Data

To examine trends in ruler duration in both the Christian West and the Muslim world, we have compiled two primary data sets. Both data sets focus on the duration of rule for the highest ranking ruler available in an independent governing unit. In many cases, this individual carries the title of king, but might also be called a prince, sultan, tsar, amir, khan, or lord. The governing units include kingdoms, duchies, principalities, tribal confederations, and city states.

The first data set draws on existing compendia of rulers across medieval dynasties. Islamic Dynasties (Bosworth Reference Bosworth1996) and Dynasties of the World (Morby Reference Morby1989) represent the best collections of information on the duration of rulers in the Islamic world and Christian Europe, respectively. Bosworth (Reference Bosworth1996) focuses exclusively on the Islamic world, providing the most comprehensive data on dynasties in this region. Morby (Reference Morby1989) provides ruler duration for dynasties across the world with an “emphasis on Europe and on its roots in the ancient world” (Morby Reference Morby1989, vii). Our baseline data set (which we henceforth refer to as the “Bosworth/Morby data set”) includes all the rulers in Bosworth (Reference Bosworth1996) who assumed power before the year 1500 CE and those in Morby (Reference Morby1989) who assumed power before this date in the subsections 1–7 and 10 of the section entitled “Europe” and the section entitled “The Barbarian West.”Footnote ⁶

The second data set collects information on all sovereigns for political units depicted in a set of historical maps developed as part of the Euratlas project (Nüssli Reference Nüssli2011).Footnote ⁷ The Euratlas project provides the geographic boundaries of all political entities covering the earth's surface in the box between the longitudes 15 west and 50 east and the latitudes 20 north to 60 north. This encompasses all of Europe, North Africa, and parts of northwestern Asia including Anatolia and the Levant. We use the GIS shape file “sovereign states” in Nüssli (Reference Nüssli2011) at the start of t ∈ {700, 800, 900, 1000, 1100, 1200, 1300, 1400} CE to define the population of political entities. We have attempted to determine the length of rule for every monarch that assumed power in each of these political entities on the interval [t, t + 100). Next, we assigned rulers who assumed power on the interval [t, t + 100) to the map of year t. In other words, our sample is updated every 100 years to include the entities Nüssli (Reference Nüssli2011) denotes as sovereign states.

We have constructed this second data set for two reasons. First, Nüssli's maps were, to the best of our knowledge, compiled independently of the Bosworth/Morby data set and thus help alleviate concerns that our results are sensitive to a given author's definition of ruler. Second, Nüssli's maps allow us to introduce a set of geographic control variables that permit testing of alternative explanations.

Trends in Ruler Duration

This section presents the trends in ruler duration for the Christian West and the Muslim world. These results are summarized by Figure 1 which plots the moving average of ruler duration in Western Europe and the Islamic world starting in the year 700 CE using the Bosworth/Morby data set. Ruler duration in Europe overtakes leader tenure in the Islamic world sometime around 1000 CE.

FIGURE 1. Divergence Between Ruler Durations in the Christian West and Islamic Worlds

Note: The darker line denotes the 100-year moving average for the Islamic world whereas the lighter line denotes the moving average for the Christian West.

Table 1 reports the regression output associated with this relationship. We restrict the sample to rulers assuming power on or after 700 CE and before 1500 CE and run regressions of the form

where duration _itc gives the duration in power of ruler i that assumed power in year t in century c. The d_c are century dummy variables and WE _i is an indicator variable equal to 1 if ruler i assumed power in Western Europe. Throughout the empirical section, we cluster standard errors by dynasty.Footnote ⁸

The coefficients θ _c and α _c estimated using the Bosworth/Morby data set are presented in columns (1) and (2) of Table 1. Column (1) reports _c , or the mean duration of rulers in the Islamic world, by century. Column (2) displays _c , or the difference in mean duration between Western Europe and the Islamic world, by century. These estimates show that although before the year 1000 CE ruler duration in Western Europe and the Islamic world were not jointly statistically different at the 10% level, after this date one can reject the null hypothesis that leadership tenures were the same.Footnote ⁹

Columns (3)–(5) present the results using the data we have collected using the Nüssli data set. Columns (3) and (4) of Table 1 provide estimates of θ _c and α _c . Although the results obtained using the Bosworth/Morby and Nüssli data sets are similar, in the latter data set the differences in ruler duration between Western Europe and the Islamic world are not statistically significant until after 1100 CE.

In Column (5) we present our estimates of _c when geographic controls are included. These controls include the area of the political entity at the start of the century, the proportion of the entity that was part of the Roman Empire in the year 100 CE, the latitude of the centroid of the political entity, and the average agricultural suitability of the entity. The introduction of these control variables does not change the qualitative implications of the results.

Feudalism, Parliaments, and the Rise of Europe

By the 13th century parliamentary institutions had spread across Western Europe (Stasavage Reference Stasavage2010). The usual narrative describing the birth of representative, sovereign-constraining political institutions begins with the collapse of the western Roman Empire. The fiscal position of the Germanic successor states to the Roman Empire tended to be weak. Unable to fund military expenditure through tax receipts, European rulers sought other avenues for raising armies. The innovations introduced by Charlemagne marked a pivotal change. Lacking the capacity to introduce a system of tax collection, Charlemagne required landholders to contribute troops instead of funds.

This change increased the power of large landlords in two ways. First, small, independent landowners pooled their lands with those of larger landholders to avoid having to offer themselves up for military service. As individual landholders began to “aggregate up,” large landowners emerged who could ensure the cultivation of land while distributing the burden of military service across the larger body of peasants. Second and contemporaneously, European kings—like Charlemagne—required mounted troops, not just infantrymen, as a result of the introduction of the stirrup. The technological innovation of the stirrup meant that “mounted shock combat” became the norm in warfare and the large investment required to purchase a horse and armor for battle meant that monarchs needed to recruit individuals with wealth to serve as the mounted military elite (White Reference White1962).

Mounted warriors, or knights, were often compensated for their service to the king through land grants (North, Wallis and Weingast Reference North, Wallis and Weingast2009, 79).Footnote ¹² According to Mann (Reference Mann1986, 393), the primitive state of European economies left “land grants, which gave the vassal soldier a potentially autonomous power base” as the only option for cash-strapped monarchs. European barons operating in the feudal system entered battle with their own, privately financed equipment, archers, and associated infantry. Such individuals often enjoyed opportunities to increase their land holdings or other forms of advancement as a result of their fighting. Together, the methods of military recruitment that emerged in medieval Europe came to be known as the feudal system.Footnote ¹³ The net result of these innovations was the creation of a landed aristocracy in Western Europe.

Strayer (Reference Strayer1970) provides a particularly compelling discussion of the feudal system and its consequences for state development. He argues that in the weakened and cash-strapped environment of early medieval Europe, “standing armies or permanent officer corps were unthinkable” (Strayer Reference Strayer1970, 27). Feudalism, characterized by its “fragmentation of political power” (1970, 14), emerged whereby rulers would raise armies on an as-needed basis by offering inducements of land or other privileges in exchange for support. While the system might appear to work against the creation of an effective state in the short run, Strayer argues that ultimately such a system “. . .can become a basis for state-building” (1970, 15). Although he does not bring any systematic empirical data to bear on this question, he argues that there was a notable increase in Western European political stability following 1000 CE and it was this stabilization of the political scene that allowed for an economic revival that included higher levels of agricultural production, population growth, and a revitalization of long-distance commerce (Strayer Reference Strayer1970, 19).Footnote ¹⁴ He finds that during this period of deepening political stability, the basic components of the modern state began to appear in Europe (Strayer Reference Strayer1970, 34).

The stability of European monarchs evolved hand in hand with both increased economic opportunities and growing constraints on the executive. Europe's more stable political environment contributed to the rise of towns and a nascent commercial revolution that became apparent beginning in the 12th century. Perhaps more importantly, the nature of elite military recruitment under feudalism led monarchical abuses to be self-limiting. Barons—who served as vassals to the king—had the military means by which to rebel and demand satisfaction of their grievances (Breay Reference Breay2002). The independent military power of the barons allowed for a degree of bargaining strength vis-à-vis the monarchy as barons could either rebel against the king or support an opposition figure who might meet their demands in exchange for support.

While increasing monarchical constraint did not translate immediately into the emergence of mature parliamentary institutions, in the years following the introduction of feudal reforms, sovereigns convened assemblies of their vassals to discuss, negotiate, and approve military activities (van Zanden et al. Reference van Zanden, Buringh and Bosker2012). Over time, these assemblies became institutionalized and an expected element of the political scene. By the 13th century, more complex legal systems began to emerge in Europe (Stacey Reference Stacey and Abulafia1999). The long-term impact of this progression is described by Downing (Reference Downing1992, 3):

The promulgation of the Magna Carta in 1215 and eventual establishment of an English parliament populated by knights and barons in 1265 serve as a useful example for how this process played out in a particular country case. Under feudal institutions in England, the king had the right to demand “military service. . .whereby kings expected their vassals to contribute either in men or in money to armies” (Holt Reference Holt1992, 30). Military service was a source of “widespread and perennial acrimony” between the king and his vassals (Holt Reference Holt1992, 78). The Magna Carta reflected a greater acceptance of baronial demands on the part of the monarchy (Holt Reference Holt1992, 189).Footnote ¹⁵ The Magna Carta laid the groundwork for future demands to limit the power of the monarch in England. Over time, a coalition of English elites established credible constraints on the executive with the Glorious Revolution of 1688, allowing for property rights and security from arbitrary taxation that ultimately encouraged economic growth (North and Weingast Reference North and Weingast1989).

In England and beyond, feudalism represented a meaningful fragmentation of political authority.Footnote ¹⁶ Kings—while the technical heads of government in their respective territories—had ceded considerable strength to local strongmen who enjoyed both public and private power, including control over public goods provision and land and rental income (Bisson Reference Bisson1994). In the face of divided and decentralized political power, how were European sovereigns able to increase their length of rule? Our argument is that more consensual government—as it emerged in Europe, with roots in medieval feudalism—enjoyed an advantage in terms of political stability. Forced by economic weakness to bargain and negotiate with local elites, European monarchs developed forms of political organization that exhibited a flexibility which trumped forms of more absolutist rule. These governmental forms contrast sharply with political organization and military recruitment in the Islamic world during the same time period.

The Islamic Equilibrium

Political development in the Islamic world provides an important comparison to institutional evolution in Western Europe.Footnote ¹⁷ Like the Latin West, Muslim states ruled over some of the wealthiest Roman provinces and had access to the institutional heritage of ancient Greece, Rome and, in some cases, the Germanic states. Muslim states also controlled some areas that eventually reverted to Latin control.

Yet, feudalism—with its complex system of interlocking economic and military rights and obligations—did not emerge in the Islamic world. Despite being largely agrarian, no “landed aristocracy or gentry” materialized (Crone Reference Crone, Raaflaub and Rosenstein1999, 322) nor did nascent parliamentary institutions develop. How were Muslim rulers able to circumvent the emergence of the type of landed aristocracy that proved so critical to constraining monarchs in Europe? This section argues that Muslim reliance on mamluks—or military slaves imported from non-Muslim lands—weakened state-society relations and hindered the development of impersonal political institutions. Thus, while European rulers were negotiating with local gentry to raise armies for matters of defense, Islamic rulers bypassed local elites by creating highly skilled armies of foreigners who had no ties to the existing gentry and swore allegiance directly to the sultan.

Historians of the medieval Islamic world have come to describe the introduction and eventual widespread adoption of mamluk institutions as a uniquely Islamic phenomenon. A mamluk is generally described as a military slave, though the term also refers to such individuals after their emancipation (Irwin Reference Irwin1986, 3).Footnote ¹⁸ Mamluks might better be defined as elite military slaves given the fact that they were typically well trained and generously paid.Footnote ¹⁹ As such, mamluks were not prototypical slaves, but rather military elite who might serve in positions like provincial governor or treasurer (Irwin Reference Irwin1986, 4). After introduction by the Abbasids in the 9th century, mamluk armies were quickly adopted by numerous Muslim polities (Ayalon Reference Ayalon1994a, 25) and spread across the settled areas of the Islamic world (Crone Reference Crone2003, 79). Indeed, mamlukism became a defining feature of Muslim polities for a period of more than 800 years (Crone Reference Crone2003).

Who became mamluks? The most sought after mamluks were of tribal origin imported from areas “marginal to the settled Islamic world” (Crone Reference Crone2003, 78) like the Caucasus and Transoxania (present-day Uzbekistan, Tajikistan, and Kazakhstan).Footnote ²⁰ Imported as children, mamluks often underwent years of training which sought to both imbue them with military skills but also to encourage their loyalty to the sultan (Pipes Reference Pipes1981, 9).Footnote ²¹ In order to stem the “corruption” of military slaves from the luxuries of settled life, mamluk institutions created a disconnect between the soldier slaves and local society. To deter the mamluks from being able to develop ties to either the local population or home-born troops, the mamluks were kept in “strict isolation” (Crone Reference Crone, Raaflaub and Rosenstein1999, 319). Mamluks typically bore their Turkish names, even after their conversion to Islam, and predominantly used Turkish when speaking to one other with often superficial knowledge of the local language (Ayalon Reference Ayalon1994b, 16–17). Mamluks were mainly married to female slaves from their countries of origin rather than local women (Ayalon Reference Ayalon1994b, 16). The sons of mamluks (who did not enjoy mamluk status themselves) more frequently married women from the local population thus offering one opportunity for assimilation into nonmamluk society (Ayalon Reference Ayalon1994b, 16–7).Footnote ²² A mamluk, then, was characterized by both his “personal dependence” on his master, the sultan, as well as his “cultural dissociation” (Crone Reference Crone2003, 74) given both his foreign origin and the development of practices that kept him highly removed from the local populace.

There is no consensus in the existing literature as to why mamlukism emerged and spread throughout the Islamic world. In particular, why didn't medieval Muslim sultans use indirect rule as we observe emerged in Europe? Patricia Crone offers perhaps the most compelling explanation for why mamlukism arose in the Islamic world. Crone compares the Abbasids explicitly to the Carolingians—their contemporaries—who also faced the challenge of creating a polity for which their past experience offered no model. She writes, “both fell back on private ties, and in both cases, the outcome was political fragmentation. But because the fiscal and administrative machinery survived in the east, the Abbasids could simply buy the retainers they needed, and so they lost their power not to lords and vassals but to freedmen [i.e., manumitted mamluks]” (Crone Reference Crone, Raaflaub and Rosenstein1999, 326, emphasis added).Footnote ²³ This suggests that the superior economic position of the Muslim rulers allowed them to import the military support that they needed rather than to develop a system of feudalism where a king delegated land—and political power—to local lords.Footnote ²⁴

Imported military slaves were thought to be “safest to rely on” by a sultan (Marshall 1977, 399), offering the best defense of the ruler's interests (Lapidus Reference Lapidus1973). Indeed, according to one observer, the “principal deterrent to the sultan's overthrow was the strength and loyalty of the royal mamluks” (Dols Reference Dols1977, 148). If a “well-controlled” mamluk army could bring political stability to a polity, an uncontrolled one was a potential source of regime breakdown and disintegration (Crone Reference Crone2003, 84). In some cases, sultans found themselves “imprisoned” by their own “praetorian guard” (Lapidus Reference Lapidus1973, 37–38) who were successful at usurping power from within (Pipes Reference Pipes1981, 91). Military slaves came to threaten the very dynasties that had trained them, eventually establishing their own slave sultanates (Dale 2010, 16; Pipes Reference Pipes1981, 23).

Perhaps more pernicious than the direct challenge mamlukism posed on ruler stability was the indirect impact of mamlukism on state-society relations. Military slaves who “had no roots in or commitments to local communities” were responsible for collecting taxes, maintaining order, and controlling important resources (Lapidus Reference Lapidus1973, 39). Mamluks would typically hold a temporary, nonhereditary deed to land, called an iqta’, while living in urban areas far from their agricultural holdings (Borsch Reference Borsch2005, 26–32). Islam became unique among civilizations for the extent to which government service ceased to be associated with land ownership (Crone Reference Crone2003, 87). While military slaves enjoyed the ability to serve as tax collectors as part of the iqta’ system, “slave soldiers were no barons” as the iqta’ did not invest the soldiery with land in a way comparable to the European fief (Crone Reference Crone2003, 87; Finer Reference Finer1997, 674). Sultans—reliant on their mamluk coterie for enforcing economic and political control—found themselves “alienated from the mass of their subjects” (Lapidus Reference Lapidus1973, 37–38).

The provision of military service in medieval Europe, then, was highly decentralized in contrast to the mamluk system where military slaves constituted a centrally located and ethnically distinct caste. Mamluks were unable to transform themselves into a “hereditary landed baronage,” in part because of the “impossibility of transmitting mamluk status to one's children” (Finer Reference Finer1997, 676; Fukuyama Reference Fukuyama2011, 218; Irwin Reference Irwin1986, 8). Historians have argued that attempts to bequeath land holdings or title to offspring met with opposition from other mamluks who, as a collective, would have been hurt by such a move (Ayalon Reference Ayalon1987, 208). Thus, while Western Europe saw a strengthening of lords who were responsible for defense of the land (Duby 1974), the Muslim world saw a deterioration in the bargaining strength of the aristocracy as control of the means of violence became dominated by a caste of military slaves. The relative bargaining strength of the gentry vis-à-vis the ruler has proven to have profound implications for the development of executive constraint and the creation of impersonal economic institutions.

Theoretical Implications of the Historical Narrative

The historical narrative above describes a pattern of political divergence in the Christian and Muslim worlds beginning in the 8th century following the end of Roman hegemony in the Mediterranean basin. Rulers in Christian Europe were forced to enter into forms of consensual rule with their local elite. The “feudal complex”—as this system came to be known—rolled out across continental Europe to places like England, Spain, and Scandinavia along Carolingian lines; feudal institutions expanded less evenly to Eastern Europe. In the Islamic world, on the other hand, mamluks were “not readily convertible into a rural nobility” (Anderson Reference Anderson1979, 506). While the political power of the European landed aristocracy increased over time, leading to a gradual transfer of power (e.g., control over monetary rents, public goods, and the legal system) from the sovereign to his “vassals” (e.g., the landed aristocracy), a similar process was not observed in the Islamic world.Footnote ²⁵

The result was the emergence of a set of political institutions and norms in Christian Europe which have been associated in the contemporary literature with forms of executive constraint. Constraint on the sovereign did not emerge without considerable local pressure and contestation. Nor did fully developed parliamentary institutions materialize immediately following the introduction of feudal reforms. Rather, medieval assemblies increased in importance over time and came to serve as a “logical extension of the traditional presentation of auxilium et consilium—aid and advice—by the vassal to his overlord” (Anderson Reference Anderson1979, 411). A sovereign's ability to tax without consent diminished under the relative strength and influence of local notables. The ability of European elites to guard against abuses of the executive increased during the medieval period while comparable developments were absent in Muslim polities. Figure 3 is a schematic of our argument.

FIGURE 3. Comparison of Institutional Development in Christian West (left) versus the Islamic World (right) Following the Decline of the Roman Empire

More abstractly, it can be argued that feudalism had three main effects. First, it decreased the rents, broadly construed, flowing to the sovereign. Second, it increased the rents flowing to the sovereign's potential rivals, the nonruling elite. Third, it increased the geographic decentralization of political power. Holding the probability of successfully overthrowing the sovereign constant, the emergence of feudalism should lead to a decrease in the probability of successful revolt. In other words, feudalism reduces the payoff to a successful revolt by decreasing the wedge between what the sovereign and his rivals receive with the status quo. Since feudalism makes revolt less enticing for the sovereign's rivals because they have both more to lose from an unsuccessful rebelion (and less to gain from a successful one), fewer revolts will be successfully staged “in equilibrium.”Footnote ²⁶

Is it possible that the probability of successfully overthrowing the sovereign significantly declined as political power flows to his vassals, overwhelming the effects of the decrease in the wedge described above? Here the third premise (geographic decentralization of political power) is important. It seems unlikely that collective action among the landed nobility living on decentralized fiefs would have been easier than for mamluks, or other elites, concentrated in dynasty capitals. This view is supported by Stasavage (Reference Stasavage2010) who has argued that the costs of travel and communication before the 19th century were significant and served as an important barrier to elite coordination. This perspective is also supported by historians who have argued that coups are more common in polities with centralized, salaried armies than for “land-based” military officers (Wickham Reference Wickham2005, 120–121). For Muslim sultans, mamlukism left rulers vulnerable to “discontents, mutinies, and ultimately, usurpations” (Finer Reference Finer1997, 710).

Carolingian Origins of Europe's Political Transformation

Our theoretical narrative draws on historical evidence suggesting increased constraint on the sovereign originates with the Carolingian dynasty. Carolingian governance relied—to a greater degree than its predecessors and peers—on forms of aristocratic collaboration that encouraged discussion, negotiation, and monarchical dependence on elite consent rather than coercion (Nelson Reference Nelson and Reuter1999).

One empirical implication of our argument is that if the introduction of feudal modes of social organization was associated with decreased likelihood of monarchical overthrow, we should see a break in the trend line in ruler duration beginning during the Carolingian period. To empirically test this claim, we limit our sample to non-Muslim Western Europe and investigate the origins of the increase in European leadership tenure. In order to further extend our sample in Western Europe, we supplement the Morby data set used in the second section with data—also drawn from Morby (Reference Morby1989)—from the Imperial Roman period (which begins in 27 BCE).

We calculate the Quandt likelihood ratio (QLR) statistic for the following specification:

where duration _t is the average tenure of rulers who assumed power in the year t and D_t (τ) is a dummy variable equal to one if t > τ. Equation (2) allows for one break in which there is both a mean and a slope change. We use 15% trimming and find that the F statistic testing the null hypothesis that both β₁ and β₃ are zero is maximized in the year 790 CE. In this year, the F statistic is well above the 1% cutoff of 7.78. In other words, the data identify a break in ruler duration in Western Europe in the year 790 CE, which approximately corresponds to the midpoint of Charlemagne's reign.Footnote ²⁷ Figure 4 presents the results in graphical form. The upper graph shows mean ruler duration in each year (denoted with grey dots) and the fitted values of Equation (2) with the break date in the year 790 CE. The lower graph plots the F statistic within the central 70% of the sample. The vertical line denotes the estimated break date in 790 CE.Footnote ²⁸

FIGURE 4. Political Stability and the Trend Break

Note: Vertical line marks the estimated break date in 790 CE. The upper graph provides data on yearly duration and the fitted values. The lower graph details the F-statistic testing for a break in both the level and trend of ruler duration in Western Europe.

Geography and the Spread of Carolingian Institutions

If increases in constraints on the sovereign originated in the Carolingian Empire, increases in ruler durability should have initially been most notable in these areas. To test this prediction, we limit the sample to Western Europe and estimate a specification of the form

where the variables are as defined in Equation (1) and Carol _i is the percentage of a polity's landmass that belonged to the Carolingian Empire in the year 800 CE. We provide our estimates of _c in column (1) of Table 2. These estimates show that during the first two complete centuries following the reign of Charlemagne, rulers in the successor polities of the Carolingian Empire remained in power for longer than other rulers in Western Europe. Around the year 1100 CE, however, durations converge across Western Europe. These results are consistent with the hypothesis that the increase in ruler stability originated within the boundaries of the Carolingian Empire spreading to the rest of Western Europe by approximately 1100 CE.

Political Divergence and Constraints on the Sovereign

We have hypothesized that the increase in ruler duration observed in Western Europe is a reflection of increasing constraints on European sovereigns. In this section, we investigate the empirical support for this hypothesis. To do this, we relate ruler duration to the existence of a representative assembly where such assemblies are believed to serve as a potential constraint on the sovereign.Footnote ²⁹ We create an indicator variable Parliament _i equal to 1 if either van Zanden et al. (Reference van Zanden, Buringh and Bosker2012) or Stasavage (Reference Stasavage2010) record the polity as having at least one parliament meeting on the relevant interval.Footnote ³⁰ Equipped with this metric, we limit the sample to Western Europe, drop all rulers who assumed power prior to the year 1100 CE and run the following regression:

Estimates associated with Equation (4) are presented in column (2) of Table 2 using the Bosworth/Morby data set and in column (3) of Table 2 using the Nüssli data set. Both data sets provide evidence for a positive relationship between parliament meetings and ruler stability, particularly for the first two centuries after parliaments emerge.

The evolution of ruler duration within dynasties provides an additional opportunity to test if the Western European increase in political stability is a reflection of increasing constraints on the sovereign. We begin by creating a variable measuring the order in which a ruler falls in the dynastic chain. This metric is equal to 1 for the founder of the dynasty, 2 for the next ruler, etc. If a ruler is unconstrained, we expect his time in power to be a function of his (sovereign-specific) human capital (Jones and Olken Reference Jones and Olken2005). Moreover, we expect the founder of a dynasty to be highly endowed with such capital. Inasmuch as rule remains within a given family we also expect sovereign human capital to mean-revert over time. If correct, this dynamic should produce a downward relationship between the place the ruler occupies in a given dynastic chain and his time in power. On the other hand, if a ruler is constrained as a result of feudal, or other, institutions this should mitigate the expected downward relationship between a sovereign's place in the dynastic order and his duration in power. The intuition is that when sovereigns are constrained as a result of the institutional framework, sovereign-specific human capital should matter less.

These predictions are consistent with the data linking ruler location in the dynastic chain with his duration. We examine location of a ruler in the dynastic chain with his duration by running the following regression:

where d_o is a dummy variable equal to 1 if the ruler occupies position o in the dynastic chain and WE _i is a dummy variable equal to 1 if the ruler assumed power in Western Europe. We present the results in Table 2 for the first eight rulers in the dynastic chain. In column (4) of Table 2 we present our estimates of θ _o or the mean duration of rule in the Islamic world by dynastic order on the interval [1000, 1500). The downward trend in average ruler durations following the founder is striking. In column (5) of Table 2 we present _o , or the difference of mean duration of rulers in position o in the dynastic chain between Western Europe and the Islamic world on the interval [1000, 1500). These results show that ruler durations in Christian Europe after 1000 CE did not exhibit the Islamic world's sharp downward trend within dynasty. This suggests that although dynasty founders everywhere are endowed with the high levels of human capital associated with long duration of rule, Western European polities enjoyed forms of political institutionalization that made the personal attributes of leaders less relevant.

Culture and Geography

Prominent alternative explanations for the rise of Western Europe invoke the roles of culture, particularly religious culture, and geography. Among cultural explanations, many scholars argue that Christianity fostered stable political institutions when compared to Islam. The simplest version of this theory emphasizes the time-invariant effects of Christian religious practice or belief and, therefore, would predict Christian rulers to enjoy longer duration of rule from the start of our sample.Footnote ³¹ More sophisticated versions of this hypothesis focus on the time-varying effects of Christianity, like the emergence of separation between church and state in the West (Black Reference Black2010, Huntington Reference Huntington1996, Stark Reference Stark2005). These scholars stress that whereas church and state were united in the Islamic world they were divided in Christian Europe from a relatively early date (Black Reference Black2010, Lewis Reference Lewis2002).

The origins and spread of Islamic religious practice has long been associated with desert topography.Footnote ³² To what extent can durable rule be attributed to the religion's unique geography? While the analysis presented in Section 2 suggests that the broad trends in ruler duration we report are robust to the inclusion of a variety of geographic control variables—including land suitability—in this section we explore trends in ruler duration within two geographic areas, over time. In particular, we focus on the Iberian and Anatolian peninsulas. Both regions were home to Christian and Muslim dynasties for centuries, and both regions were part of a single cultural region prior to the Muslim conquests (Iberia had a Roman and Visigothic heritage while Anatolia had been part of the Greek, Roman, and Byzantine empires).Footnote ³³

Columns (1) and (2) of Table 3 present results of regression (1) substituting the dummy variable Christian _i —equal to 1 if the dynasty was Christian and zero otherwise—for WE _i . Our estimates of _c —the difference in the mean value of ruler duration between Christian and Islamic dynasties— are presented in column (1) of Table 3 using the Iberian sample and column (2) using the Anatolian sample. For the Anatolian peninsula, we report results only for the centuries in which we have data on ruler tenure for both Christian and Muslim dynasties.

The results for the Iberian Peninsula mimic those found in the broader sample and suggest that, even after holding geography constant, Christian monarchs developed an advantage over their Muslim counterparts over time. The Anatolian case is important because it allows for the disentangling of religious culture from institutions while still controlling for geography and nonreligious heritage. The Byzantine Empire—though part of the Christian world—did not develop Western European-style feudal institutions.Footnote ³⁴ Since our causal story focuses on the development of military-institutional relations in the Islamic world (and not other aspects of Muslim culture), a failure to reject the null hypothesis in the Anatolian subsample works against explanations that argue that other aspects of Islamic culture hurt the durability of rulers.

The empirical patterns documented for the Anatolian sample provide some evidence against the simple version of the “Christianity hypothesis.” Our focus up until now on the Islamic world and Western Europe, however, does not allow us to distinguish the effects of the emergence of feudal institutions from the possible effects of simultaneously changing Christian beliefs and institutions. This is because the development of feudal institutions in Western Europe roughly coincided with the emergence of the beliefs and religious institutions—like a growing separation of church and state—thought to have been important for the political and economic development of the Christian West.

To investigate the extent to which changes in Christian beliefs and institutions can account for the documented empirical trends, in this section we incorporate data on ruler duration data from Eastern Europe.Footnote ³⁵ Eastern Europeans were predominantly Christian by the 11th century but divided along Latin, or Western, and Orthodox, or Eastern, lines (Shephard 2008).Footnote ³⁶ In addition, the extension of feudal institutions to Eastern Europe was uneven.Footnote ³⁷ Latin areas seem to have largely adopted feudal institutions by the start of the 14th century (Rowell Reference Rowell and Abulafia2008) whereas Orthodox areas, like Russia, did so to a lesser degree if at all.Footnote ³⁸ For example, Vernadsky (Reference Vernadsky1939, 322–323) writes that it is only in the Lithuanian areas of far-western Russia—which happened to be Catholic, not Orthodox—that Carolingian-style feudalism emerges between the 14th and 16th centuries. In the Latin areas, the lag between the adoption of Christianity and the development of feudal institutions provides an opportunity to investigate the extent to which the introduction of Latin Christianity (in the absence of feudalism) led to an increase in ruler duration. Since Orthodox regions generally did not develop feudal institutions, these regions allow for an additional test of the impact of Christian religious belief and practice.Footnote ³⁹

We begin by restricting the sample to Muslim and Orthodox Christian political entities to investigate the extent to which time-invariant effects of Christian practices (i.e., Christian “core beliefs”) are driving the differences in ruler duration.Footnote ⁴⁰ To do this, we estimate a regression of the form

in which Christian _i is an indicator equal to 1 if the polity is (Orthodox) Christian. Our estimates of _c are presented in column (3) of Table 3. There is no statistically significant distinction in ruler duration between Orthodox Christian and Muslim dynasties. This provides additional evidence against claims that Christian beliefs drive the divergence in ruler duration between the Christian and Islamic worlds.

Could it be that the divergence in ruler duration between the Christian West and the Islamic world is driven by the time-varying effects of Christianity? Since ruler durations in Western Europe had diverged from those in the Islamic world by 1100 CE we would expect leader tenures in Catholic areas of Eastern Europe to both diverge from Orthodox areas of Eastern Europe and to converge to western levels around this date.

We investigate the empirical relevance of this prediction in two waves. First, we restrict the sample to Eastern Europe and estimate a regression of the form

where Orthodox _i is an indicator variable equal to 1 if more than half of the political entity's landmass was Orthodox. The results of Equation (7) are provided in column (4) of Table 3 and show that prior to 1300 CE, there was not a statistical difference in ruler stability between Orthodox and Latin areas in Eastern Europe. After this date—and roughly coinciding with the introduction of feudal institution in Latin Eastern Europe—there was a sharp divergence in ruler stability between Catholic and Orthodox areas. The fact that this divergence occurred roughly 200 years after the introduction of Latin Christianity into Eastern Europe is not consistent with the “sophisticated” Christianity hypothesis.

Second, we limit the sample to the Latin Christian world and run a regression identical to Equation (1). The estimated _c values provide the mean difference of ruler duration between Western and Eastern Europe within the Catholic world. These estimates are presented in column (5) of Table 3 and show that on the interval [1100, 1200) ruler duration in Western Europe was significantly higher than in Catholic Eastern Europe. Furthermore, on the interval [1200, 1300) mean ruler duration in Western Europe was higher than that in Catholic Eastern Europe although the difference is not quite statistically significant at the 10% level.Footnote ⁴¹ After this date ruler tenure in Eastern Europe converges to levels in Western Europe. This result is consistent with the hypothesis that feudal institutions were the key innovation that led to longer ruler duration in Western Europe since ruler duration in Catholic regions of Eastern Europe do not seem to have converged to those of Western Europe until after the introduction of feudal institutions around 1300 CE.

Finally, an additional alternative hypothesis related to geography involves differential state size in Western Europe and the Islamic world. Stasavage (Reference Stasavage2010, 625) offers one possible explanation for why small European polities “were able to survive despite threats from much larger neighbors.” He finds that geographically compact polities—with lower exogenous monitoring costs—could maintain representative parliamentary institutions to a greater extent; an implication of this finding is that smaller polities, as appear to have been more common in Europe, enjoyed better institutions and were, thus, more likely to enjoy stability. Although state size is likely endogenous, we have included it as a control variable in column (5) of Table 1 to show that differential state sizes do not drive our result. Indeed, the historical evidence presented above suggests that the small size of states may have been driven by the emergence of a landed aristocracy across Western Europe. In addition, while the number of political entities in Western Europe increased following the fall of the Carolingian Empire, the Islamic world saw a comparable increase in political fragmentation with no commensurate increase in ruler durability. Similarly, there is no robust or consistent correlation between the area of a polity and the length of ruler tenure in that polity.Footnote ⁴²

Differential Demography and Family Practice

Could it be that differences in the cultural demographies and family practices of the Islamic world and Western Europe drove patterns of ruler duration? The permissibility of polygyny in Islam creates the possibility for a relatively large number of male heirs who might each enjoy some legitimate claim to succession. If monogamy led to fewer potential rivals, then this implies that the Christian world should have witnessed a constant advantage in political stability over the Islamic world. We have shown that ruler duration in the Christian world did not uniformly diverge from duration in the Islamic world.

In addition, it is not clear if Christian monarchs necessarily had meaningfully fewer potential successors. Ermakoff (Reference Ermakoff1997) argues that until 1100 CE, European kings engaged in family forms that were quite similar to polygyny in the Islamic world. European monarchs would regularly repudiate their wives and marry again while the existing queen was still alive. Such a scenario did, in some cases, create the conditions for political instability and even civil war (Nelson Reference Nelson and McKitterick1995, 401).

Second, the historical record also suggests that monogamy alone was not sufficient for durable rule. In the Byzantine Empire, “absolute monogamy was the norm and raison d'etre of Christian marriage” (Meyendorff Reference Meyendorff1990, 100). Yet, the Empire suffered from significant political instability. One historian, who analyzed the period from 1025 to 1118 CE, suggests that this instability was a function of centralized rule “built into the political structure” (Angold Reference Angold, Luscombe and Riley-Smith2004, 223). The Ottoman conquest of Constantinople provides an additional opportunity to consider whether Christian monogamy might be driving our finding. After the 1453 CE conquest of Constantinople, the Ottoman Empire controlled many formerly Byzantine territories. Ottoman sultans were notorious for fathering children by multiple wives and concubines. After the conquest, however, ruler stability in the Ottoman Empire remains very close to the Byzantium mean of about 12 years.

If European rulers increasingly passed political power to their eldest sons through norms of primogeniture, this might explain why Christian monarchs survived in office longer than their Muslim counterparts. Although primogeniture clearly has its roots in medieval Europe, the practice of passing power and wealth to the first-born son only became prominent in the 13th century, spreading across Europe up through the 17th century (Bertocchi Reference Bertocchi2006). For example, disputes regarding succession in Scottish and Burgundian monarchies in the 13th and 14th centuries imply that primogeniture was still being established as an institutionalized practice. This is well after the break in the political trend identified in the data on ruler duration and suggests that primogeniture may have emerged endogenously from increases in political stability and constraints on the sovereign since potential rivals to the monarchy could directly constrain the sovereign regardless of his identity.

One might worry that increases in European ruler duration may have been driven by increases in perceptions of the legitimacy of passing rule from father to son. To investigate this claim, we construct a dummy variable son _itc using Morby's data set that is equal to 1 if a ruler was the son of the previous ruler. While this does not test the primogeniture conjecture, since we do not have data on the birth order of the successor, it does provide evidence regarding the comparability of the importance of familial ties on succession. Equipped with this metric, we limit the sample to Western Europe and the Islamic world and estimate the regression:

The estimated α _c is presented in column (6) of Table 3 and do not reject the null hypothesis that rulers were as likely to be sons of the previous ruler in the Islamic world and Western Europe. This result casts doubt on the importance of passing rule from father to son in generating the observed results.

Finally, one might believe that the reason rulers survived in power longer in Western Europe compared to the Islamic world was that life expectancy in Europe was increasing relative to trends in Muslim polities. However, empirical evidence presented in the second section suggests that increasing ruler durability in Europe was being driven by a decreased rate of ruler overthrow rather than some other factor. In addition, there exists no historical qualitative or quantitative evidence to suggest life expectancy would have been shorter in Islamic polities. Muslim physicians during the medieval period were highly sophisticated, discrediting theories of humorism, setting up some of the earliest dedicated hospitals, and making tremendous advances in pharmacology. Borsch (Reference Borsch2005) shows that the disease environments of the East and West were quite similar during this period with the bubonic plague, for example, impacting both Muslim and Christian cities. In England, medieval life expectancy for the higher ranks of society (i.e., those who inherited land) was between 46 and 49 years (Jonker Reference Jonker2003). Estimated life expectancy for monks of Christ Church priory in Canterbury during the 15th century was between 46 and 54 years; monks at this time tended to be relatively well fed, with better medical care than the general population (Hatcher Reference Hatcher1986). Studies of life expectancy of influential religious scholars in 11th century Muslim Spain show that the most prominent within this occupational group lived between 69 and 75 years, on average (Shatzmiller Reference Shatzmiller1994, 66).

Military Technology and Vulnerability to Outside Invasion

A common narrative for why Islamic polities became less stable over time links back, either directly or indirectly, to the Muslim world's vulnerability to invaders from a less developed periphery. This narrative takes one of two forms in the existing literature. The first is that Muslim political and economic decline can be explained by the Mongol conquests of the 13th century.Footnote ⁴³ The divergence in political stability between Christian Europe and the Islamic world predates the Mongol invasions, however. Nor can the success of the Mongols be considered strictly exogenous as weaker, less stable polities may also be more subject to successful foreign invasion.

A more complex rendering of this narrative suggests that susceptibility to invasion from the periphery made Muslim societies less likely to develop the types of military technologies that proved to be useful for development over the long term. Chase (Reference Chase2003) and Hoffman (Reference Hoffman2011) argue that places with nomadic (i.e., barren) peripheries did not adopt firearms because this type of weaponry was not effective against nomadic invaders. Places that adopted firearms most effectively, like Western Europe and Japan, enjoyed important political and economic advantages (Chase Reference Chase2003). While our empirical tests do not consider this hypothesis directly, our inclusion of a variety of geographical control variables, including the latitude of the entity and an area's agricultural suitability, address some predictions associated with this theory. In addition, the oldest depiction of a firearm dates to the 12th century and firearms were not widely adopted until centuries later. The break that we identify in ruler duration, again, precedes the rise of firearms as a critical military technology.