Window of Opportunity: War and the Origins of Parliament

Abstract

Two important puzzles characterize the development of pre-modern Eurasian polities. First, most rulers convened councils of nobles, but only European monarchs expanded them to create parliaments. Second, war was common throughout Eurasia, but only in Europe did it correlate with the formation of parliaments. We advance a new argument about the emergence of parliaments that accounts for both stylized facts while integrating the literature highlighting the rulers' need to finance wars with that emphasizing the importance of the medieval communal revolution. Using novel data, we document a ‘no communes, no parliaments’ rule: monarchs established parliaments only after they had fostered the creation of self-governing towns (aka communes). We also show that war was a significant predictor of parliamentary births across medieval Europe – but only during a window of opportunity that opened after a polity had experienced the communal revolution.

Introduction

Throughout early medieval Eurasia, rulers convened councils of nobles for various purposes (Ahmed and Stasavage 2020). Yet, only in Europe did these councils eventually expand to include representatives of non-noble segments of society, thereby creating parliaments. The emergence of European parliaments is thought to be a vital part of the state-making process leading to modern forms of parliamentary democracy (Stasavage 2010; van Zanden, Buringh, and Bosker 2012). Moreover, by improving property rights protections, parliaments may have also promoted Europe's remarkable economic take off (Acemoglu, Johnson, and Robinson 2005; Cox 2017). A clear understanding of how parliaments first emerged is thus key to explaining modern patterns of political and economic development.

One traditional explanation for the formation of parliaments highlights the need of medieval European rulers to finance wars (for example, Dhondt 1950; Finer 1997; Graves 2001; Sidgwick 1903; Tilly 1994). According to Finer (1997, 1,026), the ‘one perennial, common factor’ in the emergence of parliaments was that ‘kings and princes wanted to make war’. Similarly, Tilly (1994, 24) sees the establishment of parliaments as the ‘outcome of bargaining … for the wherewithal of state activity, especially the means of war’. While influential, this perspective generally overlooks the fact that war was common across many parts of pre-modern Eurasia, where parliaments never emerged (for example, Dincecco and Wang 2018; Dincecco et al. 2022).

Another school of thought emphasizes the role of medieval Europe's communal revolution (Angelucci, Meraglia, and Voigtlander 2022; Greif 2008; Mitterauer 2010; Pirenne 1946; van Zanden, Buringh, and Bosker 2012). For example, Mitterauer (2010, 135) argues that ‘[urban] self-government [was] a precondition for [towns’] representation in territorial and imperial Estates’, the key move in the creation of parliaments. However, this perspective often downplays how military pressures incentivized rulers to strike political bargains in the first place.

In this paper, we offer a new argument about the formation of parliaments that integrates both of these important literatures. While each literature is long-standing, ours is the first effort we know of that explicitly considers how they relate to one another. Medieval European rulers faced two political constraints that their counterparts elsewhere in Eurasia did not, thereby opening a ‘window of opportunity’ during which they had incentives to broaden their noble councils into parliaments. First, after the collapse of the Roman Empire, European rulers had lost central bureaucracies capable of collecting taxes, thus forcing them to enlist the help of intermediaries (Stasavage 2020, Ch.5). Second, during the medieval communal revolution, circa 1000CE to 1300CE, many European towns acquired rights of self-government and simultaneously became crucial intermediaries in urban tax collection (Weber [1921] (1958)).¹ Thus, in the early modern period, Europe's parliamentary window of opportunity opened with the communal revolution and shut when rulers managed to create central tax-collecting bureaucracies.

We argue that, during this window, rulers faced the dual challenge of collecting taxes to finance war via intermediaries and integrating self-governing towns into their decentralized tax administration systems, and they had strong incentives to expand their noble councils into parliaments. Rulers in need of funds – generally for military aims – found it more cost-effective to bargain collectively with self-governing towns. Therefore, they summoned non-noble urban elites to attend previously exclusive noble councils.

When rulers did not face the two political constraints noted above, war should not have induced them to form parliaments. For example, since imperial bureaucracies already collected taxes, the war did not create any tax-administrative reason for Chinese emperors to add merchant elites to the noble councils they consulted. When rulers did face both political constraints, a war-induced need to consult with tax administrators was an important (even if not the only) mechanism for promoting parliamentary births.

To empirically evaluate our account, we identify parliamentary birth years for all Christian polities that were sovereign (or semi-sovereign) in 1200, exceeded 5,000 square kilometers in physical size, and survived at least 100 years (that is, until at least 1300). To our knowledge, the resulting sample of forty-four territorial monarchies is the most comprehensive database of its kind – approximately twice the sample size in Stasavage (2010) and Kokkonen and Møller (2020), and one-third larger than that in van Zanden, Buringh, and Bosker (2012).

Using our database, we first document a ‘no communes, no parliaments’ rule: the medieval formation of communes preceded all the parliamentary births in our sample. Moreover, we show evidence that the timing of the communal revolution in each polity correlated with the timing of the subsequent emergence of parliament and that the towns first summoned to parliament had rights that made their active cooperation in collecting taxes valuable.

We next analyze how war promoted parliamentary births in the aftermath of the medieval communal revolution. Previous work examined how the war impacted the frequency of parliamentary meetings (Abramson and Boix 2019; De Magalhaes and Giovannoni 2022; Grzymala-Busse 2023; Kokkonen and Møller 2020; Stasavage 2016). By contrast, we study whether military pressures induced rulers to create new legislative institutions in the first place – a key implication of Tilly's (1994) classic thesis that ‘war made the state’.

To measure military pressures, we code each polity's annual participation in nearly 850 pre-modern conflicts (using Jaques 2007), producing one of the most comprehensive databases available. Our panel regressions indicate that a polity that averaged one conflict per year over a quarter-century had a per-year probability of a parliamentary birth 5.60 percentage points higher than a polity that never fought.

By including polity fixed effects, we control for each polity's initial economic, institutional, and socio-cultural features (for example, initial ruler strength) and fixed geographical features (for example, waterway access, terrain ruggedness, soil quality, physical size, and island status). In addition, time-fixed effects control for common time shocks (for example, economic trends, population dynamics such as the plague, and technological innovations). For robustness, we also include polity-specific trends to account for time-varying unobservable features specific to each polity. Further, we control for several time-varying observable features, including each polity's trade potential, years of ruler successions, lack of male heirs at the ruler's death, and hereditary succession arrangements (primogeniture). Similarly, our results are robust to excluding polities one by one, adding city-states to the polity sample, coding alternative parliamentary birth years (in cases where historians disagree), employing alternative conflict data (from Kokkonen and Sundell 2020), taking 10-year and 20-year data averages, and employing fixed effects at a higher level of aggregation.

Overall, our investigation helps to resolve two key puzzles. The first is why the war had different effects on state-building across pre-modern Eurasia, even though it was common throughout. The second is why parliaments only emerged in medieval Europe – even though noble councils were a widespread phenomenon throughout Eurasia – and did so only during the window of opportunity, during which we argue that rulers had strong incentives to undertake institutional reform.

In the rest of the paper, we proceed as follows. Section 2 develops our argument for the formation of parliaments. Section 3 documents the ‘no communes, no parliaments’ rule. Section 4 analyzes the impact of war on the establishment of medieval parliaments across Europe. Section 5 performs robustness checks. Finally, section 6 concludes with additional remarks about why we would not expect war to have had similar impacts on political development elsewhere in Eurasia.

Governance Rights in Medieval Europe

We highlight two uniquely European political constraints under which war should have motivated rulers to expand their noble councils to include non-noble members. First, circa 1000CE, European rulers neither possessed central bureaucracies capable of collecting taxes nor had they the wherewithal to create them (Stasavage 2020, Ch.5). Accordingly, they were forced to farm out tax-collecting rights and responsibilities. The first recipients of such rights and responsibilities were the banal lords, who attended their ruler's councils (Mitterauer 2010, Ch. 4). By contrast, rulers elsewhere in Eurasia had central tax-collecting bureaucracies and armies and never allowed hereditary nobilities to take root (Mitterauer 2010, Ch. 4; Stasavage 2020, Ch. 6). Second, after the medieval communal revolution (circa 1000CE to 1300CE), self-governing towns with important roles in tax administration emerged in all major European polities. By contrast, such towns never emerged elsewhere in Eurasia (Weber [1921] (1958)). As more European town corporations acquired rights of governance and tax administration, securing their cooperation in collecting taxes became just as important as securing the cooperation of princely tax farmers. Accordingly, rulers began to call ‘the burghers into the councils of prelates and nobles with whom they conferred upon their affairs’ (Pirenne 1946, 220).

Why did the establishment of communes promote the subsequent formation of parliaments? One school of thought highlights the need to coordinate newly autonomous government units. Greif (2008, 30–1) argues that ‘administrative power leads to political representation because constitutional and other rules specifying rights are inherently incomplete contracts’. Similarly, Angelucci, Meraglia, and Voigtlander (2022, 3,446–7) invoke ‘a literature in both economic history and organizational economics that connects administrative autonomy to [the need for] centralized coordination’. Many details of the extraordinary taxes that funded the medieval war machine needed to be organized, and parliaments were an efficient way to make such arrangements (Bates and Lien 1985, 56; Pasquet 1964, 225; van Zanden, Buringh, and Bosker 2012, 847).

Another complementary line of thinking emphasizes rulers' principal-agent problems. Rulers needed to secure the active cooperation of tax administrators since the latter could cause substantial delays or shortfalls in revenue collection simply by dragging their feet, and rulers lacked the bureaucratic capacity to detect and punish this foot-dragging. The potency of foot-dragging as a counterweight to royal power was widely recognized. In Spain, for example, a popular maxim held that the king's commands should never be challenged as to their legality but could be indefinitely delayed in their implementation (Fernández Hevia 2001). In England, elites opposed the creation of royal bureaucracies in any area not subject to parliament's financial oversight, viewing the resulting opportunities for foot-dragging or non-implementation as crucial bulwarks against royal tyranny (Cox 2018). To secure active cooperation, rulers could offer various goods upfront and/or promise to share the spoils of victory. Either way, the expected profit of the proposed war(s) effectively constituted the surplus over which the ruler and the tax farmers might bargain. Parliaments were the fora for conducting necessary negotiations (Barzel and Kiser 2002).

Rulers who never allowed their cities to become self-governing did not face tax-based incentives to add urban representatives to their noble councils. By contrast, rulers who sanctioned autonomous cities quickly faced tax-based incentives to bring them on board. Thus, one can view parliaments as path-dependent creations stemming mainly from prior princely decisions to create self-governing towns.

Empirical Predictions

Our argument produces two main empirical predictions. First, European rulers should have summoned urban representatives to their pre-existing noble councils – thereby transforming them into parliaments – only after they had fostered the creation of communes. This is the ‘no communes, no parliaments’ rule. Second, once the communal revolution began, the necessity to finance war should have prompted European rulers to establish parliaments (by adding urban representatives to their noble councils). We will use these two predictions to guide our empirical analysis below.

No Communes, No Parliaments

Across Eurasia, no non-noble urban elites were ever summoned to join a ruler's councils absent pre-existing communes. This was clearly true outside Europe, where neither self-governing cities nor sovereign parliaments ever emerged. This section provides evidence that the ‘no communes, no parliaments’ rule held within Europe.

Polity Sample

To document this rule, we identified all European polities that satisfied the following three criteria: (1) the polity existed as a sovereign (or semi-sovereign) non-tribal entity in either Latin or Orthodox Christendom in 1200; (2) it exceeded 5,000 square kilometres in size in 1200; and (3) it survived at least 100 years (that is until at least 1300). We first created a master list of potential sample polities based on Stasavage (2010), van Zanden, Buringh, and Bosker (2012), Wikipedia, and Euratlas (Nussli 2010). Using this master list, we then determined whether each polity satisfied our criteria. This process yielded the forty-four territorial monarchies listed in column 1 of Table 1. We explain why relaxing the third criterion –including shorter-lived polities that survived a minimum of 50 years (that is, versus 100 years as in the benchmark) – would not significantly impact our analysis in Appendix B.

Table 1. The communal revolution and parliamentary births in Medieval Europe

Note: For construction methods and source materials, see text and Appendices B and C.

Parliamentary Birth Years

Definition

In his classic work Medieval Parliaments, Antonio Marongiu defines a parliament as ‘an independent body … containing members of three estates (the clergy, the nobility and the cities …), whose main functions are the granting of taxes and the participation in realm-binding legislation …’ (as quoted in van Zanden, Buringh, and Bosker 2012, 837). Following Marongiu, we define the birth year of a parliament as the first year in which a tax-controlling council included non-noble urban elites. Operationally, we count a parliament as tax-controlling if and only if it had the right to approve or reject some important forms of extraordinary taxation by a majority vote. Since most European councils already had authority over taxation and already contained the high clergy and nobility, our definition focuses on when urban elites joined.

An alternative would be to count all fiscally powerful councils as ‘parliaments’ irrespective of the presence of urban elites (Abramson and Boix 2019; Kokkonen and Møller 2020). If we followed this approach, we would include the birth of what Marongiu (1968, 45–51) calls ‘pre-parliaments’ in our analysis.² Such bodies emerged when monarchs conceded formal tax powers to their noble councils. Here, however, we focus on another important innovation – expanding membership in councils that already controlled taxation to include non-noble urban elites.

Examples

We rely on historical accounts for each polity to identify parliamentary birth years. Appendix B describes our sources and justifies our coding decisions. Column 2 of Table 1 lists the parliamentary birth years in our polity sample. We now provide three illustrative examples of how we proceeded, using the cases of England, Spain (León), and France. The legal historian Paul Brand (2009, 10) writes the following about England:

The period of just under a century which begins with King John's granting of the Magna Carta in 1215 and ends with the death of Edward I in 1307, is a significant one in the early history of parliament. It is the period when the term ‘parliament’ first comes to be used for the special occasional meetings of the king's council to which a larger group of the king's subjects were summoned … The term ‘parliament’ (parliamentum in Latin, parlement in French) was not used before the thirteenth century for the occasional special meetings of the king's council to which a wider group of participants was summoned to provide general advice to the king and to give consent on behalf of a wider national community to royal taxation and legislation.

The italics are ours. They highlight the parts of the quotation that correspond with our definition of parliamentary birth years. Brand makes two important points here. First, it was only between 1215 and 1307 that a parliament first existed in England. Second, a key role of this parliament was taxation. While Brand indicates that a ‘larger group of the king's subjects’ was called to such meetings, he does not yet specify which subjects he means. For this, we turn to another quotation of Brand's: ‘The earliest evidence of knights being summoned to attend parliament as representatives of individual counties come from 1254; the earliest evidence of burgesses being summoned to attend as representatives of their towns or cities only from 1265.’ (Brand 2009, 11)

This quotation indicates that ‘burgesses’ – that is, urban elites – first joined parliament in 1265. We thus date the birth of parliament in England to this year.

To further illustrate our definition of parliamentary birth years, we turn to León, home to the earliest parliament in medieval Europe. The historian Stanley Payne (1973, 82) writes: ‘In 1188 Alfonso IX of León faced major problems in consolidating his rule over an internally divided and disorderly kingdom while facing mounting financial demands.’ To deal with these issues, he summoned representatives of leading towns to meet with aristocrats and church officials at a royal assembly. As a result, he proclaimed a brief royal charter promising justice and recognizing local laws as well as the need to establish greater order. At a subsequent meeting, he gained approval for debasing coinage to increase royal purchasing power.

This quotation makes several key points with respect to our definition of parliamentary births. First, ‘representatives of leading towns’ – by which Payne means urban elites – did not join the king's royal council until 1188. Second, an important rationale for summoning urban elites was to meet fiscal demands. Here, the king received approval for currency debasement – effectively, new taxation. We thus date the birth of parliament in León to 1188.

France provides a final illustration of our definition of parliamentary births. The historian Antonio Marongiu (1968, 98) writes:

As relations with the Pope grew tense and bitter and Boniface VIII threatened the King with the gravest sanctions, Philip played his trump card and summoned the barons, prelates, and envoys of the cities to Paris on 15 February 1302. The writ of summons informed them that the King wished to treat and deal with them about the grave problems of the moment … The terms … expressed the substance of the great principle that ‘quod omnes tangit ab omnibus approbari debet’. This assembly is traditionally described as the first reunion in France of the ‘estates general’, of a parliamentary assembly or institution.

Here, the king first summoned ‘envoys of the cities’ – indicating urban elites – in response to ‘grave’ international relations. The Latin phrase ‘quod omnes tangit ab omnibus approbari debet’ translates as ‘that which affects all must be approved by all’, which we interpret in terms of consent to taxation. For such reasons, we date the birth of a parliament in France to 1302.

Challenges

The discussion above raises several points that merit clarification.

First, by ‘urban elites’, we mean non-noble urban elites, typically merchants, whether they represented communes or not. Were we to define parliaments as bodies containing elected representatives of self-governing towns, then the ‘no communes, no parliaments’ rule would follow by definition. Our less restrictive definition avoids this.

Second, urban elites throughout Eurasia could seek ennoblement and subsequently join their ruler's councils. However, we would not count this traditional route to political power as creating a parliament. Instead, our definition focuses on the expansion of membership in councils beyond the high clergy and nobility so that even urban elites who had not been ennobled could participate in council affairs.

Third, our coding strategy does not require that parliaments regularly convened after their first appearance. This reflects this paper's aim of evaluating whether war induced institutional innovations. Analyzing the frequency with which rulers turned to such institutions once they were invented or adopted is a separate (and complementary) exercise.

That said, weighing each parliamentary birth by the frequency of meetings that followed over the next half-century or century might be useful. Unfortunately, data on meeting frequency is not available for all of the parliaments in our sample. In principle, this means that our analysis may include ‘still births’ in which an institutional innovation was attempted but not subsequently kept to. In practice, though, stillbirths rarely occurred. Parliamentary births, as we define them, were typically followed by meetings over the next century. As in the study of technological innovations, we also believe there is ample value in documenting first uses, even if later usage is not uniform.

Town-Polity Matching and Communal Birth Years

We rely on NUTS territorial units to match towns to sample polities (Eurostat 2015). The virtue of this approach is that it always captured towns that lay within the historical territorial nucleus of each polity over time. In turn, we can produce a balanced panel of town-polity matches across European towns in the years between 1000 and 1600 (Bosker, Buringh, and van Zanden 2013). Appendix C describes this matching method in greater detail.³

Drawing on this town-polity matching, we determine that, when the medieval communal revolution affected each sample polity, rulers may have had important tax-based incentives to summon urban representatives to their noble councils, thereby creating a parliament. Column 1 of Table 1 displays the century in which the first commune was established in one of the polity's major towns (according to Bosker, Buringh, and van Zanden 2013).

Timing of Births of Communes and Parliaments

Table 1 documents two ways in which communes were important harbingers of parliaments. First, we observed that urban representatives were only summoned to attend parliament if their towns had previously acquired communal rights. Thus, the ‘no communes, no parliaments’ rule holds. Second, we observed that polities experiencing earlier onsets of the communal revolution tended to establish parliaments more quickly. For example, in polities whose first major town became a commune in the eleventh century (six cases); twelfth century (seventeen cases); thirteenth century (thirteen cases); or fourteenth century (one case), respectively, the median year in which they first convened a parliament was 1315; 1326; 1397; and 1523. Figure 1 depicts the positive correlation between the timing of the communal revolution and parliamentary births across medieval Europe.⁴

Figure 1. The communal revolution and parliamentary births in Medieval Europe

Notes: See text and Appendices B and C for construction methods and source materials.

Our argument suggests that the towns represented in the earliest parliaments should have generally been self-governing since cooperation with such towns would have been essential to the collection of taxes. Anecdotal evidence supports this claim. For example, in Castile, Flanders, Hungary, Portugal, and Scotland, only self-governing towns were called to parliament.⁵ In the Duchy of Brabant, approximately 75 per cent of urban areas represented in parliament in the fourteenth century were self-governing. The rest had liberties such as the right to hold markets (Damen 2018, 153–9). In England, most towns represented in parliament were self-governing (Angelucci, Meraglia, and Voigtlander 2022, 3, 452–4).

Large cities with rich merchants appeared throughout Eurasia, as did noble councils. Rulers throughout Eurasia had the power to expand their noble councils to include non-noble urban elites. Yet, this only happened in Europe. Moreover, they did so only within the window of opportunity we identified, during which time rulers had strong incentives to expand their councils to include the new species of corporate tax farmers that arose with the communal revolution.

Parliamentary Births

The previous section documents the ‘no communes, no parliaments’ rule derived from our argument. Given this empirical regularity, it is obvious that war prior to the medieval communal revolution does not predict parliamentary births, as no such births took place before there were communes. In this section, we analyze how the war impacted parliamentary births in the aftermath of the communal revolution. Each polity enters our sample when it experiences the communal revolution, thereby opening a window of opportunity for parliamentary birth, and exits when it establishes a parliament.⁶

Methodology

We use linear probability models to follow a general trend in the literature (see Angrist and Pischke 2009). This approach allows us to directly tackle unobserved heterogeneity that may cause bias via fixed effects (which are not as straightforward to implement in non-linear models; see Farbmacher and Tauchmann (2021). Moreover, we prefer the linear probability model to logistic regression due to ease of interpretation. To evaluate the impact of war on the births of parliaments, we estimate the following model using OLS:

(1)$${\rm Parliament}\;{\rm Birt}{\rm h}_{it} = \beta \,{\rm Conflict}\;{\rm Participatio}{\rm n}_{i( {t-p, t-1} ) } + {\boldsymbol \;}\alpha _i + \lambda _t + {\boldsymbol X}_{it}^{\rm ^{\prime}} \Phi + \varepsilon _{it}$$

here, Parliament Birth_it is a binary indicator that takes the value 1 for the year t in which sample polity i created a national parliament. To reflect the absorbing nature of this variable, it turns to missing for all subsequent years.

Conflict Participation_{i(t−p,t−1)} equals the average number of military conflicts per year in which polity i participated over the previous p years. In the benchmark specifications, p = 25 (that is a quarter-century). For robustness, we vary p = 5,   10,  15,  20,   30,   35 (to be reported in Section 5). Since this variable exploits information about the conflicts in which a polity participated in each year, it is more accurate than a dummy variable approach that computes the share of years in which a polity participated in a conflict. Still, for robustness, we employ the share of years of conflict as an alternative variable of interest (to be reported in Section 5).

Including polity fixed effects α _i allows us to control for each polity's initial economic, cultural, institutional, and social conditions (for example, initial ruler strength), along with controlling for the fixed geographical features of each polity (for example waterway access, terrain ruggedness, soil quality, physical size, and island status). Farbmacher and Tauchmann (2021) study the bias to which linear probability models with unit fixed effects are prone. However, they employ a short panel (that is T = 5) and evaluate bias as the number of observational units increases asymptotically. By contrast, we evaluate a long panel of several hundred years with a fixed number of units (that is, thirty-seven). Thus, we follow precedent (for example, Belloc, Francesco, and Galbiati 2016) and include unit fixed effects in our benchmark specifications. For robustness, however, we show that the main results remain similar in sign and significance if we exclude polity fixed effects and replace them with fixed effects at a higher level of aggregation (that is, by region, to be reported in Section 5).

Year fixed effects λ _t control for common time shocks, including economic trends, population dynamics (for example, due to the plague), and technological innovations.

${\boldsymbol X}_{it}^{\rm ^{\prime}}$ is a vector of time-varying control variables, including trade potential, years of ruler successions, lack of male heirs at the time of a ruler's death, and hereditary succession arrangements (primogeniture). We describe these controls in greater detail below. The standard errors are robust and clustered at the polity level.

If we were to perform our analysis on only the observations before the communal revolution in each polity, we would find that war had the same effect on parliamentary formation in Europe as it did elsewhere in Eurasia – none at all. This is because, while there were wars, there were no parliamentary births before the communal revolution. Thus, we restrict the sample to polity-year observations following the century in which the first commune was established in each polity. However, the main results hold if we include polity-year observations before the communal revolution (to be reported in Section 5).

Threats to Inference

Threats to inference in Equation 1 take the form of selection, omitted variables, and reverse causation.

With respect to selection bias, we rely on broad criteria to determine our polity sample (see Subsection 3.1). Thus, selection bias in favour of the most prominent surviving polities is unlikely to drive our results. Further, we show that our results are robust to single outliers by excluding polities one-by-one (to be reported in Section 5).

Regarding omitted variables, unobservable features may have impacted both war participation and parliamentary births. On the one hand, polities with warlike rulers may have been more willing to grant representation in parliament to urban elites to shift the balance of political power in their favour and away from noble councils. This would raise the coefficient on conflict participation in Equation 1, thereby biasing our empirical prediction that war motivated the formation of parliaments. On the other hand, polities with warlike rulers may have been more reluctant to share political power via parliaments – a medieval version of the modern finding that autocrats tend to fight more than democratic rulers (Lake 1992). Again, this would depress the coefficient on conflict participation in Equation 1, thereby biasing against our empirical prediction.

Our analysis uses unit fixed effects to help account for each polity's baseline propensity to fight and establish a parliament. However, for robustness, we also include polity-specific trends to account for time-varying local unobservable features (for example, economic trends, population dynamics, and technological innovations) specific to each polity. Further, we control for several time-varying observable features, including each polity's trade potential (Abramson and Boix 2019; Acemoglu, Johnson, and Robinson 2005), years of ruler successions (Kokkonen and Møller 2020), lack of male heirs at the time of a ruler's death (Acharya and Lee 2019; Kokkonen and Sundell 2020), and hereditary succession arrangements (Kokkonen and Sundell 2014, Kokkonen and Sundell 2020) (to be reported in Section 5).⁷

With respect to reverse causation, our analysis employs prior (that is, lagged) conflict participation in predicting parliamentary births. Thus, reverse causation can only occur if a ruler's anticipation of their future decision to form a parliament influences the current decision to enter a war. In our view, there is not much difference between a ruler entering into a war, incurring costs, and then establishing a parliament (without having planned it previously) and a ruler entering into a war and planning to form a parliament in the future to help finance it. In both scenarios, the ruler's belief that a parliament can help improve wartime funding drives any observed correlation between war and parliamentary births.

While we employ a wide range of techniques and robustness checks to address various threats to inference, we do not claim that our within-polity estimates identify the causal impacts of war on parliamentary births. Instead, political innovations were the result of complex interactions between a variety of factors. Although we do not treat our estimates as causal in nature, our analysis nevertheless highlights new and robust data patterns about the relationship between war and the creation of parliaments.

Data

Parliamentary birth years

We have described the data and construction method for our dependent variable, Parliament Birth_it, in Subsection 3.2.

War

To Systematically investigate the role of war, we code each polity's annual conflict participation according to Jaques (2007). This compilation covers all historical conflicts worldwide, with written documentation from at least two independent sources and a consensus among sources on the main conflict details. Jaques provides a one-paragraph description for each conflict, including the date, major participants, and the conflict type (for example, land battle). In addition, we include all conflicts in which at least one sample polity actively participated, regardless of type and whether they took place in Europe or elsewhere (given that all such conflicts should impact the ruler's demand for ready money). The resulting database contains 848 unique conflicts over the 1000–1599 period. We employ this database to compute our variable of interest, Conflict Participation_{i(t−p,t−1)}, as described in Subsection 4.1.

In our view, Jaques is the best source to construct the conflict participation data for this analysis. Jaques spans our entire sample period from 1000CE to 1599CE. This improves on two other well-known works on global historical conflicts: Brecke (1999) and Clodfelter (2002). Brecke does not include conflicts prior to 1400, while Clodfelter does not include them prior to 1500. Nevertheless, we use the available conflict data from Brecke and Clodfelter as part of our robustness checks (to be reported in Section 5).

To our knowledge, there are three recent Europe-specific databases, in addition to ours, that document conflicts in medieval and early modern Europe: Acharya and Lee (2019), Kokkonen and Møller (2020), and Kokkonen and Sundell (2020). Kokkonen and Sundell provide the most expansive coverage of the three. Their data spans twenty-eight polities from 1000CE to 1799CE in an unbalanced panel. By contrast, Kokkonen and Møller only include sixteen polities, while Acharya and Lee stop in 1500. Our data still improves on those of Kokkonen and Sundell since we include external and internal conflicts fought by sample polities worldwide. By contrast, Kokkonen and Sundell focus on civil wars fought within Europe.

Nevertheless, we employ the Kokkonen and Sundell (2020) conflict data for robustness. Kokkonen and Sundell construct their data from five different conflict compilations: Dupuy and Dupuy (1970), Luard (1986), Clodfelter (2002), Phillips and Axelrod (2005), and Kohn (2013). Additionally, they add data from Brecke (1999). Thus, employing the Kokkonen-Sundell data for robustness enables us to account for several alternative sources of historical conflict data (including Clodfelter and Brecke, as described above).

A major drawback of this approach, however, is that we lose a great deal of data after merging ours with those of Kokkonen and Sundell (2020). Namely, more than 50 per cent of our sample polities and nearly 50 per cent of our total observations are lost. There are two reasons for this. First, many polities that are part of our sample are not part of the Kokkonen and Sundell data. Further, their data constitutes an unbalanced panel. In four cases, their data for a given polity does not begin until after the year in which a parliament was first established while, in another case, their data ends before the parliamentary birth year.

Main Results

Table 2 presents our main results. In column 1, we show that a polity's participation in the war over the preceding twenty-five years significantly increases the likelihood of the establishment of a parliament. The coefficient on Conflict Participation_{i(t−25,t−1)} is 0.055.

Table 2. War and parliamentary births

Notes: The estimation method is OLS. The unit of analysis is polity-year. The sample period is 1000–1599. Robust standard errors clustered at polity level in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.

In our regression framework, unit fixed effects control for polity-specific features that do not vary over time, while time-fixed effects control for widespread shocks in a specific year. However, there may still have been unobserved time-varying confounders in different polities that influenced both war and parliamentary births. To help control for potential confounders of this sort, we include polity-specific linear trends in column 2. The result for Conflict Participation_{i(t−25,t−1)} remains highly significant, with a magnitude similar in size to the previous column.

In column 3, we add a control for a key time-varying observable feature: the trade potential of each polity. We proxy for this variable by computing the natural logarithm of the average Christian urban potential across all major towns in each polity per century, in line with Bosker, Buringh, and van Zanden (2013). This variable measures the natural logarithm of the distance-weighted sum of the population of all other Christian towns in Europe, the Middle East, and North Africa per century for each major town within a polity, which we then average across.⁸ Again, the main results are robust.

In our most stringent specification – which includes fixed effects, polity-specific trends, and the control for trade potential (see column 3) – we estimate that a polity that averaged one conflict per year over a quarter-century had a per-year probability of a parliamentary birth that was 5.60 percentage points higher than a polity that never fought. To provide a sense of what this magnitude means, say that the peaceful polity had no chance of establishing a parliament; in this case, the belligerent polity's probability of convening a parliament at some time in its quarter-century of conflict participation would have been 76 per cent.⁹

Robustness

We have already described the robustness checks involving fixed effects at a higher level of aggregation (that is, by region; see Appendix Table A.1), including polity-year observations prior to the medieval communal revolution (Appendix Table A.2). The main results continue to hold across both of these checks. Therefore, we now perform several additional robustness tests.

Exclude Polities One-by-One

Appendix Figure A.1 reruns the main specification (column 1 of Table 2) after excluding each polity one by one. Again, no single outlier polity drives our results.

Alternative Parliamentary Birth Years

There are eight cases in which historians disagree on the exact year in which urban elites first gained representation in the ruler's noble councils. Therefore, in Appendix Table A.4, we rerun the main specification after replacing the benchmark birth years with the alternatives one at a time (as listed in Appendix Table A.3). The main results remain robust.¹⁰

Alternative Conflict Participation Years

Appendix Table A.5 reruns the main specification after we vary the share of years in which a polity participated in a conflict over the previous 5, 10, 15, 20, 30, or 35 years, respectively (that is, versus p = 25 in the benchmark). Again, the main results are robust. The coefficient on Conflict Participation_{i(t−p,t−1)} increases in value as we extend p from 5 to 35. This is consistent with the notion that persistent conflict created larger fiscal hangovers for rulers (Stasavage 2011), thus making parliamentary establishment more likely.

Alternative Conflict Participation Variable

Appendix Table A.6 reruns the main specification when we take the share of conflict years over the previous five to thirty-five years as the alternative variable of interest. Again, the main results continue to hold overall.

Data Averages

Appendix Table A.7 reruns the main specification using 10-year and 20-year data averages. The main results remain positive and significant for the 20-year averages. The point estimate for Conflict Participation_{i(t−25,t−1)} in the specification with the 10-year averages remains positive in sign but misses statistical significance (the p-value is 0.151).

Include City-States

Since our argument concerns the addition of urban representatives to noble councils, the main regression analysis centres on territorial monarchies. In Appendix Table A.9, we repeat the main analysis for an extended polity sample that includes the years of parliamentary establishment in six city-states in Italy and one in Germany (as listed in Appendix Table A.8). The main results continue to hold.

Alternative Conflict Data

In Appendix Table A.10, we rerun the main specification after employing the Kokkonen and Sundell (2020) conflict data to recompute Conflict Participation_{i(t−25,t−1)}. Recall from Subsection 4.3.2; a major drawback of this approach is that we lose more than 50 per cent of the polity sample and nearly 50 per cent of the total observations after merging our database with theirs.¹¹ Nevertheless, the main results, as shown in column 1 (which employs Kokkonen and Sundell's benchmark conflict data constructed from five different compilations) and column 2 (which adds the conflict data from Brecke 1999), remain robust.

Controls for Successions, Heirs, and Primogeniture

Kokkonen and Sundell (2020) also provide data on three time-varying observable variables which may have impacted both war participation and parliamentary births: the years of ruler successions, the lack of male heirs at the time of a ruler's death, and hereditary succession arrangements (that is primogeniture). To our knowledge, the Kokkonen and Sundell data provide the most expansive coverage available for these three variables (even if we continue to lose more than 50 per cent of the polity sample and nearly 50 per cent of the total observations after merging databases).¹² In column 1 of Appendix Table A.11, we rerun the main specification after controlling for the years of ruler successions. Following Kokkonen and Sundell (2020), we proxy for this variable with rulers' years of natural deaths (that is, due to accidents or disease). In column 2, we control for the years of ruler deaths in which there was a lack of male children. In column 3, we include both variables plus the interaction term Natural Deaths × Lack of Male Heirs. Finally, in column 4, we add the control for the presence of primogeniture (that is, the custom of having the eldest son inherit rule) plus the interaction term Natural Deaths × Primogeniture. The main results continue to hold.

Note that the positive relationship between succession crises and parliamentary births that Kokkonen and Møller (2020) uncover for medieval Europe should be subject to the same sorts of political constraints that we have highlighted here for war. Elsewhere in Eurasia, there was no such relationship between succession crises and parliamentary births. Only when the second and final political condition was met (that is, the existence of communes) would one expect towns to become involved in royal successions, with parliaments called as convenient fora in which the new actors could have their say.

Conclusion

In this paper, we have advanced a new argument about the historical emergence of parliaments in Europe that integrates the literature highlighting rulers' desire to finance interstate wars with that emphasizing the medieval communal revolution. The key political constraints that determined whether rulers seeking to finance war had incentives to expand their pre-existing noble councils to form parliaments were a lack of tax-collecting bureaucracies and the prior delegation of governance and tax-collecting rights to at least some major town corporations. We have argued that Europe, alone among Eurasian regions, had a window of opportunity – opening with the medieval communal revolution and shutting with the creation of tax-collecting bureaucracies in the early modern period – during which these two political constraints were met. Within this window, we show that wars significantly predict parliamentary births. Moreover, in response to war, rulers were more likely to summon non-noble urban elites to join previously exclusively noble councils.

In our view, war promoted the establishment of parliaments in two steps. Given the first political constraint (that is, the lack of tax-collecting capacity), the war drove rulers to sell governance and tax-collecting rights to towns. When combined with the merchant elites' increasing incentive and ability to outbid landed nobles for such rights, the result was the communal revolution – the process by which hundreds of towns across Europe acquired rights of self-government, thereby meeting our second and final political constraint.

The next step in institutional development was to bring the new class of urban tax farmers and justice administrators into the ruler's pre-existing noble councils. However, separate negotiations with each town were expensive and invited foot-dragging and free riding. Thus, many rulers preferred to summon urban elites to great councils that grew out of the noble councils they had already been convening. Consistent with this argument, we have shown that parliaments were only established in polities where communes had previously emerged.

Elsewhere in pre-modern Eurasia, we do not expect war to have had similar impacts to those that we have documented here. Although other parts of Eurasia experienced ‘warring states’ eras (for example, Hui 2005), and while tributary relationships and alliances might be viewed as akin to a market in governance rights, the delegation of administrative and judicial rights to incorporated towns was uniquely European (Weber [1921] (1958)). Moreover, because the European corporation differed significantly from corporate organizations in the Islamic World (Kuran 2010) as well as from kin-based collective action in Asia (Greif and Tabellini 2017), the possibility of forming political institutions that resembled communes was not typically on the menu of possibilities anywhere outside of Europe. Thus, warring-states periods exclusively involved competition between rulers and nobles and ended with the emergence or re-establishment of imperial hegemony.

By our account, Europe's pre-modern window of opportunity for parliamentary births should have closed when rulers became powerful enough to build their own tax-collecting bureaucracies. Consistent with this expectation, the establishment of central tax bureaucracies during the early modern period strongly correlates with the suppression of parliaments (Cox, Dincecco, and Onorato, 2022).