1.1 Introduction

Democracy seems to be in a state of crisis. Across the world, politics is increasingly dominated by elitist, corrupt, or authoritarian politicians, and the past decade has seen the rise of populism and extremism in politics (Norris & Inglehart, Reference Norris and Inglehart2019, Berman Reference Berman2021). Traditional parties are losing their members and therefore must focus more on short-term survival and rent-seeking (van Biezen et al. Reference van Biezen, Mair and Poguntke2012, Mair Reference Mair2013); meanwhile, campaign finance makes elections exclusively accessible to elite candidates (Cagé Reference Cagé and Camiller2020). While there is a debate as to the extent of democratic backsliding in established democracies,Footnote ¹ it is clear that citizens’ distrust in both politicians and representative democracy is at an all-time high (Cain et al. Reference Cain, Dalton and Scarrow2003, Van der Meer Reference Van der Meer2017). While this “democratic fatigue” or “democratic malaise” (Van Reybrouck Reference Van Reybrouck2016, Reuchamps & Suiter Reference Reuchamps and Suiter2016) stems from several causes, the overarching theme is clear – voters feel neither included nor represented, in a system that is dominated by disconnected party elites.

Political lotteries have dominated both the academic and policy discourse as an alternative to elections to improve inclusion and representation. Recently, a number of academics have advocated for a system of “lottocracy” where the election of legislators is replaced with selection by lot (Guerrero Reference Guerrero2014, Gastil and Wright Reference Gastil and Wright2019, Abizadeh Reference Abizadeh2021). Other scholars have argued for the introduction of randomly chosen consultative or decisional deliberative institutions, such as citizens’ assemblies or mini-publics, to supplement or as part of a system-wide reform of democracy (Landemore Reference Landemore2017, Reference Landemore2020). In all cases, lottocratic selection is advocated as a way to strengthen citizen participation in democratic decision-making, reduce polarization, and combat misinformation among citizens (Dowlen Reference Dowlen2009, Van Reybrouck Reference Van Reybrouck2016, Dryzek et al. Reference Dryzek, Bächtiger and Chambers2019, Gastil and Wright Reference Gastil and Wright2019, Landemore Reference Landemore2020).

Selection by lot is also gaining traction in the policy world.Footnote ² There are very few cases today where lotteries are used in an institutionalized manner to select representatives or to hand over legislative powers to unelected citizens – notably, Ostbelgien (the third and smallest, German-speaking community of federal Belgium), the city council of Paris, and the borough of Newham in London have created permanent bodies chosen by lottery.Footnote ³ However, a more common and perhaps better-known practice is that of randomly chosen citizens assemblies, initiated all over the world at both the local and national levels, designed to provide nonbinding policy recommendations to governments. Citizen assemblies or mini-publics are now a frequently used political institution worldwide. Notable examples include the Irish Constitutional Convention and Citizen’s Assembly in 2016, the UK Citizens’ Assembly on Brexit in 2017, or the Citizens Convention for Climate in France in 2020. In fact, the OECD has identified no fewer than 566 citizens assemblies or other deliberative processes from 1979 to 2021 (OECD 2021), and there is an even larger catalog of democratic innovations available at https://participedia.net/. Deliberative mini-publics are even going digital – recently the technology company Meta launched a series of online Community Forums that brought together thousands of randomly chosen social media users from thirty-two countries for a weekend-long deliberation, conducted completely online (Fishkin et al. Reference Fishkin, Siu, Chang, Ciesla and Kartsang2023).

So why are political lotteries a solution proposed to reinvigorate democracy? In short, because by their very nature, lotteries equalize opportunity of access to decision-making. The selection of political representatives by lottery, also known as sortition or lottocratic selection, can be incorporated into democratic governance in several ways. Legislative or governing bodies could be chosen entirely or partially by lot, or lotteries can be used to assign leadership, committees, or other political posts. Lotteries can also be used to select citizens to participate in institutions of deliberative democracy, such as in Citizens’ Assemblies or deliberative polling, designed to provide recommendations on policy. Including different perspectives may even improve the quality of decision-making, creating a sort of wisdom-of-the-crowds effect. These types of lottocratic institutions could be permanent or temporary, and could be structured to have either advisory or authoritative powers.

In this Element, we focus on the historical use of political lotteries as a permanent feature of governance, to select representatives or officers in assemblies with legislative, agenda setting, or governing power. While there are many ways to incorporate lotteries in democracy, many modern initiatives are temporary, advisory, and with everyday citizens; instead, we wish to focus on how political selection by lottery affects the characteristics and behavior of political representatives or officers with governing power. Recent calls want to make lottery-based selection, and more generally deliberative institutions, permanent. Yet these proposals are theoretical, and there is very little empirical evidence upon which to base these recommendations. If lottery-based selection is here to stay as a permanent feature of democracy, we need to better understand how they work over time, in a setting with elite politicians and high stakes.

Fortunately, selection by lottery is far from a new idea – lotteries were used in politics in Europe for millennia; from Athens in 300 BCE, to the medieval Italian city-states, to nineteenth-century parliaments of France, Denmark, and many others. It is only in the last 200 years that political culture has shifted toward equating democracy with elections (Stasavage Reference Stasavage2020). Lotteries disappeared with the advent of electoral parliamentary democracy, because of both the development of a theory of political legitimacy that linked legitimacy to consent via elections and the interests of the leading actors of the time (Manin Reference Manin1997). Today, elections and other forms of appointment also appear to best serve the challenges of participation in large, diverse polities who are more or less structured by political parties and career politicians. So why incorporate lotteries back into democracy? Can lotteries improve representation and inclusiveness? Can lotteries prevent elite capture? Can randomly chosen deliberative groups improve policymaking? Can sortition truly replace elections as a method of political selection?

In this introduction, we now review the recent calls for political lotteries, and our historical political economy approach, before providing a roadmap of the Element.

1.2 Recent Calls for Political Lotteries

The case for sortition often begins as a case against elections. Social scientists argue elections currently suffer from two key flaws: a lack of representation and capture by elites (Engelstad Reference Engelstad1989, Dowlen Reference Dowlen2009, Zakaras Reference Zakaras2010, Guerrero Reference Guerrero2014, Abizadeh Reference Abizadeh2021, Landa and Pevnick Reference Landa and Pevnick2021). If elections fail to give a broad subset of citizens the chance to win office, then this leads to a lack of political equality and poor representation. If elected officials are beholden to special interests and campaign finance, and citizens cannot hold these politicians accountable, then this ultimately leads to elite capture of political institutions.

These flaws were emphasized by Belgian writer and journalist David Van Reybrouck in 2016, when he attracted wide attention with an original, short book in which he advocated Against Elections. He argued that people across the world increasingly suffer from what he calls “democratic fatigue,” where citizens in democracies are frequently asked to express their votes but feel they have little influence over actual policy outcomes. While modern democracy is defined by elections, Van Reybrouck discusses that historically democracy was designed to preserve the power of the elite at the expense of the masses. This fundamental tension is now driving democratic fatigue – he writes, “Elections are the fossil fuel of politics. Whereas once they gave democracy a huge boost, much like the boost oil gave the economy, it now turns out they cause colossal problems of their own.”

Van Reybrouck’s claims were radical but influential, and as we will show, many others’ remedy is to bring back the random selection of political representatives from the general public. There is a long and robust academic literature on the theoretical benefits and challenges of sortition (e.g. Elster Reference Elster1989, Manin Reference Manin1997, Landemore Reference Landemore2020), as well as modern-day experiments with participatory and deliberative democracy (e.g. Mansbridge Reference Mansbridge1980, Fishkin Reference Fishkin1995, Reuchamps, Vrydagh, and Welp Reference Reuchamps, Vrydagh and Welp2023). Yet Van Reybrouck, after reviewing experiences with temporary and issue-based citizens’ assemblies and citizens’ parliaments (including the G1000, the Citizens’ Assembly in Belgium, that he himself helped organize), rejects the use of deliberative democracy in temporary assemblies as a solution. He argues that the “democratic fatigue syndrome” cannot be solved by neo-parliamentarian or anti-parliamentarian structures because they cannot address the essential problem of representative democracies: elections (Van Reybrouck, Reference Van Reybrouck2016). For sortition to be used most effectively, it must be used as a tool of political selection for permanent assemblies.

The recent calls for lotteries in democracy operate on a similar, ambitious scale. We now review a number of influential academic proposals. Some propose a radical new form of government called “lottocracy” in which adult citizens would be randomly selected to serve as lawmakers for existing parliaments. Others argue for hybrid, bicameral parliaments, involving both an elected and a sortition chamber. A different set of scholars propose to restructure democracy by involving citizens in frequent and permanent deliberative mini-publics, emphasizing both non-electoral forms of representation and the importance of deliberative democracy. The next sections briefly highlight these recent calls to demonstrate the variety in both scope and rationale for why lotteries may solve democracy’s problems.

1.3 Historical Political Economy Approach to Evaluating Lotteries

It is challenging to empirically evaluate recent calls for lotteries in politics. The overwhelming majority of proposals have been theoretical. And while there is increasingly good observational and experimental data on citizens’ assemblies, these bodies are temporary, consultative, and low stakes. In order to assess the promises and pitfalls of lottocratic selection as a permanent and salient feature of democracy, we need empirical data from cases where representatives in permanent political assemblies were selected by lottery. Today, this data is almost impossible to generate – researchers are unlikely to convince a modern parliament to randomly select its members or change its selection rules, because political parties and other organized groups have a vested interest in controlling political selection.

This motivates our historical political economy approachFootnote ⁵ – we can use historical cases to analyze the causal effects of lotteries used in governing bodies. History provides a wide range of use cases – we can see how governing bodies selected exclusively by lot operated in ancient Athens, or how random leader selection worked in medieval Italy; we can also consider how lottery-based procedures operated in nineteenth-century France, and how deliberation in random groups helped write the 1848 constitution in Denmark.

Due to the wide variation in historical context, governance structures, and type of lottery-based procedures, each case is very different and supported by different types of data, and our outcome variables are often not directly comparable or able to be pooled in a single analysis.

As a result, in this Element we do not use the set of cases to test a single hypothesis. Instead, we provide a framework to guide our analysis that lets us present general yet theoretical predictions, use different data and research designs for each country, and then summarize patterns from our findings across cases. For example, the individual characteristics of candidates and what qualifies as an elite politician is defined very differently in ancient Athens and nineteenth-century France, yet our framework can still broadly assess how lotteries affected elite access to power in both cases.

It is important to note, however, that our case selection is driven by two factors: the canonical cases of lottery use in governing bodies (in that over time, these cases drove institutional spillover that led to the use of lotteries in other countries) and the fact that these are the historical cases for which extensive micro-level data has been collected. While our set of cases is not exhaustive, it is influential, and these are also the cases that employed lottery-based procedures either for a substantial set of time (in legislative procedures or governance) or for a significant event in history (in the case of constituent assemblies). Our cases were not selected for their outcomes based on our framework; our framework instead provides us a way to learn from existing canonical cases.

Thanks to innovations in early democratic republics that incorporated lotteries, we can analyze historical evidence on everything from the practical implementation of sortition to how it changed political incentives in the actual process of governing. This allows us to make a new contribution to the literature, in that we can go beyond comparing the “contemporary empirical reality” of democratic systems with some “future ideal of sortition” (Malleson Reference Malleson2018) and instead evaluate real-life cases with actual governing consequences. We also help to better place these historical cases in context – as Sintomer (2018) notes, while cases like ancient Athens are invoked as a model for lottocratic governance, more recent experiments with political lotteries are in the vein of deliberative democracy, not radical self-governance.

Finally, we also address one limitation of the existing literature on sortition, a critique made by Dowlen (Reference Dowlen2009). He argues that more attention has been paid to who lotteries select, and less on the “form and function” of the lottery process. He writes, “The general picture is that those who address the specific issue of the selection of political officers by lot tend to advocate sortition without necessarily examining how it works or what it, as a mechanism, brings to the process of selection.” Since political sortition has been used in different ways throughout history (Manin Reference Manin1997), in this Element we synthesize a wide range of empirical evidence on the real-life practice of lotteries.

1.4 Element Outline

This Element responds to recent calls to incorporate lotteries in democratic selection, by analyzing historical cases of their implementation. We review different types of lottery-based selection used in the past and present a simple framework to guide our synthesis of the historical cases. We focus on the rationale behind and benefits of lotteries – to prevent elite capture, equalize access to power, and improve deliberation – and then the details of their implementation. Throughout the Element, our goal is to assemble empirical evidence, and we rely both on prior quantitative work and our own original, micro-level datasets on lottery assignment and legislative outcomes. Each section then focuses on a different historical case – premodern Athens in 300 BCE and medieval Florence, the constituent assembly of Denmark in 1848, and the National Assembly of France during the Third Republic. We then conclude by extending our findings to modern-day lottocratic bodies and citizen’s assemblies. The uses of lottery-based selection – and its political consequences for selection, policy, and broader political outcomes – deserve more theorizing and empirical testing from social scientists. This Element is our first synthesis of recent historical political economy contributions to this broader research agenda.

In this first section, we have reviewed recent calls to incorporate political lotteries once more in politics, which motivates our empirical study of how lotteries worked in the past. Section 2 thus presents a framework by which to analyze the use of selection by lottery throughout history. We identify three dimensions by which to assess the effect of lottocratic selection—representation, elite capture, and deliberation—and contribute by outlining key details regarding the implementation and execution of sortition.

Sections 3–6 then feature our historical cases. In Sections 3 and 4, we discuss the premodern uses of political lotteries, in the contexts of ancient Athens and medieval Florence. Both regimes chose to select political representatives or leaders by lottery, but employed other institutions of selection and auditing in implementation that we argue are often overlooked. In Section 5, we look at the use of lotteries to create deliberative groups in the Constituent Assemblies of France and Denmark. Using unique micro-level data from Denmark in 1848, we analyze how this affected constitution drafting in a context of high stakes with long-term consequences for the political regime. Section 6 reviews the use of lottery-based procedures, to choose political officers within an elected assembly, a procedure that was used in multiple cases in nineteenth-century Europe. We use original data from the case of the French Third Republic, to study how lottery-based selection interacts with political parties.

Finally, the Conclusion examines how our findings speak to uses of lotteries and deliberative mini-publics in modern democracies. By looking at the primary cases of lotteries used over the centuries, we can juxtapose recent calls with how these lotteries worked in practice. Ultimately, we argue that the study of historical cases, focusing not only on the democratic theory behind lotteries but also on the specific implementation, can inform recent calls for use of lotteries today.

2.1 Defining Concepts

A lottery generates a fixed number of outcomes with given probabilities, such that appointment by lottery is a mechanical form of decision-making. Its outcome is “arational” (e.g. Elster Reference Elster1989, Dowlen Reference Dowlen2008), in the sense that no individual agent can be held responsible for its outcome (Tridimas Reference Tridimas2011). Lotteries can be employed in a wide variety of ways, or combinations with decision-making which is not arational. In our cases, we encounter three types of lottery-based political selection:

• Sortition – the selection of a body of political representatives or a political office directly by lottery

• Lottery-based procedure – the incorporation of a lottery into a multistage selection process combining lottery and election

• Deliberative committees – the use of a lottery to create groups for deliberation within a legislative body

Table 1 details our cases and the type of lottery used.

Sortition is associated with Athenian democracy, and is also known as “selection by lottery,” “selection by lot,” “demarchy,” or “lottocracy.” It is method of political selection by which individuals are randomly selected from a predefined pool. Today, it is most commonly known as the method for jury selection in Anglo-Saxon countries, though sortition could be used to select entire legislative or governing bodies, committees within such bodies, or any type of consultative assembly. In this Element, we review two premodern cases where political representatives were chosen using lottery: Athens in fifth century BCE, where the Council of 500 (and other political and judicial officers) were selected by lottery, and the Florentine Republic in the fourteenth to fifteenth centuries, where a lottery was used in the selection of the highest office in city-state government.

It is important to note that understanding selection by lottery first involves defining the population. Some authors (e.g. Dowlen Reference Dowlen2008) use the term “sortition” to indicate that no preselection of the population has taken place, either by election or by another rule-based procedure such as eligibility requirements. However, throughout history, selection by lot has been employed in a wide range of cases where the pool has been defined as a function of the time or context, and we see value in studying the random selection of citizens from such predefined pools – even if most literature does not call such random selection “sortition.” In addition, later in this section we also make a distinct argument that the benefits of sortition are always a direct function of how the pool is constructed.

A lottery-based procedure we define as one where randomization is incorporated at some point in a multistage selection process of political officials; here, either choosing a set of random groups by lottery and then having these groups choose political officers (typically by election), or using election to determine the lottery pool, then randomly choosing officers from this pool. Lottery-based procedures were used in several European countries, in both the early and late modern periods: a frequent application comes from nineteenth-century Europe, where parliaments were divided into random groups and then these groups elected members of powerful parliamentary committees (Cirone and van Coppenolle Reference Cirone and Van Coppenolle2018, Reference Cirone and Van Coppenolle2019). In this Element, we first look at the case of the Florentine Republic, which combined election and lottery to select the Standard Bearer of Justice, the leader of government at the time. We also focus on the case of the French Third Republic (from 1870 to 1920), which used a lottery-based procedure to help select representatives to the powerful budget committee; we also discuss the use of similar procedures across Europe.

Lottery-based procedures have been absent from recent calls in favor of sortition, even though they were used more frequently than pure sortition (i.e. selection in which bodies were entirely chosen by lottery from some population). We argue these are important, yet overlooked, methods of selection. While both sortition and lottery-based procedures exploit the element of chance found in randomization, our expectations of the political outcomes are different. Selection using only sortition can reap the benefits of a random draw, while lottery-based procedures are somewhat attenuated (because there is an additional step in the selection process, that involves election). As we will discuss in future sections, while the randomization in the first stage of a lottery-based procedure helps improve access to power and reduce elite capture, the procedure retains the element of election in the second stage (and so may again become more vulnerable to elite capture or corruption, but could positively select on expertise or other characteristics).

Finally, we also examine the use of deliberative committees. Lotteries can be used to randomly draw groups within elected parliaments, who then debate drafts of constitutional or legislative documents. It is important to note here that the lottery is creating deliberative groups, but not selecting the political decision-makers directly. The pool of representatives had already been chosen via election, so a human-choice element was definitely retained. We study this in the context of the 1848 Constituent Assembly in Denmark, where assembly members drafted the constitution in randomly assigned groups.

The rationale for incorporating randomization in political selection is multifaceted, and there are both normative and practical benefits. The literature on sortition is extensive (see Elster Reference Elster1989, Dowlen Reference Dowlen2008, Reference Dowlen, Delannoi and Dowlen2010, Manin Reference Manin1997), and in the interest of space, we will not summarize it here. Instead, in this Element, we use empirical and historical data to examine three of the primary arguments in favor of political lotteries: equalizing access to power, preventing elite capture, and improving deliberation. We also consider implementation. The following sections present our framework to focus on the aspects of lotteries that can be compared across cases and can inform their modern use.

2.2 Equalize Access to Decision-making Power

One of the primary arguments for political lotteries in a democratic setting is that they can help equalize or distribute access to power across citizens or groups, and therefore improve representation. Elections have long been known to be “aristocratic,” because they can be dominated by subsets of an elite population (Engelstad Reference Engelstad1989, Manin Reference Manin1997, Guerrero Reference Guerrero2014, Abizadeh Reference Abizadeh, Gastil and Wright2019). If electoral democracy is having a crisis of representation, it is because politicians and leaders not only consist primarily of the elite but are also increasingly incentivized to serve powerful and wealthy interests, instead of representing a broad range of citizens (Van Reybrouck Reference Van Reybrouck2016).

In contrast, lotteries are often assumed to be egalitarian. By definition, using a lottery means all citizens in the selection pool have equal chances of being chosen to hold political office (or a fair chance when the lottery is weighted, according to the criteria used for stratification); the probability might be small, but selection is exogenous to all observable and unobservable characteristics. In practice, this gives citizens equal chances to serve in political office, regardless of individual resources or connections. Egalitarian selection is better maximized with a universal or near-universal selection pool, from which the representatives are drawn. Importantly, most modern advocates of political lotteries also emphasize the importance of rotation, combined with sortition. With frequent rotation, political power is accessible to all over time (Landemore Reference Landemore2020). Even if the chance of selection is small, it is equally distributed (and thus an improvement over election-based forms of selection).

Equalizing access to power also mechanically improves representation. Lottocratic selection is often designed to help “ensure some form of statistical representativeness of the population affected by the issue” (Curato et al. Reference Curato, Farrell and Geissel2021). Note that one random draw does not necessarily ensure a representative snapshot. In fact, modern lottocratic selection often relies on the use of stratified sampling, or quota sampling, which weights the random draw based on demographic characteristics of the population (such as gender or geographic location). Such sampling techniques can result in a pool of representatives chosen that resembles a “microcosm” of society. In this way, political lotteries induce more diversity in selection or in the characteristics held by minorities in the population from which individuals are drawn, than when the appointment process is controlled by established elites who form part of the majority. Another advantage of “sortition as representation” is that it can lead to more representative policy (Gastil and Wright Reference Gastil and Wright2019).

But did selection by lot actually lead to broader participation of citizens in politics, or were the individuals who volunteered to participate in the lottery a small set of educated, powerful elites? What were the benefits of sortition over time? We argue lotteries equalized access to power and improved representation. To answer these questions, in all sections we assess empirical evidence on the characteristics of representatives chosen by lotteries and compare this to historical accounts of or data on the composition of both the elite class and population at large. While political hierarchies and the diversity of society vary by country and time period – from tribes in Athens, to guilds in Florence, to the elected representatives of France and Denmark – we can use the concept of elites and citizens across cases to comparatively assess how lotteries affected access to power.

2.3 Prevent Elite Capture

The second argument in favor of lottocratic selection is that it can prevent elite capture of the democratic system (Dowlen Reference Dowlen2009). A fundamental flaw of elections is that they generally favor well-resourced and well-networked candidates, who, once elected, may only represent a minority of society – either because they represent the preferences of only the elite or because they are indebted to wealthy or special interests (Zakaras Reference Zakaras2010, Landa and Pevnick Reference Landa and Pevnick2021). Elections are also subject to influence by powerful interest groups and, at the extreme, are susceptible to corruption by political candidates and representatives. In turn, elite capture and corruption can negatively affect policy.

Advocates argue sortition can serve as an anti-corruption device (Dowlen Reference Dowlen2009, Bagg Reference Bagg2024), for a number of reasons. First, selection by lot can help insulate democracy from undue influence in the electoral selection of representatives. Because members are chosen at random and do not need to run for office, representatives do not need to rely on powerful interests to finance campaigns (Landemore Reference Landemore2020). Lotteries also reduce incentives to target private over public interests, while in office – sortition with a sufficient pool ensures almost certainly the same politicians will not be in power the next term, eliminating any ability to exchange government resources or votes for future electoral support.

Preventing elite capture via lotteries can also provide stability to democratic systems. More generally, the chance element in selection by lot, if executed correctly, also means the results of the selection process cannot be known in advance. This can reduce corruption, elite capture, and electoral manipulation. Historically, lotteries helped break up concentrations of power in the early stages of political consolidation and protected nascent democratic institutions (McCormick Reference McCormick2006, Dowlen Reference Dowlen2008, Stone Reference Stone2011).

But did selection by lot prevent politics from being dominated by a subset of elites? Are political lotteries enough to prevent corruption, or does the use of a lottery just shift conflict over office to other aspects of democratic politics? This requires looking at civil conflict in Athens, control by family dynasties in Florence, and the actions of anti-mass-democracy forces in nineteenth-century Europe. We argue that across our historical cases, lotteries improved governance by preventing elite capture and providing stability. Here, we assess the empirical evidence on how lotteries undermined corruption and anti-system elite coordination and prevented self-interested elites from capturing political (and often newly democratic) institutions at the expense of the regime.

2.4 Open Decision-making

A different argument in favor of lottery-based selection is that it can improve policymaking. Academic debates surrounding sortition typically rely on appeals to self-government and radical democracy, but the “new wave” of democratic experiments using lotteries rely instead on the logic of deliberative democracy (Sintomer 2018).Footnote ⁶ The rationale for direct, deliberative democracy is that all individuals can equally participate in political debate and decision-making. This ability is arguably lacking in modern democracy, where deliberation occurs among people in clearly defined parties who structure the conversation. Therefore, electoral democracy is not very conducive to deliberation or “open decision-making” (Landemore Reference Landemore2020).

How can the use of lotteries in selection help improve decision-making? Electoral democracies only sample the wisdom of a narrow subset of the population (namely, the one that wins elections), while lottocratic selection involves a wider range of individuals. Landemore argues for the importance of technological change in reinventing democracy. Experiments with novel democratic designs have been inspired by revolutions in our understanding of how people aggregate information, by developments in information technology, and by globalization. If previously electoral representation was thought to be necessary for selecting the smartest individuals to contribute to the conversation, we now better understand the importance of the variance in information brought by very different individuals (Landemore Reference Landemore2020). Many argue that this, in turn, improves deliberation (Mercier and Landemore Reference Mercier and Landemore2012, Landemore Reference Landemore2013, Abizadeh Reference Abizadeh2021). Lotteries promise to open access to the debate and could therefore potentially improve the deliberative quality of the decision-making process. Skeptics draw on other recent research to argue precisely the opposite (see Farrell et al. Reference Farrell and Schwartzberg2023 for an overview of the debate).

Recent research has found a number of benefits to randomly selected citizens who deliberate in citizen’s assemblies (Dryzek et al. Reference Dryzek, Bächtiger and Chambers2019, Farrell et al. Reference Farrell, Suiter and Harris2019, Renwick et al. Reference Renwick, Dynes and Johnstone2019), yet there is much less research on how deliberation affects serving politicians (Flinders et al. Reference Flinders, Ghose and Jennings2016, Farrell et al. Reference Farrell, Suiter and Harris2019). Deliberative committees, or the random assignment of citizens and serving politicians to legislative committees within an assembly, are only now being implemented (e.g. in Belgium) and so we have yet to see how open-decision-making works in governing bodies. In particular, since such committees involve seasoned politicians, we might expect different effects.

In Section 5 we focus on the effects of deliberative committees on legislation in governing assemblies in determining high-stakes policy. We are able to use unique micro-level data and exploit the case of Denmark’s Constituent Assembly in 1848, to assess the empirical evidence on the role of deliberation in randomly assigned groups on legislative activity. We use this natural experiment to analyze how random group assignment and exposure to opposing viewpoints affected roll call voting on pro-democratic amendments, which ultimately shaped Denmark’s constitution.

2.5 Implementation

We also emphasize in this Element that “the devil is in the details” – more specifically, that the implementation and specifics of lottery-based selection matter. This is perhaps not surprising, but often absent from media coverage surrounding lotteries. Our perceptions of lotteries also might fundamentally change, because of these details – for example, while ancient Athens is often invoked as a model case, the lay reader might not realize that not only was the eligible pool from which lots were drawn heavily restricted, but individuals chosen were heavily vetted before and after service (Taylor Reference Taylor2007, Carugati et al. Reference Carugati, Ober and Weingast2016). Or France successfully implemented lottery-based procedures for decades in the nineteenth-century Third Republic, but only during the period where political parties were weak (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2019).

Here, we aim to make a contribution by comparatively assessing both prior research and our own empirical datasets. Implementation varied widely by country, but documenting this variation demonstrates how contemporary policymakers approached the challenges of lottery-based selection. Not only can we learn from these implementation and case-specific details, but they might indicate the conditions under which lotteries can be useful today. We leverage unique variation in our historical cases – implementation rules varied as a function of the historical and political context of the time, and such innovations (or failures) can help us understand the functioning of political lotteries. We consider three elements of implementation: (1) the composition of the lottery pool, (2) accountability devices, and (3) the role of elites and political parties.

First, we consider to what extent there were restrictions, formal or informal, on the pool from which citizens were randomly sampled. Democratic ideals would assume the pool is universal, or at the very least representative of society at large. Otherwise, there is a fundamental selection problem; even if a group is selected by lottery, if the pool is highly restricted, then the representative and diversity benefits of sortition vanish.

Second, we document to what extent there were any methods of accountability for political representatives selected by lottery, in our historical cases. In particular, we look to see if there were any institutions that audited randomly chosen representatives before, during, or after service; or if there were any consequences for office holders who made poor decisions or abused their power in office (Manin Reference Manin1997). We simply note the absence of accountability in cases when lottery-based selection was incorporated in already existing chambers, for these featured almost no institutions to audit politicians or randomly selected groups.

Finally, we consider the role of political elites and political parties. We establish in each case the initial rationale for the adoption of the lottery, and (if applicable) the context in which lottocratic procedures were removed. In the premodern cases, such choices were made by political elites; but in nineteenth-century Europe, the adoption and dissolution of lottery-based selection were done by fledgling parties. We analyze to what extent the rationale for adoption lines up well with democratic theory’s highest ideals, in that institution builders hoped to be inclusive and fair; as well as study whether the reasons behind lotteries related to civil conflict or stability. In either case, understanding how political lotteries were adopted or dismantled can help us understand to what extent they were able to equalize access to power and prevent elite capture, as well as understand if and why this selection rule was retained.

This analysis is also informed by recent failures of modern political parties to work with lottocratic governance. A lack of perceived legitimacy of lottery selection is obvious from the results of democratic experiments that were recently conducted. There is also an unresolved question of how the legitimacy of differently selected assemblies should be balanced. We could, for example, observe this in France where the decisions made by the randomly drawn “Convention Citoyenne pour le Climat” were watered down by the elected parliament. A similar problem emerged both in the Canadian and the Icelandic examples of Citizens’ Assemblies designing constitutional reform. In particular, political parties refused to endorse the constitutional suggestions made by a randomly drawn citizens’ assembly in the Canadian referendum in 2011, and their proposals were subsequently rejected by the population in popular referenda in 2005, 2007, and 2009 (Fournier et al. Reference Fournier, van der Kolk, Carty, Blais, Rose, Fournier, van der Kolk, Carty, Blais and Rose2011). Finally, our focus on elites also speaks to a gap in modern debates about the role, if any, for political parties in lottocratic governments (Landemore Reference Landemore2020).

3.1 Athens in the Fifth Century BCE

The use of lotteries in politics predates the consolidation of democracy in Athens, in the fifth century BCE. There is evidence they were introduced by SolonFootnote ¹⁰ in 594 BCE, as detailed in the Athenian Constitution, but these reforms were interrupted by war and the establishment of a tyrannical regime. When Kleisthenes in 507 redefined the rules regarding citizenship and the new powers of the Assembly of the Citizens of Athens, selection by lot was a fundamental part of the design of the Boulè – ordinary citizens were selected by lottery for the Council of 500, as well as other executive, legislative, and judicial posts. As result, selection by lot was widely incorporated in the broader institutional structure (which also involved election), and remained in use until 322 BCE (Dowlen Reference Dowlen2008).

The specifics of each lottery varied by governing institution. The Council of 500, called the Boulè (βουλη`), was a deliberative council selected by lottery from a group of native, male, but not necessarily propertied citizens.Footnote ¹¹ The Council prepared the agenda for the assembly of citizens,Footnote ¹² and helped implement laws and conduct administration. The lottery was in fact a stratified lottery, whereby fifty men were drawn from each of Athens’ ten tribes who were drawnFootnote ¹³ from a combination of groups who inhabited different districts (demes) of Attica (Taylor Reference Taylor2007, Dowlen Reference Dowlen2008). The People’s Court served as a court of appeals for citizens. Yearly, 6,000 citizens, equally from each tribe, swore the oath (Hansen 1999), among which several were then chosen by lot to serve as jurors on any given day (Carugati Reference Carugati2019). Jurors heard individual cases (also assigned by lot) but also served as an executive check on decisions of the Assembly and cases of corrupt officials. They also reviewed the eligibility of citizens selected by lot to serve in other institutions, and reviewed them upon leaving office (Tridimas Reference Tridimas2011). Finally, 9 Archons and 600 executive magistrates responsible for a variety of functions were also selected by lot (Hansen 1991).

At the time, Athenian society was fairly egalitarian, in that all native adult males had the right to participate in politics, and many political decisions were voted on by citizens directly. But notably, they refused to use sortition for all offices. Elections were used for offices that required specific expertise – this included strategoi (commanders of army and navy), military officials, financial officials, and other administrative jobs (Taylor Reference Taylor2007). But still, non-elite citizens were also guaranteed access to politics, because sortition was used in the main governing bodies. This system meant that Athenian citizens spent a considerable amount of time participating directly in government; scholars have estimated that in the fourth century “over a third of all citizens over eighteen, and about two-thirds of all citizens over forty, became councilors, some of them twice” (Hansen Reference Hansen1999, Ober Reference Ober2008).

3.2 Discussion

While very different historical contexts, premodern cases can provide insight into the use of lotteries today. Both Athens and Florence chose to use such procedures to mitigate conflict and prevent elite capture. However, the use of sortition in these cases was much less representative than commonly assumed; both cases are notable for severely restricting the pool from which officers were drawn. Given the historical context, this still improved representation, though this demonstrates that lotteries could be subject to manipulation through the pool of entrants. Self-selection was also a factor in both cases, which is an issue for modern-day deliberative mini-publics (Lancelle-Webster and Warren Reference Lancelle-Webster, Warren, Reuchamps, Vrydagh and Welp2023, Cirone, Flanigan, and Redman 2025).

We can also learn from other details surrounding implementation. The first insight is that frequent rotation was an important element of implementation. As Tridimas (Reference Tridimas2019) writes, “Sortition implied that any citizen might hold office; rotation implied that every citizen might hold office at some time,” and this allowed citizens to both govern and be governed in turn. Another insight comes from the Athenian case, where the practice of scrutiny played an important role in political selection. Dokimasia was designed to ensure those selected for office were fit to serve, and was a system of screening implemented alongside the lottery. Current calls for lottocracies have focused on providing support and presentations by experts to randomly selected participants in deliberative mini-publics (Gastil and Wright Reference Gastil and Wright2018, Reuchamps, Vrydagh, Welp Reference Reuchamps, Vrydagh and Welp2023), but there have been few detailed calls for screening or auditing participants in lottocratic bodies.

Finally, the premodern cases also demonstrate the ability of elites to manipulate, and even undermine, the use of lotteries in the selection process. In Florence, the Medici dynasty successfully corrupted the procedure such that draws were no longer random. Lotteries may only be a useful tool of political selection if elites support their use.

4.1 Lottery-Based Procedures in Florence

The Florentine Republic is particularly known for its long-term use of lottery-based procedures, during a period that corresponded with the height of Florence’s influence in politics, republicanism, and its role as the center of the Italian Renaissance. Florence already used political lotteries to select political officers in 1291 (Manin Reference Manin1997). The first ordinances that established the use of political lottery and scrutiny date from 1328 (Manin Reference Manin1997). As the power of the Medici family grew in the mid-fifteenth century, the procedure became co-opted. Political lotteries were again used after their fall from power in 1494 and after Savonarola until 1512 (who, as a Dominican cleric, was against the use of political lottery and in favor of rational decision-making in line with one’s religious consciousness). Lotteries were used again from 1527 to 1530, or in the second republican system of Florence, as these two time periods together are sometimes called (Manin, Reference Manin1997).

Between the thirteenth and sixteenth centuries, Florence was arguably the birthplace for the modern-day conception of a “republic” (Boutier and Sintomer Reference Boutier and Sintomer2014) in addition to being a major economic hub of trade and industry. Public deliberation, elections, and ballot voting were all developed in the city that was a focal point of the Renaissance. While most well-known for the dominance of its powerful families, elite society was organized around a set of occupational guilds, where membership also dictated citizenship and the franchise. Membership in guilds was also necessary to hold a position in the government. As a result, during this time, the Florentine Republic more resembled an institutionalized oligarchy with limited franchise based on membership of occupational guilds, and balancing the power between these guilds was part of the reason lottery-based procedures were used.

As we discuss in this section, the Florentine Republic used a lottery-based procedure that combined lottery and election (via electoral scrutiny) to select the city’s leader and government. Candidates wishing to run for office had to be chosen by a preselection committee, and then from this pool of accepted candidates the city’s executive was chosen by lot, stratified by guild and geography. Here we focus on how lotteries were used in the selection of the city-state government. The highest office was Standard-bearer of Justice (Gonfaloniere di Giustizia), part of a nine-member priorate (the Signoria). The Priorate drafted and executed laws, while the Standard Bearer was the highest executive office and responsible for internal security. Florence was dominated by powerful families, arranged geographically across four quarters of the city; the Signoria seats were distributed across these four quarters, and the Standard Bearer office rotated across each quarter, to ensure geographic balance of power. These offices were also distributed across the major and minor guilds (McCormick Reference McCormick2011).

The selection process for these offices was a multistage lottery-based procedure (called Tratta), that took place every two months (Sintomer Reference Sintomer2012). The entire process was weighted by allocating offices as a function of membership of the guild association. For the Signoria, a subset of citizens eligible was decided by a process of “electoral scrutiny” by which a special committee would determine via majority voting which citizens were eligible (believed to be a function of past political experience, status, and other factors; see Abramson Reference Abramson2024 or Kent Reference Kent1975). From this pool, names were randomly drawn out of a bag to select members of the Signoria. For the Standard Bearer, a similar process was used, but a separate scrutiny was held to ensure the candidate pool consisted only of the most elite members of guilds (Najemy Reference Najemy1982). Once all offices were selected, candidates were excluded if they were in tax arrears, had criminal records, had recently served, or if they were directly related to recent office holders.

4.2 Discussion

The use of lottery is well-documented in Italian republics such as Venice and Florence in the fifteenth and sixteenth centuries. While Athens was often invoked as a democratic ideal, experimentation in a hierachical society defined by guilds let to the strategic use of lottery-based procedures that incorporated an element of randomization but allowed for a multistage selection process to preserve guild influence. They chose to use such procedures to mitigate conflict and prevent elite capture. However, as a result the use of sortition was much less representative than commonly assumed, thanks to severely restricting the pool from which officers were drawn. Given the historical context, this still improved representation, though this demonstrates that lotteries could be subject to manipulation through the pool of entrants.

Finally, the premodern cases also demonstrate the ability of elites to manipulate, and even undermine, the use of lotteries in the selection process. In Florence, the Medici dynasty successfully corrupted the procedure such that draws were no longer random. Lotteries may only be a useful tool of political selection if elites support their use.

5.1 French Constituent Assembly in 1789

Deliberative policymaking in random groups within a representative body had its origins in the pre-democratic Assembly of Representatives of the Third Estate in 1789. The Assembly of the Third Estate, one of the bodies of the Estates General, was a large body of 600 deputies. A system of committees was created, which divided the chamber into a set of smaller groups (bureaux) that would be frequently redrawn. These groups were chosen by selecting members in order from the list of election districts (baillages) with deputies in alphabetical order. This is actually a form of systematic random sampling, where samples are selected at series of preset intervals.

The bureau-system was then institutionalized in its successor, the French Constituent Assembly of 1789. The Assembly was even larger – over 1,200 members, from the three different estate assemblies – and French elites again relied on bureaux as an organizing principle. The bureaux had a number of functions, and these functions changed over the course of the assembly. The groups played an important role in discussing the initial draft of the constitution (prepared by a separate Constitution Committee), both via an exchange of views and preparing information and drafting sections (Frederiksen Reference Frederiksen1936). The bureaux also considered motions and elected the president and the secretaries of the assembly (Point du Jour, vol III, p. 87, July 2, 1789 (Session of July 1)).

Why did French elites rely on a lottery-based selection procedure? One rationale for this procedure was to equalize access to the debate. As the Marquis de Ferrière wrote, “I do not speak at all in the general Assembly, and in that I am like eleven hundred and ninety who do not say a word, some because they have nothing to say, others because they do not want to. In the Bureaus I express my opinions, I am listened to there; there one can do some good” (Frederiksen Reference Frederiksen1936). Frederiksen (Reference Frederiksen1936) provides a comprehensive historical analysis of the bureau-system, and noted that the smaller venue allowed individuals from all backgrounds (military, judiciary, clergy) to speak their mind, when otherwise in an assembly of 600 they would be silent.Footnote ²¹ The bureau provided a means for deputies to exchange ideas, and allowed for persuasion, which could change political attitudes (Dawson Reference Dawson1978).

The creation of the bureau-system in France also sheds light on why lotteries are useful in times of uncertainty, to prevent elite capture. Deputies sought to incorporate committees into the rules of procedure to expedite the assembly’s work, but also realized the necessity of breaking up geographic or provincial loyalties (Frederiksen Reference Frederiksen1936, Dawson Reference Dawson1978). Both the Assembly of Third Estate and French Constituent Assembly were created amid the existence of strongly opposing political factions, and across coalitions of nobility, clergy, and peasants, before the institutionalization of parties occurred to help structure democratic governance. Given the unpredictability in election and naming conventions, this method of selection meant members could not manipulate or anticipate group composition. This is why Sièyes discussed the role of lottery before the French Constituent Assembly first met, and arguably also why rotation initiated by rule-based assignment was adopted to subdivide the unwieldly large assembly in smaller deliberative committees, in order to open up the debate and frustrate factional interests (Frederiksen Reference Frederiksen1936).

The implementation of the bureau-system was retained for use in the constituent assembly, and so the pool consisted of all the deputies already elected to serve in 1789. There was no screening or auditing of deputies or committees. Particularly in nascent institutions, creation of committees by lottery is fair and can prevent corruption or coordination by anti-democratic elites. When the bureau-system was introduced, Mirabeau noted, “However imperfect these provisional rules may be, however vicious may be the meetings by Bureaus, in spite of all the changes demanded it has not been possible to do better in a moment when as yet we have only a feeble knowledge of the tactics of political assemblies and feel the urgent need of some discipline” (Point du Jour, vol. I, p. 298, June 7, 1789, Session of June 6). Mirabeau, however, could not anticipate the future use of these procedures – lottery-based selection would continue to perform a sanitizing and organizing role, and would be used during the Third Republic (as we discuss in the next section).

Finally, the case of the French Constituent Assembly also provides evidence that the random assignment of deputies led to high-quality legislation. In fact, the bureaux also played an essential role in developing one of the greatest documents in constitutional history — the Declaration of the Rights of Man (Dawson Reference Dawson1978). This document was a preamble to the ultimate Constitution, and defined and guaranteed rights such as liberty, property, security, and freedom from oppression. Yet initially, the Constituent Assembly was having difficulty coming to agreement on the text of the Declaration of Rights; various drafts had been sent to the bureaux and soundly rejected (Frederiksen Reference Frederiksen1936). Amid these difficulties, some bureaux created their own drafts for consideration by the Constituent Assembly. The foundational draft of the famous Declaration of Rights actually came from the sixth bureau (Dawson Reference Dawson1978) – the text was first provided by this group, and then slightly modified and adopted by the chamber. The Declaration of Rights was born of the discussion of deliberative, random groups.

5.2 Danish Constituent Assembly 1848–1849

We now turn to an empirical investigation of the effect of deliberative policymaking in the Danish Constituent Assembly in 1848. In the mid nineteenth century, Denmark was a multiethnic kingdom that consisted of mainland Denmark, inland duchies, and overseas territories that had been ruled by an absolutist monarchy since 1665. Yet revolutionary pressures across Europe, namely, the wave of revolutions in 1830 and 1848, had emboldened Danish liberals seeking change. To prevent revolution, a succession of kings voluntarily initiated the transition to a constitutional monarchy, beginning with a Constituent Assembly in October 1848, which then culminated in the adoption of the “June Constitution” on June 5, 1849. Even though Denmark experienced peaceful democratization, during the constitution drafting there was sharp internal conflict between groups wanting radical democracy and those wishing to preserve the power of the king (Neergaard Reference Neergaard1892).

Institutionalized parties had not yet developed in Denmark, but historians and Danish scholars have clearly identified three ideological movements, which we term blocs, corresponding to the Right, Center, and Left (Olsen Reference Olsen1972; Pedersen Reference Pedersen2014; Hansen Reference Hansen2016; Hansen, Klemmensen, and Ralud Reference Hansen, Klemmensen and Ralud2018). Though these were the first manifestations of democratic parties, our case benefits from the fact that blocs had clear positions on the extent to which the constitution should be liberal and democratic.

The Conservative Right (Højre) of the assembly consisted of the aristocratic, landed elites, who wanted to preserve the power of the monarchy and drastically limit suffrage. The National Liberals were a diverse group of centrists, representing the educated elites of the cities. They were pro-democracy (restricted to the educated classes), pro-suffrage, and limiting power of the king, and had pushed for the constitution (with support from farmers groups). However, both the conservative landed proprietors and bourgeois liberals faced a significant threat from the radical left, originating in the agricultural classes. The Left (the Society of the Friends of the Peasants) was a highly organized movement of well-to-do smallholders and farmers that promoted radical, “pure democracy” with the broadest suffrage possible.

Thus, the main cleavage running throughout the constitutional debates, which we explore in our analysis, revolved around the level of freedom in the constitution draft – a tradeoff between expansion of political power to everyday citizens, and powers retained by the constitutional monarch. Importantly, groups were also reasonably cohesive (Hansen, Klemmensen, and Ralud Reference Hansen, Klemmensen and Ralud2018). We estimate ideal points as an insightful summary statistic to describe revealed preferences in the constituent assembly, employing OC non-parametric estimation using all forty-four roll call votes of the Danish constituent assembly (Poole et al. Reference Poole, Lewis, Lo and Carroll2012).

Figure 1 plots the ideal points of each member, as well as their party bloc affiliation from the constituent assembly elections. We can clearly distinguish three ideological blocs that are mostly ordered on a single dimension (hence, our findings are consistent with those of Hansen Reference Hansen2016). It is likely that the first dimension represents the ideological left–right scale, which at the time largely indicated preferences for more or less democracy, while a second dimension could be related to the urban–rural divide (Hansen Reference Hansen2016).

5.2.2 Equalizing Access, Preventing Elite Capture, and Implementation

Equalizing access to power was one rationale for the use of deliberative committees. Notably, the historical record shows us that Danish parliamentarians knew of the use and consequences of deliberative groups. The minutes of the Constituent Assembly reference both the French system of bureaux and its use in other countries, when discussing the adoption of such procedures. For example, Algreen-Ussing, a leading conservative member, argued that discussions should happen in the afdelinger because it allows for participation from a broad range of members;Footnote ²² and Ørsted concluded it allows for dialogue across politicians who normally would not speak to each other, and so the system should be tried.Footnote ²³ Thus, qualitative evidence shows that the draft constitution was deliberated in randomly assigned groups of legislators in order to ensure equality and diversity in the debate.

Elite capture in this historical case can be defined as the ability of pro-monarchical, conservative elites to draft the constitution in a way that advantaged the monarchy and undermined the ability of other political groups to have access to power. Given that the king appointed additional like-minded deputies – after the election for the Constituent Assembly had taken place – to ensure elite interests dominated, this was a real concern. But were deliberative committees successful in exposing individual deputies to different viewpoints, and preventing the pro-monarch (and anti-radical democracy) elites from dominating the process?

Using our unique, micro-level data on deputies and afdelinger assignment, we can look at how the lottery provided random variation in group composition. Figure 2 demonstrates how party bloc varied across groups as a result of the lottery; for example, the Left-wing deputies had a plurality in afdelinger group 4, but were the smallest faction in afdelinger group 3. It was, therefore, the case that, thanks to the lottery, landed elites were not concentrated in a way where they could capture even one committee – let alone the entire process. In addition, in all groups, members were exposed to crosscutting viewpoints in their deliberation (as opposed to primarily caucusing within their party blocs, as we might have expected in an early nineteenth-century chamber). Since the primary mechanism behind studies of deliberative democracy is exposure to different viewpoints (Goldberg et al. Reference Goldberg and Bächtiger2023), our data demonstrates important variation across the groups.

Figure 2

Composition of randomly assigned afdelinger groups, by party bloc

In terms of implementation, groups to debate the constitution were drawn from a hat. There is no historical evidence to suggest that the lottery was not random or was not implemented correctly. The pool from which the committees were drawn consisted of all deputies elected in the constituent assembly, and so (like in the French case) there was no screening or auditing of deputies or committees. The selection process was an administrative tool, used to constitute the groups, which stayed constant throughout the rest of the session.

5.3 Empirical Study: Open Decision-making in Denmark

Current research on the use of deliberative groups emphasizes their potential for policymaking, but what do we know about deliberative policymaking? How does policy deliberation in randomly assigned groups affect legislator behavior? Modern studies show that deliberation in random groups of mixed partisanship can moderate everyday citizens’ opinions and reduce polarization (Sunstein Reference Sunstein2002; Fujiwara and Wantchekon Reference Fujiwara and Wantchekon2013; Grönlund, Herne, and Setälà Reference Grönlund and Setälà2015; Gastil Reference Gastil2018; Dryzek et al. Reference Dryzek, Bächtiger and Chambers2019; Gastil Reference Gastil2018; Farrell, Suiter, and Harris Reference Farrell, Suiter and Harris2019; Strandberg, Himmelroos, and Grönlund Reference Strandberg, Himmelroos and Grönlund2019; López-Moctezuma et al. Reference López-Moctezuma, Wantchekon, Rubenson, Fujiwara and Pe Lero2022), but they study regular citizens. Other studies have assessed deliberative effects through opinion change in surveys or participant observation, and citizens are rarely asked to vote on policy issues (Farrar et al. Reference Farrar, Green, Green, Nickerson and Shewfelt2009; Esterling, Fung, and Lee Reference Esterling, Fung and Lee2015). One recent experiment (Zelizer Reference Zelizer2022) with present-day legislators conducting (supplemental and voluntary) deliberation on a set of bills demonstrates a causal effect of deliberation on position-taking, but this study relates only to extracurricular caucusing. While group composition is an important determinant of outcomes (Mendelberg and Karpowitz Reference Mendelberg, Karpowitz and Rosenberg2007), we do not know how these theories of deliberation in random groups translate to a national scale with elected politicians, where policy stakes are high.

Our study builds on prior work by exploiting random variation in the partisan composition of groups to test the theory that exposure to diverse viewpoints – as participants debate with members from opposite party groups – changes voting behavior. Our study also contributes by testing these claims in a legislative setting. A standard legislative committee system always involves the deliberation of policy in groups. However, group composition is often dominated by governing parties, and individual members either self-select into their committee assignment or are assigned to them by party members (interfering with our ability to estimate the causal effect of deliberation). Only rarely can we exploit random assignment to committees (Broockman and Butler Reference Broockman and Butler2015, Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2018), and only one recent study, to our knowledge, tests deliberative effects of randomly selected bills in a voluntary, bi-partisan group setting (Zelizer Reference Zelizer2022). Our study builds on this nascent literature, to also estimate the causal effects of deliberation on policymaking over time in a high-stakes setting.

5.3.1 Empirical Strategy

Our primary research question is to what extent deliberative groups affected legislative behavior during the drafting process, as measured by voting via roll calls. To answer this question, we assembled a micro-level dataset based on archival records from Denmark relating to the Constituent Assembly of 1848. We have data of Danish deputies, from 1780 to 1880, generously shared by Martin Hansen and supplemented by our own research. We match this to party affiliations, the random lottery draw, and the official minutes of the constituent assembly, as well as the forty roll call votes that occurred in 1848–1849 (we use the forty votes after deliberative groups were drawn). Our sample consists of maximum 152 deputies elected to the Constituent Assembly.

We provide statistical evidence demonstrating that the afdelinger truly were random in Figure 3; this shows that individual covariates (age, education, whether they were appointed by the king, whether they were of the nobility, whether they were a political before, a deputy from Copenhagen, and party bloc affiliation) did not predict afdelinger assignment.

Figure 3

Balance in characteristics across random groups. P-values shown; characteristics with p-values smaller than 5 percent signify significant differences in a given afdeling compared to afdeling 1 (the base group). Results robust to alternating the base group.

Our treatment variable is the partisan composition of a member’s afdelinger group assignment, measured by the total number of Left or Right deputies. Our control variables measure the individual, pretreatment, and time-invariant characteristics of assembly members. To measure elite status, we include indicator variables for whether the deputy was king-appointed, a member of the nobility, or representing the capital Copenhagen. We also include the age of the deputy and an indicator variable that measures university or law school education. We also consider prior political experience, so we code whether an assembly member had served as a member of one of the consultative assemblies, established in 1834.

We also analyze the parliamentary group of each deputy, relying on Barfod’s (Reference Barfod1849) classification and Hansen’s (Reference Hansen2016) codings. Parties did not yet exist, so our analysis relies on ideological blocs that dominated the election and discussion at the time of the assembly: Left, Centre, and Right, representing the radical left and the Society of the Friends of the Peasant; the National Liberals, and the conservative factions, respectively. Membership was fixed throughout the assembly, and so we use bloc membership and roll call votes to examine party cohesion, using the standard Rice cohesion index. Rice scores are calculated as $RIC{E}_{ij}=\left|AY{E}_{ij}-NA{Y}_{ij}\right|$ for group $i$ on vote $j$ , averaged across all votes to obtain the index (Rice Reference Rice1928, Carey Reference Carey2009). We do this for both party and deliberative group.

Our dependent variables measure different types of voting behavior – voting in support of pro-democratic amendments and voting with one’s ideological bloc. First, if deliberation in diverse groups matters, then one way this might manifest would be in an increased likelihood to vote in favor of pro-democratic amendments. In these deliberative groups, Conservatives and Centrists were randomly exposed to new, radical democratic viewpoints of the Left. To measure this, we restrict the roll call vote sample to eighteen votes relating to democratic governance or king’s powers, and we recode votes such that the ProDemocracy variable takes the value of “1” if the deputy voted in favor of a pro-democratic reform and “0” otherwise. We exclude absences and abstentions. These votes covered topics like age and property requirements for suffrage, the dissolution of the parliament, powers of the king, and the electoral law.

In our estimation strategy, we take advantage of the fact that groups were constructed using a lottery, and so group assignment is orthogonal to other explanations for voting behavior. We provided statistical evidence demonstrating that the afdelinger truly were random (see Figure 3). We then employ a simple specification that estimates the effect of random group composition on voting for constitutional amendments, during the year of the Constituent Assembly, using a legislator-vote panel dataset:

$Outcom{e}_{it}={\beta }_{1}TotalNumbe{r}_{n\left(\neg i\right)}+{\lambda }_i{X}_i+{\gamma }_t+{\epsilon }_{it}$

where $Outcom{e}_{it}$ is a member’s vote in support of a pro-democratic amendment at bill $t$ , or vote against the Party Bloc on an amendment $t$ . $TotalNumbe{r}_{n\left(\neg i\right)}$ is a variable that measures, within each Afdelinger n (where n is between 1 and 5), the number of members of a party bloc (excluding the individual legislator $i$ ). We construct two versions of this variable, looking at either the number of pro-democratic Left or pro-king Right deputies in a specific Afdelinger. Then $X_i$ is a vector of individual covariates, ${\gamma }_t$ a set of dummies for each roll call vote, and ${\epsilon }_{it}$ is the error term. We cluster standard errors by individual. Importantly, as afdelinger group is randomly assigned, we can estimate the causal effect of deliberative group composition on behavior. Therefore ${\beta }_1$ the coefficient on $TotalNumbe{r}_{n\left(\neg i\right)}$ can be interpreted as the effect of adding an additional member of a certain party to the deliberative group, on voting behavior.

5.3.2 Open Decision-making Results

We first examine to what extent deliberative group assignment induces cohesion, as a function of moderation of preferences. If constant discussion in deliberative groups changed attitudes, we might see group members exhibit high levels of cohesion within their assigned afdelinger. We construct Rice index scores for the forty deliberated constitutional roll call votes, to measure deliberative group cohesion. These vary from 0 to 1 per group and vote, and average out at 0.41. In contrast, average RICE index per vote for party blocs on all 44 votes was higher – 0.55, 0.60, and 0.69 for the Right, Centre, and Left. Hence, party cohesion scores are higher than deliberative group cohesion scores. Yet this effect might vary over time – if deliberation is inducing cohesion, we might be able to better track this by looking at cohesion over the sequence of votes.

Figures 4 and 5 present these Rice cohesion scores in chronological order of roll call votes, by afdelinger group (Figure 5) and by party (Figure 4), as well as a fitted trend line. Vertical line indicates when afdelinger groups were drawn. Figure 5 demonstrates there is no such clear increasing trend line that is the same in all of the five groups. In contrast, there is such an increasing pattern for party blocs in Figure 4, showing that parties perhaps coordinated slightly better as the Constituent Assembly progressed.

Figure 4

RICE scores by party bloc

Figure 5

RICE scores by afdelinger group

To explore directional effects, we now examine the group composition with respect to both ideological extremes in the case – the left, in favor of radical democracy, and the right, in favor of retaining the powers of the monarchy. Table 2 presents regression results of pro-democratic voting on group composition, party bloc, and individual pretreatment characteristics, and we find no group composition effects. We look at the share of Left Deputies, but also consider the share of Right Deputies, as our treatment variable. Increasing the percentage of Left Deputies, or alternatively the Right, has no statistically significant effect on voting behavior. Instead, we find statistically significant effects of party bloc on voting behavior – this is the case even though political parties were new.

Table 2Regression results. Dependent variable is voting for pro-democratic amendments. Columns 1 and 2 look at the number of Left Deputies as the treatment variable, and Columns 3 and 4 use the number of Right Deputies. Alternate specifications include pretreatment, individual controls (king-appointed, nobility, on the constitution committee, political experience, Copenhagen, education, and age).

Table 2 Long description

The table presents the results of regressing the dependent variable voting for pro-democratic amendments in the Danish Constituent Assembly. Columns 1 and 2 look at Number of Left Deputies as the treatment variable, and Columns 3 and 4 use the number of Right Deputies. Alternate specifications include pre-treatment, individual controls (King appointed, nobility, on the constitution committee, political experience, Copenhagen, education, and age). Models include party bloc and bill dummies.

We also consider heterogeneous treatment effects, in that group composition might have differentially affected members of different parties; if this is the case, our previous results might be averaging over the true effect. Figure 6 splits the sample to look at effects by party, and similarly finds no statistically significant effect of group composition on voting behavior. This demonstrates that even fledgling party groups exert significant influence on members, even in early stages of democratization.

Figure 6

Main results by party bloc subsample. Dependent variable is voting for pro-democratic amendments. Specifications include bill and afdelinger dummies as well as individual controls (king-appointed, nobility, on the constitution committee, political experience, Copenhagen, education, and age).

As discussed earlier, prior research has consistently found effects of deliberation in mixed ideological groups on preferences, attitudes, and even voting behavior. We find no such effects on behavior, even in a setting where deliberation happened for an extended duration across multiple policy areas, and where other determinants of voting (such as party discipline) were theoretically weak. One limitation of our study is power – while we can exploit the random assignment of the treatment variable, and the voting on amendments over time, we are limited in our sample size (both in the total number of deputies and in the historical length of Constituent Assembly) and thus our ability to detect anything but large treatment effects. Another limitation is that Constituent Assembly voting might not generalize to typical legislative behavior. However, our study still sheds important light on understanding the potential role of deliberation in an actual legislative context.

5.4 Discussion

The Danish case shares much in common with modern-day citizens’ assemblies, frequently used today to debate constitutional reforms, and deliberative committees. In this section, we explored the use of lotteries in constituent assemblies. The French Constituent Assembly in 1789 and the Danish Constituent Assembly of 1848 provide insight into the rationale and use of lottery-based groups in representative assembles. It shows that deliberation in random bureaux was used to provide structure in a time of uncertainty. It also demonstrates the main benefit was deliberation and exposure to different viewpoints, across class and partisan affiliation.

In this section we also provided original evidence on the Danish 1848 case. We exploited a natural experiment that introduced lottery-based deliberation in the 1848 Constituent Assembly of Denmark. We analyzed if policy deliberation in randomly assigned groups has a causal effect on voting behavior, both in supporting the inclusion of democratic amendments and partisan voting. We also examined deliberative group and party cohesion. Overall, we fail to find strong support for the hypothesis that deliberation in mixed groups alters voting behavior, which adds to a growing literature studying the modern implications of deliberative democracy.

There are several reasons why prior findings from research on citizens’ assemblies may not replicate in a legislative context. Effects may be attenuated when the stakes are high, or when politicians know their behavior could shape the democracy for years to come. Legislators also may individually update from deliberation, but instead vote in a way that represents their constituents. Importantly, while we find no deliberative effects, we find statistically significant effects of party affiliation, even in a fledgling democracy. This suggests any changes as a result of deliberation may not be strong enough to upset partisan affiliation.

Though a historical case, the Danish Constituent Assembly provides a natural experiment by which we can learn about the real-life effects of deliberation, in a foundational setting. While modern studies of deliberative policymaking have found that deliberative groups change attitudes of members who are everyday citizens (Carpini, Cook, and Jacobs Reference Carpini, Cook and Jacobs2004, Dryzek et al. Reference Dryzek, Bächtiger and Chambers2019) and legislators who voluntarily deliberate (Zelizer Reference Zelizer2022), these effects might not travel to national arenas that involve repeated high-stakes decision-making and partisan coordination. Further, the introduction of deliberative decision-making is increasingly being advocated as a democratic reform, and our results indicate more research is needed in a legislative context. Finally, even partisan preferences have significant power in early democratization, regardless of innovative institutional procedures.

6.1 Lottery-Based Procedures in Nineteenth-Century Europe

As mentioned in the previous section, the origins of the modern use of lotteries can be found in the French Revolution and the Constituent Assembly. The use of randomization and lotteries to subdivide the legislature into smaller groups called bureaux continued in the First Legislative Assembly of 1791 (Frederiksen Reference Frederiksen1936), in the different republics and empires in France up to 1870, and then also in the French Third Republic.

But outside of France, and as an example of institutional spillover, other European countries adopted lottery-based procedures (sometimes making explicit reference to the French practice). The Danish Constituent Assembly of 1848 is one example, as discussed in the last section, but lottery-based procedures to subdivide legislators into smaller, deliberative committees, to assign legislative tasks, or to select individuals to committees, were used in different parliaments in the Netherlands from 1815, Imperial Germany from 1871, and in the Austrian Empire from 1875. Table 3 details the use of lotteries in Europe.

The rationale behind the adoption and dissolution of lottery-based procedures seems to have been similar across cases. In the long nineteenth century, there was high distrust among political actors, as there were various, sometimes unknown, individuals supportive of the ancien régime who tried to work against the democratic experiment in favor of the established monarchies. It is precisely in this context of high distrust that individuals will be most likely to agree to a political lottery, which removes the human choice element from the selection (Elster Reference Elster1989). It is also precisely in such a context that the political lottery will be most effective in opening up access to new actors. We argue that the primary concern was to avoid one organized faction, such as the revolutionary or the anti-democratic faction, influencing the debate and decision-making process (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2019).

Importantly, lottery-based procedures were not necessarily a temporary solution – they remained for decades in the cases of France, Austria, and the Netherlands, and were stable institutional rules as part of regular democratic policymaking. The widespread use across European cases over time demonstrates the utility of lotteries in day-to-day legislative politics.

6.2 Lottery-Based Procedures in the French Third Republic

The French Third Republic was established after military defeat, after a succession of regimes following the violent revolution of the eighteenth century. Civil conflict was rife, and monarchists opposed republicans in how they viewed the future of the country. The National Assembly, the elected lower chamber of the bicameral parliament, was formed in 1871 and existed until 1940. Parties were not legalized in the French Third Republic until 1901. Before then, politicians could hold multiple affiliations to various parliamentary groups. While mass organization was banned, the legislative groups did start to organize and direct legislative activities within the chamber. For most of this period, a lottery-based procedure was used to choose members of the most powerful standing committee – the budget committee.

To shed light on the effects of lottery-based procedures in legislative committees, we undertook a comprehensive study of the powerful budget committee in the French Third Republic after 1870. The data in this section comes from two studies of lotteries in France (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2018, Reference Cirone and Van Coppenolle2019). Our dataset is a panel of 1,330 deputies by year, from 1881 to 1914. Draws took place before the nomination of the budget committee up to 1909. We collected information on bureaux lists, committee appointments, legislative activity, and district-level electoral outcomes for each deputy. We also conducted detailed biographical research to collect data on deputy qualifications, occupations, and careers. This data allows us to look at the effects of the lottery-based procedure across decades of legislative activity.

6.2.2 Representation and Capture in the Budget Committee

The rationale behind the incorporation of a lottery in the selection procedure stems from the fact that the members of the French parliament were wary of factions; as mentioned previously, political parties would be banned from organizing until 1901. Legislators were also concerned with the skill, competence, and motivation of the committee members who would be selected – all of which are certainly required for the highly technical issues studied in detail by the budget committee. But did the lottery-based procedure open access to and diversify the powerful budget committee?

We use our empirical data to show this was the case. In Cirone and Van Coppenolle (Reference Cirone and Van Coppenolle2018), we analyzed our micro-level yearly panel of deputies from 1881 to 1914 and we studied the types of legislators who were selected to the budget committee. In nineteenth-century France, society was roughly divided into the elite, middle class, peasants, and workers, though only the elite and middle class typically served as deputies. The Third Republic was characterized by the rise of the middle-class politician, coming from the grand bourgeoisie of lawyers, doctors, professors, and bankers, at the expense of landed and elite politicians.

We first conducted balance checks and placebo regressions to ensure that no observable and theoretically relevant characteristic predicted group assignment (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2018). The French adopted an important detail concerning implementation, in that starting in 1894, the bureaux lists were only drawn up two hours before selections were to take place, to protect the budget committee from corruption and to impede coordination across bureaux (Gooch Reference Gooch1928). As a result, the lottery to select the bureaux groups was not manipulated in obvious ways, even as it took place repeatedly over time.

We also collected data on deputy characteristics that would indicate a member of the elite political class or rising middle class, and also look at other demographic data and the ideological bloc (party family). We then compare the composition of the budget committee to the legislature across the entire period during which the bureau-system was in place, from 1887 to 1910, in Figure 7. Under the lottery-based procedure, compared to the chamber average, committee members were younger, more centrist, and more often middle-class.

Figure 7

Deputy characteristics: Chamber versus budget committee under bureaux

(Cirone and Van Coppenolle 2019)

But it might have been the case that there were norms or unobserved explanations for why the budget committee composition was different than the chamber during this period. To disentangle this issue, we also exploit the fact that the lottery-based procedure was abolished in 1910, and control over selection of budget committee members was given to party groups. We conducted a statistical analysis to look at who gets selected to the budget committee in the 1906 term (under the lottery-based procedures) to who gets selected in the 1910 term (under partisan selection). Controlling for individual characteristics, like age, wealth, experience, and partisanship, we found a statistically significant difference in selection after the lottery was abolished (see Cirone and van Coppenolle Reference Cirone and Van Coppenolle2018). Once parties obtained this advantage, they once again favored more traditional elites, and more loyal party members.

6.3 The Dismantling of Lotteries in France and Europe

Political lotteries played an important role as part of lottery-based procedures during European democratization. Such lotteries helped to bridge the gap by organizing legislative activities while parties were still emerging. Yet lottery-based procedures, even if they were in place for a long time, were typically dismantled or sabotaged once parties became stronger and the party system stabilized. That is because organized parties with stable party leadership prefer to control selection, by election or appointment.

We defined a political party as a group with a given label and a stable membership that acts collectively to achieve policy or office goals (Mainwaring and Scully Reference Mainwaring and Scully1995, Levitsky and Cameron Reference Levitsky and Cameron2003, Lupu and Riedl Reference Lupu and Riedl2013). In new democracies, parties are often fluid groups with changing labels, platforms, and memberships; party factions that undermine a party’s stability can suddenly arise, and very quickly gain support from voters or legislators (Webb and White Reference Webb and White2007). Therefore, when parties are weak, party factions can easily dominate parliamentary or constitutional committees. As long as the lottery-based procedure is in place, the inherent uncertainty ensures that such factions cannot dominate selection. Yet once party membership stabilizes, parties move to eliminate these legislative inefficiencies. Party leadership prefers control over elections or nominations in order to be able to offer selective incentives to their members. In this way, they can encourage legislators to toe the party line, which helps the group improve cohesion and achieve its collective goals in the legislative arena.

In the French Third Republic, parties were legalized to exist outside the legislature in 1901, competed with party labels in 1906, and in 1910 dismantled the lottery-based procedure. In Cirone and van Coppenolle (Reference Cirone and Van Coppenolle2019) we examine the partisan incentives behind institutional reform using an analysis of the roll call vote on the 1910 reform that dismantled the lottery-based procedure. Here, we focus on party instability as an independent variable, which we measure using the proportion of deputies within a party who switched.

The right and the left party groups had a stable membership by 1910, which meant few deputies in that group were part of a different group in the previous legislative term. In stark contrast, within the centrist party group there were many smaller factions who housed members who had recently switched. Party disloyalty is an important predictor of how deputies voted over the 1910 reform which abolished the lottery-based procedure and gave control over selection to parties. The better organized parties of the extremes, with stable membership, voted overwhelmingly for the reform, with over 90 percent of their deputies supporting reform. They were the ones most clearly disadvantaged by the lottery-based procedure, and they wanted to control more seats as well as who would fill these posts. In contrast, the centrist party group was divided over the reform, therefore voting against the lottery-based procedure.

In France, the reform happened in 1910 and not earlier, because at this point party membership had sufficiently stabilized for parties to prefer control over appointments to further the collective goals of their individual members. Thus, party system institutionalization is what explains the move away from the lottery-based procedure.

As we have shown in France, the dismantling of the bureau-system coincided with party consolidation. Evidence from other countries is suggestive (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2019), and also points to an important role of lottery-based procedures in regulating legislative activity, until parties are cohesive enough to take over.

6.4 Discussion

The use of lottery-based procedures in the French Third Republic can shed insight into the broader use of lotteries in parliament. First, even a lottery-based procedure (as opposed to full lottocratic selection) can equalize access to power and improve representation. In the French Third Republic, the members of the budget committee were more likely to be middle class and centrist politicians. Once the lottery-based procedure was dismantled, and parties retained control of committee assignments, elites dominated once more.

Second, the two-step selection procedure might retain unique advantages. While the randomization in the first stage of a lottery-based procedure helps improve access to power and reduce elite capture, it retains the element of election in the second stage. Candidates who are those most willing, motivated, interested, and capable to serve can still compete for the position in the second stage. That creates very different incentives for legislators compared to sortition, where a random draw cannot select on these characteristics. We found that those chosen by a lottery-based procedure sponsored budget legislation and were to further their careers by serving as ministers in the future; this selection procedure did not return suboptimal candidates. Here we demonstrate that introducing political lotteries need not imply a loss of expertise, all the while opening up the selection process to a more diverse range of candidates.

Finally, the role of parties in modern political selection must be considered.Footnote ²⁴ The modern literature on lottocracy recognizes a lacuna in that there is no proposed solution on how to incorporate political parties into parliamentary politics in the absence of elections (Landemore Reference Landemore2020, Landa and Pevnick Reference Landa and Pevnick2021), with the exception of hybrid parliaments with a sortition chamber that preserve political parties in the elected chamber (Gastil and Wright Reference Gastil and Wright2019). It is also the case that parties have begun to incorporate lotteries at lower levels of selection, such as using a lottery to select candidates to stand in the election from within the party; certain political parties in France, Greece, and Mexico have used such measures recently (Sintomer Reference Sintomer, Gastil and Wright2019, Poertner Reference Poertner2023).

But our evidence demonstrates that any calls for lottocracies must also theorize a place for parties. In European democratization the “arational” element in the selection procedure was only acceptable as long as factions were weak and highly distrustful of each other and was also quickly dismantled once parties became strong and wanted control over selection to reward loyalists (Cirone and Van Coppenolle Reference Cirone and Van Coppenolle2019). Lottery-based procedures must be acceptable to and supported by political parties, or else they will be undermined or dismantled.