Authoritarian footprints in Central and Eastern Europe

ABSTRACT Central and Eastern Europe is the last world region to transition towards democracy. Today, it shows alarming signs of de-consolidation, most prominently in Hungary, Poland, and Serbia. This article assesses whether these observations form part of a systematic pattern across the region. It relies on newly-updated objective data from the Democracy Barometer for the period between 1990 and 2016. It revisits evidence for the three most prominent explanations of democratic backsliding in the region: the rise of populist parties, the incapacity of the European Union to secure democracy once pre-accession incentives weaken, and the global financial crisis.


Introduction
Never before in history has a greater extent of the world been under democratic rule than today. In spite of this, political observers are deeply concerned with the global state of democracy (Merkel 2015). In particular, many accounts highlight problems in the last regional cluster of countries to democratise: the post-communist countries in Central and Eastern Europe (CEE). After an initially rapid transition towards democracy in the 1990s, the region has shown worrying signs of de-consolidation, and has, in recent years, even been diagnosed as being on the edge of an authoritarian backlash.
Recent political developments have further stoked fears about democratic backsliding 1 in the CEE region, even resulting in assessments that they may herald the beginning of a new, reversed wave of democratisation (Diamond 2015;Levitsky and Ziblatt 2018;Mounk 2018;Snyder 2018). In particular, political parties with illiberal programmes have entered government not only in Poland, Hungary, and Serbia, but also in Bulgaria, Slovakia, and most recently in the Czech Republic (Rupnik 2007(Rupnik , 2018Enyedi 2016;Kelemen 2017;Mounk 2018). In some cases, they have subsequently exploited the weakness of the media and judiciary to strengthen partisan control over the state, for example in Macedonia and Serbia (e.g. Esen and Gumuscu 2016;Bieber 2018). The region's international and economic environment further reinforce these warnings: While external actorsin particular the European Unionhave been important for promoting democracy in the region in the past (Schimmelfennig and Sedelmeier 2004;Agarin, 2020), they seem unable to guarantee the same incentives in the long run (Carothers 2015). In addition, the repercussions of the global financial crisis have weakened both democracy at home, as well as the capacity of external actors to implement their democracy agenda (Armingeon and Guthmann 2014;Morlino and Quaranta 2016).
Many of these pessimistic accounts are based on qualitative assessments of single countries (e.g. Enyedi 2016; Krekó and Enyedi 2018;Pehe 2018) or on the observation of trends across small sets of cases. These often include the same set of cases: Mečiar and Fico in Slovakia, Orbán in Hungary, and the Kaczynskis in Poland (Brusis 2016;Bustikova and Guasti 2017;Grzymala-Busse 2017;Kelemen 2017;Luce 2017;Rupnik 2018; but see Cianetti, Dawson, and Hanley 2018). In contrast, the few existing quantitative investigations of recent democratic developments have either focused on Latin America (Huber and Schimpf 2016b) or they include only a few EU members from Central and Eastern Europe (Huber and Schimpf 2016a;Huber and Ruth 2017). 2 Studies with global samples face difficulties of distinguishing between potentially strongly diverging processes such as the entrenchment of authoritarian systems or backsliding in the quality of democracy (e.g. Lührmann et al. 2018).
Contributing to this debate, we identify and assess the state of the quality of democracy in all 19 democracies in Central and Eastern Europe. 3 In addition to taking stock of recent changes, we assess their associations with the most prominently-cited explanatory factors behind purported democratic backsliding in the region. In particular, we investigate three widely-cited explanations: first, the increasing role of anti-elite, populist parties in government; second, the European Union's fatigue with efforts to sustain improvements in the quality of democracy once an applicant state has joined the Union; and, third, the financial crisis of 2008, which has deeply affected the economies of CEE countries and which has had serious repercussions not only for their political systems, but also for the capacity of the European Union and its democratic agenda.
Our regional focus on Central and Eastern Europe offers three advantages: First, as earlier research has found important differences in the effect of populism on democracy between consolidated and weak democracies Rovira Kaltwasser 2012a, 18-26, 2012b), the recent transition of the region towards democracy offers a set of similar cases for comparison. Second, it also offers crucial variance for the most prominent explanations for the purported backsliding. On the one hand, Central and Eastern European democracies have experienced a strong influence of populist actors (Grzymala-Busse 2017) from all ideological families, thus enabling a more systematic assessment of their relationship with democracy than regions dominated by populists of the right (Western Europe) or the left (Latin America). On the other, it also offers the opportunity to assess the role of external factors, in a region where EU accession processes are linked to a political agenda of promoting liberal democracy (Schimmelfennig 2005). Third, the investigation of gradual changes in the quality of democracy enables us to focus on the most crucial process in the region, leaving aside inter-regime transitions that are more relevant in other contexts.
For our analysis, we rely on the Democracy Barometer dataset. This has two advantages, First, it is based overwhelmingly on objective or survey data and thus allows us to avoid the known problems of expert surveys (e.g. Silva and Littvay 2019)often recruited among academics with strong liberal-democratic norms. Second, it is suited for an assessment of the Quality of Democracy in a fine-grained, disaggregated fashion, along nine principles. We have extended the data to cover the period until 2016 across 70 democracies, including 19 in Central and Eastern Europe (Merkel et al. 2018).
Our results indicate that the widespread public perception of a deep crisis of democracy in Central and Eastern Europe may be exaggerated. While improvements in the quality of democracy that started in the 1990s have indeed faltered, and while some democratic functions are under threat in some countries, this does not appear to constitute a general trend towards the deterioration of democratic quality across the region. However, according to our data, in several countries authoritarian leaders or governing parties have left a significant mark on two functions of democracy: the rule of law and the freedom of the press. In particular, concern seems appropriate about press freedom, transparency, competition, and the rule of law in Moldova, Bulgaria, the Czech Republic, Slovakia, Ukraine, and Hungary. Vice-versa, European integration has been able to foster some principles of democracy in the course of the EU accession process, but the EU has no powers in preventing post-accession backsliding under populist governments.
In the rest of this paper, we proceed as follows. First, we introduce the term "quality of democracy", discussing both its conceptualisation and operationalisation with the newlyupdated Democracy Barometer dataset. Second, we use this dataset to empirically trace recent trends in the quality of democracy in all 19 countries in Central and Eastern Europe, identifying both countries displaying major transformations, as well as the democratic functions most strongly affected. Third, we take advantage of a large set of new data sources to trace the importance of the three suspected drivers behind recent democratic reversals in the countries of the region: populist actors in and outside government, post-EU accession fatigues, and the 2008 financial crisis. In our last section, we conclude and compare our findings to the results of expert-based assessments.

The quality of democracy
Central to our purpose of tracing recent democratic developments in Central and Eastern Europe is the concept of quality of democracy. Before starting our empirical endeavour, this section introduces its conceptualisation and subsequently presents how we operationalise it using a newly-updated version of the Democracy Barometer.
The concept of quality of democracy only applies to political regimes which fulfil minimal democratic criteria. It allows us not only to distinguish to what degree these regimes fulfil ideals established in democratic theory, but also to identify potential democratic deficits (cf. Diamond and Morlino 2004;Morlino 2004). Hybrid regimes, which at best display only a democratic facade, do not fulfil the minimal democracy criteria, rendering the concept of quality of democracy inappropriate in these cases. This caveat is especially important for our study periodafter the end of the Cold Waras it brought about the rise of political regimes which conduct regular elections, but are only superficially democratic (Diamond 2002). These countriesregional cases thereof are Belarus or Russiaare not captured by our concept of the quality of democracy and a priori excluded from all analyses. Merkel's (2004) and Diamond and Morlino's (2004) concepts of "quality of democracy" build on mid-range concepts of democracy which go beyond the electoral principle and also incorporate further liberal functions of democracy, such as the rule of law or an active political society. This more extensive, yet procedural concept explicitly foresees a number of trade-offs between the realisation of different functions of democracy found in specific types of democracy. For example, while majoritarian democracies put an emphasis on stable governments and the responsiveness of democracy to the median voter, consensus democracies emphasise a broad and inclusive system of representation (see, for instance Lijphart 1999;Powell 2000;Bochsler and Kriesi 2013).
The specific concept of quality of democracy which we use in this paper is based on an encompassing set of operationalizable functions of democracy. At its core, it relies on the functions of democratic control and decision-making. These include institutions which make the government accountable to citizen preferences (through competitive elections), which make this process inclusive (for example, through universal suffrage and inclusive representation), and which encourage broad participation in it. These, in turn are embedded in democratic rights and freedoms, as well as checks and balances, including assurances of free political debate, basic civil rights (personal liberty, freedom of expression and of information, freedom of religion, economic and social rights), the rule of law, limitations to the power of governments, provisions for transparent government procedures, and horizontal checks on government (parliamentary procedures, independent courts, bicameralism, federalism, central bank independence). A further indispensable function is provided by pluralist intermediaries between state and society, including the media, the public sphere, and civil society. Finally, to maximise the quality of democracy, governments should not be subject to extra-constitutional actors, such as the military, but should have actual power to govern. All functions combine both formal constitutional rules (establishing the legal environment for democracy), and democratic practices (for example, effective electoral competition, or the absence of state-based or private repression of journalists).
Most of the existing literature on democratic backsliding relies on measures of democracy that broadly differentiate between democracies, authoritarian regimes, and intermediate cases on one or two dimensions. The most commonly-used measures are the Polity IV, the Freedom House, and the V-Dem index of democracy (Freedom House 2018;Lührmann et al. 2018), although a small number of articles rely on other sources (Morlino and Quaranta 2016;Börzel and Schimmelfennig 2017). On all of these indices, Central and Eastern European democracies usually obtain stable ratings at the top end of the scale and, with few exceptions, were already doing so years before EU accession. This article aims for a more nuanced assessment which allows us to incorporate specific transformations of the quality of democracy into our study.
In order to identify such changes in the quality of democracy, we rely on the Democracy Barometer dataset (Bühlmann et al. 2012). Mirroring our concept, the Democracy Barometer conceives of democracy as based on individual freedom, equality, and actual degree of control the government possesses over the state. It builds on a total of 105 indicators primarily from objective sources and representative surveys rather than expert assessments. They are grouped along nine primary functions of democracy: individual liberties, rule of law, the public sphere, competition, mutual constraints, governmental capabilities, transparency, participation, and representation. Each indicator is coded so that the majority of its values, and the aggregate scores for each function, lie between 0 and 100 (see Bühlmann et al. 2012;Merkel et al. 2018, for details).
3. Trends in the quality of democracy in Central and Eastern Europe, 1990-We start our empirical analysis with a descriptive assessment of the quality of democracy in the 19 countries of Central and Eastern Europe, beginning with the year of democratisation (usually 1990 or shortly thereafter), up to 2016. Except for Belarus and Russia, all Central and Eastern European countries can be considered democracies, according to Polity IV (minimal Polity IV score of 6). 4 Figure 1 plots the overall trends for democracy in Central and Eastern Europe by democratic function. There is a clear positive trend since the 1990s, with the region making advances with regards to competition, transparency, and the public sphere. However, this upwards trend has slowed down markedly since around 2000, with some democratic functions even appearing to show decreases. Furthermore (similar to the new democracies in Latin America (cf. Bochsler and Juon 2018)) the countries in Central and Eastern Europe clearly have not attained a similarly high level of democratic quality as the established democracies in Western Europe. Most remarkably, there is a deep crisis of the rule of law (and to a somewhat lesser degree, the public sphere) in Central and Eastern Europe. However, to the extent that these reversals can be captured by our indicators, relying on objective data rather than expert-coding, our findings contradict recent claims by academics and political analyses of a widespread backsliding against the quality of democracy in the region (Bustikova and Guasti 2017;Rupnik 2018). Indeed, no such general trend can be observed on any of the nine functions of democracy which we assess, at least as regards the time period until 2016 which is included in our analysis.
Figure 2 unpacks these regional averages to reveal country-specific trajectories, showing separate trends for each of the 18 Central and Eastern European countries for which we have data (excluding Kosovo, due to missing data). 5 Reflecting the regional average, the quality of democracy has remained stable or improved on most functions of democracy in most countries, in particular in two Baltic States (Latvia, Lithuania), in Poland, and in Sloveniathough in each of the countries with significant (enduring) weaknesses. 6 However, there are temporary fluctuations in some of the indicators (especially those measuring government capacitythis is partly related to measurement issues). The subsequent analysis focuses primarily on important instances of backsliding, rather than on persistent democratic shortcomings. As regards particular democratic functions, the most systematic development, observable in several countries, is the erosion of the rule of law. Our measure for the rule of law reached its peak in the late 1990s, with several countries establishing a truly independent judiciary and with landmark judgments signalling to governments that the courts constituted a genuine limit to government power (Bugarič and Ginsburg 2016), though suffering from significant informal infringements elsewhere (e.g. Bugarič and Kuhelj 2015). However, since the 2000s, we observe a erosion of the rule of law in countries such as Slovakia, Ukraine, Bulgaria (since 2000), Romania (around 2012Romania (around -2013, and in Albania (since 2011). The underlying reasons for this are not only institutional changes, but also government infringements on the judiciary in practice, including replacements of members of the judiciary with party loyalists (Bugarič and Ginsburg 2016). Other symptoms of this process are significant backlashes against judicial independence and an immense drop in public confidence in the judiciary and police.
A further noteworthy development is the drastic reduction of political competition in three countries in the 2010s. In all three, this was related to the surge of a new, dominant leader and political party. In Hungary, Viktor Orbán and his Fidesz-KDNP alliance won a landslide election victory in 2010, gaining a 68% supermajority in parliament (cf. Sata and Karolewski 2020). In Romania, the Romanian Social Liberals of Victor Ponta became dominant after the 2012 elections (59% of the seats). In Serbia, the Serbian Progress Party won 29% of the seats in 2012 and has dominated politics since 2014 with a 63% majority in parliament. While the changes in these countries are suggestive, they clearly do not represent an overall regional trend.
On other democratic functions, the developments are even more heterogeneous, with recent decreases appearing to reflect country-specific circumstances rather than a regional trend: First, there has been a recent drop in participation in several countries, in particular in the Czech Republic after 2001 and in Hungary after 2014, driven by the curtailing of suffrage through new electoral rules. Second, there has been an erosion of transparency in Hungary since 2009 and in Serbia since 2015. Both countries have reduced press freedom (Huszka 2018). In Hungary the transparency of government communication has also suffered. Third, individual liberties deteriorated in Ukraine and in Moldova in 2014, culminating in riots. They have also been under threat in Slovakia since 2005 with de-facto violations of political liberties and reported cases of torture. Fourth, the public sphere has deteriorated in several countries. Croatia experienced an erosion of its civil society and of economic interest groups in the 2010s (as indicated by declining membership numbers). In Estonia, the press has become increasingly politically unbalanced since 2002.
Overall, our analysis finds both parallels and differences in the development of the quality of democracy in the countries of Central and Eastern Europe. It does not support the notion of an overall regional deterioration in the quality of democracy, however. On the one hand, the upwards trend of the 1990s has certainly stopped, and, as regards several specific countries in the region, has reversed direction. However, the general trend of democratic backsliding, alleged by country experts, is not reflected across countries and functions, at least not in the objective indicators on which the Democracy Barometer dataset relies and at least not in the time period we are able to consider (until 2016). In other parts of the region, the quality of democracy has even improved, although this process was usually similarly limited to some democratic functions. While our data thus contradicts pessimistic judgements for the whole region, the time period covered does not allow us to consider more recent alleged impairments of checks and balances and the liberal rights regime through constitutional changes and non-constitutional infringements of individual and minority rights between 2017 and 2018 (Sandurski 2019).

The three processes behind democratic changes in CEE
Having mapped the heterogeneous democratic developments in the CEE region, we now proceed to analyse the processes behind them. In particular, we consider the three most prominently-cited factors behind purported democratic backsliding in the region: first, populist-authoritarian parties in government and in opposition; second, reduced incentives to uphold liberal democratic principles after EU accession; and, third, political repercussions of the global financial crisis of 2008. In order to do so, we have collected data for each of these three factors and descriptively investigate their associations with democratic changes in our sample. In our main text, we focus on the four functions of democracy for which our data show the highest variance: Individual liberties, rule of law, competition, and transparency. 7 While sample limitations prevent us from conducting a causal investigation and from considering interrelated effects, our systematic bivariate investigation nevertheless enables us to trace patterns and serves as a plausibility test for the three explanations. The discussion highlights the most influential cases driving these developments, as well as the heterogeneity between individual countries.

Process 1: authoritarian-populist governments
The first widely-cited regional driver of democratic changes, especially of recent erosions of the quality of democracy, are authoritarian leaders, linked to the emergence of strong populist parties. In Central and Eastern Europe, they played an important role long before populism spread over the continent. Some of these parties were populist with an emphasis on identity politics, in particular the nationalist parties in the Western Balkans (e.g. the Croatian Democratic Union, HDZ, in Croatia; the Serbian Radical Party, SRS, in Serbia; or the Internal Macedonian Revolutionary Organisation, VMRO, in Macedonia). In other countries, authoritarian leaders were political entrepreneurs combining populist ideology with centrist or mainstream political positions. Some rose quickly and became the new political elites, such as the former Bulgarian king Simeon Sakskoburgotski (prime minister 2001-2005), or newcomer political parties in several of the Baltic states (Sikk 2012;Hanley and Sikk 2016). However, it was only after the transformation of Viktor Orbán's Fidesz into a populist-authoritarian party in Hungary and its rise to power in the 2010 elections that populism was discussed as a threat to liberalism and democracy.
We define populism as a "thin ideology": populists juxtapose a supposedly "homogeneous people" with an elite class they portray as corrupt and as dominating politics, the economy, society in general and/or the media (Mudde 2004). This defining feature of populism can be combined with diverse "thick" ideologies (e.g. identity politics/nationalism, socialism, conservative thought) or issue positions, for example those of the far-left (March 2011), the far-right (Mudde 2011;Huber and Schimpf 2016a), and the political centre ( van Kessel 2015). Thereby, the defining features of populism are anti-liberal, anti-pluralist and anti-democratic, and it can transform or damage democracy (Mudde and Rovira Kaltwasser 2012a). Furthermore, it also closes the door to external actors promoting democracy (Carothers 2015). 8 To identify populist parties, and in line with our definition of populism, we relied on existing sources 9 (Hawkins 2009(Hawkins , 2013March 2011;Mudde 2011;van Kessel 2015;Hawkins and Silva 2016;Schimpf 2016b, 2016a), and combine them to maximise coverage, 10 partly using temporal extrapolation. 11 Based on an underlying partylevel coding (which included all parties with at least 2% of the seat share in the first chamber of parliament), we then created our country-level variables for populist parties in government and opposition, discarding country years where we could not account for the populism status of a significant share of parties. 12 Figure 3 identifies the footprints of populists in opposition (left panel), in government as partners or majority parties (middle), and in electorally dominant governments (right). 13 It shows the trends for (uninterrupted) years of populist activity, with the number of years that populist actors have been in government or in opposition on the X-axis. Multiple populist time spans within the same country are included as separate cases. The trend estimates in Figure 3 and all subsequent figures (in red) are based on bivariate models with period-fixed effects; we interpret them when they show a substantially important change. 14 Populist governments with at least two-thirds of parliamentary seats have a particularly dominant role, which in many countries corresponds to the threshold for changing the constitution.
Populist footprints leave three observable marks on the quality of democracy. First, we find a drop in the rule of law in four countries right after populist administrations enter office. This is the case for Vasile Tarlev (Moldova, 2001), Mirek Toplánek (Czech Republic, 2006), Robert Fico (Slovakia, 2006), and Boyko Borisov (Bulgaria, 2009. In Ukraine, the rule of law is already extremely weak at the beginning of the second Timoshenko administration in 2007, and only drops slightly further (see Figure 1). However, as the comparative analysis also shows, there does not appear to be a homogeneous relationship between populist governments and developments of the rule of law. Instead, there is considerable fluctuation in the trend, and the presence of pure downward trends is rather the exception than the rule.
Second, under some populist cabinets, transparency deteriorates. This is the case for Ivo Sanader in Croatia (2003), Robert Fico in Slovakia (2006), Viktor Orbán in Hungary (2010) and Beata Szydło in Poland (2014). Again, this authoritarian footprint is not homogeneous. Other populist administrations, have boosted, rather than weakened the transparency of rule, e.g. the right-wing populist cabinets of Adrian Năstase and Călin Popescu-Tăriceanu in Romania (2000Romania ( -2007. This goes back to a period when the Romanian government's primary agenda was access to EU membership. Third, competition drops sharply when populist governments take office in Hungary (after 2010), Bulgaria (after 2009 and and Moldova (in 2001). In Hungary, the victory of the Fidesz was accentuated by an electoral system with partly majoritarian features, which created a huge discrepancy in the representation of the governing party and the oppositionthe Fidesz party won 68% of the seats with only 53% of the votes. In all these cases, the drop in our measure of competition is partly due to the effective dominance over the party system by populist parties, and partly because of changes to electoral rules and rules on party funding.
While the populist critique of the state of liberal democracies aims to strengthen vertical accountability at the expense of strong horizontal checks and balances, we find that this critique in some of the countries disguises the real impact of populists on democracy. When populist governments do change the institutional setup of democracy, they also target aspects of democracy which are crucial for citizen control over their government, such as party funding rules or freedom of the press. However, in our sample, populism is neither a sufficient, nor a necessary condition for any of the erosion in the quality of democracy: the rule of law is also seen to suffer under some of the non-populist governments, e.g. Albania (2010)(2011)(2012)(2013)(2014)(2015)(2016) or Hungary (1999)(2000)(2001)(2002). Freedom of the press and government transparency can similarly also suffer in non-populist politics, e.g. Poland (1996Poland ( -1999 or Hungary (1999)(2000)(2001). For all other functions of the quality of democracy, the trends under populist governments are even more heterogeneous. The same applies for periods where populist governments govern with a twothirds majority in parliament, or for those where populists emerge as viable opposition forces. 15 These patterns do not allow us to reject the explanation that populist rule has transformed democracy in Central and Eastern Europe. However, they show that populist governments differ significantly with regards to our research question, and that the relationships between populist rule and the quality of democracy are strongly contextdependent (cf. Agarin 2020). Across countries, the targets of populist governments vary, and in several contexts our data has not picked up alterations in democratic practices, at least not yet. In particular, the populist-authoritarian backslides remain limited to countries with only weak EU membership perspectives (Moldova, Ukraine), or to periods after CEE countries have acceded to the European Union.

Process 2: EU political integration and post-accession backsliding
According to a second line of argument, declines in the quality of democracy in Central and Eastern Europe are directly related to processes occurring in post-EU-accession periods. In this reasoning, before granting membership, the EU is able to render any deepening of European integration conditional on political criteria. Among other factors, the Union demands that prospective members improve their democratic, human rights, and minority protection credentials, and that they enshrine the rule of law (Pevehouse 2002;Dimitrova and Pridham 2004;Freyburg and Richter 2010;Kolev 2020). While these effects might be less important for liberal political parties, with a democratisation agenda, EU conditionality could play a peculiar role in countries with a mixed government between liberal and authoritarian parties (Schimmelfennig 2005). Once a country has achieved membership, as is the case for 11 countries in the region, this conditionality mechanism vanishes. While the Union has a limited range of sanctions at its disposal, they are not very effective, politically costly, and difficult to activate, requiring a unanimous decision after a lengthy procedure.
Recent regional developments illustrate this reasoning. The EU disciplinary proceedings against Poland and Hungary, activated in 2017 and 2018 due to their systemic threats to the rule of law, have remained largely ineffective. As long as the two countries' governments back each other, the necessary unanimity for disciplinary sanctions cannot be reached. In addition, the financial crisis of 2008 has put the success story of democratisation, European integration, and liberal markets into question, especially in Central and Eastern Europe. There, popular hopes that democracy would also deliver economic welfare gave way to disappointment, also spawning doubt in democracy as a political system (Mishler and Rose 1996;Brusis 2016;Bochsler and Hänni 2019).
As regards previous evidence, several qualitative studies come to the conclusion that the EU cannot prevent the erosion of democracy and the deterioration of the rule of law, media and academic freedom in some of its new member states (Kelemen 2017;Enyedi 2018). However, there is no systematic evidence for a drastic anti-liberal turn after countries access the European Union. Börzel and Schimmelfennig (2017) do not find any significant and systematic trends in the level of democracy, once new member states have accessed the Union, although other studies indicate that after this crucial moment, the speed at which new members implement pro-democratic reforms slows down (Levitz and Pop-Eleches 2010). However, Levitz and Pop-Eleches' analysis only includes two post-accession years, and stops before the global financial crisis of 2008, while Börzel and Schimmelfennig rely on means across countries in the region, assuming homogeneity in the preand post-accession effects.
Different from these analyses, we employ our new multi-dimensional measure of the quality of democracy, tracing heterogeneous developments, and distinguish between periods of conditionality and post-accession. In particular, we assume that conditionality is in play only before countries make the step towards a higher level of integration, while post-accession fatigue only becomes a factor after countries have reached a more intensive degree of integration. To assess the role of conditionality and EU accession, we use the four levels of integration identified by Börzel and Schimmelfennig (2017): Association Agreements, weakly credible accession perspectives, credible accession perspectives, and EU membership. 16 Figure 4 assesses the development of democratic functions in the time period including the 5 years before and the 10 years after the four steps of EU integration, focusing again on the most volatile dimensions of the quality of democracy (see appendix B2 for a full assessment). In brief, it indicates that EU integration is a success story for the quality of democracy before membership is achieved, although the trajectories become heterogeneous thereafter.
A first insight is that significant advances in democratic qualities occur right before countries are offered weakly credible accession perspectives, which is due to important reforms passed at these crucial points in time. These advances are related to several democratic functions. One case, which drives this effect in particular, in the field of competition, is Croatia. It has reformed its electoral system, while Albania improved its rules on party funding in 2000. Competition also de facto improved in Romania in 1992, in Macedonia in 1998, and in Albania and Croatia in 2000-2001 In the fields of individual liberties, the rule of law and the public sphere, the three Baltic States (Estonia, Latvia, Lithuania) passed their post-Soviet constitutions in 1992/1993, or reinforced pre-Soviet constitutions with more expansive provisions. In particular, they introduced or reinstated constitutional rules concerning the protection of individual liberties, and rights and liberties related to public communication, the independence and professionalism of the justice system and public trials. Albania experienced a significant improvement in the area of law and order in 1998, after public security was reformed in 1997. In contrast, transparency improved in several countries (Croatia, Estonia, Latvia, Serbia and Montenegro, Slovenia) right after they were granted weakly credible accession perspectives, rather than before.
A second major improvement across multiple cases and dimensions occurs in the years before countries are offered credible accession perspectives: Representation became more inclusive in Poland in 1993, in Albania in 2001, and in Macedonia in 2002. Political competition intensified in Romania in 1993, in Latvia in 1996, in Lithuania 1997, and in Bosnia-Herzegovina in 2001. In many places, constitutions were also amended at these points in time, guaranteeing freedoms related to the public sphere. This concerned, in particular freedoms of association, assembly speech and/or of the press, for example in Latvia in 1997, in Lithuania and Poland in 1993, and in Slovenia in 1994. Furthermore, the public sphere improved de facto in Croatia in 2002. Similarly, transparency increased due to improvements in press freedom (legally and de facto), and due to party finance disclosure rules in several countries (Poland in 1991and 1993, in Slovenia in 1994, Estonia and Latvia around 1993-1995, and Lithuania by 1997. Hungary lifted legal restrictions on freedom of information between 1990 and 1992. Croatia lifted restrictions on freedom of information around 2000-2003, and introduced effective press freedom, both legally and in practice. In neighbouring Serbia-Montenegro, this step occurred simultaneously, but was less successful: in 2003, press freedom was under threat, and corruption prevailed.
In contrast to advances due to accession perspectives, EU membership itself is not systematically associated with improvements in the quality of democracy, with our measures failing to pick up any homogeneous trend. Three governments became more transparent in the years before their countries became EU members. Slovakia in 2001 and Poland in 2002 lifted restrictions on freedom of information. Slovakia also improved the legal environment for press freedom and Romania introduced legislation for transparent party financing in 2003.
In addition to these improvements, we also find evidence for several instances of democratic erosion occurring at different stages of the EU accession process. Some countries experienced important relapses in some of their democratic functions in the aftermath of EU accession, for example the aforementioned cases of Hungary, Slovakia, Bulgaria, and of the Czech Republic. However, such backslides on some democratic functions occur not only in EU members, but also in countries with Association Agreements (e.g. Moldova, Ukraine) and those with future membership perspectives (Albania, Serbia). Among EU member states, they occur in those governed by populist governments (e.g. rule of law and individual liberties in Bulgaria, Slovakia), as well as in those by non-populists (same functions, in Poland, 2009, or Romania, 2010-2012. Hence, while association agreements are often associated with an improvement in some of the democratic functions analysed, EU membership is neither a sufficient, nor a necessary condition, not even in conjunction with further factors analysed in this article, for changes in any of our functions of democracy. Once countries become EU members, their pathways are very heterogeneous.
In sum, on the ladder of European integration, the most important positive steps for the quality of democracy occur in the initial stages where countries aim for either weak or strong commitments in order to achieve integration into the Union. Less important is the last step -EU membershipas well as other forms of association, which do not come with the promise of later membership. The improvements are based on the logic of conditionality: reforms with regards to individual liberties, minority protection, rule of law, transparency, etc. occur before, and not after reaching a new stage of integration. Once a new step of integration is reached, the EU instruments provide little leverage, unless the EU in in a position to reward democratic progress with even further integration.

Process 3: the consequences of the financial crisis
A third factor that is an often-cited driver of alleged regional erosion of the quality of democracy is the global financial crisis of 2008. In particular, it has spawned significant concerns about how young democracies can cope with economic turmoil (Bermeo and Bartels 2014;Kriesi and Pappas 2015). The severe economic consequences which a number of democracies in the region have suffered in the wake of the financial crisis have direct and indirect implications for political regimes. First, they can lead to a collapse of trust in political institutions and dissatisfaction with democracy (Armingeon and Guthmann 2014). Second, social unrest can precipitate the collapse of governments. Third, the preoccupation of European Union institutions with stabilising the financial sector and the common currency reduces the Union's capacity to deal with democracy-related issues and precipitates enlargement fatigue, thus curtailing the enlargement effect on the quality of democracy in not-yet member states.
In order to assess these arguments, Figure 5 plots four democratic functions before and after 2008. However, we do not find a reversal of the regional trends after the crisis for any of the nine functions (see also appendix, figure B3). Four countries (Bosnia and Herzegovina, Bulgaria, Croatia, Hungary) experienced important backslides in individual liberties after the financial crisis, partly because of civil and political rights violations by states, reported by Amnesty International and US State Department reports. In two countries (Bulgaria, Slovakia), the rule of law suffered, and in Moldova the pre-existing crisis of rule of law deepened further. In Albania and in Hungary, competition deteriorated, and the governing parties became more dominant. However, beyond these individual cases, our more systematic assessment of political developments across the region corroborates Ekiert's (2013) observation that the financial crisis has not initiated an illiberal turn in Central and Eastern Europe. Instead, this occurred only where authoritarian parties in power managed to channel the grievances against the liberal institutions.

Discussion
Since the 1990s, the countries of Central and Eastern Europe have seen an unusually fast transition towards, and a consolidation of, democracy. Has this trend now started to reverse and has the region moved into a period of de-consolidation of democratic rule, a period of defective democracies with illiberal features? Both the academic and the public debate have centred around such a purported anti-liberal backslide (Rupnik 2018). Prominent explanations for the perceived erosion in the quality of democracy include the rise of populist-authoritarian parties that dominate the national governments in several countries of Central and Eastern Europe (Brusis 2016;Bustikova and Guasti 2017), the inability of the European Union to enforce democratic standards after countries have been granted membership (Sedelmeier 2014;Kelemen 2017), and the economic and political consequences of the global financial crisis of 2008 (e.g. Bermeo and Bartels 2014;Brusis 2016).
This paper offers a new, systematic assessment of this alleged illiberal turn of Central and Eastern European democracies. It compares the trajectory of the quality of democracy across all democracies in Central and Eastern Europe, relying on a newly extended dataset covering nine dimensions of the quality of democracy, which is based on objective indicators. It examines the trends in the quality of democracy in light of the three above-mentioned explanations, analysing both the overall regional trend as well as subsets of countries. In brief, while it finds correlational evidence consistent with purported democratic effects of populist governments and the EU accession process, it does not detect any systematic changes associated with the financial crisis.
As regards the former, the rise of populist actors, the article finds that Central and Eastern Europe has seen the meteoric rise of newcomer political parties long before this trend spread to Western Europe (Sikk 2012). Some of these parties ran with a populist programme, and by 2016, every one of the 17 countries of the region for which we have measures of populism available experienced at least one episode of a populist party in government. Populists view themselves as correctors of democracy. In their eyes, the institutions of horizontal accountability, and some liberal rights are tools in the hands of corrupt elites. They decry what they perceive to be the excessive protection of minority rights, weakening the majoritarian principle of democracies (Houle and Kenny 2018). Instead, they want to strengthen vertical accountability, strictly following the "will of the people"a combination of majoritarian and plebiscitary ideas.
The analysis of the quality of democracy in Central and Eastern Europe finds mixed evidence for these purported effects of populists in government. In some cases, they are clearly associated with changes in the quality of democracy: The most important patterns thereof which the Democracy Barometer reports are a reduction of political competition and moves towards less transparent governments during populist rule. While our objective indicators identify, to a considerable extent, the same backsliding democracies as expert-based data (Lührmann et al. 2018(Lührmann et al. , 1327, the insights gained from our analysis (limited to 2016) rebuff the alarmist view of a deep and widespread crisis of democracy. Different from studies of other regions (Huber and Schimpf 2016b), we identify several populist governments that have not left any mark on the quality of democracy, while others have even made changes that reduce future opportunities to exert popular control. However, our empirical assessment is also not in line with the view of populists as a corrective (Huber and Ruth 2017). Developments in the quality of democracy are not uniform in the region, and the observable footprints of populist-authoritarian governments on the quality of democracy appear highly contextdependent.
As regards the process of European integration, the article overall confirms the assessment that it advances the quality of democracy in the region. This appears to occur right before the countries take steps on the ladder of European integration, in particular in the earlier stages of the process of European integration, when countries are aiming for either weak or strong commitments for integration into the Union. However, after countries have taken steps towards European integration, they can de-consolidate, owing to a lack of disciplinary instruments. Such steps backwards have occurred in EU member states (Hungary, Slovakia, Bulgaria, and of the Czech Republic), but not only there. In many ways, the deterioration of the public sphere and of transparency of government action under the Vučić presidency in Serbia (outside the EU), closely resembles the deterioration of press freedom under the Orbán administration in Hungary. In Macedonia and Serbia, governments have profited from the fragility of the political and legal institutions and the media system. They have used democratic weaknesses to consolidate their partisan control of the state, while dismantling liberties and horizontal controls (e.g. Esen and Gumuscu 2016;Bieber 2018). Political pluralism in Serbia and Hungary has been erased by dominant ruling parties, which have further consolidated their rule through changes to electoral laws and rules on party funding.
In spite of indicating a crucial role for both populism and the EU accession process, the article cannot confirm an overall regional trend that is attributable to the two. While drastic cases, such as Hungary and, more recently, Poland, that highlight both the influence of populist actors and the weakness of EU instruments are given extensive media and academic spotlight, they do not seem representative for the region. Our assessment indicates that there is relatively limited backsliding until 2016, which can be observed by relying on objective quantitative data (though some important anti-constitutional moves have taken place since, esp. in Poland, cf. Sata and Karolewski 2020). They are counterbalanced by improvements to the quality of democracy in other cases.
None of the three explanations this article has investigated is either sufficient or necessary on its own to explain erosions in the quality of democracy. If the quality of democracy in Central and Eastern Europe is understood as a competition between the European Union and populist-illiberal leaders, then the score remained undecided as of 2016: The accession of CEE democracies to the EU means that the EU loses its trump cardconditionality. However, the quality of democracy remains stable in most of the new EU member states. While the global financial crisis of 2008 slowed down enlargement, and therefore weakened the power of the conditionality card regarding non-members, it was not followed by a genuine backslide. The incapacity of the Union to prevent major backsliding in member-states (Hungary, Poland) and non-members (Serbia, Ukraine) might, however, be read as a green light to populist-authoritarian governments across Europe to take authoritarian measures and compromise the EU's capacity to sustain the quality of democracy. In this way, rather than only exhibiting unsystematic, localised impacts, the authoritarian footprints left by populist actors may be felt more strongly across the region in the years to come. Notes 1. We define "democratic backsliding" or "erosion of the quality of democracy" as a gradual deterioration of democracy or some democratic principles (Waldner and Lust 2018), i.e. changes within a political regime of democracy. Authoritarian backlashes are understood as a process of de-democratisation. 2. But see Börzel and Schimmelfennig (2017), who study post-accession dynamics. However, their analysis of regional and sub-regional means does not allow the identification of heterogeneous country developments. 3. Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Macedonia, Moldova, Montenegro (2006), Poland, Romania, Serbia (2006, Serbia andMontenegro (2001-2005), Slovakia, Slovenia, the Ukraine. As for Kosovo many indicators are missing, it is only included in parts of the analysis. 4. We also include Bosnia and Herzegovina, which Polity IV does not rate as a democracy due to the international supervision active there. We do so because Bosnia and Herzegovina has fully developed domestic institutions. 5. In addition, the figure also shows periods of populist parties in governing cabinets. The operationalization for this factor is discussed in section 4.1 below. 6. For instance, Latvia suffers from a lack of inclusive participation and representation. A weak civil society, and informal practices and a deteriorating public confidence in the justice system weaken the rule of law in Latvia, Lithuania and Slovenia. (Bugarič and Kuhelj 2015). 7. In our supplementary material, appendix A provides corresponding results for all nine functions. 8. Alternatively, populism is seen as a "corrective" to the current state of democracy, eliminating allegedly corrupt elites' control of the state (through horizontal accountability, and some of liberal rights), simultaneously freeing it from the grip of minorities, and instead strengthening vertical accountability (Ruth and Hawkins 2017), revitalising participation, including the participation of lower classes (Anduiza, Guinjoan, and Rico 2019). 9. Where the sources coded presidential candidates (in particular, Huber and Schimpf 2016b) or chief executives (in particular, Hawkins 2009Hawkins , 2013, we first looked up these individuals' party affiliations with a quick web-based search. 10. The sources differ slightly in their definition of populism, and we combine sources that identify the populist left, right, and the centre. Where several sources exist for the same party, we accept a single identification as populist as sufficient. Most sources directly code political parties in a dichotomous way. Where previous sources used a continuous measure of populism, we define a cut-off point (0.5 for the measures provided by (Hawkins 2009(Hawkins , 2013Hawkins and Silva 2016). 11. We extrapolated the eight sources for each party in our dataset across the maximum time period , regardless of their time coverage. Following our sources, our aggregated measure allows for changes in a particular party's populism status over time. We did not extrapolate across a party year if at least one source codes a change in the populist status at that point. For example, van Kessel (2015) codes the Hungarian Fidesz party as populist only from 2006 (and as not populist before). Consequently, we did not extrapolate any other source's measure (for which some of the election periods were missing) across this time period. 12. Even though we combined a large number of sources to obtain as broad a coverage as possible, we were still unable to code populist status for a number of mostly smaller or more recently-emerged parties. In order to keep a maximum of information in, we kept country years where at least one governing party is populist (and where, according to our measure, the maximum degree of populist power access is thus already achieved, no matter how any missing parties are coded), and country years for which we have populist status data of at least two-thirds of the governing coalitions' seats (i.e., if we lacked data on a small coalition partner. If the size of the uncoded coalition partner was at least 33% of the total seat share of the governing coalition, we discarded the country-year). Out of a total of 988 parties that form the basis of the country years in our analysis, we have populism data for 861. 13. Defined as governments with a seat share of 67% or larger in parliament, under participation of populists. We follow Huber & Schimpf's argument according to which the presence of at least one populist party in government suffices to decisively influence policy outcomes. For simplicity's sake we do not consider the special case of surplus cabinets where populist government parties have an arguably smaller influence (Huber and Schimpf 2016a). 14. See model results reported in Online Appendix B, calculated for all functions with at least ten separate periods. 15. We find significant coefficients for populist governments or oppositions in our full models (appendix, tables B1a and B1b), e.g. on competition or constraints, or representation. Where these coefficients are minuscule and are attained in specifications without controls for timevariant confounders, we do not consider them as sufficient evidence for a discussion in the paper. 16. We omit Neighbourhood Policy. Our sample contains too few country years for this weakest form of integration.