From Empire to Nation-State: Explaining Wars in the Modern World, 1816-2001 Author(S): Andreas Wimmer and Brian Min Source: American Sociological Review, Vol

From Empire to Nation-State: Explaining Wars in the Modern World, 1816-2001 Author(s): Andreas Wimmer and Brian Min Source: American Sociological Review, Vol. 71, No. 6 (Dec., 2006), pp. 867-897 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/25472435 Accessed: 05/01/2010 12:02

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review.

http://www.jstor.org From Empire to Nation-State: ExplainingWars in theModern World, 1816-2001

Andreas Wimmer Brian Min University of California, Los Angeles University of California, Los Angeles

on war The existing quantitative literature takes the independent nation-state as the self

evident unit of analysis and largely excludes other political types from consideration. In contrast, the authors argue that the change in the institutional form of states is itself a are major cause for war The rise of empires and the global spread of the nation-state the

most important institutional transformations in the modern age. To test this hypothesis,

the authors introduce a new data set that records the outbreak of war infixed geographic territories from 1816 to 2001, independent of thepolitical entity in control of a territory. wars are more Analysis of this data set demonstrates that much likely during and

because of these two transformations. For the transformation to the modern nation-state,

the authors confirm this hypothesis further with logit regressions that control for

variables that have been robustly significant in previous research. The results provide

support for the main mechanisms that explain this time dependency. Modern nation

states are ruled in the name of a nationally defined people, in contrast to empires, which a across or govern to spread faith the world, to bring civilization to backward people, to

advance the world revolutionary cause. The institution of the nation-state thus introduces incentives for political elites toprivilege members of the national majority over ethnic

minorities, and for minority elites to mobilize against such political discrimination. The over resulting power struggles the ethno-national character of the state may escalate wars can into civil wars. Interstate result from attempts to protect co-nationals who are

politically excluded in neighboring states. The reported research thus provides a

corrective to mainstream approaches, which exclude ethnic and nationalist politics as

factors that would help understanding the dynamics of war.

lives and memories of most contempo a postcolonial civil war. The list of current con The raries have been shaped by the destructive flicts is long and includes entries from power of war. Entire generations fought and Afghanistan to Zaire. "Events of the recent past died in the two world wars. Throughout the have once again clearly demonstrated that the developing world, many remember the armed world is not yet ready for perpetual peace" struggles for independence or have experienced (Hintze 1975:215). This statement is as obvi

Direct to Andreas correspondence Wimmer, advice regarding research design and data coding; of of Department Sociology, University California Michael Hechter, Stathis Kalyvas, Jack Katz and Los 264 Angeles, Haines Hall, Box 951551, Los Rob Mare for offering us opportunities to present this CA Angeles, 90095-1551 ([email protected]). paper to various audiences; Lars-Erik Cederman, The authors thank Nicole Busse, Wesley Hiers, Michael Ross, Nicholas Sambanis, and Christopher Veronika Lenarz, Ani Sarkissian, and Nusrat Sheikh Winship for generous comments on a first written ver for excellent research of this assistance; Christopher sion article. Four outstanding anonymous Lars-Erik Xiao Indra Blattman, Cederman, Chen, reviewers helped to improve and clarify the argu de Felix Havard Rob ments. Soysa, Elwert, Hegre, Mare, All responsibility for errors of thought or fact Lillian Min, John O'Neal, and Michael Ross, for remains with the authors.

American Sociological Review, 2006, Vol. 71 (December:867-897) 868 AMERICANSOCIOLOGICAL REVIEW

ously true today as when itwas when penned Our study intends to overcome three basic 100 years ago by Otto Hintze inwhat is perhaps limitations of mainstream quantitative research the first sociological treatise on warfare. on war. First, most current conflict scholars Why do wars break out?Where and when are focus on short-term dynamics, such as the they most likely to occur? Despite the early impact of democratization over five years interest by Hintze and others of his generation, (Mansfield and Snyder 1995) or the effects of sociology has subsequently left the study of independence over two years (Fearon and Laitin these questions to other disciplines. Sociologists 2003). Recently, scholars have assembled ever have discussed war as a cause for other phe more disaggregated data sets to explain where nomena of interest to them, but rarely as an and when wars are fought (Levy 1998), includ explanandum in its own right. Thus, a long and ing detailed war event histories that decompose respectable tradition in comparative historical a civil war into various battle episodes (Raleigh sociology, stretching from the early work of and Hegre 2005) or studies of war theaters at the Ratzenhofer (1893) to Tilly (1975) and Centeno regional level (Buhaug and Rod 2005). (2003), analyzes the role of war in the making Meanwhile, macropolitical perspectives have of the modern, sovereign state inWestern largely been abandoned. Europe and beyond. Another tradition was ini The once prominent long-wave theories of tiated by Theda Skocpol's (1979) well-known war sought to explain the periodic recurrence of book, inwhich she looked at how wars helped world wars as a consequence of economic cycles to bring about the great revolutions of the past stretched over six decades (Goldstein 1991), or centuries, such as those of Russia and China. of the rise and fall of hegemonic powers that Meanwhile, research into the causes of war has compete for preeminence in world politics remained almost the exclusive domain of his (Modelski and Morgan 1985; Thompson 1988). torians and political scientists.1 Although there is no doubt that the past two cen Yet sociology has much to offer for the study turies have seen several such global wars involv of war. This article shows that that the macro ing the major power centers of the world, most historical processes traditionally studied by researchers now recognize that they do not fol comparative sociologists need to be taken into low a clear pattern of periodicity. In other words, account to arrive at a proper understanding of there are no cycles of a uniform length between wars. why, where, and when wars break out. We assert global an that two such processes play especially With this article, we attempt to revitalize this important role in the history of modern war: the macropolitical perspective by examining pat expansion of empires during the 19th century terns of conflict over the span of decades rather and the spread of nation-states across the world than across a handful of years as inmainstream during the 20th century. contemporary research. In contrast to long we That empire building and nation-state forma cycles theory, however, do not conceive of as an nor as a tion are major driving forces of war might not the globe integrated system, sin as a in unit of but rather as an arena come surprise to scholars well versed gle observation, world history or those familiar with the political for the discontinuous diffusion of institutional development of a particular region. It represents, forms. however, a new insight for the highly specialized Second, the standard units of analysis in quan literature on wars that has emerged over the past titative political science are existing independ several decades. Furthermore, considerable ent states, which are treated as continuous and once enter methodological obstacles need to be overcome for comparatively stable entities they over this hypothesis to be evaluated in quantitative the international community of states. This terms. To that end, we created a new data set of looks the fact that their institutional shape and wars from 1816 to 2001 that identifies conflicts territorial extension may change dramatically in all of theworld's territories, including most pre over time, not least as a consequence of war.2 colonial and all colonial lands.

2 For a general discussion of the problems associ with the of constant units?when 1A partial exception is Randall Collins's theory of ated assumption geopolitics (see summary inCollins 1995). they may de facto merge or split?see Abbott (1998). FROM EMPIRETO NATION-STATE 869

Tsarist Russia of 1846, for example, is con state, and thus are most likely to break out when to same unit as the Soviet Union ceived be the these institutional principles are contested. More of 1926 and the Russian Federation of 2006. precisely, our model suggests that the mecha Even more standard country disconcerting, nisms relating nation-state formation towar are year data sets exclude those parts of the world's similar: wars over territory inhabited by co surface and that are not population governed by nationals on the other side of a state border independent states?still more than half of the (commonly called irredentist or revanchist wars) globe by 1900. With such data sets, we cannot follow the same of nationalist that observe the consequences of macro-institutional logic politics transformations such as the colonization of the may drive civil wars, as majorities and ethnic world in the 19th century or the shift to the minorities compete for control over the state nation-state form during the 20th. (Weiner 1971). The entrenched division of labor To overcome these difficulties, our data set between scholars of interstate wars and schol uses fixed territories as the unit of geographic ars of civil wars may be an obstacle to the devel analysis, independently of whether a territory is opment of an encompassing understanding of part of an internationally recognized inde war, as many political scientists have recently pendent state. By relating each territory's con observed. flict history to its history of institutional change Our thus the of over a 200-year period, we are able to identify project explores prospects the macrohistorical tradition in the a recurring pattern: the likelihood of civil and revitalizing war. interstate wars is highest during the two insti study of To be sure, we conceive of this as tutional transformations that have shaped the a complement, rather than an alternative, to of modern political landscape the world. The mainstream lines of research. Many of the war data associated with a shift to a terri problems generating factors identified in the recent liter data set are considerable because pre tory-year ature are indeed important to understanding the independent territories rarely have been units of conditions under which political conflict may data generation and recording. However, does develop into full-scale war, remain virulent over this justify blinding out theworld of empires and time, or way to the of peace. the entire period of colonial domination? By give possibility we These factors include at the doing so, greatly reinforce the methodolog political instability ical nationalism that plagues so much of the centers of power (Fearon and Laitin 2003), local social sciences (Wimmer and Glick Schiller feuds that feed wars independently of the larg to 2002) ?the tendency take the nation-state for er political dynamic (Kalyvas 2005), the avail and to conceive of the social world as resources granted ability of natural to support warring an of nation-state societies without assemblage parties (Ross 2004), a specific dynamic of rival asking how this came about and what the con ry thatmay lead states to fight each other on the sequences of this particular form of political battle field (Vasquez and Leskiw 2001), and organization might be. This article iswritten as various mechanisms which make autocracies an invitation for other researchers to join us in more than democracies the effort to overcome these methodological war-prone (Levy 1998). Our to limitations and to improve on the 200-year data approach may help understand at what set we have created for this project. point along the long-term history of institu the Third, literature has largely overlooked tional transformations political tensions are between civil wars and inter possible linkages likely to build, allowing the preceding factors to state wars beyond occasional acknowledgments come into play. To put it differently, whereas our of military strategic spillover effects (Most and approach focuses on when wars occur, the afore Starr 1980;Wilkenfeld 1973; but see Davis and mentioned research ismore suitable for explain Moore 1997). Although we are not able to exam ing where it However, an ine the precise nature and extent of this rela happens. developing account that var tionship, our results suggest thatmany civil and encompassing integrates these interstate wars are outcomes of the same ious levels of analysis and specifies the link processes. Both civil and interstate wars are between long- and short-term processes is fought over the institutional structure of the beyond the scope of this article. 870 AMERICAN SOCIOLOGICALREVIEW

FROM EMPIRESTO NATION-STATES:A definitions offered by Samuel Eisenstadt LONG-TERM,INSTITUTIONALIST (1963:10-24) and Stephen Howe (2002:13-20), APPROACH empire is defined by the following institution al features: centralized bureaucratic forms of We develop our "big picture" argument from an government, the domination of a core region institutionalist perspective, as it has been devel over peripheries, an ethnically or culturally oped in the study of nation-state formation and defined hierarchy between rulers and ruled, conflict over the past decade (Brubaker 1996; and claims to universal legitimacy?whether Hechter 2000; Mann 2005; Posner 2005; Strang referring to a revolutionary ideology (e.g., the 1990; Wimmer 2002). Building on these vari Soviet Union), a mission civilisatrice (e.g., ous cited works, we sketch out an integrated colonial or conversion framework for how institutional empires), religious (e.g., understanding the Nation-states also are transformations relate tomodern war. The basic Spanish empire). based on centralized bureaucratic forms of gov hypothesis states thatmodern wars have result ernment, but are ruled uniformly without an ed from two processes of institutional transfor institutionalized differentiation between core mation: incorporation into empire and and periphery, embrace the principle of equal nation-state formation. In the next section, we of citizens hierarchy), and gov offer a descriptive account of these two trans ity (replacing ern name a formations. We then introduce our theoretical in the of bounded national rather than some universal model and specify themechanisms linking these community princi The nation be as multieth two processes to war. ple. may imagined nic and multireligious, as in Switzerland or India, or as monoethnic and as Imperial Expansion and Nation-State monoreligious, in France and Japan. Formation, 1816-2001 All polities governed by other institutional Figure 1 shows the percentage of the world's principles were assigned to the category of area over cen cen surface occupied the past two "other." Absolutist kingdoms also know turies by empires, modern nation-states, and tralized bureaucracies, but lack the other political institutions. On the basis of the center-periphery structures and the universal

' ; / v / *>'Y

'' T, // 'v Y / / /'./ '"//, -' y y o '/ ' / / / < ''/ // / / < // Areas Y'x -> Y/--Y'// Y $ governed by // '/ Y '/ ?$ Y/-' /.

1820 1840 I860 1880 1900 1920 1940 I960 1980 2000

Figure 1. Percentage of Land Surface Governed by Empires, Nation-States, or other Institutions, 1816-2001

Note: States smaller than 25,000 km2 are excluded. FROM EMPIRETO NATION-STATE 871

and nation-state formation. ist forms of legitimacy of empires.3 In contrast empire building to nation-states, absolutist states are not based While empires replaced "other" governments the 19th also on the equality of all citizens, which makes a dif during century (see Strang 1991), theWestern colonial inAfrica ference even if a nation-state is ruled by a dic especially empires and but also the Romanov in tator with the powers of an absolutist king.4 Asia, empire Central were at the same time Feudal states, tribal confederacies (e.g., the Asia, they nation-states in theWestern hemi Sanusi of Libya), city states (e.g., Switzerland replaced by most in Latin America. before 1848), and patrimonial empires (e.g. the sphere, importantly The of the 19th Tukolor orMongol empires) all lack centralized expansion empires during bureaucracies. century does not match, however, the dramatic of their decline the 20th Note that we exclude "informal empires" proportions during when the nation-state form (Mann 2006), such as the contemporary United century, spread across the to achieve an uncon States or the dispersed hegemonic "empire" of globe finally tested as John and Hardt and Negri (2000), from our definition of hegemony, Meyer (1997) others have shown. This empire because these are not politically coher (e.g., Strang 1990) of diffusion within ent entities. Note also that following the terri process developed entirely our time horizon. In 1816, France and torial logic on which our project is based, we only Great Britain were autonomous nation-states. code the political institutions governing a par By 2001, had and ticular territory, not those of entire states. Thus, empires entirely disappeared, a handful of states were absolutist the territory of Great Britain is classified as a only king doms Saudi or nation-state, even while it was the core of a (e.g., Arabia, Bhutan, Brunei) patrimonial states (Libya). There is thus only a large colonial empire. The territory of the con moderate left- and for temporary United States is a nation-state, right-censoring problem our of the second transformation, although Guam is governed according to impe analysis whereas there is considerable left-censoring for rial principles. the first transformation because more than half The graph shows that shortly after the of the world was already under imperial rule Congress of Vienna was held in 1814/1815, when our data series begins. This is one reason roughly half of the world's surface was ruled by why our research focuses more on the second empires and half by "other" political systems. transformation to nation-statehood. In 2001, almost the entire globe was controlled The diffusion of the nation-state occurred in by modern nation-states. The two centuries waves, each triggered by the crisis of a major between tell the story of a struggle between empire and its eventual dissolution. The first wave followed the collapse of the Spanish empire. The second wave occurred afterWorld 3 In contrast to Eisenstadt (1963, chapter 1) and in War I with the breakup of the Ottoman and line with Howe we exclude the abso Stephen (2002), Habsburg empires, and the third after World lutist and of Western kingdoms principalities Europe War II,when the Middle East as well as South from our definition of We did not want to empire. and Southeast Asia were decolonized. The assign Wurttemberg before Bismarck or the Papal fourth wave followed about 1960, when the State before Garibaldi to the same category as impe or British and French colonial empires broke apart, rial China the Spanish empire. and the fifth occurred when the oldest colonial 4 Our definition of the nation-state is based on the empire, the Portuguese, finally dissolved. The broad typologies developed in political sociology, sixth wave rolled over the Soviet empire during rather than on the regime types of political science the early 1990s. These six main waves of nation We thus assume that the (e.g., democracy, autocracy). state creation are discernible in 2 a difference between nation-states and absolutist states Figure (see even definition of nation-state creation later). asserts itself within the same political regime type: modern dictators such as Idi Amin cannot rule same as in the way Louis XIV; they cannot rely on A Long-Term, Instttutionalist Model of dynastic legitimacy, but instead must show that their Modern War government benefits "the people" (e.g., of Uganda), for instance, by expelling Indian traders as "parasites" Why should these two institutional transfor from the national home. mations?empire building and nation-state for 872 AMERICAN SOCIOLOGICALREVIEW

1991-1995

1956-1960 I

? 1916-1920 1946-19501 I 1821-18254 I ? III I

0 Y i^f f 1 T ?|>r ,?,?j?,?,?,?j?, \?J?1?i?\ j 1?j?1?1?1?J |?1?1?1?j?1?1?1?j?????r^ 1800- 1821- 1841- 1861- 1881- 1901- 1921- 1941- 1961- 1981- 2000 1805 1825 1845 1865 1885 1905 1925 1945 1965 1985 2001

Figure 2. Number of Nation-State Creations (Five-Year Periods), 1800 to 2000

mation?cause war? Our model assumes that contractions of states are associated with war. wars are fought over the most basic institution A study by Diehl and Goertz (1988:115) shows al principles of government i.e., the informal and that only one fourth of all territorial changes in formal rules that determine who legitimately can the world state system from 1816 to 1980 have lay claim to governmental power and what the involved some violence. According to another legitimate borders of a polity should be. It is thus study, very few of the 26 sovereign states that a genuinely political understanding of war in have disappeared from the political map since which economic interests or military-technical 1415 did so as a consequence of war (Strang feasibility play only a secondary role. From this 1991:155). Third, where wars do coincide with perspective, the history of modern warfare the creation of new states or the absorption of we appears to be not so much the result of chang existing ones, argue that they are the con ing power balances between actors or of revo sequence of institutional transformation rather as a a new lutionary conflict primarily struggle than separate causal mechanism. A prin between competing projects of state building. ciple of political legitimacy implies a new def This struggle leads to the two transformations inition of those that should legitimately be described in the preceding section. included in the territory of a state and those However, is it not self-evident that the cre that should not, as discussed in the section on new or new ation of states the absorption of existing mechanisms later. Attempts to create states states into an expanding empire will be accom or destroy existing ones can therefore be con panied by war? We should like to address this ceived as part and parcel of the fight over the potential misunderstanding early on. First, most institutional form of governance. Fourth, many of the wars in our sample are not associated with instances of nation-state formation do not go or the creation absorption of states, as we show hand in hand with the creation of new states, as in greater detail later. This is true for most civil the examples of Japan, Switzerland, France, wars and many interstate wars. Second, and Thailand, and Ethiopia illustrate. The previous conversely, not all territorial expansions and graphs thus do not display the creation (and FROM EMPIRETO NATION-STATE 873

destruction) of states, but rather, the diffusion globe and bring civilization, Christianity, Islam, of the nation-state form across the territories of or revolutionary progress to all of humanity, the globe. irrespective of the ethno-national background of If nation-state formation is not synonymous the population. Empires show an institutional with the establishment of new states, criteria are ized drive to expand their domain through con needed to establish when the institutional trans quest, even if at high military, political, and formation to nation-statehood has been economic costs. Claims to universal legitima achieved, independently of changes in a state's cy make the extension of the imperial domain borders. For the process of both nation-state a benchmark for judging the success of the formation and imperial incorporation, we iden military-political elite. Moreover, the cen tified a crucial "turning point" representing the ter-periphery structure allows for easy incor climax and successful completion of the respec poration of newly conquered populations. They tive institutional transformation. Whether this are simply added as new pieces to the ethno turning point is ever reached or not depends on national mosaic and henceforth ruled through who wins in the struggle over the institutional some mixture of direct and indirect rule. shape of the state.5 During imperial expansion, Local political units?tribal confederacies, the turning point occurs when the principle of alliances of city states, traditional kingdoms? imperial sovereignty is established in the local may resist imperial expansion and refuse to be arena. This may occur through warfare, strate "pacified" and "civilized" by the encroaching or and A shift gic marriage alliances, religious conversion, army the imperial administration. and thus is analytically distinct from war and toward imperial principles of rule not only a imperial conquest. We call this turning point implies loss of power, but also a delegitima "imperial incorporation," after which the terri tion of the very institutional rules that allowed tory is governed as a dependency of a faraway elites to struggle for and perhaps gain power. imperial center and ruled in the name of the Tribal sheiks, for example, risk losing not only over a once an spread of civilization, Islam, Christianity, or military control territory impe revolutionary progress. In the process of nation rial army starts to establish garrisons. The very a cen state formation, the turning point is reached possibility of gaining power by holding when nationalist forces manage to establish the tripetal alliance of clans together is undermined new principle of national sovereignty (thus over by the bureaucratic system of administra were coming dependency from an imperial center), tion. Similarly, bourgeois elites that organ and the territory is henceforth ruled in the name ized in city councils lose their political standing of a nationally defined group of equal citizens, and capacity for forging alliances with other once replacing the hierarchical universalism of city-states their government is incorporated as a empire. We call this turning point "nation-state the lowest tier into political and bureaucratic creation." Detailed coding rules are discussed hierarchy far removed from the reach of elite in the data section. councils. Under conditions not specified by our model (e.g., military opportunities, a given distribution The first transformation, to empire: mech resources or of between actors), tribes city anisms. What are the causal mechanisms that states may choose to fight against expansionist link imperial expansion and nation-state for empires. Once the crucial turning point is mation to war? First, we briefly discuss impe reached and a territory is governed according to rial incorporation. Empires are defined, imperial principles, the hinterland may contin according to our model, by center-periphery ue to resist imperial expansion. When the entire relations, hierarchical inclusion, and claims to territory ismilitarily subdued or "pacified," the universal legitimacy. Thus, they know no natu process of expansion is complete, and imperi ral borders and may potentially cover the entire al peace should prevail, interrupted by occa sional rebellions against higher taxes, violations of traditional rights or undue interference in 5 local or On the related concepts of "turning point," "crit political affairs, attempts of ambitious ical junctures," and "cross-roads" see Abbott (2001, provincial governors to establish their own mini and cease chapter 8) Zapf( 1996). empires and paying tribute to the center. 874 AMERICAN SOCIOLOGICALREVIEW

The second transformation, to nation out significant changes in their territorial exten state: mechanisms and intervening variables. sion. In these cases, the struggle over the insti The second institutional transformation starts tutional shape of the state leads not to an with the spread of themajor competing project secessionist wars, but instead to increased war. of state-building: modern nationalism. For the likelihood of other types of civil Once the is the insti sake of simplicity, we treat this diffusion process turning point reached, tutional of the creates further as exogenous to our model.6 The spread of the logic nation-state nationalist doctrine?that states should be gov incentives for making war, both between and within established states. Because the erned in the name of a nationally defined com newly new state is in the name of a nation munity of equal citizens?triggers the process governed defined "Staatsnation" in of nation-state formation. During this first ally people (a Meinecke's well-known the new phase, we expect a higher likelihood of seces terms), polit ical elite tend to treat members of this sionist wars than during any other period of "Staatsnation" In other words, history. In the political order for which nation preferentially. equality before the law, protection from arbitrary alist leaders aspire, each ethno-national group violence, and political participation are con should govern itself, and the government in fined tomembers of the dominant ethnic group. turn should be representative of the ethno The legal and political status of ethnic minori national makeup of the population. This stands ties now often worsens dramatically (Wimmer in stark contrast to the institutional principles of 2002; see also Noiriel 1991;Williams 1989). At empire, in which the ethno-cultural hierarchy the same time, differential treatment of indi between elites and is consid imperial subjects viduals on the basis of ethnic or national back ered a stabilizing and legitimate feature of prop ground now contradicts the fundamental er government, and political legitimacy is institutional principles of the state, which is derived from God or revolutionary progress, supposed to be based on the twin principles of and not from a nationally defined people (cf. equality and freedom from "foreign" rule. Calhoun 1997). The prospect or reality of ethnic discrimina nationalist movements will thus Many tion increases the likelihood of both interstate encounter the resistance of elites, who imperial and civil wars. To protect co-nationals living will have no chance of admitted to the new being across a state border (a legacy of the patchwork for that often are of a game power, given they settlement pattern of empires) from ethnic dis different ethnic background than the local pop crimination and to show to their own con ulation. Wars of secession an against imperial stituency that they care "about our ethnic center are thus one road to the second institu brothers," a nationalizing state's elite may be tional toward turning point nation-statehood, tempted to annex the corresponding territory in when a territory starts to be governed accord the name of "national unification" (Weiner note ing to nationalist principles. We again, 1971). Such irredentist wars over borders and as however, that other roads are traveled well. territory7 are of a different nature than wars of Absolutist states such as France and Thailand, conquest. The very logic of nationalist doctrine or alliances of city-states such as Switzerland may transform into modern nation-states with 7 Not all interstate wars in the modern era arise over

territorial issues, and not all territorial issues are the 6 of the described Various theories accounting for this process of dif consequence ethno-political dynamic discussion. But we know from fusion have been discussed in comparative historical in the preceding pre vious research that the likelihood of interstate wars sociology and political science including economic increases if such territorial and ethno functionalism (Gellner 1983), world culture argu significantly accounts conflicts in a of countries ments (Meyer 1997), political culture political emerge dyad Most (Anderson 1991), diffusion models (Strang 1990; (Huth 1996;Vasquez and Leskiw 2001). impor Davis and Moore show that states will Wimrner and Min 2005), and political modernization tantly, (1997) have levels of conflict if an eth arguments (Hechter 2000). We need not enter this higher (including war) or debate here because we are more interested in the con nic group is dominant in one state and dominated war in in the sequences of nation-state formation for modern mobilized antigovernment protest neigh than in its causes. boring state. FROM EMPIRETO NATION-STATE 875

and even impedes modern nation-states from expanding lization against such discrimination, much beyond the domains of their core ethno tually to conflict. Research by Cederman and national group. Girardin (forthcoming), using a demographic At least two factors help to predict whether proxy to capture levels of ethno-political dis the shift to the nation-state will be accompanied crimination, provides additional quantitative set by interstate war. The first factor is the ethno support for this hypothesis. Our data intro meas demographic makeup of a state and whether or duces a newly constructed variable that not it contains substantial minorities that are ures ethnic discrimination more directly majorities in neighboring states. Previous (discussed later). research has shown this to be a powerful pre Obviously, ethno-political mobilization does dictor of interstate war onset (Davis and Moore not always lead to civil war. Our model speci 1997). Cross-border ethno-national politics also fies three intervening variables. First, rich coun produces spillover effects, along the lines that tries find it easier to react to ethno-national Weiner described as the "Macedonian syn protest through a policy of power sharing, affir drome" (Weiner 1971). The second factor is mative action, or redistribution than do poor that the likelihood of interstate war will be high countries. In poor countries, state resources are er when a neighboring territory is already scarcer and alternative sources of income are involved in a war because ethnic kinships can lacking, transforming competition over state revenue a zero-sum stimulate governments to protect their cross into game (cf. Wimmer border co-ethnics from consequences associat 1997). Indeed, much empirical research finds ed with power shifts in the warring state. Such that gross domestic product (GDP) per capita is "outside interference" may not be tolerated by one of the most robust factors in predicting the neighboring government, and tensions may civil war onsets (Sambanis 2004). escalate into another interstate war. Second, different sources of state revenue Unfortunately, our data do not allow us to test have different effects on the political process, the effects of cross-border ethnic kinship net including those leading towar. Oil resources in works, although we do test for more general particular have been linked to an increased like spillover effects. lihood of civil war (Ross 2004; but see Sambanis We now turn tomechanisms and intervening 2004). Our model incorporates one of the expla variables for civil wars. The shift to the nation nations brought forward in the literature (cf. state may lead to civil war when minority mem Humphreys 2005): a state that depends not on bers mobilize against political discrimination taxes, but on oil rents may develop higher and attempt to overthrow the ethnocratic regime degrees of clientelism than tax-dependent and by force, or when they use violent force to resource-poor states. Ethnic clientelism may in secede and create a new state in which they turn feed a dynamic of ethnic exclusion that can would represent the national majority or join lead to ethno-political mobilization and even their co-ethnics in a neighboring state. Building tually to civil war. on earlier comparative historical research Third, civil wars in one country are likely to (Wimmer 1997, 2002), our model thus identi affect relations between politically relevant eth fies political discrimination along ethnic lines nic groups in neighboring countries, resulting as the central mechanism that links nation-state in an escalation of political tensions in those formation to these various forms of conflict. countries.9 Although we again lack data on rela High degrees of political discrimination along ethnic lines8 will lead to ethno-political mobi

(Wimmer 1997). Unfortunately, we were unable to find data that would measure civil society develop ment at the time of nation-state creation for a suffi 8 Our full theory endogenizes discrimination by ciently high number of territories. 9 specifying the conditions under which higher degrees Gleditsch (2003) shows that the likelihood of civil of exclusion along ethnic lines are to be expected: if wars increases if an ethnic group stretches across are a nation-states created at moment when networks the territories of two neighboring states (however, the of civil society organizations are only weakly estab results are not supported by Ellingsen 2000). This is on lished, political elites will rely ethnic and other consistent with Sambanis' (2001) finding that wars a a communal ties to build up political following in neighboring state increase the likelihood of 876 AMERICAN SOCIOLOGICALREVIEW

tions of ethnic kinship across state borders, we Summary: institutional transformations can at least determine we can and war. whether observe In summary, we expect the ceteris the effect that our spillover model postulates. paribus likelihood of violent conflict to be high In summary, the transition from empire to est near periods of institutional transformation, nation-state increases the likelihood of both when the struggle over the institutional princi interstate and civil wars because the institu ples of government is most intense. Figure 3 tional of principles legitimate government summarizes the expectations of our model. The claims to universal are change: legitimacy likelihood of war is predicted to crest at the rule in the name of the and replaced by people, turning points associated with imperial incor the realm of the state is reduced to the territo poration and nation-state creation, and to drop members of the nation. This ry occupied by to lower levels of risk during periods of insti switch in the rules of the political game provides tutional stability. The result is a double invert new incentives for the of includ pursuit power, ed U-shape. violent means. In the first of the ing by phase The model also describes which type of con secessionist wars of "national liber transition, flict should be most likely during a territory's ation" the are to be against ruling empire likely trajectory through these two institutional trans In the second wars between new fought. phase, formations. Wars of conquest will be the most states over mixed break ethnically territory may frequent type of war in territories undergoing the out, and ethnic minorities rebel might against first transition as empires replace other forms exclusion national The political by majorities. of governance. During the first half of the model that such civil wars are more predicts nation-state formation process, secessionist civil likely where ethnic discrimination is where high, wars against imperial rule should be the domi governments are poor and thus unable to accom nant war type. After the turning point, nonse modate ethno-political protest, or where there cessionist struggles over the ethno-national is on oil revenues that foster ethnic dependence character of the state and interstate conflicts clientelism and exclusion. over ethnically mixed territories will be more This model obviously does not attempt to frequent. explain all wars in the modern world. We should like to briefly discuss the charac Revolutionary wars are not driven by the poli ter of this model here. It stylizes and simplifies tics of nation building and ethnic exclusion that complex historical event chains and thus accompany the shift to the nation-state form. We describes a certain developmental pattern rather thus expect that the likelihood of war recedes than a relationship between independent vari back to a baseline rate after nation-state for ables that affect stable units in a history-free mation is complete. Nation-state formation space (cf. Abbott 1998). This does not mean, comes an to end once struggles over the ethno however, that our model is not causal or cannot national character of the state and its borders be tested using standard regression methods. with neighboring states are settled. This may be The model is causal because it specifies the reached through a stable, institutionalized mechanisms through which institutional trans arrangement of power sharing (as in the Swiss formations and warfare are linked. To be sure, case), through a series of ethnic cleansings in these mechanisms are historically specific and civil and interstate wars (as in Eastern Europe), irreversible: during some phases of the overall or through a successful policy of assimilation process, wars cause institutional shifts are that achieves the nationalist dream of homo (and thus independent variables), whereas in other geneity in other ways (as in France). During this phases, wars are the consequence of this shift late stage in the process, a posthegemonic state (and are thus dependent variables). The model based on the idea of multicultural diversity may can be tested with standard methods because we finally emerge. can measure institutional shifts independently from warfare and thus determine whether their temporal follows the predicted pat ethnic civil war much more significantly than noneth relationship nic civil wars. For further statistical evidence of the terns under ceteris paribus conditions. We also can test whether the mechanisms insti "contagion effect" of ethnic conflicts, see Gurr linking (1993:181), and Lake and Rothchild (1998). tutional shifts to war hold up empirically by FROM EMPIRETO NATION-STATE 877

Probability of war

A , ,

Non-secessionist 1^ 1^ Wars of Wars of Secessionist civil wars X*"^\ ^^^s\ conquest / ! \ conquest civilwars / ! \ and / I \ / I \ inter"statewars

Turning point I: Turning point II: Politicaldevelopment '" Imperial incorporation Nation-state creation territory

Figure 3. A Stylized Historical Model of Institutional Change and War

including intervening variables in the regression ning of our data set in 1816. All our informa equation. tion thus had to be coded accordingly, as the fol lowing examples illustrate. An income estimate A NEW DATA SET for "Bosnia" in 1915 relates to the territory that in 1991 became Bosnia, independently of the To tests of the we had perform adequate model, fact that this territory in 1915 was a part of the to create a new data set. Our that were theory required Austrian-Hungarian empire. Wars coded we fixed territories as units identify geographic as occurring on the territory of its major bat of because this would allow us to analysis only tlefields, defined, according to conventions in determine whether in the institutional changes the field, as a conflict with more than 1,000 form of over a are indeed governance territory battle deaths. If colonial subjects rebelled related to the onset of war. If we took inde against Her Majesty's government inwhat today states as our units of observation, such pendent isKenya, the war is attributed to the territory of institutional shifts would be to trace. impossible Kenya, and not to theUnited Kingdom, of which We thus from the standard country departed Kenya was a part at the time the rebellion set to year data collect and code data for fixed or occurred. Episodes of nation-state creation units both for our dependent variable geographic imperial incorporation also were attributed to (onset of war) and our independent variables. individual territories: when Yugoslavia was This section describes the most cod important founded in 1921, each of the individual territo ing principles. More details can be found at ries that belonged toYugoslavia at the time was www. soc .ucla. edu/faculty/wimmer/data. coded as experiencing a nation-state creation. If the different territories forming a state are gov Units of observation. Which territorial units erned according to different institutional prin of observation would be adequate for the task ciples, the coding for the individual territories at hand? We used the division of the world's also differs. The territory of Great Britain was states in 2001 as a territorial grid, extending governed according to nation-state principles these fixed geographic units back to the begin throughout the period under consideration, 878 AMERICAN SOCIOLOGICALREVIEW whereas the territory of current Ghana went our data set contains less than a handful of cases through imperial incorporation in the 19th cen for which our territorial logic makes little sense, as as war tury and began to be governed a nation-state such the Russo-Japanese mentioned ear much later, in 1960. lier. The shift to a territorial logic allows one to hold A second possible objection is that our units a spatial unit constant and observe over time of observation are not independent of the causal how it is governed, whether by an empire, amod processes we observe. The 2001 grid of states ern nation-state, or some other form of polity. We indeed has resulted partly from the past 200 can thus determine whether periods of institu years of war associated with empire building and our tional transformation from one type of gover nation-state formation. However, explanan nance to another are indeed more likely to be dum is not the territorial shape of states, but accompanied by war. Before we proceed to a rather the occurrence of war. That we observe description of the data set, we note two possible this occurrence through a grid that is the result objections to our choice of observational units. of future wars would be problematic only if + First, the territories do not always correspond these wars fought at t x were influencing the to process units (i.e., the units within which the state of variables at t,which at least according relevant political dynamics unfold). The Russo to Humean notions of causality is quite impos Japanese war, to give themost striking example sible. from our data set, was entirely fought on the ter of current China, and is thus attributed to ritory The war data set. A series of data China rather than to the of Russia or problems territory had to be overcome once we shifted from a data Japan, where the decisions of going towar were set that contains only sovereign states to one that made. covers the entire globe over the past 200 years. we should determine, at each in Ideally, point We invested considerable effort to create rea time, within which units the relevant political for wars, our that led to a war and then sonably complete codings depend processes unfolded, ent variable. Our was the code all other information with reference to starting point widely used Correlates ofWar data set. these units. most of the territories in (COW) project However, We first had to find and code wars that occurred our data set did represent relevant political enti in territories excluded from the COW country ties during most of the period covered by our list because they were not part of theWestern data set. As mentioned earlier, empires are char state system as defined by COW?a limitation acterized by the domination of peripheral units that has been mostly overlooked by researchers by a political center. The peripheries are gov who use the COW data (other problems are dis erned as separate political units, whether cussed by Sambanis 2004). We searched for through indirect rule or, much less frequently, wars missing from COW's data set for much of through direct administration. Bosnia under the 19th-century Latin America, Central Asia before Habsburgs and Egypt while still under the nom the Russian conquest, and the like. Our major inal sovereignty of the Sultan were indeed units sources were Richardson (1960), who provid within which many political decisions were ed the basis for COW's original data set; a made, including the decision to go to war with detailed historiography of wars of the modern neighboring territories (as in the Egyptian con world (Clodfelter 2002); and a number of online of or to raise arms the quest Sudan) against sources such as onwar.com. We are confident imperial center (as in Bosnia throughout the that our new war data set is reasonably complete, 19th century). Because almost all current states with the exception of some areas in precolonial that were part of former empires represented ] Africa and Central Asia.1 We also updated the such imperial provinces, our approach may be considered reasonable. For territories that have es we observe: the 2001 grid of states is indeed part shifts in state war not experienced any dramatic ly a result of the past 200 years of associated with boundaries (e.g., France, Japan, Switzerland, empire building and nation-state formation. However, Ethiopia), the problem does not exist.10 Overall, our explanandum is not the territorial shape of states, but rather the occurrence of war. That we observe this

occurrence at time t through a grid that is the result 10 wars at t + x would be A second possible limitation is that our units of of the fought problematic wars were related to the observation are not independent of the causal process only if these future causally FROM EMPIRETO NATION-STATE 879

list to 2001, relying on Gleditsch et al. (2002), actors involved. Our institutionalist model sug and followed some of the revision to the COW gests that the aims of warring parties may data set proposed by Gleditsch (2004). change according to the institutional environ We then added locational codes for all wars ment. We thus reclassified all wars according to in our database so wars could be assigned to one a simple typology of war aims (for a similar of the fixed territories. In the COW data set, effort, see Holsti 1991, chapters 1 and 12). wars are attributed to states, independently of According to this typology, war participants their actual territorial extension. Thus, COW can fight for domestic power (civil wars) or to as codes a war in early 20th-century Morocco enlarge the power of the state vis-a-vis other a French war because Morocco was part of the states (interpolity wars). Civil wars are subdi French empire.12 Most of the information vided depending on whether the participants regarding the location of battlefields was col try to establish a new independent state (seces or over an lected from Clodfelter (2002). sionist war) gain/retain control exist Finally, COW's classification of wars depends ing one (nonsecessionist civil war). Interpolity on the status of actors within theWestern state wars are subdivided into wars of conquest, occur a rec system. "Imperial wars" between which aim at the permanent incorporation of the ognized state actor and an actor which is not part territory and population of the enemy state, or of that system (a tribe, an independent king interstate wars, inwhich the balance of power dom). A "colonial war" is fought between a between states is at stake and participants are not recognized state actor and a nonstate actor which trying to absorb the enemy state completely. is part of the system. "Interstate wars" take Codings depend on intentions rather than out place between independent system actors, and come: when secessionists fail to establish their a "intrastate wars" are waged between state own state, the war still is classified as seces actor and a domestic nonstate actor. sionist. When conquest is successfully resisted, Because our units of observation are fixed thewar is nevertheless coded as one of conquest. geographic territories rather than actors, we had The type of war is therefore not dependent on to come up with a new typology of wars that who won. would be independent of the character of the We dealt with the difficult coding problem of changing or conflicting war goals as well as we could. Table 1 gives an overview of this new classification scheme and the COW categories state of variables at t, something Hume tells us is that contain the greatest overlap with our new impossible. It should be noted that the definition of 11 types. We that some of the wars guess following may these war types is independent from our defi have reached the 1,000-battle-death threshold: the nitions of political systems. Indeed, all types of wars Yoruba states in among precolonial Nigeria, wars can occur in premodern states, in empires, the civil wars in Ethiopia and Afghanistan during the and in modern nation-states, which is not to middle of the 19th century, the wars connected to say that we expect them to be equally likely. Buganda expansion in Uganda in the precolonial to the our model claims that era, and the wars between the khanates of Central Asia Quite contrary, institutional frames influence the aims before Russian conquest. Unfortunately, we were not political able to gather reliable data on battle deaths, and thus that actors pursue, including the aims of warfare, did not include them into the current version of our and that certain types of war should therefore data set. be more frequent in certain types of political sys 12Sambanis discusses this and (2004) problem tems. For this claim to be nontautological and one other solution: to take the entire terri possible nontrivial, definitions of war must be inde tory of an as the unit of observation. However, empire pendent from definitions of political systems. this would create a difficult data problem because Itmay thus be useful to briefly illustrate this GDP and other figures would have to be averaged independence for interpolity wars, in which it over the entire empire. Furthermore, important dif might appear to be more problematic. Interstate ferences in the living conditions?including human wars may occur between rights and democratic participation?of the "moth competing empires to snatch from each such erland" and the colonies would disappear from the trying territory other, as picture. Sambanis does not discuss the possibility of when the Romanov and the British empires territorially defined units of analysis. fought over control of Afghanistan, however, 880 AMERICAN SOCIOLOGICALREVIEW

Table 1. A New War Typology

Wars between Independent Polities Civil Wars

Types and Secessionist Non-Secessionist Subtypes ofWar Wars of Conquest InterstateWars Civil Wars Civil Wars Description Expansion of state Fight between Fight against the Fight between territory, states over political center groups, at least permanent incor- borders and with the aim of one of which poration of new territory, regional establishing an represents the territories and hegemony (but independent state; central govern populations; without aim of resistance against ment, over - resistance against permanent such independ- domestic power such expansion. incorporation as ence by the relations, degree in wars of con- political center of autonomy of quest) provinces or ethnic groups, tax burden, dynastic COW War Type with succession etc. wars Greatest Overlap Imperial extra-state Inter-state wars Intra-state Intra-state wars and wars colonial extra state wars = Note: COW Correlates of War.

without Britain aiming at absorbing the entire or a territory legally became a protectorate territory of the Russian empire into the realms or colony (whichever came first), we coded of Her Majesty (as in a war of conquest). this as the year of imperial incorporation. We Interstate wars also occur between city states coded as the year of nation-state creation the vying for control of regional trade, between date when a territory began to be governed on nation-states over disputed territory (howev the basis of a written constitution that identi annex er, without attempting to the neighbor fied a national group as the sovereign of the ing state), between absolutist states, and state, whether the nation was defined in mul between tribal confederacies. Wars of con tiethnic or monoethnic terms. National sov quest have historically occurred between ereignty has a domestic and an external empires (e.g., the dismembering of the component. Domestically, the constitution an assorted of Ottoman empire by group should establish a nationally defined com Western World War imperial powers during I), munity of equal citizens and foresee an insti the United States con nation-states (e.g., tutional representation of this community (not the Dominican or Nazi quering Republic necessarily a freely elected parliament). Eastern to Germany conquering Europe gain Externally, national sovereignty means control between "Lebensraum"), city-states (e.g., over the foreign policy decisions of the coun Athens and and tribes. Sparta), try. The criteria had to be cumulatively ful filled. The United States and South Africa, to were as THE INSTITUTIONAL TRANSFORMATION DATA give two examples, treated nation set. To test our model, we had to determine states only after the abolition of slavery and the end of after introduction of when the turning points in the two processes apartheid (i.e., of institutional transformation occurred in equality for all members of the nation). We or calendar time, independently from the actual thus excluded "Herrenvolk"-democracies from our definition of occurrence of war or border changes. We slaveholding republics established the following coding rules. As modern nation-states. Saudi Arabia, Bhutan, soon as a territory was effectively adminis and Brunei still lack popular representation or a was to be ruled to tered by an empire, garrison estab today and continue according lished that controlled the territory militarily, dynastic principles. FROM EMPIRETO NATION-STATE 881

6.0%

?>5.0% I A

I Wars and nation-state g?\ JL

* 8 * Wars and imperial i ,0%V^ V\/ %**V ^ l? vi*? r? . incorporation *, ^ !,

0.0% I-{-1-1-1-;-1-1-1-1-?-1-1-j-1-1-;-;-1-1-1 -10 -8 -6 -4 -2 0 2 4 6 8 10

Decades to/from turning point

Figure 4. Rates of War Relative to Imperial Incorporation and Nation-State Creation (10-Year Moving Averages)

a can Note that territory experience various DISCOVERINGTHE PATTERN: episodes of imperial incorporation and of TEMPORALVARIATION INWAR RATES nation-state creation. include the Examples Rates of War around the Two current of which has territory Poland, gone Transformations through various partitions and imperial annex This new data set allows us to test whether wars ations, and the multiple nation-state creations have been more frequent during the two periods that a territory such as today's Croatia has of major institutional transformation that have experienced over the past 100 years. We thus characterized themodern world. Figure 4 shows arrived at a data set of 484 distinct wars, strong support for this claim. We calculated 77 wars of 111 interstate including conquest, rates of war onset for each year before, during, were wars, and 296 civil wars, 109 of which and after the two turning points by tabulating the secessionist civil wars and 187 were nonse number of territories at war versus those at cessionist civil wars. Our data collection peace in any given year relative to each trans formation. the x-axis is made of the efforts involved more than 156 territorial units, Thus, up years before and after a territory experienced 140 of which were incorporated into an empire either imperial incorporation or nation-state at some point (92 during the temporal range creation. It does not record chronological time, of this 1816 to and 150 of project, 2001), but rather shows the transformation clock of which nation-state creation. We experienced the individual territories, the year of the turn use this reduced set of 150 territories slightly ing point being set at zero. The y-axis shows the for all subsequent analysis. percentage of territories in which a war broke 882 AMERICAN SOCIOLOGICALREVIEW out in a particular year before, during, or after not associated with any significant change in the these two transformations. We depict the war boundary of the polity. Conversely, some such rates associated with the two transformations territorial changes (e.g., the Louisiana Purchase separately. in the United States, various territorial acquisi The dashed curve shows the rate of war as a tions through royal intermarriage, and the trans 10-year moving average in relation to the year fer of colonies between imperial powers as of imperial incorporation?excluding territo spoils of war) had nothing to do with either ries that were never part of an empire.13 The imperial incorporation or nation-state creation. wave pattern thatwe expect is clearly visible: the Moreover, three-fourths of border changes since rate of war rises dramatically during imperial 1816 have occurred peacefully (Diehl and expansion and is roughly twice as high at its Goertz 1988:115). turning point as in the decades that follow.14 Second, half of our war types and 61% of The solid curve shows wars as they relate to wars in our data set were not fought, at least not the second transformation, to the nation-state, exclusively, over borders. Nonsecessionist civil are this time including all the territories of the wars, both in empires and in nation-states, to our con world. Again, the pattern conforms styl entirely unrelated to border changes. They ized model. The rate of wars increases sharply stitute 38% of all wars in the sample. Interstate as events move closer to the date of nation-state wars may ormay not be fought over territory and creation. Roughly speaking, more than twice the may or may not lead to boundary changes. In number of territories were at war in the imme other words, the empirical pattern that our two diate years around nation-state creation com graphs show must be the result of a more com pared with several decades before or afterward. plex political process than simply a fight over Even at this high level of aggregation, these state borders. graphs show that violent conflict does not occur at a uniform rate across time, and that there is Rates of Onset for Different Types of a of systematic pattern temporal dependence War thatmust be caused by some underlying mech or Our historical model makes anism beyond chance randomness. stylized specific the of war that should To what extent are these findings nontriv assumptions about types with different in the trans ial? Do the two graphs simply tell us that when be associated phases formation We look at these focus ever polities change boundaries, this may process. by on the transformation to nation-statehood for happen through war? We come back to our ear ing war data are available over a lier argument that the two turning points? which longer range of time and offers a substantial number of obser imperial incorporation and nation-state creation?are not the same as the extension or vations.15 5 our contraction of state boundaries. First, in quite a Figure confirms expectations regarding wars. Wars of are less fre number of cases (e.g., Switzerland, Thailand, interpolity conquest once a is a nation Japan, France), the shift to the nation-state was quent territory governed by state and conforms to standards of statehood that gradually establish an uncontested hegemony during the period under consideration. Our 13 with of We also experimented moving averages results confirm the findings of Strang (1991): not 20-, 5-, and 1-year periods. The main pattern does change. 14 Almost half of the wars that constitute a second 15 peak 70 to 100 years after imperial incorporation The number of observations is too small to be war and consist of nationalist wars of secession, which clear significant for most subtypes years during to nation It is sufficient to that the ly relate to the second transformation, imperial expansion. say peak war related to wars of statehood, in many territories. The picture is not in the probability of is mostly assumes. distorted by the fact that our war data start in 1816 conquest?just as our model These become were less after and are no only, whereas many territories incorporated gradually likely incorporation 40 after After into an empire centuries ago. A graph that includes longer significant years incorporation. an the civil wars and interstate wars occur only territories that became part of empire for incorporation, same to time in the random our model first time after 1816 displays the pattern (not from time pattern shown here). predicts. FROM EMPIRETO NATION-STATE 883

WARS OF CONQUEST INTER-STATEWARS

3%- 3%- I ; &> 2%- ! 2%- t_ Shaded areas 1 11 a represent if j| t\ 95% confidence P 11 f I ?><* * interval '1 |- | ! I%~j jjli f '""*J* / IV l*Wrm , .?. s_ 0%-JitCXiaip offlfT?I Jl 1.1 0% ,.,i.TSr Iflf .,,!,..?.? ! ' ' ? ' ' ' ? * ' ' ' ? ' ' ' ' ' >L I i I I ! I I I I I 1 I I I I I I I I I I I O -10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10 *-> I CIVILWARS (SECESSIONIST) CIVILWARS (NON-SECESSIONIST) ? 3%~ 3%~

8 2%~ N 2%- !

H ,,AW IW [:; i 1 1 > i 1 t 1 1 . /v^VnJ_ij[. i . . 1 1 i . . 1 I I I I ; j j f 1 j j j , j j 1 * 1 j j j 1 .10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10

Decades to/from nation-state creation

Figure 5. Nation-State Creation and Types of War (10-Year Moving Averages with 95% Confidence Intervals)

few states, once recognized as nation-states, the modern nation-state form may bring about. have been attacked by other states. The most We were surprised, however, that the overall prominent exception is the attempt by Nazi decline in the war rate is not as steep after Germany to establish an imperial polity. The rate nation-state creation as we had anticipated. We of interstate wars rises dramatically around the had expected that after two generations, the time of nation-state creation?potentially conflict potential that nation-state formation because of the "Macedonian syndrome" of irre entails would have been absorbed through seces dentist claims on the territory of neighboring sion, ethnic cleansings, or various institutional as states?and drops steadily throughout the cen mechanisms such power-sharing formulas, tury thereafter. The war rates for secessionist federalization, or a regime of minority rights (cf. wars show a dramatic peak immediately before McGarry and O'Leary 1993). At closer exam nation-state formation, conforming exactly to ination, the civil wars beyond 100 years after our model. Secessionist wars are much less nation-state creation all are revolutionary wars likely once a nation-state is established and the in Latin America (e.g., the cristero rebellion in nationalist principle of legitimacy has been Mexico, the "dirty" wars inArgentina and Chile, established. the sendero luminoso rebellion in Peru) and in The pattern for nonsecessionist civil wars is Spain (the civil war). more complex. As we expect, the rate rises from In general, Latin America's civil wars are not an average of 0.4% in the years between -150 well described by our model. Despite high and -10 to nearly 2% immediately after the cre degrees of ethnic exclusion inmany countries ation of a nation-state. According to our model, (of theAmerindian population in Peru, Bolivia, this results from the struggles over the ethno Ecuador, Guatemala, and Mexico, or of the national distribution of power that the shift to black populations in many former plantation 884 AMERICAN SOCIOLOGICALREVIEW

societies), none of the civil wars have been splines, or restricted cubic splines, are con fought along these lines. The pathways leading structed of piecewise cubic polynomials such to revolutionary civil wars?not covered by our that the function, its derivative, and its second model?seem to be traveled with considerably derivative are continuous at each of the speci higher frequency in Latin America than else fied knots. In addition, the spline function is where. This finding shows up also in the regres constrained to be linear beyond the end points, sion models that we present in the following simplifying their calculation. Splines are an sections. Ifwe exclude Latin American territo efficient way to estimate nonlinear time depend ries from Figure 5, the war rates do indeed ence, which can be traced easily using the esti decline to what we suspect is their baseline after mated spline coefficients, as we will show later 60 years (not shown here). in this article. Although the patterns of war described in In addition to tracing time dependency, we this section generally conform well to our coded the variable that tracks the central mech model, they might be generated by mechanisms anism linking nation-state formation to war in other than those on which our model is built. The our model: political discrimination against following two sections evaluate this possibility minority groups. We generate a political dis by controlling for other factors previously shown crimination variable from theMinorities at Risk to be strong predictors of war onset. We again (MAR) data set collected by Ted Gurr and col focus on the transformation to nation-statehood laborators.17 Unfortunately, MAR's data begin and look at onsets of civil war and interstate war. only in 1950 and are coded for independent We first introduce our variables along with countries only, so we can test this variable only on a our hypotheses of their expected effects. subset of data after remapping the MAR codings to conform to our territorial VARIABLESAND HYPOTHESIS logic.18 As discussed in the theory section, the degree Dependent variables. Our dependent variables of political discrimination along ethnic lines are civil war onset and interstate war onset. varies under the influence of other factors. First, They are coded 1 in the territory-year that a the governments of rich countries can afford to war begins and 0 otherwise. Because we are discriminate less and co-opt ethno-political interested in war onsets, and because years in protest movements through redistribution. We which a war is already ongoing are not gener expect per capita GDP to be correlated nega ally at risk for another onset, we drop all ongo tively with the likelihood of war. All our GDP war ing years when conducting regressions. and population data come from Maddison

Testing the institutionalist model. Our state creation variable. We tried alternate numbers of main hypothesis, as described earlier, is that knots and locations, but there were no major changes the likelihood of civil wars and interstate wars in the results. to nation-state cre 17 is highest in periods closest The MAR dataset has been faulted for selection ation. This can be contrasted with the null of the dependent variable (Fearon 2003:196) because that war onset is as at hypothesis just likely any it includes only groups that have shown some mini point in time across the institutional history of mal degree of political mobilization and/or have been on Our a territory. To estimate time dependence the substantially discriminated against. assumption year of nation-state creation, we follow the is that all ethnic groups not coded by MAR have neither been discriminated nor advice of Beck, Katz, and Tucker (1998) and use against politically natural cubic with three knots con mobilized. splines 18 For each MAR codes a dis structed as a function of the number of years to group, political crimination score for ranging from and from nation-state creation.16 Natural cubic five-year periods 1 (underrepresentation addressed by affirmative action) to 4 (an apartheid type situation). We create a a single score for each territory by calculating pop 16 of the scores for all minor The splines were created in Stata 9.2 using the ulation-weighted average within that and the score RC_Spline package (http://fmwww.bc.edu/repec/ ity groups territory scaling to 0 to Territories with no MAR bocode/r/rc_spline.ado) with three knots placed at the range from 100. a 0. 10th, 50th, and 90th percentiles of the time to nation groups received FROM EMPIRETO NATION-STATE 885

(2003), who offers the best available GDP esti various periods of the hegemonic cycles that mates for the 19th century and full data for all Modelski andMorgan identified and that should territories except the Soviet and Yugoslav suc show significant differences in the likelihood of cessor states from 1950 onward. Second, not war. only the amount of revenue, but also its source Democratic peace theory refers to the risk of matters. Oil-rich territories should have a high war between democratic dyads. It assumes that er probability of civil war than countries that rely two democratic countries will not go to war cre on taxes for revenue generation because oil with each other, whereas the likelihood of war ates economic rent-seeking behavior that fare between an autocratic and a democratic ethnic clientelism and discrimina encourages country is highest. To make the democratic tion. To measure the of we impact oil, generate peace hypothesis testable with our monadic ter an oil variable based on an production per capita ritory-year data set, we constructed interac historical data various that (Mitchell years) tion variable between a territory's democracy includes annual oil estimates for a production indicator variable and the percentage of its cross-section of comprehensive oil-producing directly contiguous neighbors that also were we construct a states and colonies.19 Finally, democratic. variable to test the effects that our spillover Whereas democratic peace theory refers to model Ethnic civil wars in one ter postulates. static dyads of countries, other models have tend to stir tensions and civil wars in ritory up related war to the process of democratization. territories that harbor similar eth neighboring Mansfield and Snyder (2005) maintain that nic groups, and interstate wars of an irredentist societies experience a higher likelihood of inter nature have a tendency to draw in other states state war in the early stages of democratization that seek to influence the ethnic balance of because the nationalist spirit conjured up when power in their neighborhood. We counted the power shifts to the often is channeled were con people number of wars that ongoing in any toward the outside. Our regime type data are tiguous territory during the same or any of the unfortunately not fine-grained enough to test preceding three years and thus constructed a this theory in a way that does full justice to the neighboring war variable. details of its Still, all There are a of other variables?not argument. democratizing range will be situated somewhere between related to our model?that have been identi regimes autocracy and democracy. Such "anocracies" fied by various authors as crucial to under should thus be more prone to war with their standing the dynamics of war. We have included neighbors than either full democracies or autoc themost robust or theoretically interesting ones racies. and present them briefly below. The democratic civil peace theory states that democracies are better able to solve internal Other independent variables. to According disputes. Autocracies, on the other hand, can the of Modelski and hegemonic cycles theory suppress rebellions by the use of force or by interstate wars occur Morgan (1985), during threatening massive violence. Civil wars should the transition from one world to the hegemon therefore be less likely in democratic and auto next.20 We introduced variables for the dummy cratic societies (Muller andWeede 1990).21 All 19 our variables are based IV Most published studies (e.g., Fearon and Laitin regime upon Polity 2003) use either a dummy variable for oil exporter or calculate the share of oil exports to GDP. However, a an per capita figure represents improved opera wars and small-scale wars involving only two coun as tionalization because it is not dependent on the tries principally different. Because hegemonic strength of other economic sectors, as are the per power is by definition global in its reach, its effects centage of GDP figures, and the risk of collinearity on the conflict behavior should be the same for all is reduced considerably, as compared with dummies states. (cf. Humphreys 2005). 21 A number of quantitative studies have con 20 Originally, the hegemonic cycles theory was firmed this so-called democratic civil peace argument supposed to explain only the major "system wars," (Ellingsen 2000; Hegre et al. 2001). Reynal-Querol such as the two world wars. However, as Pollins (2002), and Sambanis (2001) found similar results, (1996) has shown, there is no reason to treat system but for ethnic wars only. 886 AMERICAN SOCIOLOGICALREVIEW

data, and use the widely adopted cutoffs of+6 person country inwhich the likelihood of war and -6 to identify democracies, autocracies, fare is necessarily 0 and contrast this with a and anocracies. In addition, we created an anar world-state inwhich all wars would be count chy category for territories with no central gov ed as civil wars. Similar thought experiments wars. ernment (including years of interregnum).22 could be made regarding interstate as a Fearon and Laitin's (2003) much discussed Population size is thus included simple wars are control variable. "insurgency model" maintains that not we included ameasure of ethnic and driven by questions of political legitimacy, but Second, fractionalization. The literature is by military opportunity. If government forces are religious inconclusive the role of ethnic frac weak and disorganized, and ifmountainous ter regarding in meas rain allows rebels to hide and retreat, ambitious tionalization explaining civil war.23 The urement leaders will be able to organize a rebellion in captures demographic characteristics, but is a rather indicator of the whatever name: national liberation, fewer taxes, poor political of as Cederman or more self-government; elimination of class dynamics ethno-nationalism, and Girardin have oppression; religion; or straightforward self (forthcoming) argued. enrichment. We included a measure of moun Because it is political discrimination, not ethn that lies at the tainous terrain, previous regime change (which odemographic heterogeneity, heart of ethnic conflict, the number or demo weakens the government vis-a-vis rebels), and shares of ethnic groups should not be change in the repressive capacity of govern graphic to civil war. We include the ment as variables to test theirmodel. The moun expected predict fractionalization measure as a control variable, tainous terrain data are adopted from their data but do not expect significant results. set. Regime change in the previous two years is defined as any shift between regime types (e.g. from anocracy to autocracy, from anarchy to TWO REGRESSIONMODELS democracy), and the repressive capacity is prox The problems finding data on all of these inde ied the percentage in the number of by change pendent variables were considerable once we soldiers relative to the average over government moved further back in time beyond 1950 and, the decade. We thus assumed that rebel previous even more importantly, once we attempted to forces react to the in the number? change find data for former colonies or precolonial rather than the absolute number?of govern areas of the developing world. Still, we were able ment soldiers (taken from COW data). to create a data set with a reduced set of inde Fearon and we should Following Laitin, expect pendent variables (including time to nation rebels to descend from their mountain retreats state creation, regime type, neighboring wars, and as soon as the of the fight military strength and oil production) for all of the world's terri government has been reduced. tories and years. Once we included GDP and we included two control variables. Finally, population, the two most robust variables in the a more First, larger population represents quantitative civil war literature, we had to drop as opportunities for warfare, the following almost half of our observations, excluding most thought experiment illustrates. Imagine a one of Africa and Central Asia before 1950, most of

22 we 23 Collier and Hoeffler For territories not represented in Polity IV, Bates (1999) and (2000) were as autoc find an inverted Sambanis proceeded as follows. Colonies coded U-shape relationship. a linear in that racies because several test codings of individual (2004) reports relationship samples colonies showed that one would never arrive at an include low-intensity civil wars and longer time shows that ethno-reli anocracy score for a colony (for a different approach, spans. Reynal-Querol (2002) is when a see Fearon and Laitin 2003). Dependent territories of gious polarization (which highest popula same score tion is into two of has a linear classic land-based empires received the split groups equal size) were effect on war hazard rates. finds as the imperial center. Independent territories Ellingsen (2000) sup no for a linear to fractionalization coded as "anarchy" if they had central government, port both relationship as and a to In Fearon as "autocracy" if they could be classified traditional U-shaped relationship polarization. fractionalization nor states such as emirates, or as "anocracy" if they were and Laitin's (2003) study, neither elite democracies such as the Swiss confederation. polarization is confirmed. FROM EMPIRETO NATION-STATE 887

19th-century Latin America, and pre-1870 on civil war onset for a range of models.24 All we dis Eastern Europe. When include political specifications include natural cubic splines with we are crimination and ethnic fractionalization, three knots to trace time dependence relative to confined to post-1950 data for independent ter the year of nation-state creation.25 We present ritories only. However, across this wide range of specifications of the regression model, themain results remain as discussed 24 remarkably stable, All logit estimations were performed in Stata 9.2. in the . following. Because standard errors are likely to be correlated for observations within territories, we specify the robust and cluster options to correct the standard errors. Civil War Onsets 25 Our expectation in this project is that temporal a Table 2 presents estimates from logit regressions dependence exists primarily as function of institu

Table 2. Logit Analysis of Civil War Onset

Model 1 Model 2 Model 3 Model 4

1816-2001: Dependent Variable: 1816-2001: Dropped 1816-2001: Civil War Onset All Observations Observations'1 Excluding Some Territories6 Post-1950e Nation-state Formation Time to nation-state creation Spline 1 .0215 .0145 .0202 .028 (.0061)*** (.0042)*** (.0058)*** (.0167)* Spline 2 -.0194 -.0205 -.0374 -.2961 (.0055)*** (.0057)*** (.0124)*** (.0939)*** Political discrimination3 .0147 (.0054)*** Control Variables

Civil wars in neighboring territories .0902 .0529 .0541 .0472 (.0172)*** (.0265)** (.0261)**(.0382) Regime typea Democracy -.3448 -.0349 -.0474 .0702 (.2860) (.2385) (.3170) (.5003) Anocracy .1442 .2657 .0666 .1718 (.1801) (.2112) (.2900) (.3240) Anarchy 1.1079 1.2117 1.2749 2.1135 (.4166)*** (.3778)*** (.4399)*** (.7508)*** Previous regime changea .6935 .6077 .6194 .8081 (.1965)*** (.2082)*** (.2568)** (.3963)** log(%mountains) .26 .3326.3081 .2796 (.0633)*** (.0767)*** (.1157)*** (.1678)* Oil per capita8-0 .0317 .146 .2098.3145 (.0498) (.0652)** (.0672)*** (.0651)*** log(population)ab .3029 .3335 .3759 (.0566)*** (.0871)*** (.1074)*** GDP per capita3^ -.0003 -.0003 -.0002 (.0001)*** (.0001)** (.0001)** Previous change in army sizea -.4475 -.2234 -.3436 (.2732) (.3264) (.9309) Fractionalization Ethnic -1.3875

(.6808)** Religious -.0663

_(.8371)

(Continued on next page) 888 AMERICAN SOCIOLOGICALREVIEW

Table 2. (Continued)

Model 4 1_Model 2_Model 3_Model 1816-2001: Dependent Variable: 1816-2001: Dropped 1816-2001: Civil War Onset All Observations Observations'1 Excluding Some Territories6 Post-1950e

Regional dummies -.3034 -.5368 -.4878 Eastern Europe (.3762) (.3017)* (.3815) Asia .3777 -.2092 -.2364 (.4368) (.3386) (.3902) North Africa andMiddle East .2733 .2449 .3579 (.4016) (.4045) (.4860) Sub-Saharan Africa -.0168 .1231 -.0647 (.3802) (.3942) (.4602) Latin America .3337 .8245 .842 (.3740) (.3657)** (.4609)* Constant -4.3228 -7.0585 -7.007 -7.6666 (.4675)*** (.7421)*** (1.0259)*** (1.1546)***

Observations_24,779_13,707_6,554_2^031_ Note: Huber-White robust standard errors in brackets. a Lagged one year. MnlOOO's. c In metric tons. d Dropped observations due to missing data for most of pre-independent Africa, some 19th century Latin America, pre-1870 Central Europe, some small territories. e Excludes territories with no data on key independent variables in period of nation-state creation. *p

four models, gradually increasing the number ing does not change: the spline functions show of independent variables to minimize the num that across all models, the likelihood of war ber of dropped observations resulting from miss depends significantly on the time to/from state ing data. We therefore include all variables for creation. All else being equal, the likelihood of war in which we have no missing data inModel 1, and increases the decades before nation-state at moment a new introduce the variable that results in the high creation, peaks the nation-state is and recedes thereafter. est number of dropped observations inModel formed, slowly 4. The results demonstrate that even after con How does this basic finding relate to the literature on civil war? It is trolling for a range of important independent existing certainly consistent with Fearon and Laitin's variables and despite variations in the sample (2003) study, which finds that the odds of civil war onset are size as a result of missing data, our main find more than five times higher in the first two years after independence than in the other tional time and not calendar time. When we include postindependence years. They suggest that this war splines, defined as a function of a calendar-year vari supports their insurgency model of civil able for civil war onsets, the calendar-year splines are because they suppose that the departure of the We thus with our civil war insignificant. proceed imperial army provides insurgents with an analysis by focusing only on potential time depend opportunity for rebellion. However, our results ence to nation-state creation. there is evi However, show that the likelihood of civil war increases dence of dependence on calendar time for interstate already well before independence, when the war onsets. In addition to the splines for depend still is in Their ence on time to/from nation-state creation, we include imperial army firmly place. "new state" variable thus half of a set of four hegemonic cycle phase dummies to pick captures only up the possibility of calendar-time fixed effects. the relevant temporal relationship between the FROM EMPIRETO NATION-STATE 889

formation of nation-states and the odds of civil Central Asia as well as much of 19th-century war.26 Latin America and Eastern Europe because very From Model 1 onward, we control for inten little time-series data exist for these regions of sity of neighboring civil conflict, regime type, the world. That said, population is positive and prior regime change, mountainous terrain, and significant across all models, confirming that oil production. The neighboring civil conflict territories with larger populations have a high variable is highly significant and stays thatway er risk of civil war. Consistent with much of the in all our models except Model 4, consistent empirical literature on civil war that uses income with our argument that ethno-national strug as a proxy for various causal mechanisms, we gles in one territory often spill over into neigh find poverty to be a very strong predictor of civil boring territories. Closer investigation shows war onset. According to our interpretation, this that much of this effect is driven by the strong is because the governments of richer territories spillover effects from nationalist wars of inde can react to ethno-political mobilization with pendence (results not presented). This confirms redistribution and co-optation. A recent change our stylized historical model, according towhich in the size of a government's army is not statis nationalist movements spread from territory to tically significant, contrary to the central argu territory and increase the likelihood of armed ment of the insurgency model. This comes as no we struggle against an imperial center. Also con surprise because have already seen that anti sistent with our expectations (and that of other colonial nationalists rose in arms even when explanatory models), we find that oil resources imperial armies were still in place, a fact that is increase the risk of civil war. The results gain overlooked when only independent countries high significance inModels 3 to 5 when many are included in the data set. Overall, our find of the observations of the pre-oil era are ings show the weakness of approaches that fore dropped. The regime variables show no sup see no place for institutional factors and port for the democratic civil peace hypothesis. explicitly reject the idea that the lack of legiti Conditions of anarchy and instability after a macy may contribute to the dynamics of insur regime change are closely related and are both gency. positive and significant, as Fearon and Laitin's In Model 3, we test the robustness of our insurgency model predicts.27 findings by dropping all territories for which we From Model 2 onward, we control for popu lack data on independent variables in the 5 lation size, per capita GDP, and change in the year period centered on nation-state creation. In size of the government's army.We have to drop this way, we make sure that our results are not observations for much of pre-1950 Africa and biased by what statisticians call "left trunca tion." It could be that our results are biased because we include observations for territories 26 only several decades after they became modern Yet another explanation would focus on the post nation-states without any data on war patterns colonial power vacuum in new states and their insti before nation-state creation. Even after tutional weakness, more specifically their incapacity dropping 73 territories for which we lack to politically control armies and militias. However, pre-nation-state no we would again be leftwondering why the likelihood creation data, there is major change in the sig war of civil should increase in the periods before a nificance levels of the variables that we have new, institutionally weak state is founded. The power reported for preceding models. vacuum and institutional weakness argument is not In Model 4, we include the political dis incompatible with our approach: both reinforce a crimination variable to test whether the central toward escalation of conflict when tendency princi mechanism linking nation-state creation to civil ples of political legitimacy shift during the transition war holds. For lack of data, themodel excludes from empire to nation-state. 27 all observations before 1950 and in dependent In contrast to previous research, we find that territories, such as colonies. We also include eth anocracy is not a strong predictor of civil war onset. nic and fractionalization indices, for We believe this is because most definitions clump religious which data are available that reflect the together all midrange Polity scores into an anocracy roughly constellations after 1950. category, including years of interregnum. We believe ethno-demographic it is sensible to separate out interregnum years as The political discrimination variable is strong years of anarchy. ly significant. Higher levels of discrimination 890 AMERICAN SOCIOLOGICALREVIEW

against minorities increase the risk of civil war, predicted rate of war drops off rapidly. This pat suggesting that nation-state formation may lead tern of time dependency is basically the same to wars because of the ethnic discrimination for the other regression models (figures not that it often entails. As expected, the degree of shown).28 ethnic and religious diversity is generally not significant. Interstate War Onsets We also ran a model (results not shown) that Table 3 results from a of includes only those territories that became presents logit analysis nation-states after 1950 and for which we had interstate war onset for our full data set of ter ritories from 1816 to 2001. The data in the years before and after nation-state territory-year structure of our data set is not suited for creation. Analysis of this restricted sample? ideally of interstate wars because studies that basically sub-Saharan Africa, Indochina, and analysis the territories that once formed the communist use dyadic data sets routinely show that coun attributes do not war as well as world?allows for closer scrutiny of potential try-level explain truncation bias in our earlier models. Some the characteristics that describe country pairs. We conduct our with the modest aim right-hand truncation will remain, it should be analysis of for our model and to noted, because we do not observe patterns of war suggesting plausibility encourage future tests research and peace that have not yet occurred, especial using dyadic designs. ly in the recently founded states of the former To see whether the likelihood of war depends communist bloc. In general, the results do not on the stage in the process of nation-state for suggest a large truncation problem in Table 2. mation, we use natural cubic cre Time to nation-state creation remains partially again splines ated as functions of years to nation-state significant, and political discrimination against creation. Model 5 includes independent vari ethnic groups is as powerful a predictor of war ables for which we have data the entire as inModel 4. covering globe from 1816 to 2001. Model 6 adds addi Taken together, these results, from multiple tional control variables, which results in the model specifications that control for a variety loss of observations those for of variables and examine different combina many including pre-independent Africa, much of 19th-century tions of geographic regions and periods, pro Latin America, and pre-1870 Central Europe. In vide strong support for our institutionalist both models, the signs and significance of the model of civil war. Simply looking at the coef spline coefficients are as predicted, tracing out ficients of the variables, however, tells spline wars an increase in during the years approach us little about the shape that the time depend ing nation-state creation, followed by a decline ency takes and when the likelihood of war afterward. Model 7 includes only the 76 terri onset is likely to be highest and lowest. To tories for which we have data on all independ facilitate interpretation of logit coefficients, we ent variables in the five-year period centered on graphed predicted probabilities for each year nation-state creation, thus reducing possible relative to the year of nation-state creation truncation problems. The spline coefficients while other variables were held at their mean are no longer significant at standard levels, but or modal values. Figure 6A presents the pre the signs and relative magnitudes remain sim dicted probability curve for Model 2 along ilar to those of earlier models, suggesting that with confidence intervals. although the smaller number of observations The graph conforms well to our theoretical expectations and the descriptive results pre sented in previous sections. There is a marked increase in the likelihood of civil war onset in 28 We also ran Model 1, for which we do not have the years around nation-state creation. The risk data, for the different cohorts of of civil war in the second decade after significant missing peaks territories that became modern nation-states, usual nation-state creation, with civil wars starting same ly because they emerged from the empire. The in 1.5% of territories. The risk of war is 25% results (not shown here) demonstrate that with the than in the window higher average 50-year exception of the first largely Latin American cohort beginning in the decade before nation-state of the first half of the 19th century, the pattern of time creation. Outside this high-risk window, the dependency is stable and strong for all cohorts. FROM EMPIRETO NATION-STATE 891

A. CIVILWARS (Model 2) B. INTER-STATEWARS (Model 6)

- ~ 3.0%i 3.0% 1 1 i

2.5%~! 2.5%-; ^JHSi

2.0%~ , 20%- 2 :-$?^KIm, ^MI^^^SIMY J^H^^M^,I YMY^^^^^^Bmli

0% i i 1 i i 1 1 i I 1 1 1 1 i 1 t 1 0% T*n.~* 1 i ' 1 ' i ' 1 * 1 ' 1 -16 -12 -8 -4 0 4 8 12 16 -16 -12 -8 -4 0 4 8 12 16

Decades to/from nation-state creation

Figure 6. Predicted Probabilities of Civil Wars and Interstate Wars

contains too much noise for time dependency to cies, because many democratizing polities will emerge, the general pattern appears to be con be categorized in this "anocracy" category. sistent with our argument. Democratization may conjure up an aggressive To make the spline coefficients easier to nationalism that political elites then direct interpret, we again plot the predicted probabil toward neighboring states. Their hypothesis our ities of war onset against time to nation-state complements model because democratiza creation in Figure 6B, on the basis of estimates tion often runs parallel to nation-state formation from Model 6. We note that the spline coeffi and further nourishes the irredentist claims that cients are only weakly significant, but that the the transition from empire to nation-state can shape of the calculated probability of war is as encourage. Table 3 provides some support for expected by our model. The two graphs of the dangerous democratization hypothesis. The Figure 6 thus lend support for our claim that results are weakened, however, as soon as we interstate wars and civil wars might be caused control for per capita GDP inModels 6 and 7. by a similar constellation of factors, a relation A more direct measurement of democratiza ship often overlooked in the specialized litera tion would capture Mansfield and Synder's tures. Struggles over the ethnic distribution of hypothesis in amore adequate way, and perhaps can power in emerging nation-states drive irre produce stronger results. dentist wars that extend across borders, just as The democratic peace theory, tested by an they can drive secessionist wars and ethno interaction term between democracy in a terri political conflicts within states. Conforming to tory and the percentage of democratic neigh this broad historical argument, we also find bors, is once again strongly confirmed, although strong spillover effects for interstate wars, as we our monadic research design is less than perfect did for civil wars: The fight over the ethno for capturing dyadic effects between state pairs. national character of states may draw in neigh The results show that democratic territories are boring territories populated by the same or much less likely to be attacked by?and to related ethnic groups. attack?other democracies. What do the results reported inTable 3 tell us In contrast to our findings with regard to about other theories of interstate war? The "dan civil wars, a previous regime change does not gerous democratization" hypothesis of increase the likelihood of interstate wars. The Mansfield and Snyder postulates that anocrat population variable, included as a simple con ic regimes are more likely to be engaged in trol variable, strongly influences the likelihood or war as interstate wars than democracies autocra of interstate onsets. Also expected, rich 892 AMERICAN SOCIOLOGICALREVIEW

Table 3. Logit Analysis of Inter-state War Onset

Model 5 Model 6 Model 7

1816-2001 Dropped Excludes Some Dependent Variable: Inter-State War Onset All observations Observations0 Territories'1

Nation-state Formation Time to nation-state creation Spline 1 .0162 .015 .0121 (.0047)*** (.0056)*** (.0076) Spline 2 -.0115 -.0157 -.0083 (.0046)** (.0087)* (.0108) Control Variables Inter-state wars in neighboring territories .2013 .155 .1026 (.0288)*** (.0379)*** (.0600)* Regime type3 Democracy .145 .6493 .4428 (.4020) (.3566)* (.4150) Anocracy .5704.425 .1353 (.1868)*** (.2518)* (.3523) Anarchy .8547 1.0649 1.393 (.4196)** (.5419)** (.4974)*** Democracy X % democratic neighbors -3.0307 -6.8142 -8.7757 (1.5851)* (1.9310)*** (2.8671)*** Previous regime change3 .1201 .1225 .5017 (.2487) (.2617) (.3128) log(population)ab .2878 .3212 (.0710)*** (.0682)*** GDP per capita3'5 -.0002 -.0002 (.0001)* (.0001)* Previous change in army size3 .3082 .4606 (.1726)* (.2022)** Hegemonic cycle phases Delegitimation -.4218 .3074 (.3355) (.4133) Deconcentration -. 122 .1105 (.3943) (.5646) Globalwar .27 .7798 (.3472) (.5163) Region Dummies Eastern Europe .2243.1844 .1302 (.3775) (.3577) (.4424) Asia .7738 -.0759 -.2656 (.4250)* (.4436) (.5030) North Africa andMiddle East 1.0465 .6595-. 15 (.3981)*** (.4670) (.5780) Sub-Saharan Africa -.5576 -1.1109 -1.7444 (.5147) (.5713)* (.6662)*** Latin America -.3483 -1.0545 -1.4437 (.4158) (.4765)** (.3966)*** Constant -4.7331 -6.8937 -7.2321 (.4383)*** (.7311)*** (.8066)***

Observations_26,307_14,511_7,169 Notes: Huber-White robust standard errors in brackets. 3 Lagged one year. blnl000's. c some 19th Latin Dropped observations due to missing data for most of pre-independent Africa, century America, pre-1870 Central Europe, some small territories. d Excludes territories with no data on key independent variables in period of nation-state creation. */?<.!; **/?<.05; ***/?<.01 (two-tailedtests). FROM EMPIRETO NATION-STATE 893 er territories are less likely to provide the arena dyads. Our model conceives of the global sys for an interstate war than poor territories, con tem as an arena of institutional diffusion rather sistent with the findings regarding civil wars. We than an integrated and coherent entity with its recur suppose again that richer states can afford to own logic of development. The search for solve domestic ethnic conflicts by means of ring hegemonic cycles and other properties of redistribution, thus reducing the incentives for the global system may obscure rather than illu neighboring states to intervene on behalf of minate the profound change that the institu their ethnic brethren across the border. A prior tional transformation of the system's constituent increase inmilitary personnel is associated with units has brought about. Politics in the age of an increased risk of interstate war, in line with empires followed a quite different logic than pol the expectations of rivalry theory (Vasquez and itics in the current world of nation-states, as a Leskiw 2001), which hypothesizes that states comparison between the British and the U.S. the with a history of saber-rattling and involved in occupation of Iraq may well illustrate. On strategic arms races are likely sites of future con other hand, the dyadic relationships between flict. states are shaped, if we extrapolate from our We find no support at all for Modelski's analysis, by their basic institutional form, and not their To hegemonic cycle argument, which we test in only by degree of democratization. Models 6 and 7 by adding phase dummies to the restate the obvious, empires behave differently wars are not more toward other than do and regression. Interstate fre polities nation-states, to quent during his "global war" phase than they the transition from the former the latter has are when "hegemonic peace" should prevail. It redefined the political character of dyadic rela con between of seems that more important than global tionships states, independently stellations of power are domestic institutional whether these are governed according to dem or transformations that shape the incentive struc ocratic principles not. tures for political actors, including the motives Our model also complements the established of We and opportunities to pursue political ends with findings comparative politics. confirm means. that and oil resources do military political instability matter and make a territory more prone to civil war. we wars are CONCLUSION Should conclude that prima rily driven by military opportunities to rebel Our study shows thatmacro-institutional trans (Fearon and Laitin 2003) or by the greed for oil formations do matter for a proper understand rents and diamond fields (Collier and Hoeffler ing of war. The comparative historical sociology 2000)? We argue that such military and eco of empire building and nation-state formation nomic incentives for war making are shaped provides important insights for the quantitative by larger processes of institutional transforma study of warfare, which has been dominated tion. Once nationalism has spread across a pop by approaches that take the institutional form of ulation, secessionist civil wars against an the independent modern state as a given and imperial center may represent a viable political exclude other types of polities from analysis. In option, especially if the government isweakened contrast, we demonstrate that the emergence of by political instability or feeble institutions. a this state structure provides themacrohistorical Instability, however, represents not prime context within an a which explanatory model of causal factor, but rather circumstantial trigger. war should be situated. More precisely, periods Once a nation-state is established, new incen of transition from one type of political institu tives to protest and rebel and new opportunities tion to another are much more war-prone than to pursue ethno-nationalist goals in civil and periods of institutional stability. irredentist wars are created. Oil resources, which This perspective puts the long-term political can be distributed along ethnic lines and thus can development of a territory at the center of analy be used to consolidate ethnic clienteles, provide sis and provides an important complement to further fuel for the dynamics of ethno-political established theories of war and peace. The stan competition, but they are not their primary dard international relations approach looks cause. either at systemic properties of world politics or Finally, our analysis speaks to the literature at the relationships between competing state on the end of the Cold War and globalization. 894 AMERICAN SOCIOLOGICALREVIEW

Several authors have argued that the wave of communist world that will continue to preoc civil wars that has swept over the globe since the cupy the international community. Conflicts 1970s must be related to increasing levels of emerging from previous waves, such as the globalization. Mary Kaldor (1999) maintains struggle over control of the Iraqi state, have not that the "new wars" of the 1990s have resulted yet found an adequate institutional solution and from two interrelated processes of globalization: will continue to be virulent. Our model does not a the emergence of global weapons market com give much support to the idea that such conflicts bined with the decreasing capacity of states to can be managed easily through "prevention," uphold the monopoly of violence. Amy Chua "peace making," or "reconciliation" (Wimmer (2004) postulates that successful minority 2004). Ifwars are fought over the institutional are to groups likely become victims of violence shape and territorial extension of the state, a suc at the hand of those who lose their in standing cessful intervention policy must address these the stiffer winds of The global competition. issues directly and design institutional solu notorious clash-of-civilization hypothesis tions that decrease the incentives for war-mak related increased levels of (Huntington 1993) ing in a sustainable way. We are not surprised ethnic and conflicts to the end of the religious that this task has proved to be much more dif Cold War.29 ficult than many had thought when the end of By contrast, our analysis shows that what the Cold War turned the world's attention to has been observed in recent decades sim may conflicts across the developing world. ply be more of the same old story. Although his tory never repeats itself, the same process Andreas Wimmer is Professor of Sociology at the patterns may be operating at different times and University of California, Los Angeles. His research to in different historical contexts (cf. Collier and aims understand the dynamics of nation-state for and Mazzuca 2006). The dismemberment of empire mation, ethnicity making, political conflict from and the formation of the nation-state have led a comparative perspective. He has pursued this theme across several on exam towars since the time of The disciplinary fields, focusing Napoleon. patterns both the and the of warfare in the Caucasus and the Balkans in ples from developing developed world, and pursuing various methodological and the 1990s resemble those on the Indian sub analytical strategies: anthropological field research continent in the those of Eastern 1940s, Europe inMexico and Iraq, network studies in Swiss immi and after theWorld War and so on. The during I, grant neighborhoods, quantitative cross-national return of the "Macedonian syndrome," asMyron research on wars, comparative historical analysis of Weiner (1971) has called the intermingling of Swiss, Iraqi, and Mexican nation-state formation, ethnic conflict and irredentist wars, explains and policy-oriented research on the prevention of such recurrent patterns of war much better than ethnic conflict. variant of To treat any globalization theory. Brian Min is a graduate student in the Department as a new them fundamentally phenomenon, of Political Science at the University of California, or uses brought about by the end of the Cold War Los Angeles. He formal models and quantitative increased globalization, represents yet another methods to study the comparative politics of ethnic He holds a B.A. Cornell and an example of the widespread tendency among ity. from University M.P.P the School Government at social scientists to perceive their own times as from Kennedy of Harvard University. unique and exceptionally dynamic (on "chrono centrism," see Fowles 1974). The consoling message conveyed by our REFERENCES research is that there are not many polities left Abbott, Andrew. 1998. "Transcending General Linear to break and to be transformed into a apart Reality." Sociological Theory 6:169-86. series of states. There competing nationalizing -. 2001. Time Matters: On Theory and still unresolved conflicts are, however, many Method. Chicago: University of Chicago Press. waves from the last two major of nation-state Anderson, Benedict. 1991. Imagined Communities: formation in sub-Saharan Africa and the former Reflections on the Origin and Spread of Nationalism. London: Verso.

Bates, Robert. 1999. "Ethnicity, Capital Formation, and Conflict." Working Paper No. 27. Center for 29 Harvard For empirical critiques see Gurr (1994), Russet, International Development, University, Oneal, and Cox (2000), and Chiozza (2002). Cambridge. FROM EMPIRETO NATION-STATE 895

Beck, Nathaniel, Jonathan N. Katz, and Richard Fearon, James D. and David D. Laitin. 2003. Tucker. 1998. "Taking Time Seriously: Time "Ethnicity, Insurgency, and Civil War." American Series-Cross-Section Analysis with a Binary Political Science Review 97:1-16. Dependent Variable." Journal of Political Science Fowles, Jib. 1974. "Chronocentrism." Futures 6:65-8. 42:1260-288. Gellner, Ernest. 1983. Nations and Nationalism. Brubaker, Rogers. 1996. Nationalism Reframed: Ithaca: Cornell University Press. Nationhood and the National Question in the New Gleditsch, Kristian Skrede. 2003. Transnational Europe. Cambridge: Cambridge University Press. Dimensions of Civil War. Unpublished manuscript, Buhaug, Halvard and Jan Ketil Rod. 2005. "Local University of California, San Diego, San Diego. Determinants of African Civil Wars, 1970-2001." -. 2004. "A Revised List of wars between Paper given at the conference on Disaggregating and within Independent States, 1816-2002." the Study of Civil War and Transnational Violence, International Interactions 30:231-62. San Diego, March 7-8. Gleditsch, Nils Petter, Peter Wallensteen, Mikael Calhoun, Craig. 1997. Nationalism. Buckingham: Erikson, Margareta Sollenberg, and Havard Strand. Open University Press. 2002. "Armed Conflict 1946-2001: A New Cederman, Lars-Erik and Luc Girardin. Forthcoming. Dataset." Journal of Peace Research 39:615-37. "Beyond Fractionalization: Mapping Ethnicity Goldstein, Joshua S. 1991. "A War-Economy Theory onto Ethno-Nationalist Insurgencies" American of the Long Wave." Pp. 303-325 in Business Political Science Review. Cycles: Theories, Evidence and Analysis, edited by Centeno, Miguel Angel. 2003. Blood and Dept: War N. Thygesen, K. Velupillai, and S. Zambelli. New and the Nation-State in Latin America. Pittsburg: York: New York University Press. Pennsylvania State University Press. Gurr, Ted R. 1993. "WhyMinorities Rebel: A Global Chiozza, Giacomo. 2002. "Is There a Clash of Analysis of Communal Mobilization and Conflict Civilization? Evidence from Patterns of since 1945." International Political Science Review International Conflict Involvement, 1946-97." 14:161-201. -. Journal of Peace Research 39:711-34. 1994. "Peoples Against the State: Chua, Amy. 2004. World on Fire: How Exporting Ethnopolitical Conflict in the Changing World Free Market Democracy Breeds Ethnic Hatred System." International Studies Quarterly and Global Instability. New York: Anchor Books. 38:347-77. Clodfelter, Michael. 2002. Warfare and Armed Hardt, Michael and Antonio Negri. 2000. Empire. Conflicts: A Statistical Reference to Casualty and Cambridge: Harvard University Press. other Figures, 1500-2000. Jefferson: McFarland. Hechter, Michael. 2000. Containing Nationalism. Collier, Paul and Anke Hoeffler. 2000. Greed and Oxford: Oxford University Press. Grievance in Civil War. World Bank Development Hegre, Havard, Tanja Ellingsen, Scott Gates, and Research Group, Washington, DC. Nils Petter Gleditsch. 2001. "Toward a Democratic Collier, Ruth Berins and Sebastian Mazzuca. 2006. Civil Peace? Democracy, Political Change, and "Does History Repeat?" Pp. 472-489 in Oxford Civil War, 1816-1992." The American Political Handbook of Contextual Political Analysis, edit Science Review 95:33-48. ed by C. Tilly and R. E. Goodin. Oxford: Oxford Hintze, Otto. 1975. "Military Organization and the University Press. Organization of the State." Pp. 178-215 in The Collins, Randall. 1995. "Prediction in Historical Essays of Otto Hintze, edited by F. Macrosociology: The Case of the Soviet Collapse." Gilbert. New York: Oxford University Press. American Journal of Sociology 100:1552-593. Holsti, Kalevi J. 1991. Peace and War: Armed Davis, David R. and Will H. Moore. 1997. "Ethnicity Conflicts and International Order, 1648-1989. Matters: Transnational Ethnic Alliances and Cambridge: Cambridge University Press. Foreign Policy Behaviour" International Studies Howe, Stephen. 2002. Empire: A Very Short Quarterly 41:171-84. Introduction. Oxford: Oxford University Press. Diehl, Paul F. and Gary Goertz. 1988. "Territorial Humphreys, Macartan. 2005. "Natural Resources, and Changes Militarized Conflict." Journal of Conflict, and Conflict Resolution: Uncovering the Conflict Resolution 32( 1): 103-22. Mechanisms." Journal of Conflict Resolution Eisenstadt, Samuel N. 1963. The Political Systems of 49:508-37. Empires. New York: Free Press. Huntington, Samuel. 1993. "The Clash of Ellingsen, Tanja. 2000. "Colorful Community or Civilizations?" Foreign Affairs 72:22-49. Ethnic Witches' Brew? and P. K. 1996. Rivalries and Territorial Multiethnicity " Huth, "Enduring Domestic Conflict during and after theCold War Dispiutes, 1950-1990." Conflict Management and Journal of Conflict Resolution 44:228-49. Peace Science 15:7-41. James D. 2003. Fearon, "Ethnic and Cultural Kaldor, Mary. 1999. New and Old Wars: Organized a Diversity by Country." Journal of Economic Violence in Global Era. Cambridge: Polity Press. Growth 8:195-222. Kalyvas, Stathis N. 2005. The Logic of Violence in 896 AMERICAN SOCIOLOGICALREVIEW

Civil War. Cambridge: Cambridge University California Institute of Global Conflict and Press. Cooperation, San Diego. Lake, David A. and Donald Rothchild (ed.). 1998. The Ratzenhofer, Gustav. 1893. Wesen und Zweck der International Spread of Ethnic Conflict. Princeton: Politik.Als Theil der Sociologie und Grundlage der Princeton Press. University Staatswissenschaften. Leipzig: Brockhaus. Jack S. 1998. "The Causes of War and the Levy, Reynal-Querol, Marta. 2002. "Ethnicity, Political Conditions of Peace." Annual Review Political of Systems, and Civil Wars." Journal of Conflict Science 1:39-65. Resolution 46:29-54. 2003. The World Lewis 1960. Maddison, Angus. Economy: Richardson, Frey. Statistics of Deadly Historical Statistics. Paris: OECD. Quarrels. Edited by Quincy Wright and C. C. Michael. 2005. The Dark Side Mann, of Democracy: Lienau. Pittsburgh: Boxwood Press. Explaining Ethnic Cleansing. Cambridge: Ross, Michael. 2004. "What Do We Know about Press. Natural Cambridge University Resources and Civil War?" Journal of -. 2006. "American Empire: Past and Peace Research 41:337-56. Present." Paper given at the conference Russet, Bruce ML, John Oneal, and Michaelene Cox. Comparative Social Analysis Workshop, 2000. "Clash of Civilization, or Realism and of Los Department Sociology, UCLA, Angeles. Liberalism Deja Vu? Some Evidence.' 'Journal of Mansfield, Edward D. and Jack Snyder. 1995. Peace Research 37:583-608. "Democratization and the Danger of War." Sambanis, Nicholas. 2001. "Do Ethnic and Nonethnic International 20:5-38. Civil Security Wars Have the Same Causes?" Journal of -. 2005. Electing to Fight: Why Emerging Conflict Resolution 45:259-82. Democracies Go to War. MIT Press. -. Cambridge: 2004. "What Is Civil War? Conceptual and McGarry, John and Brendan O'Leary. 1993. Empirical Complexities of an Operational "Introduction: The Macro-Political Regulation of Definition." Journal of Conflict Resolution Ethnic Conflict." Pp. 1^0 inThe Politics of Ethnic 48:814-58. Conflict Regulation: Case Studies in Protracted Skocpol, Theda. 1979. States and Social Revolutions: Ethnic edited Conflicts, by J. McGarry and B. A Comparative Analysis of France, Russia and London: O'Leary. Routledge. China. Cambridge: Cambridge University Press. John and John M. Meyer, Boli, George Thomas, and Strang, David. 1990. "From Dependency to Francisco O. Ramirez. 1997. "World Society and Sovereignty: An Event History Analysis of the Nation-State." American Journal of Sociology Decolonization 1879-1987." American 103:144-81. Sociological Review 55:846-60. Brian R. Various -. Mitchell, years. International 1991. "Anomaly and Commonplace in Historical Statistics. New York: Palgrave. European Political Expansion: Realist and Modelski, George and Patrick Morgan. 1985. Institutional Accounts." International Organization "Understanding Global War." The Journal of 45(2): 143-62. Conflict Resolution 29:391-^17. Thompson, William R. 1988. On Global War: Most, Benjamin and Harvey Starr. 1980. "Diffusion, Historical-Structural Approaches to World Politics. Reinforcement, Geopolitics, and the Spread of Columbia: University of South Carolina Press. War." American Political Science Review Tilly, Charles. 1975. "Western State-Making and 74:932-46 Theories of Political Transformation." Pp. 3-83 in Muller, Edward N. and Erich Weede. 1990. "Cross The Formation of National States in Western National Variation inPolitical Violence: A Rational Europe, edited by C. Tilly. Princeton: Princeton Action Approach." Journal of Conflict Resolution University Press. 34:624-51. Vasquez, John and Christopher S. Leskiw. 2001. Noiriel, Gerard. 1991. La tyrannie du national: Le "The Origins and War Proneness of Interstate en droit d'asile Europe (1793-1993). Paris: Rivalries." Annual Review of Political Science Calmann-Levy. 4:295-316. Pollins, B. M. 1996. "Global political order, eco Weiner, Myron. 1971. "The Macedonian Syndrome: nomic change, and armed conflict: Coevolving An Historical Model of International Relations systems and the use of force." American Political and Political Development." World Politics Science Review 90:103-117. 23:665-83.

Posner, Daniel. 2005. Institutions and Ethnic Politics Wilkenfeld, J. (ed.). 1973. Conflict Behaviour and in Africa. New York: Cambridge University Press. Linkage Politics. New York: McKay. . Raleigh, Clionadh and Havard Hegre. 2005. Williams, Brackette F. 1989. "A Class Act: "Introducing ACLED: An Armed Conflict Anthropology and the Race to Nation across Ethnic Location and Event Dataset." Paper given at the Terrain." Annual Review of Anthropology conference Disaggregating the Study of Civil War 18:401-^4. and Transnational Violence, University of Wimmer, Andreas. 1997. "Who Owns the State? FROM EMPIRETO NATION-STATE 897

Understanding Ethnic Conflict in Postcolonial Diffusion of Nationalist Discontent: A Macro Societies." Nations and Nationalism 3:631-65. Historical, Institutionalist Perspective." Paper given -. 2002. Nationalist Exclusion and Ethnic at the conference Alien Rule and Its Discontent, Conflicts. Shadows of Modernity. Cambridge: University of Washington, Seattle, Washington. Cambridge University Press. Wimmer, Andreas and Nina Glick Schiller. 2002. -. 2004. "Facing Ethnic Conflicts." Pp. 1-20 "Methodological Nationalism and Beyond: Nation a inFacing Ethnic Conflicts: Toward New Realism, State Formation, Migration and the Social edited by A. Wimmer, D. Horowitz, R. Goldstone, Sciences." Global Networks 2:301-34. U. Joras, and C Schetter. Lanham: Rowman & Zapf, Wolfgang. 1996. "Die Modernisierungstheorie Littlefield. und unterschiedliche Pfade der gesellschaftlichen Wimmer, Andreas and Brian Min. 2005. "The Global Entwicklung." Leviathan 42:63-77.