organizationally separate parties can coalesce or form an alliance before an election is held and offer a single package of policies and candidates. The integrative capacity arising out of electoral alliances depends on the breadth and depth of the collaboration between parties and the extent to which the collaborators give each other freedom to, for example, offer a different set of policy bundles, to select their own candidates, or to use their own party name on the ballot paper. The depth of the alliance can be relatively ‘thin’ and purely function to aggregate votes as to increase the chances to win an additional seat. At the other end, the collaboration can be intensive that one can speak of a single party except that an official merger has not taken place yet. Similarly, the territorial breadth of a collaboration can vary and an alliance can be concluded for a specific regional election or for all elections across the nationwide territory. In a multilevel electoral system, the breadth and depth of a party collaboration can interact in complex ways. For example, the CDU participates in elections across Germany except for Bavaria where only the CSU competes in elections but the CDU and CSU form one party group in the federal parliament (Hepburn 2008; Hepburn and Hough 2012).

The election articles on Brazil, Bosnia and Herzegovina, Kenya, and Nigeria provide new insights into the capacity of parties to integrate multilevel party systems in their horizontal and vertical dimensions. Ribeiro and Borges (2020) report on the 2018 simultaneous presidential and state elections in Brazil. The presidential election caught a lot of attention because of the ascension of Jair Bolsonaro – a radical-right and populist candidate who was eventually elected as president – broke the bipolar party competition that dominated in the six previous elections. Although the presidential elections were highly nationalized and were held simultaneous with the state elections, 2018 nevertheless recorded the highest level of incongruence since 1994.

To understand this counterintuitive outcome, one needs to look at the collaboration and coordination of parties across regional and presidential elections. The need to garnish a majority of the vote combined with the magnitude of the ‘prize’ of winning the presidency provides for huge incentives for parties to rally behind a single presidential candidate. Indeed, between 1994 and 2014 control over the presidency was fought by a left-wing alliance of parties headed by the PT (Workers’ Party) and a right-wing alliance led by the PSDB (Brazilian Social Democracy). A two-alliance instead of a two-party system has developed because Brazil is one of the most decentralized countries in Latin America which makes states an important focus of party competition. State governors are important actors, careers of most politicians are state-centered, governors control candidate selection, and subnational party branches organize party lists which do not require approval by the central party (Ribeiro and Borges 2020). Parties have different electoral strongholds and are able to win gubernatorial elections but they need to collaborate with other parties if they want to have a chance to win in presidential elections.

Ribeiro and Borges (2020) describe how Bolsonaro’s ascension in the public opinion polls before the presidential election disturbed the long-standing party collaborations because it increased the uncertainty for the electoral coalitions whether they would be able to win the presidential election. In addition, the electoral alliance of Bolsonaro, centred around the small PSL (Social Liberal Party), had a weak presence at the subnational level which provided opportunities for other parties to win the support in regional elections of voters who supported Bolsonaro in the presidential elections (Ribeiro and Borges 2020). Many parties chose to compete in gubernatorial and regional parliament elections on their own and this led to incongruence between the outcomes of the presidential and regional elections.

It is important to note that the Brazilian multilevel party system cannot be described as an aggregated or as a replicated polarized biparty system (see the previous section). The party system at the federal level is, or better was until 2018, bipartisan and polarized but similar party systems cannot be found at the regional level nor are there predominant party systems in the regions. This is a crucial observation because a comparison to the multilevel party systems of Poland, South Korea, and the USA – which are all aggregated polarized biparty systems although to different degrees – reveals that presidentialism (Brazil, South Korea, USA), electoral alliances (Brazil, Poland), nor high (Brazil, USA) or low (Poland, South Korea) regional authority are necessary nor sufficient conditions that sustain an aggregated polarized biparty system. Parties and their interactions between each other and with their voters are crucial determinants of the level of integration of a multilevel party system.

Bosnia and Herzegovina is perhaps one of the least integrated multilevel party system in the world. The two entities Republika Sprska (RS) and the Federacija Bosne i Hercegovine (FBiH) have almost fully different party systems and the party systems in the cantons of the FBiH are widely different. Either Bosnian, Croat or Serb ethnic parties dominate in the entities and cantons (except for two mixed Bosnian and Croat cantons) and party competition takes place within ethnic blocs between parties representing the same ethnic group. John Hulsey and Soeren Keil (2020) identify the factors that sustain this loosely integrated multilevel party system. First, Bosnia and Herzegovina is a highly decentralized federation whereby both the entities and the cantons have wide-ranging policy and taxing powers. The Regional Authority Index measures the authority of regions in 81 countries on a yearly basis since 1950 and the indicator ascribes the fourth highest score to Bosnia and Herzegovina (Hooghe et al. 2016). Second, electoral rules enable the fragmentation of the electoral system. Electoral districts overlap with territorial jurisdictions which are subsequently drawn to maximize ethnic homogeneity within the jurisdiction and this ensures that each ethnic group is represented by their own parties. Electoral thresholds are very low and because of the small sizes of (some of the) cantons a few hundred votes is sufficient to win a seat in a cantonal parliament (Hulsey and Keil 2020).

The combination of a highly decentralized state and electoral rules that induce fragmentation advances patronage and clientelism. Elections have become a competition over patronage networks which are tied to executive office which gives control over budgets, employment opportunities and state assets. New parties can be easily created and many of them are established not for ideological reasons but because of personal disputes between elites and because of disagreements over the distribution of the benefits from the patronage system (Hulsey and Keil 2020). The case of Bosnia and Herzegovina clearly reveals that decentralization does not necessarily help to consolidate democracy as is often thought (see, e.g. Diamond and Tsalik 1999). On the contrary, it may actually hamper the consolidation of democracy by fostering patronage and clientelism. The future of electoral democracy in the country may very well depend on whether the electoral rules will be changed. Cooperation between ethnic parties from different ethnic groups hardly takes place despite the power-sharing arrangements laid down in the constitution. Creating larger and ethnic heterogeneous electoral districts supplemented with an electoral threshold may stimulate ethnic parties to collaborate but the feasibility of changing the electoral rules can be doubted given the veto-powers of the ethnic groups over amendments of the constitution and electoral laws.

Electoral rules may stimulate party collaboration and contribute to the integration of a multilevel electoral system but electoral engineering is not a panacea. This is clearly shown by Henrik Angerbrandt (2020) who reports on state assembly, gubernatorial, national assembly, and presidential elections held in 2019 in Nigeria. Several electoral rules at the federal and state levels promote catch-all parties that receive broad voter support. Ethnically and religiously based parties are forbidden, the executive committee of a statewide party needs to have representation from all six geopolitical zones in the country, and a candidate will be elected into presidential office when the candidate wins a majority of the vote with at least 25 per cent in two-thirds of the states. Each state consists of three ‘senatorial zones’ and parties have agree to distribute and rotate the positions of governor, deputy governor, and speaker of the house between persons from the different zones. In order to get elected, governors need to win a majority of the vote whereby at least 25 per cent of the vote is won in at least two-thirds of the local government areas (Angerbrandt 2020). Although these electoral rules provide strong incentives for parties to collaborate, party competition has largely shifted from the inter-party to the intra-party arena.

Nigeria is arguably one of the most well-entrenched federal systems in Africa and state resources can be used to sustain patronage networks (Angerbrandt 2020). Regional strongmen within comparatively centralized statewide parties regularly switch party and they bring their followers with them which can lead to vote share swifts up to 40 per cent. State assembly and gubernatorial elections are held two weeks after the presidential elections and it is not unusual for incumbent governors to shift allegiance to the governing party at the federal level to gain access to patronage networks. State resources are used by incumbent governors to persuade their followers and voters to switch to the same party to which the incumbent governor switches to. Angerbrandt (2020) reports that the 2019 state assembly and gubernatorial elections were marred by several instances of violence and kidnappings of election officials and voters. In addition, the media and election observers were denied access to certain areas. At the surface, party democracy seems to work in Nigeria, the same parties compete in federal and state elections and most parties have offices across the statewide territory. However, internal party structures boil down to clientelist networks with ties to community associations, trade unions, and ethno-nationalist movements which are not connected to the party but rather function as platforms for personal power for individual candidates (Angerbrandt 2020).

Similar intra- and inter-party dynamics are reported by Dominic Burbidge (2020) for elections in Kenya. Every five years, citizens can cast five votes: for the parliament, senate, and president at the national level and for the assembly and governor at the county level. At the national level, two electoral coalitions viably compete for the presidency and these alliances are based on patronage networks between national and county politicians. Incumbent county governors mobilize voters to vote for a national electoral alliance and presidential candidates ‘buy’ the support from gubernatorial candidates by promising perks that can be handed out once a presidential candidate is elected. Significant vote switching can take place because voter support is attached to patronage networks of candidates rather than to parties (Burbidge 2020). In contrast with Nigeria, vote switching takes place between consecutive rounds of elections rather than between national and subnational elections because in Nigeria subnational elections are held after the presidential elections whereas in Kenya subnational elections are held on the same date as the presidential elections. Despite the difference in electoral timing, one can observe similar intra- and inter-party dynamics in both countries which underlines that electoral engineering is not a panacea to steer the actions of parties and candidates towards integrating the multilevel party system.

Motivations for vote switching between national and regional elections

Much of the regional election literature deals with vote switching between regional and national elections.⁶ The second-order model literature compares vote shares won in regional elections to previously held national elections and expects negative vote share swings for parties in national government and positive vote share swings for small parties and parties in national opposition (Reif and Schmitt 1980; Schakel and Jeffery 2013). These vote share swings are an expression of voter dissatisfaction with the performance of parties in national government which is articulated in the vote choice in regional elections. The nationalization literature is interested in the territorial heterogeneity of the vote across the nationwide territory (often referred to as static nationalization) and this literature has noticed that vote share differences across regions are larger for regional than for national elections (Henderson and Romanova 2016; Schakel 2013). This implies that a different motivation underlies vote choice in regional and national elections. Supporters of regional parties may vote sincerely in regional elections and vote for a party which is closest to their preferences but these voters may vote strategically in national elections and vote for a party which has a higher chance than the regional party to gain seats in national parliament and/or to become part of the national executive.

The election reports on Denmark and Sweden reveal considerable levels of vote switching between national, regional and local elections which consequently exposes our lack of understanding of what motivates voters to switch parties across electoral arenas (i.e. voting in multilevel election systems). Ulrik Kjaer (2020) finds that 24 per cent of the Danish voters vote for a different party in regional elections and 31 per cent vote for a different party in local elections than they would have voted for if a national election would have been held at the same day. Linda Berg and Henrik Oscarsson (2020) observe remarkably similar vote switching percentages for voters in Sweden where 28 per cent of the voters switch their vote between national and regional elections and 31 per cent between national and local elections. These vote switching percentages stand out because significant levels of vote share switching are not expected in these countries. Both Denmark and Sweden are unitary and ethnically homogenous countries where regions have far less powers compared to regions in regionalized and federal countries and where there are few and very small regional parties. In addition, local and regional elections are held simultaneously across the statewide territory in both countries and in Sweden local and regional elections also coincide with national elections. In other words, both Denmark and Sweden are typical cases where we expect to find highly nationalized party systems.

At the aggregate statewide level, the regional and national vote is very similar and comparable party systems can be found at the regional and national levels. Total volatility between the 2018 regional and national elections was 5.7 per cent in Sweden (Berg and Oscarsson 2020, Table 3) and 8.6 per cent between the 2017 regional election result and national election outcomes based on vote intentions expressed by close to 5,000 voters in a regional election survey held in 2017 in Denmark (Kjaer 2020, Table 2). These percentages stand in stark contrast to the level of vote switching observed at the individual level. Survey analyses in Sweden (Berg and Oscarsson 2020) and Denmark (Kjaer 2020) clearly reveal that a quarter to a third of the respondents (would) cast a different vote in the regional election than for the national election.

What can explain this tremendous degree of inter-level vote switching? The second-order election model does not seem to give a lot of traction because this model would expect significant vote switching from parties in national government to small parties and parties in national opposition. In addition, based on the second-order model one would expect similar voting to occur in local and regional elections. Kjaer (2020, Table 4) reveals that 7.3 per cent of the Danish voters (would) cast a different vote in national, regional, and local elections, 12.1 per cent cast a unique local vote, and 11.2 per cent cast a unique regional vote. This suggests that different motivations underly local, regional, and national vote choice, at least for a quarter to a third of the electorate.

The nationalization of elections literature draws our attention to regional parties which does not provide much analytical leverage regarding vote switching in Denmark given the virtual absence of regional parties, e.g. the Schleswig Party received 0.8 per cent of the vote in Southern Denmark (Kjaer 2020, Table 2). In Sweden, regional parties can explain a significant part of vote switching but only in specific parts of the country. In three out of 21 regions ‘healthcare parties’ attract more than 10 per cent of the regional vote (Berg and Oscarsson 2020, Table 3). Healthcare parties can be described as single-issue parties established to counteract (proposed) scaling or closing down of health care services such as maternity wards, intensive care, emergency departments, and hospitals. Healthcare is the most important competence of the regions (both in Sweden and in Denmark) and Berg and Oscarsson (2020) report that 79 per cent of the voters are aware that health is a regional competence and 42 per cent of the voters indicated welfare/healthcare as one of the three most important issues for their party choice. It is therefore probably not too far stretched to assume that quite a number of voters motivate their regional vote choice on the basis of the performance of parties in regional government. However, in both Denmark and Sweden, vote switching occurs across all regions and most of it takes place between statewide parties. We need more regional election survey research to reveal whether vote switching between regional and national elections takes place in countries beyond Scandinavia where regions do not have the competences for health policy and to tax personal income. Also, as Kjaer (2020) notes, we also need to establish whether voters switch their vote because they do not care about regional elections. We are in dire need of theory and hypotheses that can explain the significant levels of vote switching.

Discussion

The high level of vote switching reported for Denmark (Kjaer 2020) and Sweden (Berg and Oscarsson 2020) reveals that the need for a ‘multilevel voting model’ (Wyn Jones and Scully 2006) is not restricted to regionalized and federal countries such as Australia (Bolwell and Eccleston 2018), Canada (González-Sirois and Bélanger 2019), Germany (Müller 2018), India (Schakel, Sharma and Swenden 2019), Italy (Massetti 2018), Switzerland (Bochsler 2019), and the UK (Middleton 2019). We need to know what drives voters to make distinct evaluations for different levels of government, also for unitary countries with powerless regions and electorally weak regional parties. A key question to address in this research is what determines the balance between factors arising out of regional and national electoral arenas (Schakel and Romanova 2019). Knowing what drives voting in multilevel election systems is also important for how we evaluate the role of regional government in multilevel democracies. When regional elections are about national issues and regional voting is driven by factors originating from the national electoral arena one may ask the question whether holding regional elections has an added value for electoral democracy. The high level of vote switching between national and regional elections found in Sweden and Denmark may indicate that second-order election and nationalization of elections research has significantly underestimated the extent to which regional elections are about regional government. Thereby, regional elections may play a larger role in multilevel democracies than previously thought.

Election and party scholars are well aware that the consolidation of democracy depends, among other factors, on the democratic attitudes of the political elites (Chabal and Daloz 1999; Elklit 1999) and on the question whether competition between political parties is based on their proposed policies and platforms or on the perks that political candidates can allocate to voters (Kitschelt 1995; Mainwaring and Torcal 2006). A multilevel party system perspective exposes an additional factor that may be relevant in an evaluation of the state of electoral democracy in a country: the extent to which parties collaborate in various types of elections (vertical integration) held across the regions in the statewide territory (horizontal integration).

In federal, entity, and cantonal elections held in Bosnia and Herzegovina, most parties receive their support in few regions and voter support is largely based on clientelism (Hulsey and Keil 2020). In contrast, in Brazil, Kenya, and Nigeria most major parties are competitive in various regions in the country but only when one looks at national elections. At the subnational level, most parties have clear territorial strongholds and most regions are dominated by regional strongmen. Regional parties and strongmen form alliances in presidential elections in order to be able to win a presidential election. Party collaboration is based on clientelism and patronage rather than ideological and programmatic proximity. The willingness of regional parties and strongmen to support a presidential candidate depends on the perks and local collective goods that the presidential candidate can offer to them (Angerbrandt 2020; Burbidge 2020; Ribeiro and Borges 2020). As a result, electoral alliances and parties are unstable and regional parties and candidates often shift between electoral alliances. In addition, party elites and candidates often establish new parties and alliances between consecutive national elections but also between national and regional elections. Party alliances at the national level may be durable regarding their name and lead party/candidate but very flimsy when it concerns the party members. Hence, a focus on national elections and statewide parties/alliances may significantly overestimate the extent to which democracy is consolidated.⁷ Decentralization can make regional electoral arenas attractive for parties to compete across the statewide territory at both the national and regional levels and thereby may contribute to the integration and consolidation of multilevel democracy. Patronage and clientelism often result in weakly organized parties and frail collaboration between parties which undermines effective opposition, limits opportunities for candidates to stand in elections, and decreases the choice offered to voters. Decentralization in combination with patronage and clientelism can significantly hinder the consolidation of multilevel democracy because decentralization can proliferate and entrench patronage and clientelist networks by multiplying the spoils attached to executive office across regions.

A multilevel electoral system perspective also uncovers highly different party systems at the national and subnational levels in the USA, South Korea, and, to a lesser extent, in Poland. The multilevel party system in these countries can be described as an aggregated polarized biparty system where a polarized biparty system can be found at the national level, i.e. two parties (or two-party alliances) both win equally-sized vote shares and the parties (alliances) are split by an ideological divide. However, predominant party systems can be found in the regions, i.e. one single party (or party alliance) obtains a majority of the vote and controls the executive (Gagatek and Tybuchowska-Hartlińska 2020; Jung and Kang 2020; Melusky and Richman 2020). The party systems at the nationwide level arise out of the aggregation of regional predominant party systems but one party controls one half whereas another party controls the other half of the regions. This is also an important observation for the question of the consolidation of multilevel democracy because a highly competitive national electoral arena can be complemented by highly non-competitive subnational electoral arenas. In addition, South Korea reveals how easily an aggregated polarized system can transform into a single-party multilevel hegemony where one party obtains a majority of the vote and controls the executive at the national level and in the regions across the country.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Trond Mohn Foundation [grant number 812468].

Notes

1 See https://www.arjanschakel.nl/index.php/arore for more information about the annual review of regional elections and a call of interest to write an election article or election report for the next annual reviews.

2 We use the terms nationwide and statewide interchangeably throughout the introduction to denote the national level.

3 These three concepts are not mutually exclusive. For example, regional parties that participate in only regional elections contribute to a less integrated multilevel party system as well as a less integrated multilevel election system.

4 Poland falls in between the ideal types of aggregated and replicated polarized bipolar party systems. Some regions are closer to a predominant party system (e.g. Podkarpackie where PiS won 52 per cent of the vote and 76 per cent of the seats) whereas other regions resemble more a bipolar party system (e.g. Śląskie where PiS and PO respectively won 31 and 29 per cent of the vote and 49 and 44 per cent of the seats) (Gagatek and Tybuchowska-Hartlińska 2020, Tables 2 and 3). In addition, PiS and PO are not split across an (major) ideological divide because these parties can respectively be positioned as a right and centre-right party.

5 Lelieveldt and Princen (2011, 172–173) identify a fifth role for parties, i.e. they ‘form a liaison between the state on the one hand, and voters and civil society on the other’. This role is significantly impacted by the European Union multilevel electoral system: ‘Because European political parties are confederations of domestic parties, they hardly form direct liaisons between the EU institutions and citizens or civil society’ (Lelieveldt and Princen 2011). In multilevel party systems that are contained within countries, voters can cast a vote in both national and subnational elections and thereby parties form a direct liaison between voters and subnational and national institutions.

6 We use the term vote switching to include voters who change their preferred party over time (party switching) and voters who change their vote because they are confronted with different vote processes (e.g. proportional versus majoritarian rule) and/or party offers across elections (split-ticket voting).

7 The V-Dem project (https://www.v-dem.net/en/) includes indicators that tap into party-voter linkages, party competition across regions, and subnational party control which provide good proxies for the level of consolidation of a multilevel democracy. However, these indicators also may overestimate the level of consolidation to the extent that linkages between politicians within the same party or between regional and statewide parties can be clientelist, can decrease the level of competition across regions, and can increase the subnational control exercised by (sub-)national elites.