Introduction

Existing studies on government formation and party system stability in Europe suffer from a number of limitations. First, they suffer from a “survival bias” because they neglect historical cases, that is, cases that ceased to function at a point in time. More specifically, they tend to be confined to the post-Second World War period. Second, they usually provide a partial view of European party politics because they tend to exclude Eastern European cases, or include them only in their post-Communist phase. Non-EU member post-Soviet countries are virtually always excluded. Third, they tend to ignore micro-states even if, like San Marino, they have a very long democratic history. Fourth, to the extent they have information on the partisan affiliation of government members, they record affiliation with electoral blocs, disregarding the fact that electoral blocs are often composed of separate political parties.Footnote 1

The aim of the Who governsFootnote 2 dataset is to provide comprehensive and highly detailed information on the partisan composition of European governments, and to match these data with information on those aspects of party politics that can either help to understand the dynamics of the governmental arena or are under the direct influence of the composition of governments. Most of the variables represent fundamental and well-established dimensions of party politics, such as the number of new parties or the fragmentation of the party systems, but some, most importantly party system closure, are more novel. All the variables have been designed so that they can be applied to a maximum number of cases across time.

Database

The starting point of the dataset is 1848 and the geographical scope stretches from the Atlantic to the Urals. All democratic states within these temporal and spatial confines are covered. Democratic refers to countries displaying (1) a score of 6 or higher in the Polity IV index (Marshall et al. 2019), (2) universal (at least male) suffrage elections, and (3) governments formed by, or relying on, parliamentary support, rather than by the exclusive will of the head of state.Footnote 3 By states we mean countries recognized by either the United Nations or the Council of Nations,Footnote 4 including micro-states.Footnote 5 As a result, altogether 48 European democracies feature in the dataset, ranging from the earliest one, the Second Republic in France, to the most recent Kosovar Republic (see Table 1), reaching beyond any other existing European database in terms of the number of cases considered (Casal Bértoa 2021).

Table 1 Regime transition and democratic elections and governments in Europe since 1848
Full size table

The number of systems covered supersedes the number of countries because some countries had more than one political system. Whenever a major internal rupture in democracy took place, one that gave rise to a new party configuration following a non-democratic period, we counted with a new system. In the case of France, for example, we register four separate democratic systems.Footnote 6 Turkey had a similarly fragmented history, with substantially new political configurations developing after the various coup d’état’s.Footnote 7 In some cases, the consecutive cases are separated only by a few years, but in some instances, like Estonia or Latvia, more than half a century lapsed between the two democratic systems.

When establishing the dividing lines between different democratic periods, we had to make choices. If the rupture was not major (e.g. Portugal in 1917, Greece in 1935) or if it was imposed externally (e.g. during the Second World War in the Nordic region or the Low Countries), we considered democracy to be continuous. These are relatively uncontroversial decisions. It is somewhat more questionable whether the constitutional rupture in French democracy at the end of the 1950s or the collapse of the party system in Italy during the early 1990s constitute end points of political systems. In the former case, we decided affirmatively, and therefore we treat the Fourth and Fifth Republics as separate cases, in line with the logic of Polity IV which identifies an authoritarian period between those two Republics. Since no similar authoritarian interlude was recorded in Italy, and because the constitutional and political transformation was less radical, we treat Italy as one continuous case since the Second World War. Because the dataset displays the data in a disaggregated fashion, other scholars can revisit and modify these decisions.

Governments

The database contains information on cabinet duration (i.e. dates of formation and termination), the names of the various ministerial offices as well as of the individuals appointed to occupy them, and the partisan affiliation of each minister at the time the particular cabinet was appointed.Footnote 8 Junior (i.e. deputy) ministers are not included.

In line with Müller and Strøm (2000), the beginning/end of governments is defined by one of three criteria: (1) change in the partisan composition of the coalition; (2) change of the head of government; and (3) new elections. The database records the partisan composition of cabinets at the time of appointment. If a party leaves a cabinet and the remaining parties do not nominate ministers to the vacant positions within two days of that party’s departure, then the next government appears in the dataset as identical with the previous government, minus the resigned ministers. Similarly, simple reshuffles of ministerial responsibilities are not recorded.

In the case of electoral coalitions (e.g. Solidarity Electoral Action in Poland, Coalition Party and Rural Union in Estonia, Reforming Movement in France), the database contains information about the partisan affiliation of the ministers belonging to the different parties within the alliance. In those instances when two or more parties merged to form a new one (e.g. the Italian Democratic Party, the Icelandic Social Democratic Alliance or the Slovak Democratic Coalition), the partisan affiliation of the ministers belonging to the original parties is also recorded. Since this is a more granular approach to party composition than the one typically followed, the governments reported in our dataset do not necessarily coincide with those recognized by constitutional experts (although in most cases they do). Thus, for example, Poland is typically considered as being ruled by one government during the 1997–2001 period, premiered by Jerzy Buzek. But we distinguish five cabinets in this time-span because the party composition of the Solidarity Electoral Coalition (AWS) changed, in the meantime, five times.Footnote 9

The dataset also records non-partisan, extra-parliamentary, acting, in exile, ad interim, emergency, caretaker, presidential, royalist, technocratic, ecumenical, national unity, liberation or war governments (as of December 2020, altogether 196 cases). Examples include Doumerge (France 1934), Papen or Schleicher (Germany 1932), Černý (Czechoslovakia 1920 and 1926), Lloyd George (UK 1919) or, more recently, Letta (Italy 2013), Thanou-Christophilou (Greece 2015), Brega (Moldova 2015), Dimitriev (North Macedonia 2016), Cioloş (Romania 2015), Gerdzhikov (Bulgaria 2017), Bierlein (Austria 2019), Wilmès (Belgium 2019) governments. We also include cabinets that were nominated by the head of state, but failed to gain the necessary parliamentary approval, like the Pawlak (Poland 1992), Topolanek (Czechia 2006), Passos Coelho (Portugal 2015) or Babiš (2017) cabinets and interregnum “authoritarian” cabinets (e.g. Kondylis in Greece 1935; or Pais in Portugal 1917).

Following Casal Bértoa and Mair (2012), the first (“founding”) governments in our database are those created by the first free election taking place in a country after regime collapse, following independence, or after a revised constitution was approved by an interim Constituent Assembly (see also Müller-Rommel et al. 2004).

Party systems

Next to governments, the database contains information on various aspects of party politics. The average age of parties, the disproportionality of electoral systemsFootnote 10, and the number of new partiesFootnote 11 are relatively standard and simple indicators, but in this case they had to be calculated for every single year (e.g., party age) or for every election (e.g., new parties and disproportionality).

The established party dominance variable considers the percentage of votes given to parties at each election while enhancing the weight of those parties that have already achieved at least 3 per cent support during past national elections.Footnote 12 In this way, the longer a party has been around in the political history of a country, the greater the weight attached to its electoral results. The average of the party age and of the established party dominance variables is used to express the overall level of institutionalization of individual parties.Footnote 13

The rest of the variables, electoral and legislative fragmentation, electoral volatility, polarization and party system closure, tap party system dynamics. The fragmentation variables are calculated using Laakso and Taagepera’s (1979) index. The electoral volatility variable is based on Pedersen’s (1979) formula, following Bartolini and Mair’s (1990) coding rules. Polarization is represented by the percentage of votes obtained by “anti-political-establishment” parties, as defined by Abedi (2004).Footnote 14 The dataset includes a full list of such parties per country and per election.Footnote 15 Finally, party system closure is measured using Casal Bértoa and Enyedi’s (2016) index.

Data collection and database summary

The official governmental websites were the starting points for collecting information on governments. In the absence of such data, multiple secondary sources (e.g. electoral lists, newspaper articles, relevant internet-sites, secondary literature, yearbooksFootnote 16) and the advice of country experts (see their list on https://whogoverns.eu/about/) were solicited. Concerning contemporary era governments, the information was completed with the help of interviews with politicians and government personnel during fieldwork trips to more than 16 countries between 2013 and 2016.Footnote 17

As far as the party politics indicators are concerned, they were calculated on the basis of electoral results found in Nohlen and Stöver (2010), Rose and Mackie (1991), the NSD Database (2020) and the different countries’ electoral commission official websites. The data file complements other datasets (Emanuele 2015, 2016; Gallagher et al. 2011; Döring and Manow 2020; Gallagher 2019; Bormann and Golder 2013) having smaller temporal or spatial focus and partly different variables on party politics.Footnote 18

Altogether our database currently comprises 172 years (from 1848 to 2020),Footnote 19 68 different historical democratic periods,Footnote 20 including 81 political regimes (i.e. 50 parliamentary, 27 semi-presidential, and 4 presidential),Footnote 21 753 electionsFootnote 22 and more than 1817 parties and 1586 cases of government formation.Footnote 23 Table 1 displays when a party system was formed, when the first democratic (breakaway or founding) elections took place, when the so-called founding cabinet was inaugurated, and when the last cabinet was appointed in those countries where democracy collapsed.

The data reveal that the average duration of governments (for the 1848–2020 period of “democratic” polities, excluding current governments) in Europe is 501 days. In the dataset, even governments that have lasted less than a day are recorded, for example the Breisky government in Austria, the Hodza V government in Czechoslovakia, the Fernandes Costa (or the so-called Constitutional Junta) cabinet in Portugal, or the Korac government in Serbia. These were typically caretaker governments. The longest lasting cabinet in the covered period was Joseph Bech’s first government that survived altogether 2095 days. If we are, however, to exclude elections, then Helmut Kohl’s CDU-CSU-FDP government was the longest, altogether more than 15 consecutive years (5688 days). The longest serving Prime Minister was a Swede, Tage Erlander, who was in office for 8394 days, that is, for 23 years, followed by Éamon De Valera in Ireland with 7446 days. Four out of the seven longest serving heads of governments served in Luxembourg.

The aggregation of governmental data show that cabinets tend to last longest in Malta, where the average has been 1362 days. The most short-lived governments functioned in inter-war Portugal, lasting on average not more than 97 days (Fig. 1).

Fig. 1
figure 1

Source: Casal Bértoa (2021)

Duration of governments in days and according to party systems, in Europe (1848–2020).

Full size image

A historical perspective

The data allow for comparative cross-sectional studies of how the various aspects of party politics are related to the characteristics of the governments and to each other. The most distinctive feature of the dataset is that it makes possible the construction of long-term historical trends, allowing for the application of different benchmarks and different theoretical assumptions.

Figure 2 shows how European party systems have changed along four relevant dimensions: parliamentary fragmentation, polarization, party system closure and electoral volatility. Volatility represents the electoral arena, fragmentation stands for the parliamentary arena, and closure primarily expresses tendencies in the governmental arena. Polarization, measured by the percentage of votes collected by anti-establishment parties, represents predominantly the discontent of the electorate, but it also reflects developments in the parliamentary arena. The figures are decade averages, and the data prior to 1900 are not shown due to the low number of cases.Footnote 24 The variables have been standardized in order to make the results comparable.

Fig. 2
figure 2

Source: Casal Bértoa (2021)

Standardized party system indicators in Europe over time (1900–2020), decade averages (For three cases (Greece I, Portugal I, and Spain I), we use the number of seats, and not votes, to calculate volatility and polarization, as data on electoral results were not available).

Full size image

The data reveal a curvilinear temporal pattern as far as electoral volatility is concerned. The literature focuses on the recent rise of instability, and the data confirm that there has been indeed a dramatic rise in volatility figures since the 1960s onwards. But the graph also demonstrates that the current level of instability is still somewhat below the figures of the early twentieth century. As opposed to the U-shaped distribution of electoral volatility, polarization has been mainly on the rise during the examined decades, with some temporal surges in the 1910s and 1950s. Together, these results confirm the prevailing view that in terms of party politics the current era is one of extreme polarization and volatility.

The fragmentation figures also show a tendency towards systemic decomposition since the 1960s, but in this regard the current period is less exceptional: The typical European party systems in the 1920s and 1930s were even more fractionalized. Finally, in case of party system closure, where high values indicate more stability, the initial turmoil was followed by consolidation, ending in a somewhat more open, that is, less stable, era. Closure behaves differently from the other parameters in the sense that it indicates a rather tranquil format of party politics between the 1960s and the 1980s. It seems that the governmental arena exhibits more inertia than the electoral or the parliamentary stage. Another difference is that in terms of closure the 2010s brought some degree of stabilization. In these years, the European averages moved upward due to the fact that the Eastern European new democracies developed more predictable party relations than during the 1990s or the 2000s.

The latter fact reminds us that the changes above show not only cross-temporal developments within the systems but also changes in the pool of cases. The latter can be filtered out by exclusively focusing on the nine “core” countries that operated competitive multiparty democracies across most of the analyzed period (Belgium, Denmark, Ireland, Luxembourg, Netherlands, Norway, Sweden, Switzerland, and UK). Within this group of cases, the dynamic appears to be somewhat different (Fig. 3). In terms of volatility and fragmentation, one sees an almost linear trend: These core countries started concentrated and with stable electorates and then gradually moved towards more instability on both dimensions. The support of anti-establishment parties in these cases was particularly high during the first two decades of the twentieth century, but for the later period we see the familiar pattern of increasing discontent. In contrast, after the turbulent decade of the 1910s, the average closure scores increased decade by decade. This tendency towards ever more predictability has stopped, however, in the 2010s.

Fig. 3
figure 3

Source: Casal Bértoa (2021)

Standardized party system indicators in “core” Europe over time (1900–2020), decade averages.

Full size image

From the ideal–typical vantage point of consolidated and concentrated party systems, low volatility, low fragmentation, low polarization and high closure can be interpreted as signs of stability and structure. The four reviewed factors highlight the particularly stable and structured nature of European party politics in the 1950s.

In the last part of this research note, we focus on party system closure, as the most novel variable of the dataset. In this case, the scores are calculated differently than for the other variables in the sense that the values assigned to particular years are aggregations from the entire past and present of the party system (Casal Bértoa and Enyedi 2016, 2021). Since longer time spans tend to have a larger portion of years when no disruption takes place, party systems with a longer past are likely to top the rankings. This makes sense: A system can be, and probably should be, considered to be characterized by stability if its political landscape was stable across many decades. But our dataset also allows to investigate the “performance” of systems by holding the time-frame relatively constant. If, for example, one is interested in contrasting Sweden with Slovakia, then one can disregard what happened in Sweden prior to the fall of the Berlin Wall, and compare the systems within the same historical context.

In general, the data allow for the possibility to divide up history into separate portions, and only work with information collected from the particular period. In line with this approach, we contrast, below, the ranking of party systems in the post-Second World War (Fig. 4) and the post-Berlin Wall periods (Fig. 5), as they both cover around 3 decades of democratic history.

Fig. 4
figure 4

Source: Casal Bértoa (2021)

Party system closure in the post-Second World War (1945–1973) period.

Full size image
Fig. 5
figure 5

Source: Casal Bértoa (2021)

Party system closure in the post-Berlin Wall (1990–2020) period (Figure 5 does not include Armenia II (2018–2020) as only one election has taken place so far.)

Full size image

The colors in Fig. 5 separate those countries which had already democratic party competition in the 1945–1973 period (in grey) from those that did not (in black). Note that the political competition in Greece, Turkey, and (partly) in France happened under a different party system than during the post-1990 period.

The contrast of the two graphs shows that many of the systems that were closed half a century ago excel in terms of predictable interactions in the governmental arena in the post-1990 period as well, despite the fact that we have “re-started the clock” for measuring closure in 1990. Switzerland, UK, and Sweden are cases in point. But the relative position of other systems (cf. Malta and Austria), changed, showing that systems can institutionalize and de-institutionalize not only in absolute, but also in relative terms. Figure 5 also draws attention to the newcomers. The majority of these cases are concentrated in the open end of the scale, but some of them produced highly predictable party systems. Democracies born during Huntington’s (1991) “Third Wave of democratization,” Spain and Portugal, proved to be particularly successful in consolidating their party relations, but some of the post-Communist cases, especially Montenegro, also ended up among the most stable cases.

Figure 5 works with the latest data, but it disregards a large segment of the history of many systems. While in calculations that utilize information from the entire timespan (Casal Bértoa and Enyedi 2021), the UK ends up as one of the most institutionalized systems, after Switzerland and in a virtual tie with Malta and Portugal, and is relegated to the twelfth position, in the post-1990 data the UK is only at the eighth place and Portugal appears among the top performers. This is so because after 1990 party politics in the UK was less stable than during the past, while in case of Portugal the reverse pattern applies. But not all systems are so vulnerable to the scope of the empirics used to calculate the scores. Switzerland and Malta, for example, are the most stable European party systems, whichever method of measurement is chosen.

Further applications

While we have focused here on party system development, the dataset has multiple other applications. Given the fact the most basic units of the dataset are individuals, the data allow one to study the partisanship and the governmental careers of a large number of politicians. The data also show the scope of independents in governments, revealing the fact that a large percentage of the more than 3000 non-partisan ministers are concentrated in a few Eastern European countries, especially post-Soviet systems. One can contrast entirely partisan governments with those that were headed by independent PMs or those that had no partisan ministers at all. It allows scholars to analyze patterns in the time of cabinet formation, government duration, and government stability, and it provides variables that can help to explain these patterns.