Russia's invasion of Ukraine on 24 February 2022 came as a rude shock to most western observers. Yet, leading up to the assault, Vladimir Putin had justified it in a series of explicit statements, including an almost hour-long speech three days earlier. Aggrieved by the Russian people's lost unity after the collapse of the Soviet Union, the Russian president presented a nationalist reading of history according to which Ukraine rightly belongs to Russia.Footnote 1
Why was Putin's revisionist narrative not taken more seriously? If rational-choice scholars are prone to dismiss nationalist claim-making as instrumentalist “cheap talk,Footnote 2 realists typically write it off as irrational “hypernationalism.Footnote 3 Mainstream approaches to nationalism analyze nationalist narratives, but their main focus on criticizing these narratives has left their consequences under-studied.Footnote 4
While acknowledging that nations are indeed both modern and socially constructed communities that emerged after the French Revolution, we argue that some historical legacies are more consequential for modern nationalist identities and claim-making than assumed by conventional modernist scholarship. According to our “constrained” constructivist perspective, nationalists typically legitimize their claims by mixing historical facts and fiction, sometimes dating back hundreds, or even thousands, of years. Contemporary examples include Russian revisionism,Footnote 5 the conflicting territorial claims of the Israelis and Palestinians,Footnote 6 China's wish to “reunify” Taiwan,Footnote 7 and Turkey's Ottoman nostalgia.Footnote 8 Such narratives are examples of “restorative nationalism” because they make the case for restoring a past, idealized “golden age.”Footnote 9
Thus, rather than merely debunking nationalist narratives as myths and fabrications, we have to take them seriously. If millions of people share these views, aggressive nationalist projects cannot be written off as irrelevant, even if they are divorced from historical facts or common norms. As illustrated by Russia's invasion of Ukraine, the nationalist worldview has caused massive violence and is likely to do so in the future as well.
Capturing the ethno-nationalist worldview with historical data on ethnic groups and state borders, this study addresses whether the availability of plausible historical “golden ages” has made a real difference for conflict processes in post-Napoleonic Europe. Thus, we attempt to explain how two centuries of domestic and interstate conflict in Europe are systematically linked to almost a millennium of political history.
Geocoded and disaggregated data help us trace how historical border change from CE 1100 on affects conflict patterns in Europe over the past two centuries. Information on historical state borders allows us to identify past polities that could have formed plausible bases of modern territorial claims ultimately leading to armed conflict, both within and between states. Such retrospective projections hinge on more or less imagined links between modern ethnic groupings and their distant ancestors “owning” these polities.
Our objective is to test whether nationalists’ perceived loss of political power and/or national unity compared to some putative golden age correlates with an increase in the risk of conflict. To do so, we identify all past polities that spatially overlap with a settlement segment of an existing or aspiring ethnic nation within a contemporary state from 1816 to 2017. We then assess whether any of these historical states contained significantly larger shares of the ethnic group's total contemporary settlement area than the present-day state does (to capture cases of lost unity) and whether the historical polity was ruled by plausible ethnic ancestors (to code lost independence). The main assumption is that geographic overlap between what are perceived as modern ethnic homelands and historical states makes sweeping claims about ethnic descent and historical ownership more credible.
With this empirical setup, we study the structural preconditions of nationalist claim-making, rather than the ideological narratives themselves. Furthermore, our macro-historical approach forces us to rely on observational data and correlational analysis rather than stronger inferential methods. We deal with risks of omitted-variable bias by using fixed-effects estimation, and we evaluate plausible alternative explanations.
All in all, we find robust evidence that national groups with plausible claims to historically lost home rule and/or unity are more likely to attempt to rectify the situation through rebellion. Likewise, militarized interstate disputes and territorial claims become more likely in irredentist constellations where the leading nationality in one state has co-nationals abroad who have been cut off from a more unified and independent historical polity and are now ruled by an ethnically distinct host government. The effect of lost golden ages holds only for governments that explicitly promote nationalist views and is not driven by alternative mechanisms unrelated to nationalist ideology. Thus we find that constructed modern nationalisms have more violent consequences where they are not entirely fabricated but rely on selected historical facts that are reframed to serve contemporary political goals.
Literature
Pioneering contributions to the dominant modernist school focus primarily on debunking cases of historical fabrication.Footnote 10 By stressing the constructed and historically contingent nature of modern nations, most recent studies of violent nationalist conflict have followed in the footsteps of these seminal contributions.Footnote 11 Rather than seeking their origins in historical legacies, this literature focuses on how modern states shape national identities through nation-building policies, thus conceptualizing not only nationalism, but also nationalist conflict, as fundamentally modern phenomena.Footnote 12
The most prominent theoretical opposition comes from the “ethno-symbolist” school, which, unlike primordialism, accepts the modernity of nations, but insists that national identities derive from premodern ethnic cores.Footnote 13 Rather than dismissing ethno-nationalist narratives, these scholars consider them to be truly consequential in today's world. Despite their suggestive antimodernist criticism, however, ethno-symbolists have weakened their own case by insisting that ethnic cores are “necessary” for the emergence of modern nations.Footnote 14 In contrast, statistical studies of nationalism tend to follow strictly modernist principles in portraying nationalism as the product of nation-states while also using backward-projected contemporary units of analysis that cause hindsight bias.Footnote 15
Statistical investigations of how long-term historical legacies influence modern conflict patterns do exist, but this literature says more about the long-term persistence of violence and state structures than about the link between nationalism and political violence. For instance, Scott Abramson and David Carter show that territorial claims in interstate disputes in Europe after the French Revolution tended to follow prerevolutionary precedents based on historical state borders.Footnote 16 Yet their study covers only interstate claims, views appeals to ethnicity in mainly instrumentalist terms, and does not identify specific subsets of precedents that are particularly well suited for nationalist claim-making. Recent studies covering sub-Saharan Africa highlight the impact of precolonial ethnicity and statehood on contemporary civil conflict, although mostly without reference to nationalism.Footnote 17 Nadav Shelef does address nationalist legacies, yet limits his sample to the impact of “homelands” on interstate disputes.Footnote 18 Using experimental-survey evidence from China, Songying Fang and Xiaojun Li find that “historical ownership” increases support for uncompromising stances in interstate disputes.Footnote 19 Similarly, ethnic minorities in post-communist countries with a history of prior statehood are more prone to launch violent separatist campaigns.Footnote 20 More generally, lost unity has made ethnic groups more likely to engage in civil conflict since World War II.Footnote 21
Research on ethnic groups’ shorter-term legacies of autonomy and statehood shows that they increase the risk of civil conflict. Following pioneering work by Ted Gurr,Footnote 22 recent studies find that lost autonomy makes ethnic entrepreneurs more likely to claim and fight for independence, an effect driven by both grievances and opportunity-related factors.Footnote 23 Other studies have shown that ethnic groups that have recently been “downgraded” through exclusion from executive power in their states are considerably more likely to rebel against the government.Footnote 24
Our study draws inspiration from these recent research streams but advances beyond them in several respects. First, to do justice to nationalist narratives, we present data that go all the way back to the Middle Ages, which by far surpasses the time horizon of previous literature on nationalist grievances.Footnote 25 Second, while most work on long-term legacies has restricted itself to the persistence of conflict and state institutions, our analysis highlights how specific historical reversals are exploited by ethnic nationalists. Third, we show that their grievances are not limited to power losses but extend to fragmented groups seeking to reclaim prior unity. Finally, the current study analyzes both civil and interstate conflict, in contrast to virtually all previous research, which does one or the other.
Improving on the prior literature in these four regards through an integrated analysis enables us to test systematically whether historical structures that could have legitimized nationalist claims increased the risk of armed conflict in post-Napoleonic Europe.
Theoretical Argument
Whether they control their own state or not, leaders of ethnic nations are the main actors in our account of nationalism and conflict in post-Napoleonic Europe. Such actors reason and behave in accordance with the principle of nationalism, which prescribes congruence between the state and the nation.Footnote 26 Deviations from this principle can be expected to generate grievances that, combined with resources and opportunities, increase the probability of violence.
In their attempts to address incongruence, leaders of stateless segments seek to overcome alien rule, while those leading nations that already enjoy state power strive to reverse division by incorporating their kin through state expansion. Alien rule tends to generate grievances that increase the risk of rebellion against the government, typically through secessionist violence.Footnote 27 Divided ethnic groups can trigger interstate conflict if there is competition over which state will lead the unification process. Finally, the combination of alien rule and division characterizes irredentist configurations that may involve civil or interstate conflict, or both. In this case, the actor constellation features a triadic relationship between an entrapped group segment exposed to alien rule by its host government, which in turn may get involved in an interstate dispute with a revisionist kin state.Footnote 28
The key insight of our argument is that both types of revisionist claims do not stem merely from general unhappiness with the current configuration of state borders but are reinforced through historical comparisons. While they are eager to compare their status to other nationalist groups at any point in history, nationalists are particularly obsessed by the historical trajectory of their own nation. Indeed, nationalist mobilization derives major inspiration from stylized, and often embellished, accounts of the nation's history.Footnote 29
Nationalist leaders select their justification from a wealth of historical material, typically mixing facts and myths. These stories often take on a life of their own and may even lead to “rhetorical entrapment.”Footnote 30 Along with poets, philologists, and linguists, historians played a pivotal role in the crafting of these narratives as “national history.”Footnote 31 These national narratives tend to follow a restorative three-step logic:Footnote 32
Golden age. The first step is to identify an idealized “golden age” in the nation's glorious past that is characterized by political and military power, unity, and freedom from foreign influence, and that stands in stark contrast to the current vantage point of a shrunken, oppressed, and/or divided nation.Footnote 33 As mentioned, European nationalists tend to search for such periods in the Middle Ages.Footnote 34 Some histories, however, are much more recent, such as the collapse of the USSR, and some go further back as, for example, Italian and Greek nationalists’ quest for historical greatness in classical antiquity.
Dark age. The second step describes the current period as a “dark age” of oppression and/or fragmentation that brutally interrupted the golden age, leading to collective victimization and status loss. Typically, internal decline or foreign occupation is perceived to have arrested or reversed the nation's cultural and political development.Footnote 35 Examples include the Serbs’ grievances caused by Ottoman domination following the Battle of Kosovo in 1389.
Restoration. The final step features remedial action that promises to restore the nation's greatness, for instance, through national liberation or revisionist campaigns aimed at restoring unity by reincorporating lost territory inhabited by nationalist kin. Today's populist and charismatic politicians, including most prominently Vladimir Putin, are particularly likely to make such promises.
Nationalist leaders seek to overcome alien rule or division by mobilizing for the restoration of a putative golden age in the distant or recent past. To identify sufficiently plausible golden ages, they scan their geographic region's political history. If they find a past polity that can be portrayed as having enjoyed home rule and/or national unity in contrast to the current situation, they may use it as a basis for narratives of a putative golden age.
One main assumption in this regard is that past polities that incorporated at least parts of current ethnic settlements offer more attractive historical material to nationalist entrepreneurs than distant alternatives. First, geographic overlap makes questionable claims about historical continuity and ethnic ancestry more plausible. Rulers of “local” historical states or empires are more likely to have spoken a proto-version or dialect of the language that later came to be seen as defining the cultural boundaries of a modern nation. Second, geographical overlap between past political borders and current ethnic settlements facilitates linking past configurations to a leader's present-day ethno-national goals. Achieving autonomy over or incorporating territory already inhabited by national kin populations appear as more pertinent and achievable goals than resettling allegedly lost territories further away that are now populated by other groups.
We further assume that any past period, however recent or distant, short or long, will suffice as raw material for a potentially convincing narrative.Footnote 36 Ultimately, the goal of revisionist action is to restore national dignity in the form of unity, home rule, or both.
Our main theoretical claim is that nationalist leaders who can rely on such a golden-age polity are more likely to act on revisionist claims than those who are deprived of any “usable history.” Contrasting their group's current predicament with a supposedly more favorable situation in the past facilitates revisionist mobilization through two main mechanisms. First, such comparisons make current deviations from national incongruence appear unnatural and unjust, which fosters grievances and makes nationalist leaders’ claims resonate with broader audiences. Second, historical reference points provide a clear path to restorative action, help coordinate goals and expectations, and make national independence and/or unity appear attainable.Footnote 37
This temporal logic may appear straightforward, but long-term backward projection often rests on questionable historical assumptions. Besides the obvious difficulty of uncovering specific facts in the distant past, the main problem is that the longue durée of these accounts presupposes ethnic groups’ historical continuity despite centuries of migration, intermarriage, and assimilation. Furthermore, restorative nationalists have to assume that their putative ethnic ancestors controlled the polity in question, which is especially challenging in the premodern era because it was dominated by dynastic rather than ethno-nationalist politics.
According to Patrick Geary, this backward projection amounts to dangerous “pseudo history” that views the European peoples as “distinct, stable and objectively identifiable social and cultural units.” Far from being inconsequential, he says, “this pseudoscience has destroyed Europe twice and may do so yet again.”Footnote 38 Yet our analysis in no way tries to evaluate the historical validity of this sweeping retrospective projection of ethnicity. Instead, the primary task is to study whether specific historical configurations are as dangerous as Geary suggests.
The three-step logic carries particularly acute conflict potential in cases where more than one national group claims the same territory, as illustrated by Zionist and Palestinian nationalismFootnote 39 and overlapping claims to Macedonia by the Serbs, Bulgarians, and Greeks.Footnote 40 But the mere existence of a golden-age narrative does not automatically lead to conflict.
Beyond historical grievances, several factors determine whether nationalist leaders will end up triggering conflict, including resources, organization, and opportunities.Footnote 41 But the content of the narratives also matters. To be truly effective, a narrative needs to resonate with a wider audience, which is unlikely if it is entirely invented.Footnote 42 Political activists rely on “injustice frames”Footnote 43 to construct grievances about “robbery” perpetrated by specific outgroups. Such an attribution of blame invests mobilizational efforts with considerable emotional energy.Footnote 44 Nationalist elites regularly employ restorative reasoning as a particularly effective mobilization strategy providing both “diagnosis” and “prognosis” for urgent action.Footnote 45 Whether the leaders in question truly believe this ideology or only use it instrumentally matters less for conflict outcomes. Assuming the latter, recent research shows how states advance territorial claims invoking historical precedents that serve as “focal points” coordinating and facilitating collective action.Footnote 46 Yet, this perspective portrays decision makers as relatively unconstrained in their fabrication of links to ethno-nationalist precedents.Footnote 47 In contrast, we use spatio-temporal backward projection of modern ethnic settlements to identify a more constrained set of historical polities that satisfy the criteria of restorative nationalism.
Linking Historical Configurations to the Risk of Conflict
Each of the two main types of restorative nationalist grievances, lost home rule and lost unity, corresponds to a major violation of state–nation congruence relative to an allegedly more favorable past status. If an ethnic population segment is exposed to alien rule in its host state, the probability of conflict should be higher if the segment experienced home rule in the past.Footnote 48 The same logic applies to division: if the group belonged to the leading segment within its aggregate group, but was subsequently cut off from the majority of its kin, this would also constitute a conflict-inducing grievance.Footnote 49 Finally, both conditions could apply simultaneously.
The first task is to derive the link to civil conflict. Focusing on excluded segments that rebel against their host states, we depict the main configurations in Figure 1, with each row containing two temporal phases corresponding to the status in a past “golden age” followed by the current “dark age.” The three configurations correspond to (1) lost home rule, (2) lost unity, or (3) both.
Three configurations of restorative nationalism causing civil conflict
Cases that involve lost home rule (configuration 1) typically emerge as a consequence of conquest or less violent types of amalgamation. The Croatians fighting to leave the former Yugoslavia in 1991 fall into this category. Their leader, Franjo Tudjman, was a historian and retired general who articulated a restorative claim to independence with references to medieval statehood. While some historians dispute whether there was a distinctive Croatian identity among South Slavs in the Middle Ages, Tudjman proclaimed that the “centuries-old dream of the Croatian people” had been fulfilled through independence.Footnote 50 Other historical cases include groups that rebelled to regain independence after geopolitical reversals, such as the Serbs, Greeks, and Bulgarians in the nineteenth century, who identified various premodern political entities as their respective golden ages before the Ottoman occupation.Footnote 51
Lost unity (configuration 2) may occur without any loss of home rule, if the group in question did not enjoy access to power in the first place. This scenario includes some cases following imperial retraction as, for instance, the Ossetians becoming divided as a consequence of the collapse of the USSR. As Kolstö and Blakkisrud describe it, “The South Ossetians want to heal the partition and reunite the two halves of their nation. Independent statehood is seen as merely a means to this end, and will gladly be given up the moment it has been achieved. In symbolic nation-building this is illustrated by the fact that the flag and coat of arms of South Ossetia are identical to those of North Ossetia.”Footnote 52 Yet this configuration does not occur very frequently, and on its own, lost unity can be expected to generate less conflict than settings in which restoration invokes a politically independent golden age.
In configuration 3, group segments fight the government because they are doubly aggrieved, having suffered loss of both home rule and unity. There are two main subcases (which we will call 3a and 3b), depending on whether the ethnic group loses power in all segments or merely some. Polish nationalism following the partition of Poland in 1795 captures the first subcase (3a) very well. Desiring to regain unity and independence, in the nineteenth century the Polish staged a series of rebellions against foreign rule by the Russians and the Habsburgs. Relatively recent memories of established statehood reinforced the nationalists’ determination to regain sovereignty despite the high costs imposed by the occupying powers.Footnote 53
In other cases, a segment loses home rule while its kin, with which it was formerly united in an independent state, retains power (configuration 3b). This creates a potentially irredentist situation. Harking back to Celtic times, anti-unionist nationalists in Northern Ireland hold compound grievances of this type. Leading up to World War I, nationalist mobilization radicalized against British direct rule, with reference to a “golden age” of medieval statehood.Footnote 54 After the partition of Ireland, the Irish Republic emerged while Northern Ireland remained under British rule. In its armed rebellion against the British state and its unionist settlers, the Irish Republican Army fought for Irish “home rule” and unification of the entire Irish island.Footnote 55
Having analyzed all relevant configurations generating civil conflict, we summarize our theoretical expectations in one main hypothesis:
H1. Lost home rule and/or lost unity increases the probability of civil conflict.
That is, there will be an effect on conflict if at least one of the two types of historical losses applies.
We now turn to interstate relations. Here the relevant units are states rather than segments, although state action concerns a specific co-ethnic group segment abroad. Capturing a situation of lost unity only, configuration 2b is similar to configuration 2 described earlier, but here the group segment in question enjoys sovereignty in both the golden age and the current period. German reunification in 1990 fits this situation, but as in most such cases involving merger of co-nationals, it did not produce conflict. We expect restorative nationalist mobilization to be much more effective when targeted against ethnic others rather than perceived members of the same nation. While competition between the merging units can involve violence between co-ethnic state governments (dashed arrow in Figure 2), we refrain from formulating an explicit hypothesis capturing this rare case.Footnote 56
A configuration of restorative nationalism causing interstate conflict
In configuration 3b, the combination of partly lost home rule and unity can also generate interstate conflict (bidirectional arrow in Figure 2). In this setting, the kin state of the absorbed segment advances claims in support of the segment's autonomy, independence, or outright reincorporation into its own territory. Russia's annexation of Crimea in 2014 and invasion of Ukraine in 2022 are recent examples. Bemoaning the loss of empire and the ensuing disunity of ethnic Russians, president Vladimir Putin justified the incorporation of Crimea and the support for the rebels in Donbas by appealing to restorative nationalism: “I heard residents of the Crimea say that back in 1991 they were handed over like a sack of potatoes … But the people could not reconcile themselves to this outrageous historical injustice. All these years, citizens and many public figures came back to this issue, saying that Crimea is historically Russian land and Sevastopol a Russian city.”Footnote 57
We summarize our reasoning in a second main hypothesis corresponding to each conflict type:
H2. Lost home rule and lost unity increase the probability of interstate conflict.
Empirical Approach
Although our theoretical derivation focuses on ethno-nationalist leaders, throughout the analysis, ethnic group segments serve as the main locus of agency. Rather than assuming that all segments are controlled by nationalists at each time point, we treat them as potential ethnic nations.
Ethno-nationalist conflict results from elites mobilizing in the name of an ethnically defined nation, whether existing or aspirational. This basic assumption does not require that all individuals sharing the respective ethnic marker identify strongly, permanently, or even exclusively in ethno-national terms. Instead, leaders need only to rally enough of their potential followers around their revisionist political goals.
Segments are derived by interacting transnationally defined ethnic group boundaries with state borders. This data structure allows us to measure the degree of state–nation congruence, both within and between states. Our spatial approach to ethno-nationalism is central to the identification of golden ages.
In what follows, we describe in greater detail how we operationalize lost home rule and lost unity in line with the theoretical logic explained earlier. We start by conveying the basic intuition with reference to Poland and Romania, before presenting our data and key variables.
Intuition and Examples
Actor constellations. As we have argued, ethno-nationalist conflict within and across country borders comes in two distinct forms. Intrastate conflicts are fought by politically powerless groups against an ethnically distinct ruling elite. This is illustrated by the Polish population segments under Russian, Habsburg, and Prussian rule in 1863, the year of the Polish January uprising in Russia (A, B, and C in Figure 3). Interstate nationalist conflicts involve a state-leading group, such as the Romanians in independent Romania (A in Figure 4), fighting for co-ethnic territory under foreign rule in another state (B and C in Figure 4), as in 1916, when Romania entered World War I on the side of the Triple Entente, rather than its long-standing allies Austria-Hungary and Germany, in order to gain Romanian-populated territories in Transylvania (B). As such, nationalist rebellions involve groups without home rule that may or may not be united in one country, while interstate nationalist conflict requires division between at least two states and a combination of home rule and foreign rule across country borders. Foreign rule and/or division may in some cases be sufficient to motivate ethnic rebellions against the host government or irredentist campaigns against neighboring states.Footnote 58 However, mobilization seems particularly likely where leaders can stir up more intense grievances by contrasting the already unsatisfactory status quo with a supposedly greater past in their nation's history.
Coding lost home rule and lost unity: Polish example
Lost home rule, lost unity, and interstate relations: Romanian example
Golden-age polities. Nationalist leaders, writers, and historians engage in historical fishing expeditions to identify national golden ages (see the arrows pointing back from t to t ga. in Figures 3 and 4). They search the political history of their geographic region for past actually existing polities that can be portrayed as having achieved ethnic home rule, national unity, or both. Past home rule requires that the ruling elites of the historical state can be viewed as plausible ethnic “ancestors” of the contemporary nation. Claims about past unity gain credence where the past polity contained very large shares of the contemporary ethnic nation's main settlement areas.
In the Polish example, the Polish–Lithuanian Commonwealth at its territorial apex around 1620 (dashed border in Figure 3) clearly satisfies both conditions. The historical capital was Krakow, and Polish kings and noblemen held political power. At the same time, the historical border of the Commonwealth contains the vast majority of all Polish settlement areas in 1863. The Commonwealth thus provides historical raw material to portray it as a national golden age, which is exactly what nineteenth-century Polish nationalists did.Footnote 59
As for Romania, the short “union” in 1599/1600 (dashed border in Figure 4) extended beyond 1916 Romania and comprised parts of the Romanian-speaking territories under Habsburg and Russian rule (B and C). Because the polity was ruled by a Wallachian prince, twentieth-century nationalists could refer to the “Romanian” polity in 1600 as an independent and united golden age and discovered Prince Michael “the Brave” as their national hero.Footnote 60
Combining facts, half-truths, and fiction, nationalists project their contemporary conceptions and political goals onto the selected historical polity. Doing so often involves greatly exaggerated claims about past rulers’ proto-nationalist motivations, historical populations’ group consciousness, and their continuous lineage to the present ethnic nation. The Polish–Lithuanian Commonwealth was far from a modern nation state avant la lettre. Likewise, most Transylvanian peasants were arguably indifferent about whether they were ruled by a Wallachian prince, the Habsburgs, or some Ottoman proxy. The main factual ingredient in these narratives of national golden ages is a historical polity with about the same geographical extent.
Historical comparisons and mobilization. By contrasting their current predicament with such idealized golden ages, nationalists can call for restorative action. Intrastate rebellion against ethnically distinct foreign rule becomes more likely where these historical comparisons reveal lost home rule, lost unity, or both. As interstate nationalist conflict requires political control over at least one independent state, restorative nationalism can lead to international disputes where a contemporary nation rules a state but has a kin segment under foreign rule abroad that has lost both home rule and unity as compared to the historical golden age (like Transylvanian and Moldavian Romanian-speakers under Austro-Hungarian and Russian rule). Thus, irredentist interstate conflict tends to be fought in the name of nations that claim lost unity and partially lost home rule, as in the Romanian example.
Operationalizing Lost Home Rule and Lost Unity
How can one capture lost golden ages beyond individual examples? We use new data on European ethnic settlement areas since the nineteenth century and state borders since the Middle Ages to code plausible golden-age losses from 1816 to 2017. Our data-construction pipeline proceeds in five main steps.
(1) Ethnic settlement data. Information on historical ethnic settlements comes from our newly compiled Historical Ethnic Geography data set, which is based on 200 historical ethnic maps compiled from online map collections and leading libraries such as the British Library, the US Library of Congress, and the Bibliothèque Nationale de France. From this candidate set, we selected seventy-three high-quality maps with high geographic resolution, broad spatial coverage (large subregions or the entirety of Europe), and authors of varying nationality (see Figure A1 in the online supplement).
In digitizing the ethnic maps, we standardize the identity labels across different maps and time periods using the Ethnologue language tree.Footnote 61 We then rasterize all group polygons depicting the same ethnic group in the same time period across multiple maps and combine raster-level information into a best-guess group polygon.Footnote 62 To more precisely encode temporal change in ethnic settlement areas, our procedure also takes into account hand-coded information on large-scale instances of ethnic cleansing and mass migration.Footnote 63
The main goal of our map-based approach is to capture a typically ethno-nationalist worldview rather than historical “truth.” While most map makers aspired to scientific objectivity and aimed to accurately depict existing linguistic population distributions, the actual drawing of some maps reflected implicit nationalist claims.Footnote 64 Overall, these deviations seem marginal and there is major spatial agreement between the maps, as illustrated in Figure A3 in the online supplement. In the border regions and mixed settlement areas where the maps diverge, the statistical averaging of our procedure should further even out extreme views.
More importantly, identifying potential ethnic nations from historical linguistic maps does not require the mass adoption of nationalist ideologies and principles by each and every resident of a specific ethnic segment. Nor do language standardization, mass education, or other nation-building efforts need to have already erased local dialects or attachments. Elite-level convictions and a critical mass of nationalist activists and common followers suffice to mobilize around the restoration of lost national glory. As long as our maps capture existing linguistic boundaries relatively accurately and correspond to what the relevant leaders, activists, and at least some ordinary followers perceived as the cultural bases of their imagined nations, our map-based identification of ethnic nations and past golden ages should yield a valid approximation of the structural potential for restorative nationalist grievances and mobilization. This structural approach omits some ethnic groups that neither developed national consciousness nor mobilized along nationalist lines. In our view, however, the advantages in terms of operationalization and comparability outweigh this drawback compared to further differentiating ethnic groups based on hard-to-measure and likely endogenous information on national identities and mobilization.
(2) Historical state borders. Spatial data covering state borders since 1886 come from the CShapes 2.0 data set, which offers global coverage on all sovereign states and their dependencies.Footnote 65 We extend CShapes 2.0 for Europe back to 1816, drawing on nonspatial data from the Gleditsch and Ward data set of independent states,Footnote 66 the Correlates of War project's Territorial Change data set,Footnote 67 and historical GIS data from the Centennia Historical Atlas.Footnote 68
While our analysis period starts in 1816, coding lost golden ages requires data on state borders that go further back in time. Scott Abramson's data set of historical European state borders covers 1100 to 1790 in five-year intervals.Footnote 69 His data set draws on Centennia, the Nüssli Euratlas, and other sources to identify the borders of states that exhibit direct military occupation, the capacity to tax, and a common executive. Abramson's data allow us to identify potential golden-age polities between 1100 and any year t−1 between 1816 and 2017.Footnote 70
(3) Ethnic segment years. Spatially intersecting the aggregate group polygons e with the borders of European state c in years t yields our main unit of analysis: ethnic segment years (ect) starting in 1816 (see, for example, the three 1863 Polish segments in Figure 3). For each segment year, we calculate absolute area as well as territorial shares in the country and aggregate group in which the respective segment is nested. Wherever ethnic segment or aggregate group polygons overlap, we equally divide the relevant area between intersecting polygon parts.Footnote 71
Most importantly, we assign dichotomous indicators for home rule and national unity to each segment year. Home rule is coded 1 for the ethnic segment that holds the most power in the respective country's capital. The largest ethnic segment that contains the capital serves as our first guess, which we manually corrected when necessary.
The nationalist ideal of unity requires that significant shares of an ethnic group's members find themselves in a common state. Complete national unity has hardly ever been achieved in Europe. The ethnic segment data enable us to calculate the territorial share of each segment in its aggregate group, which may or may not be composed of multiple segments in different states. We therefore define an admittedly arbitrary threshold for national unity, requiring an ethnic segment to hold at least two-thirds of its aggregate group's total territory or population.Footnote 72
(4) Coding lost home rule and lost unity. Indicators for lost home rule, lost unity, and the combination of both for each segment-year ect are derived by comparing values on the home rule and unity dummies in year t to the respective values in all potential golden-age segments between 1100 and t−1. These segments can be constructed by spatially intersecting the contemporaneous ethnic polygon of aggregate group e in year t with all country polygons between 1100 and t−1. We further restrict the set of potential golden-age segments to those that spatially overlap with segment ect, to ensure the geographic plausibility of restorative claims. Going back to the Polish example, the potential golden-age segment in Figure 3 comprises all Polish settlements in 1863 (shaded) within the 1620 borders of the Commonwealth (dashed line). This candidate segment overlaps with all three post-partition segments (A, B, and C) and thus serves as a valid historical reference point for all Polish segments in 1863.
We then assign home rule and unity dummies to all potential golden-age segments in the same vein as for the post-1816 data just described and compare the maximum across all identified potential golden ages to the current value in ect. The powerless and divided Polish segments (A, B, and C) in 1863 are affected by lost home rule and lost unity, since the 1620 golden-age segment comprised more than 67 percent of the 1863 Polish settlement areas, and we code the Polish as the most plausible “ruling group” of the Commonwealth. We repeat this procedure for all segment-years ect and code an indicator for any kind of golden-age loss (that is, lost home rule or lost unity) as well as mutually exclusive dummies for lost home rule only (configuration 1), lost unity only (configuration 2), and both (configuration 3). These indicators serve as the main independent variables in our analysis of ethnic civil wars.
For our analyses of interstate disputes and territorial claims, we slightly adapt the procedure to operationalize configuration 3b (lost unity combined with lost home rule on one side of the border). Since these analyses use country dyad-years as the unit of analysis, it is necessary to aggregate our ethnic segment data to this level. We do so by identifying all dyad-years in which the politically dominant ethnic segment in country c a (for example, segment A in Figure 4) has a powerless kin segment in country c b (B and C in Figure 4) and both of these segments spatially overlap with a potential golden-age segment (here, all 1863 Romanian settlements within the 1600 Wallachian/Romanian borders).
The requirement that both the dominant and the powerless segment in year t overlap with the politically independent golden-age segment in t ga implies at least some division due to past border change. In such cases, lost home rule and lost unity go together. We repeat this procedure for all state-ruling ethnic segments in the post-1816 data to code lost golden ages at the level of directed country dyads. In our Romanian example, this variable is coded 1 for the Romanian–Habsburg and Romanian–Russian dyads in 1916 but 0 for the reverse dyads (Habsburg–Romanian and Russian–Romanian). We use the same procedure to operationalize configuration 2b (lost unity only) involving two co-ethnic segments that govern two independent states but were part of a more unified independent state at t ga. in the past.
Again, our operationalization of nationalist golden ages does not rely on essentialist claims that our candidate segments accurately reflect any kind of ethnic population distribution in the deep Middle Ages. The accuracy of the historical border data is obviously more important, but even here, nationalists project modern notions of territoriality and neat demarcation lines onto geographically much fuzzier political units. We merely assume that the historical border data get the rough contours and spheres of influence of premodern polities right.
(5) Conflict outcomes. This study operationalizes intrastate and interstate nationalist conflict with three distinct outcome variables.
First, we code an ethnic civil war onset dummy at the ethnic segment-year level. For after 1945 we use data from UCDP/PRIOFootnote 73 and the ACD2EPR data set.Footnote 74 We manually match the post-1945 EPR groups involved in conflict to their appropriate counterparts in the map-based list of European ethnicities described earlier. For 1816 to 1945, we identify all civil wars listed in the data sets provided by Gleditsch and by Sarkees and Wayman that are fought in the name of a specific ethnic group.Footnote 75 The coding rules are the same as in the ACD2EPR data set, requiring explicit ethnic claims and recruitment from a particular ethnic group.
Second, the Dyadic Militarized Interstate Dispute (MID) data set provided by Zeev Maoz and his colleagues codes dispute initiation at the level of directed country-dyad years.Footnote 76 The main analyses use fatal MID onset as a dependent variable and therefore drop minor disputes likely unrelated to nationalist revisionism. In additional specifications, we restrict the focus to territorial MIDs because these are most relevant for our theoretical mechanisms and have been shown to be particularly prone to further escalation.Footnote 77 Territorial MIDs are all cases for which Maoz and his colleagues code the dispute-initiating country as pursuing revisionist goals pertaining to territory.Footnote 78 After all, border change and territorial revision are required for national (re)unification.
Third, we investigate territorial revisionism below the threshold of militarized disputes by using territorial claim onsets as an alternative indication of interstate trouble. We focus on identity-based territorial claims only: those in which the targeted “territory includes significant portions of ethnic, religious, linguistic, or other identity groups linked to the challenger state.”Footnote 79 Only this subset of territorial claims appears relevant for our purposes since lost unity and the goal of reunification require co-nationality between the government of the claim-making state and the population of the targeted territory.Footnote 80
Analyses and Results
This section presents our main results. We first consider models at the level of ethnic segments with civil war as the dependent variable, before turning to regression analyses of dyadic interstate disputes and territorial claims.
Ethnic Civil War
Our first set of analyses investigates civil war onset at the ethnic segment–year level. We restrict the sample to all nondominant ethnic segment years between 1816 and 2017 because our ethnic-conflict coding does not involve dominant ethnic groups rebelling against their ethnic peers in power. We present ordinary least squares linear probability models with year and, except for the first specification, country fixed effects. All models control for segment size relative to the state-leading ethnic group, a divided-group dummy for all segments with transborder ethnic kin, a national-unity dummy coded 1 for all segments comprising more than two-thirds of their aggregate groups's territory, and fixed effects for the time since last conflict and the calendar year. The sample restriction to politically nondominant segments and the controls for transborder ethnic links ensure that we capture the effects of our lost-golden-age proxies above and beyond the contemporary values of division and exclusion prominently highlighted in the literature.Footnote 81 In additional models, we extend the set of control variables and add absolute country, aggregate group, and segment size (in km2), country-level ethnic fractionalization, aggregate group-level territorial fractionalization, and distance to the capital. These extended models also include segment-specific civil war history and fixed effects for years since the last border change affecting the host country since 1816 to account for likely sources of recent instability and conflict persistence that may confound our estimates.Footnote 82
Table 1 presents our findings. In a first test of H1, model 1 shows that ethnic segments that can claim lost home rule or lost unity are significantly more likely to rebel. Adding country fixed effects (model 2) and the extended set of controls (model 3) marginally increases coefficient size and results in more precise estimates. In substantive terms, the coefficient in the most restrictive specification (column 3) implies that lost home rule or lost unity is associated with a 146 percent increase from the sample mean of 0.23 ethnic civil war onsets per 100 segment years (see row 1 in Figure 5 for a graphical illustration of effect size). Model 4 includes the disaggregated indicators for lost golden ages that correspond to configurations 1, 2, and 3, respectively. All three variables enter with positive and statistically significant coefficients. By disaggregating H1, we see that lost home rule, lost unity, and the combination of both all increase the risk of ethnic rebellion. As expected, the estimated effect of lost unity only is substantively smaller than for the configurations involving past home rule. Finally, we investigate whether premodern golden ages make a difference beyond the cases of Poland and Romania mentioned earlier. Model 5 distinguishes between relatively recent (post-1816) and historically deep (pre-1816) golden ages. Both estimates are positive and significant, suggesting that our results are not merely explained by short-term revisionism but, as theoretically expected, also reflect mobilization around premodern reference points.
Civil war onset in ethnic group segments, 1816–2017
Notes: OLS estimates of civil war onsets. The unit of analysis is the ethnic segment year. Baseline controls: segment area relative to state-leading group, transborder ethnic kin dummy, national unity dummy. Extended controls: logged country, aggregate group, and segment size in kmFootnote 2; ethnic fractionalization of country and aggregate group; logged distance to capital; war history (past years with ongoing civil war); time since last border change (FE). Standard errors clustered on country (108 AG, 50 country clusters) and aggregate ethnic group in parentheses. *p < .05; **p < .01; ***p < .001.
Overview of main results
Militarized Interstate Disputes
Turning to interstate conflict, we first focus on fatal MID initiation encoded for each directed country-dyad year in post-1816 Europe. We again run linear probability models with and without fixed effects for country c a (initiator) and country c b (target). All specifications control for important baseline variables, including relative size of c a versus c b, absolute size of c b, a neighboring-dyad dummy, and logged minimum distance between c a and c b. All models account for transborder ethnic links from the governing group in country c a to country c b.Footnote 83 As a result, the coefficient of interest again separates the effect of golden-age loss from any independent impact of contemporary ethnic division. Additional specifications expand the set of control variables, adding the share of the dominant ethnic group in c a that is present as a kin segment in c b, the share of that group in its “own” state c a, the absolute size of the state-leading aggregate groups in c a and c b across all their constitutive segments, war history, area-based ethnic fractionalization scores of c a and c b, and border duration fixed effects for both countries.
The results are summarized in Table 2 and reveal positive and significant coefficients for lost home rule and lost unity across all four specifications. The specification with fixed effects and all controls (column 3) suggests that this particular type of golden-age loss is associated with a 320 percent increase in the risk of fatal MID initiation from the sample mean (see row 2 in Figure 5). These findings confirm that lost unity combined with partially lost home rule within a country dyad makes dyadic conflict more likely, as predicted by H2. The separation of modern from premodern golden ages results in similarly sized and statistically highly significant coefficients (model 4). As theoretically expected, lost unity only does not lead to more disputes between co-national state governments. With the exception of the premodern coefficient in column 4, the estimated coefficients remain small and statistically indistinguishable from zero.
Fatal MID initiation in directed country dyads, 1816–2014
Notes: OLS estimates of fatal MID iniation. The unit of analysis is the directed country dyad year. Baseline controls: relative territorial size of state A vs. state B; logged absolute size of country B; indicators for whether governing group in A has governing or powerless kin segment in B; dummies for peace and calendar years. Extended controls: logged aggregate group size of governing segments in A and B; ethnic fractionalization of A and B; share of aggregate group governing in state A located in state B; share of aggregate group governing in state A located in own country; conflict history (number of past years with ongoing MIDs involving A and B); time since last border change involving A or B (FE). Standard errors clustered on dyad, state A, and state B in parentheses (60 country A/B, 2,954 dyad clusters). *p < .05; **p < .01; ***p < .001.
Territorial Claims
The second analysis of interstate conflict uses identity-based territorial claim onset as the outcome variable. Irredentist territorial claims can be seen as a first step toward armed confrontation but may or my not escalate to the level of violent MIDs. The unit of analysis is again the directed country dyad-year, as we know the state that claims territory from its counterpart. The baseline and control variables are equivalent to the previous MID analysis. All estimates for lost home rule and lost unity reported in Table 3 are large, positive, and significant, again supporting H2. The substantive size of the lost-golden-age coefficient in model 3 amounts to a 413 percent increase in the probability of identity-related territorial claims as compared to the average across dyad-years (row 3 in Figure 5). The temporal disaggregation into separate coefficients for historically recent and deep golden ages reveals that both matter and that, if anything, the latter category yields stronger results.
Ethnic territorial claims in directed dyads, 1816–2001
Notes: OLS estimates of territorial claim initiation (identity-based claims). See the note to Table 2. Standard errors clustered on dyad, state A, and state B in parentheses (59 country A/B, 2,820 dyad clusters). *p < .05; **p < .01.
Taken together, these results provide strong support both of our hypotheses and show that historical reference points that can be portrayed as golden ages in urgent need of restoration make a clear difference for violent nationalist mobilization in both intrastate and interstate conflict.
Scope Conditions: Systemic Instability and Nationalist Governments
While they do establish strong links between structural historical configurations and conflict risk since 1816, our analyses so far provide little insight on the timing of nationalist claim-making and mobilization.Footnote 84 In addition, there are many more cases with plausible golden-age polities in our sample than there are conflict onsets. In this section, we explore international and domestic scope conditions to more precisely pin down when and where to expect restorative conflict.
First, we expect nationalist leaders to perceive periods of systemic instability as windows of opportunity to achieve their revisionist goals. Scott Abramson and David Carter show that most of the territorial claims in Europe occur during periods in which the European great powers face domestic or international turmoil distracting them from upholding the regional international order.Footnote 85 They provide a summary measure combining information on annual changes in European great powers’ military capabilities, inflationary crises, civil wars, interstate wars, proxy wars, and shifting alliances.Footnote 86
We extend the fully specified dyadic baseline model in Tables 2 and 3 with an interaction between lost home rule and lost unity and a dichotomous indicator identifying all dyad-years with above-median values of regional instability for country A. All dyad-years involving major European powers are dropped to ensure that the outcome variables do not overlap with the great-power-based instability measure. Results for fatal MIDs and territorial claims are reported in columns 1 and 3, respectively, of Table 4. The constitutive term of our lost-golden-age variable shrinks in size and becomes statistically insignificant in both models. The interaction terms, however, indicate positive, statistically significant, and substantively very large marginal effects in years of regional instability. In line with our expectations and the findings of Abramson and Carter, nationalist leaders act on their revisionist goals primarily during strategically favorable windows of opportunity.
Golden ages and interstate conflict: timing
Notes: OLS estimates of fatal MID iniation (columns 1 and 2) and identity-related territorial claim onset (columns 3 and 4). The unit of analysis is the directed country dyad year. The baseline and extended control variables are equivalent to those in Tables 2 and 3. Standard errors clustered on dyad, state A, and state B in parentheses. +p < .1; *p < .05; **p < .01; ***p < .001.
Second, official government ideology should affect when and where lost golden ages predict revisionist interstate conflict. Throughout this article, we have argued that nationalist leaders are the most important actors in crafting restorative narratives that call for violent mobilization. As a result, available historical reference points should matter more where governments share or at least instrumentally use nationalist ideologies to legitimate their actions. If our historical golden-age proxies similarly affect conflict initiation by non-nationalist governments, we may have to worry that mechanisms other than nationalist mobilization frames and grievances are driving our results. The Varieties of Democracy database provides country-year data starting in 1900 on whether state governments promote specific ideologies to justify their rule.Footnote 87 We use these data to identify all dyad-years in which the government of potential conflict initiator A explicitly promotes nationalist ideology and interact the corresponding dummy variable with our proxy for lost national golden ages.
The results, reported in columns 2 and 4 of Table 4, suggest that lost golden ages mainly predict fatal MID and territorial claim onset when challenger state A has a nationalist government. The constitutive term of lost home rule and lost unity is significantly smaller but still significant for MIDs (column 2) and statistically indistinguishable from zero for identity-based territorial claims (column 4). Both interaction terms and relevant marginal effects are large and significant. We interpret these results as further evidence that ideological narratives and mobilization around restorative nationalist projects, rather than any alternative causal mechanism, account for the strong association between plausible golden-age configurations and interstate conflict initiation.
Alternative Explanations and Robustness
Despite the encouraging findings presented in the previous section, we lack systematic data on nationalist claims and narratives that would allow a direct test of our postulated causal mechanism. It thus remains important to address alternative explanations for the association between past border configurations and conflict. There are three main challenges in this regard, which we discuss in turn.
Structural legacies. A long history of statehood or unity may have endowed some ethnic groups with institutional legacies or social norms that facilitate political mobilization, including but not limited to conflict.Footnote 88 If this were the case, mobilization around very short-lived golden ages, as in the Romanian example, should be rare. Instead, we would expect the historically accumulated experience of home rule or unity to explain conflict. In additional specifications, we thus add two controls capturing the share of years between 1100 and t−1 with plausible statehood or unity (Tables A1, A2, and A3 in the online supplement). Our main results remain stable, suggesting that even short golden ages matter.
Persistent instability. Historical conflict may affect our golden-age proxies through border change and at the same time cause recurring instability.Footnote 89 Rather than identifying the violent effects of nationalist claims, we might just capture conflict persistence and regional clusters of instability. The foregoing analyses partially address this issue by including past conflict, border duration, and country fixed effects, but all of these terms rely on post-1816 data and may themselves be endogenous to earlier instability. We therefore add historically deeper conflict lags based on battle locations (1000–1800)Footnote 90 and interstate conflict (1400–1790).Footnote 91 The main findings remain unaltered (Tables A4, A5, and A6).
Non-ethnic revisionism. Territorial conflict between states may entail attempts to recover lost territories for political, geostrategic, or military reasons that are unrelated to irredentism and the goal of national unity.Footnote 92 To account for such ethnically “color-blind” revisionism, we identify all territories historically held by country A or B or any of their predecessor states and code the largest area ever observed that was once part of state A but is now located in state B. Controlling for this territorial-revisionism proxy in the MID and territorial-claim specifications yields almost identical results as our baseline analysis (Tables A7 and A8).
Additional robustness analysis. First, we restrict our outcome variables to the theoretically most relevant subsets of secessionist civil wars (Table A9) and territory-related fatal MIDs (Table A10). As expected, results remain similar or get stronger when focusing on explicitly territorial violence within and between states. Second, we replicate all main specifications with population-based instead of purely territorial predictors. Lost unity, relative and absolute size controls, and fractionalization scores now rely on aggregated population rasters rather than mere area computations. The results remain similar (Tables A11 and A12). Third, we use alternative segment share thresholds (0.5 and 0.9 instead of two-thirds of the aggregate group's total territory) to define national unity in the civil war models (Tables A13 and A14). Results remain robust except for the lost unity only coefficient with the 0.9 threshold. Finally, we modify the ethnic polygons underlying our units of analysis and main explanatory variables. On the one hand, we use the earliest available map of each ethnic group to rule out our findings being due to endogenous temporal change in ethnic settlement patterns (Tables A15, A17, and A18). On the other hand, we construct maximalist ethnic polygons, assuming that nationalist leaders and activists can portray all areas that at least one map depicts as populated by their ethnic group as national homelands (Tables A16, A17, and A18). Finally, Tables A20, A21, and A22 display the main models estimated with a logit link rather than with linear probability. All results remain robust to these modifications.
Conclusion
In this article, we have taken ethnic nationalists at their word, not because we share their views or believe that their claims always correspond to the truth but because they have momentous consequences. These consequences include both internal and external conflict in Europe since the early nineteenth century. In fact, a good case can be made that the nationalist narratives analyzed here contributed importantly to both world wars by destabilizing European state borders, especially those of Germany.Footnote 93
What do our findings imply theoretically? First and foremost, the results show how specific uses of history increase the risk of violence. Because mainstream constructivist research focuses on highlighting historical contingencies and inaccuracies, it has less to say about the factual component of nationalist claims. Clearly, not everything is made up in nationalists’ backward-looking narratives. Indeed, our analysis shows that the structural availability of golden ages in the nation's past is statistically associated with an greater risk of conflict in post-Napoleonic Europe. Although myth-busting undoubtedly serves important historiographic and normative functions, scholarship on nationalism needs to pay more attention to the actual impact of nationalist narratives and how nationalists made, and are still making, selective—but not random—use of history.
Furthermore, dismissing nationalist narratives as mostly fictitious and irrelevant exaggerates the extent to which the modern world constitutes an abrupt break with the past. This tendency is also present in developmental theories that relegate empires to the dustbin of history once they have been irreversibly superseded by modern nation-states.Footnote 94 This perspective is blind to the impact of imperial legacies on the argumentation of contemporary nationalists.Footnote 95 Indeed, modern multi-ethnic states may not differ sufficiently from empires to write off imperial rule as an anachronism, at least to the extent that these states try to enforce central dominance in their relations with the periphery. Indeed, neo-imperialism is particularly visible in Moscow's current war of aggression in Ukraine.
Nationalist mobilization is a fundamentally modern phenomenon that emerged in the nineteenth century, but rather than being entirely “invented” or “imagined,” many national identities derive at least partly from deep historical legacies. It does not follow, however, that premodern ethnic communities produced modern nations in a deterministic one-to-one relationship. Some “ethno-symbolist” critiques of mainstream approaches to nationalism overemphasize the continuity of premodern ethnic cores by insisting that they underpin all modern nations.Footnote 96 But while there have been cases of actual long-term ethnic persistence in specific cases,Footnote 97 our argument does not hinge on such a correspondence. As we have argued, the impact of nationalism does not require historically verified continuity from early stages of history to today's world. All that is needed is that the claims of prior statehood and unity appear plausible to key political actors and audiences.
By taking nationalists’ historical claims seriously, our study also shifts the attention from what they say and think to the behavioral consequences of their words. In this sense, we follow Mark Beissinger, who stresses the need to consider nationalisms that “bite,” and not only those that “bark.”Footnote 98 Indeed, there are few consequences that are more “biting” than warfare, although populist nationalism also poses a major threat to democracy and the rule of law.Footnote 99
Finally, it should be reiterated that dissecting the impact of nationalist narratives does not mean endorsing nationalist worldviews or policies. Our research does not imply that nationality problems can or should be solved through territorial unification of divided nations or partition of multi-ethnic states.Footnote 100 Robust evidence shows that resolving conflicts within existing borders based on power sharing and territorial autonomy offers a more sustainable path to peace.Footnote 101
It is unclear what repercussions restorative nationalism may have in the future, whether in Europe or beyond. Since World War II, there has been a decline in the scope of territorial claimsFootnote 102 and conquest.Footnote 103 Rather than contradicting the restorative logic, this trend in part reflects the increased congruence of ethnic geography and political borders in Europe over the past 200 years,Footnote 104 especially because plausible restorative claims are constrained by current-day ethnic settlements.
That said, the Russian invasion of Ukraine in 2022 signals that the liberal world order is increasingly under attack.Footnote 105 With Chinese nationalism becoming more assertive, territorial claims to Taiwan also increase the risk of armed conflict. Furthermore, it would be a mistake to believe that in the Balkans, the nationalist genie has been put back into the bottle.Footnote 106 Hopefully, however, a better understanding of restorative nationalism will make future geopolitical shocks less surprising and help policymakers counter such challenges.
Data Availability Statement
Replication files for this article may be found at <https://doi.org/10.7910/DVN/JDKA3W>.
Supplementary Material
Supplementary material for this article is available at <https://doi.org/10.1017/S0020818324000122>.
Acknowledgments
Earlier versions of this article were presented at UC Berkeley, 4 April 2024, University of Haifa, 9–12 January 2023, the European University Institute, Florence (22 June 2022), Princeton LISD and IPERG-UB Workshop on National Identity (University of Barcelona, 30–31 May 2022), the Borders and Boundaries Workshop at the University of Pennsylvania (28–29 April 2022), the Sociological Research Colloquium at University of Bern (24 November 2021), the annual convention of the American Political Science Association (Seattle, September 2021), the University of Florence (October 2021), and the Immigration Policy Lab (ETH Zürich, 9 November 2021). We thank the participants of those presentations, as well as the anonymous referees and the editors of this journal, for excellent feedback.
Funding
We acknowledge generous financial support from an Advanced ERC Grant, no. 787478, Nationalist State Transformation and Conflict (NASTAC).
