Democratic backsliding from a local perspective

Democratic backsliding describes a process by which ruling parties undermine political and civil societal institutions to tilt the playing field in their favour, effecting a form of gradual regime change which may travel through various stages (semi-consolidated/electoral democracy, hybrid regime/electoral- or competitive authoritarianism), potentially culminating in stable authoritarianism. Analysts describe a common playbook used by dominant parties: undermining the fairness of elections; discretionary use of legal instruments to undermine civil liberties; and restricting the opposition's access to public resources, the media and the judicial system (Levitsky and Way Reference Levitsky and Way2010: 7–12). There is consensus among analysts that Fidesz and PiS have both borrowed heavily from this playbook, but also that Fidesz has been more successful thus far in doing so.

The origins of Hungary's backsliding date to Fidesz's winning a constitutional supermajority in 2010. This supermajority allowed Fidesz to enact institutional and policy changes designed to lock in long-term dominance: changes to the constitution, Cardinal Laws and rule of law (Kovács and Scheppele Reference Kovács and Scheppele2018); constitutional court-packing (Kelemen Reference Kelemen2017); restrictions on media (Bajomi-Lázár Reference Bajomi-Lázár2017) and academic freedom (Enyedi Reference Enyedi2018); and attacks on civil society (Kover Reference Kover2015). Poland's PiS won an absolute majority in 2015, aided by the United Left coalition's inability to reach the 8% minimum threshold. Lacking Fidesz's supermajority, PiS has not been able to change the constitution directly. Instead, PiS has sought to undermine the rule of law to the extent it can work outside the limits of constitutionalism (Sadurski Reference Sadurski2018). It undermined the Constitutional Tribunal by packing the court and reducing its scope, before instituting measures to reduce the independence of lower-court judges (Kovács and Scheppele Reference Kovács and Scheppele2018: 196). Further echoing Hungary, the PiS-led government has limited freedom of assembly (Sadurski Reference Sadurski2019) and restricted media freedom (Stanley and Cześnik Reference Stanley, Cześnik and Stockemer2019), especially of state-funded broadcasts and publications.

Figure 1 plots the V-Dem Institute's national-level measure of liberal democracy for both countries, highlighting Hungary's earlier backsliding but also Poland's subsequent convergence. Between 2009 and 2019, Hungary and Poland's liberal-democracy scores declined by 0.36 and 0.33, respectively. This decline pushed Hungary's regime into the ‘electoral-authoritarianism’ category in 2019 (Maerz et al. Reference Maerz, Lührmann and Hellmeier2020: 915); Freedom House downgraded Hungary from ‘semi-consolidated democracy’ to ‘transitional or hybrid regime’ in 2020 (Freedom House 2020). András Bozóki and Dániel Hegedűs categorized Hungary as a hybrid regime as early as 2014 (Reference Bozóki and Hegedűs2018: 1174). Poland is still classified as an ‘electoral democracy’ by V-Dem and as a ‘semi-consolidated democracy’ by Freedom House. In Bozóki and Hegedűs's words, ‘the dismantling of democracy and rule of law in Poland is still in progress, while in Hungary the hybrid regime is already entrenched and stabilized’ (Reference Bozóki and Hegedűs2018: 1177).

Figure 1. National-Level Democratic Backsliding in Hungary and Poland (1989–2020)

Source: Varieties of Democracy v11.1, www.v-dem.net/en/data/data/v-dem-dataset-v111/.

We believe that there are enough similarities between municipal elections in Hungary and Poland to make comparing them instructive, even acknowledging differences in the degree of national-level backsliding.Footnote ¹ Certainly, the national-level regime matters to opposition candidates’ strategies in municipal elections. In a stable democracy, local parties offer programmes tailored to their voters and urge those voters to vote for their candidate; however, in authoritarian conditions or when democracy is under threat, opposition parties may urge supporters to vote tactically – that is, to back the most viable opposition candidate in order to defeat the dominant party, programmatic differences notwithstanding. Since, typically, parties ‘are conservative organizations and resist change’ (Harmel and Janda Reference Harmel and Janda1994: 278), it takes extraordinary circumstances – such as democracy itself being at risk – to induce this kind of change in party behaviour, a point that we will develop in presenting our hypotheses. The degree of backsliding is significant enough in both countries to cast local politics in the light of challenging de-democratization.

If national regime matters to subnational politics, there are also theoretical and historical reasons to believe that subnational politics matter to national regimes. As Ulrik Kjær has observed, local representation can be both a ‘“respirator” that brings a party through difficult times and makes it possible for it to gain strength and re-enter the national parliament’ and an ‘incubator’ of new parties struggling to find their footing in national politics (Kjær Reference Kjær, Blom-Hansen, Green-Pedersen and Skaaning2012: 202). This kind of impact may be even greater in de-democratizing regimes. Studying it can enrich the broader literature on how subnational levels of government can be areas of institutional divergence challenging the logic of the national political regime (Gibson Reference Gibson2012; Key Reference Key1984). This literature is often framed through the lens of subnational authoritarianism, where residual pockets of undemocratic rule persist in otherwise democratic regimes. However, in backsliding regimes the opposite can sometimes be true, and urban centres can form a more liberal, more democratic centre of opposition resistance (Diaz-Cayeros and Magaloni Reference Diaz-Cayeros and Magaloni2001; Wallace Reference Wallace2013). This is one way to view the position of cities in contemporary Poland and Hungary. It is not only that the ruling parties are less successful electorally in cities. Cities can challenge ruling parties through their economic power and ability to contradict national government policies. A recent example was the Warsaw City Council's decision to expand protections for LGBT residents, strongly contradicting PiS's programme (O'Dwyer Reference O'Dwyer2019). Making a further stand against illiberal developments in the region, Budapest and Warsaw, together with Prague and Bratislava, formed the Pact of Free Cities, collaborating to increase their leverage against national governments and seek external support (Hopkins and Shotter Reference Hopkins and Shotter2019).

There are also historical reasons for focusing on local government when analysing democratic backsliding in Central Europe. Devolving power from central to local governments was seen as key to breaking the power of the old political elites and communism's hyper-centralization (Regulski Reference Regulski, Sochacka and Kraśko1996). Because of the perceived centrality of local government in the transition from communism in the 1990s, there is significant policymaking and resource allocation at the local level. Poland's local governments have maintained among the highest levels of autonomy in Europe, particularly in terms of policy scope, organizational autonomy and voice in higher-level policymaking (Ladner et al. Reference Ladner, Keuffer and Baldersheim2019: 232). Though on par with Polish localities in the 1990s, the autonomy of Hungarian localities has declined substantially (Ladner et al., Reference Ladner, Keuffer and Baldersheim2019: 142, 338). Andreas Ladner et al. (Reference Ladner, Keuffer and Baldersheim2019: 341–345) describe local governments in Hungary as ‘bureaucratized’ and ‘subordinate’ to the centre, whereas Poland's are ‘empowered’ and engage in ‘thick coordination’ with the centre. At least some of the decreased autonomy of Hungarian municipalities is strategic. In 2010–11 Fidesz amended local election rules and Hungary's Fundamental Law governing local government (Mötv) to advantage itself structurally at the opposition's expense and recentralized many aspects of policymaking to allow itself more direct control should the opposition be successful in municipal elections (Soós and Dobos Reference Soós and Dobos2014). Though the power of directly elected mayoralties inclines both Fidesz and PiS to pursue them actively, municipalities’ weaker autonomy in Hungary may help explain Fidesz's greater success in these races. It may also be argued, however, that the greater autonomy of Polish municipalities increases their value to both the ruling party and its challengers.

Mayoral election success: similar conditions, different outcomes

Mayors in both Poland and Hungary are directly elected. In Hungary, all elections are single round, and only a simple majority is required to win a municipal mayoral election. In Poland, if no mayoral candidate obtains an absolute majority in the first round, a second-round election is held two weeks later between the top two first-round candidates. This is a key institutional difference between the two countries.

Hungary's Fidesz has been remarkably successful at the local level, steadily increasing its reach into municipal politics. In 2014, Fidesz won 70% of mayoral elections that it contested in Hungarian towns and cities, after winning only 32% in 2002 (Stenberg Reference Stenberg2018). This is no accident. Fidesz consciously manipulated city council rules and regulations to reduce oversight and fair contestation in municipal politics (Jakli and Stenberg Reference Jakli and Stenberg2021). It then used local resources clientelistically to ensure the support of voters (Mares and Young Reference Mares and Young2019). Ultimately this has helped Fidesz to overcome the traditional success of the Hungarian left in urban areas.

Local electoral politics in Poland have generally followed a different pattern. Parties have never played a predominant role in municipal electoral politics (Gendźwiłł Reference Gendźwiłł2012). While the share of candidates and elected councillors going to party-affiliated candidates had been increasing before 2010, this peaked in that year and then markedly declined following local election reforms (Gendźwiłł and Żółtak Reference Gendźwiłł and Żółtak2017: 123). Poland shifted from using party lists to single-member districts in most city council elections in 2012, although lists were retained in larger cities. While independent lists were already dominant in local elections, the reform further strengthened those lists affiliated with elected mayors (Gendźwiłł and Żółtak Reference Gendźwiłł and Żółtak2017). Party politics have not penetrated municipal elections significantly, though parties are increasingly interested in contesting local elections (Gendźwiłł and Żółtak Reference Gendźwiłł and Żółtak2014).

Table 1 shows the respective electoral success of Fidesz and PiS in mayoral elections immediately before and after taking control nationally. The denominator in the winning percentage column is the number of races that the party actively contests as opposed to the total number of possible races.

Table 1. Ruling-Party Wins in Mayoral Elections by Electoral Cycle

Source: Authors’ calculations.

As additional context on national-party penetration into subnational politics, Figure 2 presents data from V-Dem on the extent to which a ‘single party control[s] important policymaking bodies across subnational units (regional and local governments)’. The clear historical baseline in both countries is a lack of influence of any one party subnationally. In Poland, this lack of influence remains the general trend after the rise of PiS, even allowing for an uptick following the 2018 subnational elections. Meanwhile in Hungary, Fidesz's rise has coincided with a significant shift towards national party penetration of subnational governments.

Figure 2. National Party Influence at the Subnational Level in Hungary and Poland Compared (1989–2020)

Source: Varieties of Democracy v11.1, www.v-dem.net/en/data/data/v-dem-dataset-v111/.

Note: The scale of subnational party control is reversed and follows the scale:

0: In almost all subnational units (at least 90%), a single party controls all or virtually all policymaking bodies.

1: In most subnational units (66–90%), a single party controls all or virtually all policymaking bodies.

2: In few subnational units (less than 66%), a single party controls all or virtually all policymaking bodies.

Although parties have nationalized local politics more in Hungary than in Poland, the urban–rural cleavage remains especially salient in both. In national elections, conservative parties tend to do better in rural areas and smaller towns and cities, while the centre-right and left-wing opposition are more successful in urban areas. Kamil Marcinkiewicz (Reference Marcinkiewicz2018) finds significant and substantively large county-level effects on political support for Polish parties in his analysis of 2015 national election results: PiS and the Polish People's Party (PSL), a conservative agricultural party, are much less likely to be supported by urban areas, whereas greater urbanization leads to much higher support for more liberal or left-wing parties. Moreover, he finds that the urban–rural divide explains divergent political support better than any other explanation.

This comparison suggests that municipal elections may potentially impede the consolidation of dominant-party regimes. Although PiS has duplicated many of Fidesz's national-level successes – limiting democratic contestation, reducing media freedom, undermining the opposition, weakening the rule of law and mobilizing rural support in national elections – local-level races have proved beyond its reach. We now investigate why, focusing on electoral rules and candidate strategy.

Hypotheses

If the subnational divergence between Hungary and Poland described above indicates that municipal elections could impede the consolidation of dominant-party regimes in the context of democratic backsliding, the question becomes how? Two possibilities are favourable electoral rules and strategic coordination among opposition candidates. Insofar as candidate strategies are adaptations to electoral rules, the latter are theoretically a given; however, to highlight the difficulties of opposition coordination in dominant-party regimes, we begin with strategies.

A central contention of the party competition literature is that combating a ruling party requires uniting the opposition around a strong candidate: this minimizes the risks associated with splitting the anti-regime vote among opposition candidates, allowing them to form a bloc. Such coalition formation is often labelled a unified opposition strategy. It has received widespread media attention in a variety of contexts. It was implemented in Hungary's 2019 local elections, achieving several high-profile successes. Most opposition parties – on both the left and the far right – supported common candidates in the mayoral elections, leading to wins in 10 (an additional six) cities with county rights, the largest cities in Hungary (Political Capital 2019).Footnote ² Although this gave the opposition credibility and momentum, it did not re-establish fair democratic contestation (Hegedüs Reference Hegedüs2019).

Scholarly research on the strategy's efficacy is mixed. For example, Keith Poole and Howard Rosenthal's analysis of voting patterns in the US Congress describes both-ends-against-the-middle, one subtype of a unified opposition coalition, as a ‘deviant voting pattern’ (Poole and Rosenthal Reference Poole and Rosenthal2011: 6). A ruling unified opposition coalition can be especially unstable at the national level, where its disparate member parties need to maintain an effective coalition to prevent the plurality party from forming a minority government (Nikolenyi Reference Nikolenyi2009: 78–80). Such coalitions may also be difficult in a dominant-party system, where the single party seeking dominance is considered central to coalition formation (Sened Reference Sened1996: 360–361).

There are, however, reasons to think that local-level conditions may be more hospitable to this strategy. Mayors are executives and may not need to maintain a governing coalition to act on local issues; unsurprisingly, the position of mayor is the most desirable in local coalition negotiations (Laver Reference Laver, Mellors and Pijnenburg1989). This may create incentives for interparty cross-municipal cooperation, especially if contending parties agree to distribute candidates equitably across the country, guaranteeing each a reasonably similar number of mayoralties. In theory, this would allow those parties an equitable distribution of opportunities for developing national candidates and for disbursing resources allocated by local government.

There are also grounds to suggest that this strategy may work when local opposition candidates face democratic backsliding, or even outright competitive authoritarianism. As Robert Harmel and Kenneth Janda (Reference Harmel and Janda1994: 278) have argued, typically political parties ‘are conservative organizations and resist change’, which augurs badly for an atypical strategy such as ‘unified opposition’. They also argue, however, that external shocks – such as constitutional changes, changes in longer-term electoral prospects or events that challenge ‘confidence in the correctness or importance of key positions’ – may make parties open to fundamental change (Harmel and Janda Reference Harmel and Janda1994: 267, 270). We argue that democratic backsliding constitutes just this kind of shock. For vote-maximizing parties, it threatens ‘the terms of exchange in the electoral arena’, and for policy-oriented parties, it calls into question adherence to policy positions when democracy itself is at stake (Harmel and Janda Reference Harmel and Janda1994: 265). Empirical work by Jennifer Gandhi and Ora John Reuter (Reference Gandhi and Reuter2013) finds that, in multiparty authoritarian regimes with competitive elections, the emergence of opposition coalitions against the ruling party depends in part on party system institutionalization and harassment of the opposition; both are moderately high in Hungary's emergent hybrid regime. Historically, the most notable instances of this coalition type have occurred in Mexico against the PRI government (Eisenstadt Reference Eisenstadt2003) and in India, where Adam Ziegfeld and Maya Tudor (Reference Ziegfeld and Tudor2017) argue that opposition coordination plays a crucial role in defeating the dominant Congress Party. This practice has been especially durable in Mexico, albeit aided by the relative absence of stable ideology among the major national parties. Elections were most frequently lost by the PRI when subnational coalitions were formed by the largest opposition parties, PRD and PAN (Diaz-Cayeros and Magaloni Reference Diaz-Cayeros and Magaloni2001: 278–281). It has been more widespread at the gubernatorial level but is also seen locally since the 1990s (Eisenstadt Reference Eisenstadt2003; Giraudy Reference Giraudy2015). In Turkey, Orçun Selçuk and Dilara Hekimci (Reference Selçuk and Hekimci2020) suggest that opposition parties in both local and national elections have coalesced around a democracy–authoritarian cleavage made salient by democratic backsliding.

How do these cross-national examples translate to the context of local politics in Central Europe? Overall, we are sceptical of the viability of coordinating united opposition coalitions in local mayoral races. First, despite the high-profile successes in Hungary's 2019 local elections mentioned above, there is no indication of how the fractured opposition will fare in governing. Likewise, there is little indication that this strategy worked in any but the largest cities, which are already predisposed to support the opposition. Even when agreements are reached, voters may vote against such a coalition. Jennifer Gandhi and Elvin Ong (Reference Gandhi and Ong2019) experimentally demonstrate that if a less favoured opposition candidate emerges in a unified opposition coalition, individual voters may defect from the opposition to a dominant party. This risk is real in Hungary, where the far-right Jobbik party's participation is numerically necessary. In Poland's May 2019 European Parliament elections, the opposition successfully negotiated an agreement to combat PiS but was still defeated (Rae Reference Rae2019). In either case, we might consider the failure to make such a coalition – organized or de facto – would probably result in fracturing the opposition vote. Second, Central European parties are generally under-institutionalized, especially opposition parties in dominant-party systems and especially at the subnational level (Gendźwiłł Reference Gendźwiłł2012).

One way we might see evidence of such coalitions would be through a reduction in the number of candidates. The logic is simple: if fewer parties support unique candidates, we would expect to see fewer in the race. Moreover, if local-level opposition coordination is a viable strategy in post-communist circumstances, then we should observe a positive relationship between the number of candidates and the dominant-party candidate's odds of victory. Dominant-party candidates should perform better when multiple opposition candidates are competing, indicating an absence of coordination among opposition forces. A weak or non-existent relationship would indicate that strategically limiting the number of candidates is not much help at the polls. This logic will be most powerful under single-round, first-past-the-post electoral rules. As discussed below, we expect that two-round electoral rules reduce the incentives for candidate coordination, allowing for an inverse relationship between the number of candidates and the dominant party's odds of success:

Our remaining hypotheses concern conditions affecting dominant-party success in two-round systems. Extant scholarship suggests that this institutional variation should impact the possibility of opposition coordination because the second round can focus the contest into a race between the opposition candidate and the dominant-party candidate. Aurélie Cassette et al. (Reference Cassette, Farvacque and Héricourt2013) find variability in two-round French mayoral elections and suggest that different factors and electorates come into play between the two rounds. Broader studies of two-round elections at the presidential level find that having a majoritarian system, as opposed to a simple majority system, favours party alternation in government (Rakner and van de Walle Reference Rakner and van de Walle2009). A second-round election creates a clear opportunity for the opposition to overcome some of the coalitional obstacles to forming a cordon sanitaire against a threatening party.Footnote ³ If the opposition has credible fears that the ruling party aspires to a single-party dominant regime, they have greater incentives to overcome these obstacles. This leads to our second hypothesis:

As indicated above, the possibility of a second round should also affect the relationship between the number of candidates and the odds of ruling-party victory in two-round systems. The possibility of a second round limits the incentives for opposition candidates to cooperate and unify in the first round, which would lead us to expect more candidates entering races. How might this affect the dominant-party candidate's odds? In theory, at least, it could advantage the ruling-party candidate: Laurent Bouton (Reference Bouton2013) argues that the ‘Ortega effect’, in which opposition voters overestimate the chances of their candidate reaching a second round, may advantage political blocs with more unified support. Those voters may choose to support less viable candidates, assuming that they will have another opportunity to vote with their head.Footnote ⁴ On the other hand, it may hurt the dominant-party candidate's odds by the following logic. First, even if the possibility of a second round leads to more opposition candidates in the first round, vote-splitting among the opposition candidates as a group does not add to the votes needed for the ruling-party candidate to achieve an absolute majority and avoid a second round. Second, the first round may serve as a filter for opposition-candidate quality: the idea here is that the ruling-party candidate will face the strongest of the first-round candidates (Carey and Polga-Hecimovich Reference Carey and Polga-Hecimovich2006). The more candidates there were in that first round, presumably, the more likely the one that survives will be a strong candidate. Our final hypotheses encapsulate these arguments:

Data

Our data consist of panel local-level mayoral results for all municipal elections at the town level or higher in Poland and Hungary. Because Fidesz and PiS assume power nationally in different years – 2010 and 2015, respectively – we base our panel on the position of an election in the cycle of establishing regime dominance. In both cases, we include the local election immediately before the regime takes power nationally as a baseline panel (in Hungary, 2006; in Poland, 2014). We then include subsequent election results after national control is established, beginning with Panel 2. Table 2 shows the asynchronous distribution of panels across election years.

Table 2. Panel Election Alignment

We include all municipalities at the town level and higher in both countries. Our sample comprises 895 Polish and 345 Hungarian local government units. In Hungary, this includes all 345 towns (város) as of 2014. In Poland, we include all urban (miejska) and urban–rural (miejsko-wiejska) local government units (gmina).Footnote ⁵ Urban gmina comprise a single town/city, while urban–rural gmina consist of a town and villages within its orbit.Footnote ⁶ In both cases, status as a town (Hungary) or urban/urban–rural gmina (Poland) is determined administratively and not based on any particular population threshold. Data on village elections in Hungary and rural (wiejska) gmina in Poland are not included, as elections in these areas are not typically contested by national parties; instead, we focus on town and city politics exclusively.

Our dependent variable is whether the ruling-party candidate (PiS or Fidesz) wins a given municipal election. We only include cases in our analysis where PiS or Fidesz chooses to field a candidate; where they do not field a candidate, our dependent variable is coded as missing, so as not to bias the results. In other words, the dominant party should not be considered to have lost a race it did not contest. We have two independent variables of interest. To proxy pre-electoral strategic coordination, we use the number of candidates running for office in a given municipal election.Footnote ⁷ We also use a dichotomous variable indicating whether an election required a second round. Uncontested elections are included in our data. The data come from the respective national election agencies and have been consolidated into a panel data set.

We include several control variables to account for potential confounders. In national-level elections, the strongholds of PiS and Fidesz are typically described within the parameters of the urban–rural divide familiar from other examples of populist resurgence. Fidesz evolved as a party catering to rural voters and maintained this electoral strength as it achieved structural dominance (Knutsen Reference Knutsen2013), while PiS performs much worse in urban areas in national elections (Marcinkiewicz Reference Marcinkiewicz2018). Both parties are considered to have their base in small- to medium-sized municipalities that are less economically dynamic, whose residents are older on average and have lower levels of educational attainment. By contrast, the liberal opposition generally prevails in larger cities with expanding economies and younger, better-educated residents. We control for three factors often tied to populist party success. The first is municipal population, as resistance from urban centres has been prominent in challenges to dominant parties in largely (Scheiner Reference Scheiner2006), semi- (Diaz-Cayeros and Magaloni Reference Diaz-Cayeros and Magaloni2001) and un-democratic (Wallace Reference Wallace2013) contexts. The second is economic distress, operationalized as the local unemployment rate. The third is generational–cultural backlash, the idea being that support for illiberal populists is driven by a return to traditional values in the face of progressive cultural change, especially among older voters (Norris and Inglehart Reference Norris and Inglehart2019). We operationalize this as the percentage of the local population over 60 years old.

Finally, we include a dummy variable to control for the presence of a dominant-party affiliated incumbent in the race. Incumbency is generally considered advantageous for a candidate in any type of election, but research shows it to be especially valuable in mayoral contests (Freier Reference Freier2015; Holbrook and Weinschenk Reference Holbrook and Weinschenk2014). Blane Lewis (Reference Lewis2020) also finds that pre-election coalitions may not matter for incumbent mayors as much as for challengers; this indicates that an incumbent affiliated with a nationally dominant party may be especially well insulated against a united opposition coalition. In Poland specifically, Jarosław Flis (Reference Flis2018) argues that incumbency is advantageous in local elections, while in Hungary Stenberg (Reference Stenberg2018) finds that opposition incumbents make it substantively less likely for Fidesz to win mayoralties. The perceived strength of an incumbent becomes an important strategic calculation for potential candidates seeking election as mayor (Flis et al. Reference Flis, Gendźwiłł and Stolicki2018). Dominant-party incumbents may be especially advantaged in Polish and Hungarian mayoral elections due to the politicization of the media apparatus by the nationally dominant party (Turska-Kawa and Wojtasik Reference Turska-Kawa and Wojtasik2020).

Control variable data come from the respective national statistical agency's subnational data sets; population uses logged electoral year estimates. Table 3 summarizes the variables.

Table 3. Variable Specifications

Notes: Variable type: DV = dependent variable; IV = independent variable. Hungary source: NVI = National Election Office; KSH = National Statistics Office. Poland source: KBW = National Electoral Office; GUS = Central Statistics Office.

To analyse these data, we use a panel logit model since our key dependent variable is dichotomous. We use a random effects model, as there is not sufficient variation within each case across time on all variables of interest (Bell and Jones Reference Bell and Jones2015). Random effects also capture differences between cities that do not vary systematically. Where appropriate, country and cycle fixed effects are included.