How Parliamentarism Developed in Western Europe

Abstract

The description of the long-run historical development of parliamentarism has presented an empirical and methodological challenge because it is only loosely related to constitutional writings. This article offers a solution. Using a wide variety of historiography, I collect data on government terminations in eleven West European states from the establishment of national parliaments until today. To describe the evolution of parliamentarism, I apply a Bayesian learning model that estimates institutional development as the change in current expectations about interactions grounded in past experience. The result is the first long-run continuous description of parliamentarism at the country level, which suggests that parliamentarism in many cases was established later than hitherto believed. In general, it is an institution of the Postwar period. The finding that unelected heads of state in several countries influenced government terminations well into the twentieth century also has implications for ideas about democratization.

To whom is the government to be politically responsible: parliament, the head of state, or both? This is a fundamental political question that concerns control over the executive power. At different points in time, most West European countries have established a parliamentary form of government, the principles of which are still enforced. These principles state that a government that parliament does not tolerate must resign (or dissolve parliament, if possible), while a government that parliament tolerates simply has no obligation to do so. But throughout Western Europe in the nineteenth and twentieth centuries, the question of government responsibility involved an often decades-long political struggle between parliament and the head of state about control over the government and policy. This invites the question of what the long-run development of parliamentarism looked like.

Few modern comparative political scientists have focused on the history of West European parliamentarism. This is striking given the well-founded call to study the historical development of political institutions separately (Capoccia and Ziblatt 2010). It also stands in contrast to the importance of parliamentarism for West European states, both when it comes to the historical evolution of executive power and of democracy (Bartolini 2000; Caramani 2004; Collier 1999; Ziblatt 2017), and to the democratic performance of those states today (Gerring et al. 2009; Lijphart 2012; Przeworski et al. 2000).

The notable exceptions include von Beyme (2000), Przeworski et al. (2012), and Cheibub and Rasch (2022), who focus on the development of constitutionalized rules surrounding parliamentarism. One important finding with implications for the study of parliamentary history is that constitutions provided little information about the subsequent development of the ability of parliaments and heads of state to make governments resign, especially before the Second World War (Cheibub and Rasch 2022, 15). It was also during the pre-Second World War period that the actual influence of heads of state, particularly monarchs, over governments was greater. Especially for this earlier period, researchers note that parliamentarism evolved relatively free from constitutional arrangements, as the outcome of ‘an evolutionary process’ of ‘interactions between monarchs, ministers, and parliaments’ (Cheibub and Rasch 2022, 4), and that the ‘establishment of parliamentary government was often a long process characterized by periods of uncertainty in the distribution of competencies … steps forwards, and setbacks’ (Bartolini 2000, 336).

Previous research posits monarchs as the historical opponent to parliamentary government and focuses on relatively well-known historical events when monarchs yielded to parliamentary majorities, seemingly assuming parliamentarism to have been established following those events. However, such a focus omits the long-run development of the relative influence of parliaments and monarchs over governments before and after the events that have been assumed as decisive. Analyses of the long-run country-specific processes through which parliamentarism developed are currently missing. The focus on monarchs also leaves unaddressed the role of other heads of state, that is, presidents, in the struggle over government responsibility. But the possibility that parliaments share power over the government with presidents should be approached in the same way as monarchs if we want to maintain the conceptual distinction between parliamentarism and democracy.

Anchored in the independent importance of description (Gerring 2012), this paper seeks to tackle the challenges of measuring the long-run development of a core institution in West European political history. In contrast to previous research, I describe the development of parliamentarism in practice (or de facto). I explicitly define parliamentarism as government responsibility to parliament only, which allows for a study of how parliaments and heads of state have vied for the power to terminate governments, and I conceptualize parliamentarism as an institution which shapes beliefs about who may or may not successfully attempt to terminate the government. I then use newly collected data on government termination attempts and resignations in eleven West European countries from the establishment of national parliaments onward to build a data set and I estimate the development of parliamentarism as the continuous outcome over time of interactions between heads of state and parliaments over government responsibility. Following my definition, I describe the development of parliamentarism by the historical change in head of state influence over government terminations, as observed through events in which the head of state makes the government resign, prevents parliament from making it resign, or initiates a dissolution (that is, events that contradict parliamentary principles). I validate the results by anchoring them in qualitative historiography, I compare them to extant research, and I perform alternative estimations to assess their robustness. The results offer the first continuous, long-term country-level measures of the changing balance of power between parliaments and heads of state over governments since the inception of national parliaments.

This paper also makes a methodological contribution. I use a Bayesian learning model to estimate institutional change. The model estimates the underlying level of parliamentarism by measuring the development of parliamentarism as the expected probability that the government currently in office at a given point in time will be terminated without head of state involvement. This expectation is updated as new events are observed. Parliamentarism is established when available information suggests that there is good reason to believe that the head of state will not successfully terminate the government or prevent parliament from terminating it. I thus suggest measuring institutional development as changing expectations about behaviour. These expectations are anchored in the experience of observed, often conflictual, interactions between the key actors who vie for political power. Such a dynamic power struggle over time is observed outside the development of parliamentarism, for instance in the struggle over suffrage (cf. Maravall and Przeworski 2003; Rustow 1970).

My approach yields country-specific descriptions that reveal that for many countries, the development of parliamentarism was a more drawn-out process than past research has been able to show (von Beyme 2000; Przeworski et al. 2012; Cheibub and Rasch 2022). Except for the Netherlands, the UK, and France’s Third Republic, parliamentarism belongs to the twentieth century. While Cheibub and Rasch (2022, 4–7) use constitutional texts to find that parliamentarism as a stable European-wide institution is best understood as a post-Second World War phenomenon, I reach a similar conclusion by my focus on practices within countries. Moreover, I update the conclusions for when some countries established parliamentarism. These updated conclusions foremost concern Belgium (the 1950s instead of the 1840s), Denmark (the 1920s instead of the 1900s), Italy (after the Second World War instead of the 1860s), and Norway (the 1900s instead of the 1880s). The novelties owe more to the data collection and the use of a broader range of sources than to a different understanding of parliamentarism. Both this paper and previous research understand the requirement that parliament can make the government resign if parliament no longer tolerates it as crucial for parliamentarism, while the ability of monarchs (and in my case, other heads of state) to make governments resign defies parliamentarism. However, detailed or recent historiography, often in the native language, challenges standard interpretations of when monarchs lost this ability and parliaments obtained it.

The detected lateness of parliamentary practices in many countries has implications for the important question of how West European countries democratized. In Belgium, Denmark, Italy, and Norway, monarchs made governments or prime ministers resign, sometimes after outright conflict, even though the governments were tolerated by a parliamentary majority elected through universal and equal (male) suffrage. If democracy requires that unelected heads of state cannot terminate governments or prevent democratically elected parliaments from terminating governments (cf. Przeworski et al. 2000), my results provide nuance to the development of democracy in particular in these four countries, wherein heads of state could hold governments responsible after the introduction of full (male) suffrage.

Parliamentarism Defined

I define parliamentarism as a regime where the government is politically responsible to parliament only. This definition resembles others in political science and captures the key difference between parliamentarism and alternative regime types, including the presidentialism of the USA and the assembly rule of Switzerland (Cheibub and Rasch 2022; Gerring et al. 2009; Laver and Shepsle 1996; Lijphart 1992). Politically responsible means that parliament can terminate the government for any reason, not just legal infringement. I focus on government terminations over government formations, which scholars sometimes include in a definition of parliamentarism. The theoretical motivation for this focus is that in parliamentary regimes, governments that parliament can be expected not to tolerate are unlikely to form (Lento and Hazan 2022; Sartori 1997; Shugart 2008). Historically, this was often confirmed by experience, and in the Online Appendix, I discuss how parliaments asserted themselves in cases of government formations. There, I also discuss the theoretical and methodological difficulties that would arise as a result of including government formations in my study, and I argue that the change in interpretation of the results would be marginal (cf. Akirav and Cox 2018). ¹

The addition of the word only clarifies that if the government is politically responsible not only to parliament, then parliament is no longer the government’s sole principal. Parliamentary responsibility may be undermined because an actor other than parliament can terminate the government or prevent parliament from doing so. In the Western European context, the definition thus opens for a study of the historical struggle over government responsibility between parliaments and heads of state, who originally were all monarchs. The definition also implies that semi-presidential regimes are not parliamentary if the president is able to remove governments from office (cf. Schleiter and Morgan-Jones 2009). This understanding explains the differences between the present results and those of Cheibub and Rasch (2022) regarding twentieth-century Finland, the Fifth Republic of France, and the last years of the Weimar Republic.

Parliamentarism concerns government responsibility, not democracy or even (formal) representation (cf. Cheibub and Rasch 2022). Thus, the definition of parliamentarism is silent on how parliamentarians assume office. Likewise, the head of state might accede through inheritance, appointment, or election. Parliamentarism thus cuts through questions of (procedural) democracy, monarchy, and republic. By making explicit that the struggle for government responsibility was not necessarily about democracy, this decoupling helps ‘reading history forward’ (Capoccia and Ziblatt 2010). Still, the alternative responsibility arrangements must be discussed. These include responsibility to both parliament and the head of state, and responsibility to the head of state only. Both can vary in the democratic dimension.

If the government is responsible to parliament and the head of state, the regime is one of power sharing with dual responsibility (von Beyme 1971). This can be a monarchy or a semi-presidential regime as long as both parliament and the head of state have the ability to successfully decide on government terminations. The power-sharing monarchy might be constitutional, but it will, by most definitions, be undemocratic because a non-elected actor has real power over governments. The semi-presidential regime can be democratic or undemocratic, depending on parliamentary and presidential election rules. This implies that a democratically elected president’s ability to contradict parliamentary principles by making a government resign can be perfectly democratic and constitutional.

If the government is responsible to the head of state only, the regime is one of power sharing without dual responsibility, for example, a personal monarchy or a presidential republic. Here, the head of state may also be the head of government, but in any case, parliament is expected to have no control over government terminations, fulfilling other functions such as representation, legislation, and deliberation. And while the presidential republic can be a democracy, the personal monarchy cannot. Thus, the institutional outcomes in this paper, that is, parliamentarism or power-sharing between parliaments and monarchs or presidents who are elected in different ways, are captured through the actors’ ability to retain or remove a government in practice. My approach does not capture the democratic dimension of the regime and it does not focus on the degree of constitutional support for the practices.

Parliamentarism in Practice: Methodological Approach

A detailed description of parliamentary development at the country level is challenging because of the institution’s historical dependence on practices as opposed to clear constitutional provisions. This section motivates the methodological choices made to address the challenges: I discuss the application of a Bayesian learning model to the institutional development of parliamentarism. I explain the country selection and the collection of data on events that capture the ability of parliaments and heads of state to hold governments accountable from the establishment of national parliaments to the present. Finally, I clarify how governments and resignations are understood for the purposes of this paper.

My approach focuses on the ability of parliament and the head of state to make the government resign in practice. This builds on Przeworski et al.’s (2012) observation that written constitutions gave (and in some cases, still give) little information about what rules were actually observed by parliaments and heads of state, and that constitutional devices often were used creatively to enforce or avoid responsibility. Indeed, most constitutions that were adopted during the nineteenth century stipulated that ministers were ‘responsible’ but chosen and revoked by the monarch. Such formulations created ambiguity and uncertainty over government responsibility to parliament and the head of state that were settled over time in practice.

Consequently, I estimate the evolution of parliamentarism based on historiographic information about events when heads of state and parliaments either intervened to terminate governments or supported the government in resisting a termination attempt of the other actor. Parliamentarism and power sharing are thus the two potential outcomes in a tug-of-war over government control. Substantively, this resonates with the understanding that political regimes are the proximate results of interactions, often downright struggles, between (elite) actors with conflicting interests who must still find common institutional ground for conflict resolution. Within the democratization literature from Rustow (1970) to Boix (2003), Ansell and Samuels (2014), Przeworski (2015) and others, this has been a rewarding approach to studying the correlates of underlying structural developments. Theoretically, the model assumes that as government resignations occur, actors update their beliefs about who might successfully terminate the government and that with experience, beliefs or subjective probabilities will converge so that expectations become shared and uncertainty decreases (Jackman 2009, Ch. 1). Accordingly, I conceptualize parliamentarism as an institution that shapes expectations about where control over the executive authority resides (Rothstein 1996). This conceptualization is grounded in ideas that institutions shape expectations about behaviour (see Huntington 1968; Hall and Taylor 1996; Maravall and Przeworski 2003). Note that if one distinguishes such beliefs or expectations from norms as ideas of what actors should do in particular situations (cf. discussions in Finnemore and Sikkink, 1998), then the emergence and transmission of a parliamentary norm are beyond my scope (see also Breen (1999, 470–471) on how beliefs and expectations may be mistaken for norms as they are transmitted over time, and Hirsch and Dixon (2021) on how expectations can concern both behaviour and norms, which cannot necessarily be inferred from one another).

Parliamentarism is institutionalized when experience gives good reason to believe that no actor other than parliament will terminate the government currently in office or prevent parliament from terminating it. Likewise, power-sharing is institutionalized and parliamentarism can be de-institutionalized if experience gives good reason to believe that an actor other than parliament may terminate the government or prevent parliament from terminating it. ² In this way, my empirical approach also captures reversals.

For estimation, I use a Bayesian learning model developed by Scott (2017). In my application, the model estimates the likelihood that a binary variable called a non-contradictory event takes the value of 1 at any point in time, and the uncertainty surrounding this estimate. The probability and uncertainty are updated as more events occur that either contradict or do not contradict parliamentarism. Each present data point is regressed sequentially on all past data points. This procedure is repeated forward in historical time, after which a smoother is projected backward to fit the updated sequence (Durbin and Koopman 2002; Scott and Varian 2014). The model thus gives the experientially grounded (changing) belief that the government in office at a particular point in time will not experience an event that contradicts parliamentarism as a probability. Although events either contradict or do not contradict parliamentarism, the resulting measure is continuous over time between 0 and 1.

The estimated model is ${\alpha _{t + 1}} = {\alpha _t} + {\varepsilon _t}$ with ${\varepsilon _t}\sim {\cal N}\left( {0,\sigma } \right)$ , assuming a random walk trend. I choose a logit-family model because the data consists of events with binarily-coded outcomes (events may or may not contradict parliamentarism). In the base models, the $\sigma $ -parameter, which estimates the uncertainty, is set at 0.4 for each country. The starting states, that is, ${\alpha _{t = 0}}$ , are found by trial and error to reflect my understanding of constitutional arrangements, or in the British case, practices around the first year (see each country story and Online Appendix Section A). I use these constitutional arrangements, for although they relate little to subsequent developments of parliamentarism (Przeworski et al. 2012), no country historiography suggests that they were meant as dead letters at their adoption, and they provide initial focal points for expectations before observable practices can develop. ³

My empirical focus is on Western Europe because parliaments and parliamentarism originated in this part of the world, were central to its political development, and characterize most of its countries today (Myers 1975, Ch. 1; cf. Cheibub and Rasch 2022). The historical scope conditions that a country must meet is that an independent national parliament and a separate head of state exist. A national parliament is a gathering of representatives with some legislative authority over the whole polity, and that is not itself the government. Independence means that the head of state (who can be head of government also) does not substantially control its composition and work (Poggi 1978, 46–48; Verney 1959, 17–38). Countries thus exit the scope conditions if their governments permanently closed parliament. This happened when dictatorship was introduced in Austria, France, Germany, and Italy in the first half of the twentieth century. These closures were internal to the national political regimes. By contrast, I disregard both temporary prorogations and closures due to foreign occupation because these did not challenge regime expectations as much as expectations about national survival altogether. ⁴

Within Western Europe, I have selected those countries that meet the scope conditions by the second half of the nineteenth century, and that have experienced no or relatively few and short interruptions in their fulfilment of these scope conditions. They are, with periods when the scope conditions are met within parentheses, Austria (1861–1932, 1945–), Belgium (1831–), the UK (1721–), Denmark (1848–), Finland (1863–), France (1788–1939, 1945–), Germany (1871–1932, 1949–), Italy (initially Piedmont, 1848–1922, 1945–), the Netherlands (1814–), Norway (1814–), and Sweden (1720–). For the UK and Sweden, the starting points are motivated by the fact that by then it is historically conceivable that parliament could make cabinets resign. ⁵ For Finland and Norway, the parliaments were sufficiently independent from the time of their inception both to legislate and to exercise influence over the government, although the countries themselves did not have full independence (Kan 2008; Klinge 1996, 222–227, 348–353; Nordby 2018, 20–27).

Parliamentarism requires that parliament only can terminate the government, and I measure the development of parliamentarism using data that capture the changing occurrence of events that contradict this requirement. The data include the coding of all changes of government and discernible attempts by parliament or the head of state to oust the government. Parliamentarism is contradicted if the head of state makes a government resign, often by demanding a changed policy or by withdrawing royal or presidential confidence. This is so even if the head of state motivates the action by the government’s lack of parliamentary support, or if parliament fails to terminate the government and then asks the head of state to dismiss it. Such events contradict parliamentarism because they show that parliament is unable to sustain parliamentarism on its own and suggest that the head of state can be expected to take independent action for or against the government. Furthermore, parliamentarism is contradicted if government-termination attempts by parliament are frustrated by the head of state, often because the government invokes the confidence of the head of state or because the head of state simply refuses the resignation.

Lastly, there are cases in which parliamentary dissolutions contradict parliamentarism. It is certainly in accordance with parliamentarism that the government dissolves parliament if parliament no longer tolerates it. However, if the ensuing elections return a parliament that remains hostile and parliament is dissolved again, that second dissolution contradicts parliamentarism. The reason is that such a ‘double dissolution’ shows that parliamentary responsibility is not accepted unless the government faces a parliament of its liking. Furthermore, the prerogative to dissolve parliament has been a jealously guarded head of state prerogative in many countries (Cheibub and Rasch 2022, 20–25; Schleiter and Morgan-Jones 2009) and dissolutions initiated by the head of state contradict parliamentarism because they let the head of state put the government at immediate risk (cf. Verney 1959, 31–32). ⁶

In all, I register seven types of events, four of which contradict parliamentarism and three of which do not. Parliamentarism is contradicted if the head of state 1) makes the government resign, 2) prevents parliament from making the government resign, or 3) initiates a dissolution. Parliamentarism is also contradicted if 4) an election following a dissolution returns a similarly hostile parliament whereupon parliament is dissolved again. Parliamentarism requires that no such events are observed. Instead, what is observed are events that do not contradict parliamentarism. These include 5) that the government resists a head of state request that it should resign (which means that the head of state fails to terminate the government), 6) that the government resigns over a conflict with parliament, often expressed through a hostile vote, or after parliamentary election results adverse to the government, or 7) that the government resigns for other reasons. By other reasons, I mean government-internal reasons such as coalition breakdowns or the death of the prime minister, and formal reasons such as when a government resigns after a constitutional reform or the accession of a new head of state. These other resignations may not relate to parliamentarism per se. Yet, they provide important information that the government was not terminated in contradiction with parliamentarism, and their inclusion completes the link of events in the historical chain of governments.

Table 1 shows the distribution of events and the latest recorded contradictory events per country. The distinction between events that contradict and do not contradict parliamentarism is in focus, but the above differentiations between events help structure the data collection and conceptualize the interactions. All events are empirically registered, and each gets equal weight in the estimations. Successful terminations by parliament or head of state dominate, as well as resignations for other reasons. Failed attempts to make the government resign, head of state initiated dissolutions, and ‘double dissolutions’ are rare. ⁷

Table 1. Parliamentary spells, number of events, and resignations per country from the creation of national parliaments until the end of 2021

¹ Includes head of state refusals to grant the government a dissolution.

² Includes government-initiated dissolutions as a response to an adverse vote.

I collected the data using a wide range of secondary literature, following advice from Lustick (1996), Møller (2015), and Moller and Skaaning (2021). I use multiple sources. I consider conceptual consistency over what a government and a resignation are. I favour relatively recent historiographies if older work does not contribute valuable details. Preference is given to descriptive work by historians that has sufficient detail to allow me to assess the involvement of parliament and head of state in government resignations. This use of historiography accounts for many of the differences between the results below and previous research.

I count governments by the head of government. Historians rarely leave any doubt about who this is. ⁸ A change of government thus equals a change of head of government, whether all ministers resigned collectively or not. This approach has the virtue of travelling back to the times when parliaments were established, before the emergence of recognizable political parties.

As for resignations, I count instances when the head of government tenders his or her resignation and someone else attempts to form a government. If the offer to form a new government is declined or the new formation attempt fails, whereupon the outgoing prime minister returns, the resignation is only counted if it was the result of a conflict with parliament or the head of state. The consequence of this approach is to disregard the many cases where the government collapses, tenders its resignation, and then gets patched together again. Such cases may matter for an assessment of government stability, but not for the development of parliamentarism. I elaborate on the coding issues discussed here in the country illustrations below and throughout the Online Appendix. In the Appendix, I also discuss the methodological difficulties that would arise if government formations were given consideration alongside terminations, and I provide examples of why this would not substantially alter the conclusions.

The Development of Parliamentarism in Western Europe

The estimated development of parliamentarism in the different countries is visualized in Figures 1–11. The figures include uncertainty estimates and the full distribution over time of all the events on which the estimations are based. Parliamentarism is enforced when the darkest areas remain near 1. When they fluctuate below 1, the regime is one of power-sharing, and both parliament and the head of state exercise control over the executive. Around zero, the government is responsible to the head of state only. These three regimes are qualitatively different because of the fundamental difference in expectations about who may terminate a government. The gaps in the Austrian, French, German, and Italian trajectories cover periods when the governments of these countries closed the parliaments permanently. The gaps have been created after estimation because, historically, politicians in the countries drew on past experiences when the parliaments were reopened.

Figure 1. The parliamentary development of Austria, 1861–1933, 1944–. Parliamentarism is institutionalized when the dark area is near 1. Events thats contradict parliamentarism are represented as rings at 0. Events that do not contradict parliamentarism are represented as rings at 1. Periods without an independent national parliament are omitted.

Figure 2. The parliamentary development of Belgium since 1831. For an explanation, see Figure 1.

Figure 3. The parliamentary development of the United Kingdom (Great Britain) since 1721. For an explanation, see Figure 1.

Figure 4. The parliamentary development of Denmark since 1849. For an explanation, see Figure 1.

Figure 5. The parliamentary development of Finland since 1863. For an explanation, see Figure 1.

Figure 6. The parliamentary development of France, 1788–1939, 1945–. For an explanation, see Figure 1.

Figure 7. The parliamentary development of Germany, 1871–1932, 1948–. For an explanation, see Figure 1.

Figure 8. The parliamentary development of Italy (Piedmont), 1848–1922, 1945–. For an explanation, see Figure 1.

Figure 9. The parliamentary development of The Netherlands since 1814. For an explanation, see Figure 1.

Figure 10. The parliamentary development of Norway since 1814. For an explanation, see Figure 1.

Figure 11. The parliamentary development of Sweden since 1720. For an explanation, see Figure 1.

The figures display a conceptually driven continuous country-level measure of the development of parliamentarism that incorporates the changed uncertainty surrounding expectations about government terminations. They show that the uncertainty is higher when two different actors may (attempt to) terminate the government. There is a general decrease in uncertainty once experience suggests that heads of state are no longer expected to terminate governments. The decrease foremost characterizes the post-Second World War period (except in Finland until the 1980s and in France from around 1960). This coincides well with Cheibub and Rasch’s (2022) finding that the rules for no-confidence and confidence votes and dissolutions were clarified in most countries after the Second World War. It suggests a greater convergence of constitutions and practices in this period, during which the results indicate that parliamentarism can be seen as a more general West European institution.

The graphs visualize a process over time. I condense this in two ways. Firstly, Table 1 includes the years of the latest recorded events that contradict parliamentarism, along with the start and end years of parliamentarism for each country. As discussed in the country stories, the latest recorded contradictory event succeeds the establishment of parliamentarism in The UK and Sweden. Secondly, Online Appendix Figure 1 shows the development of parliamentarism in each country as a line capturing the average expectation, highlighted against those of the other countries. The lines are produced by taking the mean of the sample draws for each observation or year, and applying the inverse logit function ( $exp\left( x \right)/\left( {1 + exp\left( x \right)} \right)$ ) to these means. This discards the uncertainty estimates but it offers continuous measures that can be readily used for statistical analysis.

To assess the sensitivity of the results, the Online Appendix offers alternative estimations. The correlations between these and the main model are always above 0.9. Appendix Figure 3 shows that interpretations are unchanged if the $\sigma $ -parameter is adjusted to 0.3 or 0.5. Figure 4 in the Appendix shows that results are similar if initial constitutional provisions are understood as if only the head of state is expected to terminate the government at the starting point. The figure also shows that interpretations remain similar when discarding failed attempts to make the government resign, events that might be harder to conceptualize and find in historical sources. Appendix Figure 5 addresses the possibility that expectations are substantially altered by major constitutionalized regime changes (such as the transition from the Habsburg Empire to the Austrian First Republic or between the many French regimes) by increasing the uncertainty parameter by one decimal point for each such change. Interpretations are unaltered.

The focus on events that contradict parliamentarism has the benefit of making it less important to assess whether those events that did not involve head of state action were in positive accord with parliamentary principles or not. Appendix Figure 6 compares the main results to an estimation based only on events that involved either parliament or the head of state. It thus excludes resignations for personal, formal, or coalitional reasons (so-called ‘other’ reasons in Table 1). The only notable difference in these figures is that the Swedish parliament appears less powerful in the 1770s–1850s period when different monarchs were also heads of government. Ideas about when parliamentarism was established do not change in any case.

The results convey both familiarity and novelty. In general, novelties come from a deeper engagement with historical sources rather than from a different understanding of parliamentarism compared to other political scientists. The results give more substantiated information about the development and establishment of parliamentarism than the single years suggested in von Beyme (2000), Przeworski et al. (2012), and Scarrow (2006) and the developments captured in ready-made indices, such as the Varieties of Democracy project’s (V-Dem) confidence dimension index, which covers the same country periods as my study (Coppedge et al. 2020). That index is based on expert judgments of parliament’s ability to dismiss the government for political reasons. Its purpose is not to differentiate regime types, but it nonetheless tends to capture parliamentary states (Teorell and Lindberg 2019). It does not take into account whether the head of state also can dismiss the government, so differences between my results and the index should be expected, especially for semi-presidential states. A comparison is shown in Figure 2 in the Online Appendix. The correlation is 0.65.

Below, I illustrate the results and the coding with Belgium and Finland. These two countries cover most of the different event types in Table 1 and help clarify differences with previous research and the V-Dem index. ⁹ Moreover, Finland illustrates the difference between parliamentary and semi-presidential democracy. The Online Appendix offers similar discussions of the other countries.

Belgium

The Belgian 1831 constitution established a parliament to which the government was responsible, and it stated that the monarch appointed and dismissed ministers (Stengers 1992). In the pre-party period until 1846, most governments resigned over internal tensions, conflicts with parliament, or adverse electoral results. None of this contradicts parliamentarism. However, de Muelenaere resigned after conflict with King Leopold I in 1833 because the king formulated his foreign policy, in full accord with the constitution (Witte 2005, 94–95). In 1834, the Lebeau-Rogier government resigned due to another conflict with the king (Witte et al. 2009, 38), and in 1840, Leopold I effectively terminated a new Lebeau government by denying it the dissolution it requested to counter a hostile vote from the Senate (Velu 1966, 83–88).

The years 1846–1847 saw a sequence of events that international scholars have taken as the establishment of parliamentarism (Cheibub and Rasch (2022) and Przeworski et al. (2012) mark 1847 as the first year of de facto parliamentarism). In 1846, a Liberal party formed. After its electoral successes in 1847, the de Theux government chose to resign and Leopold I reluctantly let the Liberals form a government. For the next twenty-five years, the monarch was absent in government terminations and with the resignation of de Decker in 1857, he found that a royally appointed government had little chance of survival without parliamentary support (Gubin and Nandrin 2005, 58–61; Witte et al. 2009, 84). Figure 2 clearly suggests decreasing expectations of head of state interventions during these years. Nevertheless, the monarch returned to the scene in 1871, when Leopold II required Prime Minister d’Anethan to drop some ministers. D’Anethan refused and he and his government resigned (Gubin and Nandrin 2005, 110–112). After some electorally induced changes of government, the 1871 event was repeated with Malou in 1884 (Deneckere 2005, 44–48). Then in 1899, de Smet de Naeyer resigned when he disagreed with the king’s proposed electoral reform (Boulger 1925, 74; Barthélemy 1912, 512–530). ¹⁰

Parliamentarism was again contradicted in 1911. The new king, Albert I, let Schollaert know that he would ask for advice outside the cabinet. Schollaert took this as a withdrawal of royal confidence and resigned, while Catholic politicians retroactively took responsibility for the king’s actions (Dumoulin 2006, 67–69). Then, in 1918, Albert made de Broqueville resign because of disagreements concerning army command (Haag 1990, 650–665).

The interwar period began with the introduction of full male suffrage in 1919. Governments were short-lived, resigning because of adverse elections, coalition breakdowns, or policy difficulties (Höjer 1946), none of which contradicts parliamentarism. But in 1932, Albert asked Renkin to address parliament, whose majority he commanded, with a royal note in which Albert denounced the government’s policies. Renkin resigned rather than submit to this affront (van denWijngaert et al. 2002, 205 and 285; Witte et al. 2009, 151). Note the subtlety of this termination, which took place after the introduction of full male suffrage.

Albert’s successor, Leopold III sought to preserve royal power. In the late 1930s, he insisted that Belgium remained the power-sharing regime that its constitution stipulated that it was. In 1939, Leopold pressured the outgoing Spaak to countersign a dissolution decision, which Leopold used sometime later to arrest the government-recomposition attempts of Spaak’s successor, Pierlot (Gerard 2006, 228–229; van den Wijngaert et al. 2002, 210–211; Witte et al. 2009, 188–189). This was a head of state initiated dissolution, which contradicts parliamentarism. The next year, Leopold assumed army command and chose to face the Nazi invasion, while the Pierlot government went into exile and parliament prorogued itself until the end of the war. This led to one of the few failed head of state attempts to remove a government. In May 1940, Leopold wrote Pierlot to resign, but Pierlot withheld his countersignature and remained in office (Van den Wijngaert and Dujardin 2006, 17–19; Witte et al. 2009, 211). This brought about a break between king and government, and after the war, Leopold’s brother Charles acted as regent while Leopold remained in Germany. When a government coalition hostile to Leopold’s return was about to form in 1950, Charles prevented this by invoking the constitutional letter and dissolving parliament (Van den Wijngaert and Dujardin 2006, 143; Witte et al. 2009, 239–242).

The 1950 dissolution is the latest event I registered that contradicts parliamentarism. However, in 1960, Baudouin failed to terminate the Eyskens government and replace it with a ready alternative. When asked to resign, Eyskens invoked the constitution and replied that the government as a whole should resign, which it collectively refused to (Dujardin and Dumoulin 2008, 142–144). Since the 1950s, governments have resigned because of adverse electoral results, hostile votes, and broken coalitions, and the dissolution power has been fully assumed by the government.

As seen in Figure 2, Belgian parliamentary development shows swings back and forth over time in the power balance between parliament and head of state. These swings reflect actual involvement in government terminations rather than changes in constitutional provisions. Online Appendix Figure 2 shows that few if any of these historical swings are observed in the V-Dem index. Naturally, the swings cannot be inferred from the exact year in which Belgium became parliamentary, which is what past research has focused on. Moreover, to find that year, we must consider the entire timeline. Such a consideration reveals a lateness, which might be an empirical novelty, although the monarch’s lasting influence has long been known in Belgian national historiography (Fusilier 1960, 369; Witte et al. 2009). A lateness is also observed in the V-Dem index, which for reasons unknown assigns parliament full ability to terminate governments only during the most recent years, decades after experience suggests that parliament only can be expected to terminate governments.

Finland

The parliament of Finland, then a Russian Grand Duchy, was established in 1863. It was elected without Russian interference and legislated over many national issues. The domestic government, the Economy Department of the Senate, answered to the head of state of what over the course of the next decades came to be seen as another country. Parliament and government acquired increasingly autonomous powers from the 1880s, a development that came to be paralleled by Russian efforts to thwart it (Jussila et al. 1998; Kan 2008; Klinge 1996). Among the heads of government during the first forty years, Nordenstam died, Alexander III replaced af Forselles with von Troil (the sole event that contradicted parliamentarism in the period), and von Troil, Tudeer, and Linder resigned given the general difficulties in balancing parliamentary and Russian demands (the most detailed sources for this are Estlander 1929-1930 and Nordenstreng 1936-1937).

By 1905, Finnish parliamentary politics led to increasing conflict with the Russian Czar, who dissolved parliament that year. Senate leader Streng chose to resign and Mechelin formed a government with the intention to introduce parliamentarism. Accordingly, Mechelin resigned after an interpellation in 1908 (Jussila et al. 1998, 103; Klinge 1996, 451–452). To resist the rise of parliamentarism, Nicolas II immediately dissolved parliament. The following years included royal dissolutions and the replacement of one government with another (Jansson 1992, 41; Klinge 1996, 455–456; Seton-Watson 1967, 669), all events that contradicted parliamentarism. As seen in Figure 5, the head of state’s powers over the government increased substantially relative to parliament. This changed with the October Revolution in 1917. When the Tokoi government fell apart the same month, Svinhufvud formed a government, which he led until he resigned to act as the head of the fully independent, democratic Finnish state (Jussila et al. 1998, 111–113; Luntinen and Manninen 1993, 183–185).

Parliament’s role was strengthened both in the new Finnish constitution and in practice. The indirectly elected president had the power to call new elections, and to form and dismiss governments, which had to be tolerated by parliament (Lindman 1969; Meinander 2020, 169–170). ¹¹ Both parliament and the presidential electoral college were democratically elected. However, neither constitution nor practice provided much guidance for the new relations between parliament, the government, and the president. Thus, the interwar period involved a learning process. Most governments resigned after adverse elections, coalition breakdowns, or hostile votes (Lindman 1937; Meinander 1999, 86-92), none of which contradict parliamentarism.

On other occasions, parliamentarism was contradicted. In 1924, Kallio resigned in protest against President Ståhlberg’s decision to dissolve (Meinander 1999, 91). In 1929, President Relander dissolved parliament, arguing that ‘[W]ith regards to the future it is good, that parliament knows that the president’s right to dissolve exists not only on paper but also in reality’ (quote translated from Jansson 1992, 285). The year after, Relander pushed Kallio’s second government to resign. He then dissolved parliament again, after having let Svinhufvud form a new government (Jussila et al. 1998, 188–192). In 1931, Svinhufuvd became president and terminated the Sunila government by refusing to assent to a bill (Meinander 1999, 177–178). From then until the end of the Second World War, many governments resigned after hostile parliamentary votes or because their leaders became presidents, none of which contradicts parliamentarism.

In the post-war period, most governments were short-lived, resigning after adverse elections, hostile votes, or coalition breakdowns. However, two powerful presidents, Paasikivi and Kekkonen, strengthened their positions against parliament. In 1950, when Fagerholm customarily resigned at Paasikivi’s re-election, expecting to continue, Paasikivi seized the opportunity to appoint another government (Fagerholm 1977, 227–233). ¹² Kekkonen then made Fagerholm resign again during the so-called Night-Frost Crisis in 1958 (Meinander 1999, 354–358). In 1961, Kekkonen dissolved parliament at will (Meinander 1999, 366–370). In 1968, he benefited from his re-election to terminate the Paasio government (Jussila et al. 1998, 335). Finally, in 1971 and 1975 he decided to dissolve, decisions that were enough to make the governments resign and get them replaced by caretakers to administer the elections (Jussila et al. 1998, 337–353; Meinander 1999, 439–459; the dissolutions and resignations thus yield two data points each year).

The year 1975 marked the latest event that contradicted parliamentarism. Kekkonen’s successor, Koivisto, honoured his promises to transform Finnish semipresidentialism into parliamentarism. He declared that he would not use the constitutional letter to discontinue governments (he did, however, interfere one last time in a government-formation process in 1987; Raunio 2011, 119–121), and he abolished the so-called custom of government resignations at presidential (re-)elections. By the time Koivisto resigned in 1992, experience suggested that governments would not be terminated in contradiction with parliamentarism (Meinander 1999, 466–492).

The presidential actions described above were supported by the Constitution. None violated Finnish democracy. Nevertheless, they contradicted parliamentarism, which simply suggests that Finnish presidents made use of the power-sharing provisions in the democratic constitution. In the year 2000, a constitutional reform formally replaced power-sharing with parliamentarism, as these concepts are understood here. Online Appendix Figure 2 shows that this is the year in which Finland scores fully on the V-Dem confidence index. However, preceding changes suggest a gradual development of which the constitutional reform was a natural ex-post confirmation. Again, my findings reveal a more nuanced temporal development that relates to observed power balances.

Conclusion

Based on new data on (head of) government terminations, this paper has described the development of parliamentarism in eleven West European countries. Methodologically, the analysis underlines the importance of the recommendations regarding the use of historiography for data collection and validation (Kreuzer 2019; Møller and Skaaning 2021), and it shows the fruitfulness of studying a separate institutional development in its own right to read history forward (Capoccia and Ziblatt 2010). Moreover, I suggest that a Bayesian learning model provides a conceptually consistent way to estimate institutional change that is understood to build on practice and relate to experientially grounded expectations about action. As such, it might be applied to other settings. One potentially productive fit between the Bayesian learning model and the study of institutional development lies within the strand of historical institutionalism, which often focuses on long-run developments that incorporate the interaction and adaptive but potentially uncertain expectations of actors (Sanders 2008; Thelen 1999).

Substantively, I suggest that parliamentarism often developed more gradually and convulsively than a focus on so-called key events might suggest (but not always, as Dutch history shows). Furthermore, parliamentarism developed independently of constitutions, often preceding formal reforms, although constitutions often gave actors instruments to use in their quest for power. The gradual development and the independence of formal reforms have been suggested previously (Bartolini 2000; Przeworski et al. 2012), but this paper offers the first systematic substantiation and presentation. The descriptions also give fuller and sometimes very different pictures than do tables or discussions in previous research (von Beyme 2000; Przeworski et al. 2012; Scarrow 2006). Importantly, many countries established parliamentarism later than hitherto assumed, and in general, West European parliamentarism is an institution of the post-Second World War period. This is suggested for Europe as a whole also in Cheibub and Rasch (2022), and I reach a similar general conclusion by using historiography to study government-termination practices at the country level. The decreased uncertainty in the post-Second World War period surrounding expectations of who may terminate governments coincides well with Cheibub and Rasch’s (2022) finding that the rules for no-confidence and confidence votes and dissolutions were clarified in this period. However, concerning Belgium, Denmark, Italy, and Norway, I push forward the adoption of parliamentarism by decades in relation to previous research. These countries adopted full (male) suffrage, a criterion commonly used for democratization in the comparative literature, before their monarchs lost the ability to make governments resign. If democracy is undermined if unelected officials can terminate governments (Przeworski et al. 2000), then the development of parliamentarism in these countries may challenge ideas about how and when they were democratized.

Descriptive research has independent value (Gerring 2012). A natural next step for research is to explain the development. Among other things, an explanatory endeavour might clarify how and when actors were able to activate constitutional provisions to their benefit. The present results may also inform subsequent investigations of the timing and sequencing of different institutional reforms, such as suffrage extension, electoral reform, and party development, the importance of which was underlined already by Dahl (1971). The data can be used to study consequences of parliamentarism, such as the historical origins of aspects on which parliamentary regimes are found to perform ‘better’ (Gerring et al. 2009), and help address endogeneity and temporality concerns in analyses of artificially short time periods (cf. Boix and Stokes 2003). Moreover, I have followed Przeworski et al. (2012) and remained agnostic about norm development, and I have invoked research that states that beliefs and expectations about behaviour are something else than norms. Nevertheless, text-as-data methods could be used to study the development of (expressions of) norms concerning parliamentarism (cf. Hirsch and Dixon 2021). It is then relevant to relate such norm development to the parliamentary trajectories and the underlying events that I have presented here. It would also be valuable to explore the sequential order of the disappearance of observed government terminations by the head of state and the codification of the rules of the most important mechanisms through which parliamentarism is sustained. Hopefully, the development of Western European parliamentarism as presented here will inform such research.