The Long Shadow of Violence: Legacies of Civil Wars and Support for Terrorism in the Basque Country

ABSTRACT This article analyses how events of political violence in the distant past affect the outbreak of terrorism in the present. Civil wars leave a legacy of distrust that can persist through generations, paving the way to violent responses to perceived threats. We claim that part of the explanation for the counterintuitive terrorism phenomenon in a prosperous and relatively egalitarian region such as the Basque Country (and, potentially, other cases of terrorism) lies in the legacies of political violence in the distant past. Those communities where support for Carlism, one of the warring factions in nineteenth-century civil wars, was strong, were more likely to support terrorism a century later. The article also shows that the transmission of these legacies was robust in communities that have remained largely isolated in the century that separates the civil war from the terrorism of the 1970s.


Introduction
In the last decades, social scientists have developed a set of theories on the determinants of political violence.Among these, a recent development links political violence in the present with past traumas, institutions and cultural norms.Besley and Reynal-Querol 1 on the effect of conflicts in Africa during the modern period on civil wars in the postcolonial years, Voigtländer and Voth 2 on medieval pogroms and anti-Semitism in Germany in the interwar years and Fontana et al. 3 on German occupation in Italy during World War Two and post-war political extremism are some examples of how violence in the past extend their shadow towards the present.
In this article, we analyse how civil wars in the past affect contemporary support for terrorism.We show that civil wars leave a legacy of distrust in the communities affected by them that can persist through generations, paving the way to violent responses to perceived threats.We test these ideas in a conspicuous case of modern terrorism, the terrorist group ETA, one of the more long-lasting and successful in the Western World.Its origins lie in one of the wealthiest regions in Spain, the Basque Country, far above the threshold of economic affluence that, according to most of the literature, makes terrorism highly unlikely. 4It is also a reasonably egalitarian region without any significant economically aggravated minority.ETA was created under an authoritarian regime in 1959 but thrived during democracy.
This article contends that part of the explanation for this relatively counter-intuitive outcome lies in the legacies of political violence in the distant past.We show that those Basque communities where support for one of the warring factions in nineteenth-century civil wars was strong were more likely to support terrorism a century after, during the 1970s.The article also aims to shed some light on the content of the beliefs transmitted over time and the mechanisms of this transmission, something that has not been made explicit in most literature.On the one hand, we show that there is a connection between civil wars in the past and deep distrust towards state institutions in the present.Violence in the past has an effect on disaffection and distrust towards state institutions that facilitates the support for violent secessionist groups in certain conditions, such as a perceived threat from the centre to regional autonomy.On the other hand, we also show that the persistence of the legacies of violence was especially strong in communities that have remained isolated from external influences.

Theory and literature review
We claim that political violence in the present can partly be explained by events of political violence in the distant past.Support for terrorism in the Basque Country in the 1970s, we will argue, was primarily concentrated in those places more affected by civil wars in the nineteenth century.Thus, our paper aims to advance in explaining the phenomenon of political violence, emphasising the importance of long-term historical factors.In this sense, it addresses a set of studies highlighting the relevance of past traumas in explaining contemporary outcomes.
Most analyses on the causes of terrorism have highlighted, first and foremost, the relevance of contemporary factors, such as social, economic and political grievances.Poverty is, in this sense, an obvious candidate: terrorism, it is said, is more likely in poor countries and most terrorists are from poor backgrounds 5 or groups economically discriminated, 6 although other studies have found more mixed results. 7In the same line, De Juan and Bank show that the targeted provision of goods in a patron-client relationship reduces the likelihood of insurgency among the beneficiaries of these policies. 8Another variable that seems related to terrorism is the strength of the state.Failed or weak states are more exposed to civil wars, while moderately strong states are more exposed to terrorism, 9 although, again, not all studies concur. 10Lastly, some studies claim that terrorism is more prevalent in democracies 11 because terrorists benefit from civil liberties, 12 democratic governments are more likely to comply with some of the terrorist goals after terrorist attacks 13 or terrorism as a political strategy is fostered by political competition. 14hese analyses have advanced a great deal in understanding terrorism; we know that certain types of democracies and extreme forms of grievances could make societies more vulnerable to terrorism.However, some paramount cases (for their endurance and for the support they enjoyed), such as the IRA or ETA, are not easily explained by political regimes or economic deprivation.We claim that the existence of a legacy of violent struggles in the past can partly explain why some communities recur to political violence when they feel threatened.To advance in this argument, we examine the historical roots of terrorism in the Basque Country, an affluent Spanish region that has harboured since the late 1950s to the early 2010s the terrorist group ETA.The experience of terrorism in the Basque Country is somewhat of a puzzle.It was a wealthy region in the 1970s (around 139 percent of the Spanish average per capita GDP), 15 far above the threshold in per capita GDP above which a community is usually free from domestic terrorism 16 and with high average educational levels, something that increases the opportunity costs of terrorism in rich countries. 17Far from being an economically discriminated minority, Basques live in the most equal region in Spain, with a Gini index similar to Scandinavia.And, yet, the Basque Country has experienced four decades of terrorism that, furthermore, enjoyed the support of around a fifth of Basque voters.Part of the explanation could lie in the unique historical legacy of political violence in the Basque Country.During the nineteenth century, it was the main scenario of two bloody civil wars (the "Carlist wars") where the defence of Basque autonomy, customs and local institutions was at stake.These wars left an imprint in Basque society that partly explains the rise of domestic terrorism in the 1970s.Our research goes beyond the explanation of the Basque case, however.We aim to specify further what type of beliefs transmitted from the past could be associated with political violence, why these beliefs have lasted in some communities and which are the favourable conditions through which beliefs, attitudes or norms prevail into the present.
By stressing the role of historical variables in current events, our paper addresses literature in the social sciences that explain recent outcomes through historical legacies.Besley and Reynal-Querol 18 show that conflicts in African polities in 1400-1700 affect the probability of civil wars in post-colonial Africa.Voigtländer and Voth trace the effect of medieval pogroms in Germany on anti-Semitism in the 1930s. 19According to Fontana et al.,20 the German occupation of Italy during the Second World War explains post-war political extremism, while Fouka and Voth 21 find an effect of German massacres in Greece during the Second World War on the sales of German cars during the latest economic recession.Homola et al. 22 explain current political intolerance and xenophobia in Germany by proximity to former Nazi concentration camps, a result similar to Charnysh and Finkel 23 in the case of support for anti-Semitic parties in Poland.In turn, Lupu and Peisakhin show that descendants of Tartars victimised by Soviet deportation in 1944 hold more hostile attitudes towards Russia. 24ur article can be earmarked within this stream of works.We explain political violence in the present through the legacies of political violence in the past.One of the shortcomings of these works is that they do not explicitly state which are the beliefs and attitudes linking past events with present political violence.One of the mechanisms is the influence of these past political violence events on trust.There is some evidence that civil wars increase political extremism. 25We hypothesise that these effects of civil wars could last over time, affecting especially another variable: institutional trust.Some studies show that traumatic historical legacies can long-lasting impact interpersonal and institutional trust.Nunn and Wantchekon 26 found that slavery in the Modern Atlantic World is connected to levels of distrust in contemporary Africa, and Besley and Reynal-Querol 27 show that out-group trust is lower in local African communities that have experienced conflict in the precolonial past.Grosjean et al. finally found an association between civil wars and declining political trust in Europe and Central Asia. 28n the case of the Basque Country, civil wars in the nineteenth century that, besides its main object (a dynastic struggle between two Bourbon contenders to the Spanish throne), also included among the Carlist's secondary objectives the defence of the region's autonomy against the central state, may have created a legacy of distrust towards the government that can motivate collective action.There is abundant evidence during the nineteenth century civil wars of rhetoric of distrust against the central government in the Basque Country and of defence of their autonomous institutions and laws (fueros) by one of the factions in the war, the "Carlistas."The official Carlist journal during the second war (1872-76), El Cuartel Real, highlighted the threat that the central government posed to Basque autonomy and the duty of the Basque people to take up arms against the state: "the revolutionary spirit aims to equal the Basque provinces to the status of the other provinces of the state"; "liberals have substituted the fueros by the most arbitrary despotism"; "we are the victims of the most tyrannical and oppressive centralisation"; "the sons of the fatherland shed their blood in a hundred battles to defend their sacred fueros." 29Being or not a legacy of the nineteenth century civil wars, it is noteworthy that polls systematically show that trust in the central government in the Basque Country is significantly lower than in other parts of Spain. 30Our argument is that past episodes of violent resistance to the central power could affect the interpretation by members of the affected community of policies from the central government as a threat to its autonomy.The fact that this history is common knowledge among community members makes it easy to assume that pro-independent preferences are widely shared.
In sum, two communities that are similar in a range of variables-such as economic affluence, political regime or educational level-can react differently to an external threat-both in terms of the perception of the relative importance of the danger and in the form of collective action used to cope with it-because the existence of different legacies of violent collective action.
How do these legacies of the past subsist in the present?Community members who have experienced an episode of political violence can transmit information about this to their children.This will make part of their prior beliefs about possible threats to their community and responses to these threats.We can speculate that these priors will be reinforced if the other community members transmit them to their offspring.In this case, these priors will be more useful because they allow coordinating collective action, which is another reason for parents to transmit them to their children.In the case of Carlist values and ideology in the Basque Country, the Carlists were conscious of the role of the family and the community in their transmission between generations.In 1890, the Secretary of the claimant Carlos VII wrote to the Carlist leader (the Marquis of Cerralbo) that "what is key for the transmission of our values is the memory of the deaths in the family, the stories told by the veterans to their sons and by these as examples to their children." 31There is some anecdotic evidence of the survival of these stories into the twentieth century: some Carlist historical figures from the two nineteenth-century wars, such as the general Zumalacarregui and the "Santa Cruz priest" (a guerrilla leader), were presented in the 1970s as Basque heroes and their struggle as a precursor of ETA's fight against the state. 32According to one of the ideologues of HB, the political party attached to ETA, "the war has arrived.A long war which has not ended and that I call the war of the 150 years: the time of Zumalakarregi and the first war; of Santa Cruz and the second war (. ..)And the time of ETA." 33 The preservation of this transmission process varies depending on other variables, especially the degree of isolation of each community.External shocks such as economic modernisation, immigration, and, in general, any openness of the community to external influences can lessen the impact of those legacies in current collective action.The continuity of these legacies is less certain when the community is composed of people from other contexts less exposed to them.Another factor that can diminish the impact of the legacies of past violence is population growth.This is usually a proxy of economic development and modernisation.Besides, coordination problems can be more challenging to overcome as an initially small and relatively isolated community grows larger.Therefore, we should expect an interaction between the degree of isolation of the community and the presence of legacies of previous political violence experiences in explaining current terrorism.

Historical background
During the nineteenth century, the Basque Country was the main battleground of two civil wars (1833-1840 and 1872-1876), the "Carlist Wars." "Carlismo" was a political movement aimed at the instauration of the Spanish throne of a member of a dissident branch of the Bourbon family.After the death of Ferdinand VII (1814-1833), two rival claimants to the throne fought for the succession: his daughter Isabella and his brother Carlos.The country's mainstream political elites supported Isabella's claim.Carlos' supporters ("Carlistas") resisted this.The royal forces quickly crushed most Carlist rebellions in 1833, but in the Basque lands, Carlism largely consolidated.The army's failure to crush the Carlist rebellion in the Basque Country in its cradle led to a violent civil war that lasted seven years.Besides this ostensible cause for the war, the rival forces embodied two different ideological worldviews.Liberals-those favouring centralism, laicism and progressive reforms-supported Isabella, while Carlists defended the binomial "Church and Throne" of the traditional absolutist monarchy.Basque Carlists also defended Basque privileges, autonomous institutions and traditions (fueros) against the centralising aims of the liberal government in Madrid. 34Their ideology, thus, was a mixture of antiliberalism and anti-centralism.
At their moment of maximum expansion (June 1835), Carlists controlled most of the Basque Country. 35In these provinces, the Carlist claimant established a quasi-state with a rudimentary bureaucratic apparatus and a regular army. 36This, however, did not entirely supersede the traditional Basque institutions.The Diputación General (the executive organ of the de facto Basque autonomous government) supervised the enforcement of the Carlist state's laws, ensuring that they were following the fueros. 37fter seven years of a bloody war (according to the Correlates of War dataset, 125,000 people were killed), the Carlists were defeated.An agreement between the military commanders of both armies and ratified by the Spanish Parliament assured the maintenance for the time being of the Basque laws, subordinated to the unity of the monarchy. 38hree decades after the end of the first war, the Carlists started a second war (1872-1876).During the interwar years, the Carlist ideology had largely survived in the rural Basque Country, and several minor rebellions were attempted during the 1850s and 1860s. 39Then, in 1868, a revolution ousted Isabella II from the throne.During the following six years, Spain's political situation became increasingly chaotic.After two interregnum years, the Parliament proclaimed an Italian prince as King Amadeo I in 1870; he abdicated after three years and a Republic was proclaimed.Spain also confronted a rebellion in its main colony, Cuba, and local uprisings by anarchists and federalists.The Carlists took advantage of this situation to start a new war in 1872.Their stronghold was again the Basque Country.At the end of 1873, most of it was firmly in the hands of the Carlist Army. 40In this Second War, even more than in the First, the defence of the Basque autonomy was a priority. 41In 1875, the claimant Carlos VII swore that he would defend the fueros. 42As in the First War, the Basque institutions played a significant role in the government of the Carlist proto-state. 43The anti-Madrid rhetoric was very salient in the Carlist ideology: one of the objectives of the war, according to the official Carlist journal El Cuartel Real was to get rid of Madrid's "colonial administration." 44he defeat of Carlism in the Second War was facilitated by restoring a degree of stability in the centre.The Bourbon monarchy was restored in 1875, local rebellions were repressed, and the Cuban uprising was controlled.By 1876, the Carlists had lost all hope of winning the war, and the claimant crossed the frontier into France.With the Carlists militarily defeated, the government decided in 1876 to suppress the Basque fueros and autonomy one and for all.
Despite the defeat of Carlism in the two wars, Carlist support remained strong in the Basque Country in the 1880s and 1890s.Carlist candidates attended general and local elections, stressing as its paramount aim the restoration of the Basque autonomous legislation. 45Indeed, support for Carlism remained quite strong till at least the 1936-39 civil war; during the Spanish Second Republic (1931-1936), the Carlist movement even experienced a renaissance as a reaction to the revolutionary threat, 46 there was an increased activity of the Carlist circles beginning in 1931, and the Requeté, the Carlist paramilitary militia created in 1907, gained strength in the last years of the Republic. 47During the civil war, the restoration of the fueros was still one of the main aspirations of the Carlists. 48

Data
Our dependent variable is the support for terrorism in the Basque Country during the transition to democracy in Spain.ETA was founded in 1959-during Franco's dictatorship-but was mostly active during and after the transition to democracy.Its most deadly attacks were carried out between 1978 and 1987, during the democratic period.From its beginnings, ETA developed a social movement (KAS) that included a political party, Herri Batasuna (HB).HB was founded in 1978 and was one of the constituent parts of the so-called "Basque Movement of National Liberation," a network of organisations led by ETA which insisted on the need to carry on with the "armed struggle" until Spain recognised the right to national self-determination. 49ETA enjoyed the support of a significant minority of the Basque population; vote for the party attached to it oscillated during ETA's lifetime from a maximum of 19.6 percent in 1987 and 1999 to a minimum of 7.4 percent in 2007 (when some of their electoral lists were illegalised), with a wide geographical variation.
We measure support for terrorism through HB vote across villages in the first elections it attended during the Spanish transition to democracy.Figure 1 shows the geographical distribution of the average HB vote in the elections held in the Basque Country between 1979 and 1980.
As seen in the map, the vote for HB is concentrated in the northern provinces of Guipúzcoa and Vizcaya.Within both provinces, the most substantial concentrations of support are found in the counties of Tolosa, Gernika-Bermeo, Arratia-Nervión, Markina-Ondarroa and Goierri (most of which showed strong support for Carlism in the nineteenth century).
Our second measurement of support for terrorism is the stability of the vote for HB and its successive labels.This captures the loyalty and the continuity in the vote to the political parties attached to ETA from 1979 to the 2000s.It has been measured as the share of the standard deviation over the average support to the HB and its succeeding parties (the so-called "Abertzale Left").The geographical distribution of this indicator is shown in Figure 2.
Our primary independent variable is support for Carlism in the nineteenth century.We use three indicators for this variable.The first one is subscription to the Carlist journal in the First War, the Gaceta de Oñati.This was the most important and widely circulated Carlist journal, whose main objective was to "counter the liberal propaganda." 50Figure 3 shows the distribution of subscriptions by population across villages in the Basque Country in the first semester of 1836.
The second source for Carlist support in the nineteenth century is the vote for Carlist candidates in the provincial elections of 1890-92.This election was close to the second war, and it was the first one with universal male suffrage.As we have said, Carlism subsisted in the Basque Country despite its military defeat.In the 1890s, the movement was re-organised and modernised under the leadership of the Marquis of Cerralvo, and Carlist candidates attended elections. 51Figure 4 shows the vote for Carlism in the 1890-92 provincial elections.
The two maps show a similar pattern.Support for Carlism is widely distributed, but it is mainly concentrated in the provinces of Guipúzcoa and Vizcaya.
Our third indicator is the number of deserters of the Carlist Army during the First War.This data is only available for the province of Vizcaya. Figure 5 reflects the distribution of deserters according to a report issued in 1835 by the Carlist military authorities. 52In principle, villages with a higher number of deserters per capita will be those with the lowest support for Carlism.The figure shows a pattern similar to those of Figure 3 and Figure 4.The areas with fewer desertions coincide with those with the highest vote for Carlist candidates and highest subscriptions to the Carlist journal.
The remainder independent variables capture some of the most relevant alternative explanations of terrorism.Population Growth measures the changes in population between 1857 (the first reliable Spanish census) and 1981.This is an indicator for both economic growth (as we have seen, economic affluence may be related to support for terrorism) and the degree of exposure to exogenous influences, as a large part of the population growth in the Basque Country was through  Victims in 1936 are the number of civilian victims of Francoist repression in the Basque Country during the Spanish Civil War of 1936-39 and the first post-war years, as an indicator of grievances. 55 response to Francoist repression was at the heart of ETA's rationale for terrorist violence.56 The models also include three control variables: Elevation, Population in 1981 and Distance to the coast.The summary statistics for all the variables are shown in Table 1.
Table 2 shows some basic correlations of the abovementioned variables.Vote for Carlist candidates and subscriptions to the official Carlist journal are significantly and positively correlated with HB vote.The relationship is especially strong in the case of the Carlist vote in the 1890-92 elections.There is also a negative association between both indicators and the volatility of HB vote, suggesting that support for terrorism is more stable in those villages with a stronger Carlist heritage.As expected, our third indicator of Carlism, deserters of the Carlist Army, is negatively associated with HB vote and positively correlated with voting volatility.

Results
We have estimated standard OLS regressions with the following general form: where VHB i represents our proxy for ETA support at the village level i, Carlism i XIX , our indicators for support for Carlism in the past, and X i is a vector of control variables.We have also estimated propensity score matching to compare villages with a degree of support for Carlism above and below the median as a way to minimise the problem of unobservables, variables that could affect at the same time support for Carlism in the nineteenth century and support for terrorism in the 1970s.
The results, as shown in Table 3, go in the direction of our theoretical hypothesis.It can be discerned a historical root in the support for terrorism in the Basque Country that goes back to the civil wars of the nineteenth century.Models (1), ( 2) and (3) are OLS estimations for the effect of Carlism on vote for HB.Both the Carlist vote in 1890-92 and subscriptions to the Carlist journal are positively and significantly associated with HB vote in 1979-1980; one percentage point increase in Carlist vote increases around 0.8 percentage points in vote for HB in 1979-80 (Figure 6), while one additional subscription per 1000 inhabitants to the Carlist journal increases support for HB in one percentage point (Figure 7).Our third indicator for Carlist support, the share of deserters of the Carlist Army, also behaves expectedly: there is a negative association between deserters in the First War and HB vote.
The association between Carlism and HB vote in 1979-80 holds even though models (1), ( 2) and (3) include a whole set of contemporary variables that could explain support for HB.The variables associated with political and economic grievances do not affect support for terrorism.Elevation, a variable usually considered a proxy of state infrastructural power, is also not associated with HB vote.Propensity score matching in models (4), ( 5) and ( 6) confirm the OLS results. 57ur second dependent variable measures the stability of support for terrorism.As shown in Table 4, the results are fairly coherent with what we have seen in the previous models.The three indicators of Carlism behave as expected.Two of the three indicators are significantly associated with stability in terrorism support in the OLS models, and the three are in the propensity score matching.
How do we explain the persistence of Carlist legacies across time?We have argued that it could be affected by the degree of isolation of the community.In more isolated communities, less exposed to external influences, beliefs and preferences are likely to be more stable across time.We will next test how several indicators of the degree of isolation of local communities affected the subsistence of Carlist legacies.The first indicator is population growth.This affects the transmission of beliefs and preferences across generations to the extent that part of this growth is due to the inflow of immigrants with different beliefs and preferences.The population growth in the Basque Country since the end of the Second Carlist War was primarily driven by the influx of immigrants from other Spanish regions attracted by the rapid industrialisation of Vizcaya and Guipúzcoa. 58Therefore, the demographic evolution of Basque villages is a good indicator of their degree of isolation.Models (1), ( 2) and (3)  of Table 5 include interactions between our indicators of support for Carlism and population growth from 1857 to 1981.The models show that population growth decreases the effects of support for Carlism on HB vote, although for subscriptions to the Gaceta de Oñati, the interaction, even if it is in the right direction, is not significant.Figure 8 and Figure 9 show graphically how population growth has affected the persistence of the legacies of Carlism.In villages with high population growth, there is  a negative relationship between vote for Carlist candidates and vote for HB, and there is no effect of the share of deserters to the Carlist Army and HB vote, whereas the opposite effect shows up in villages with low population growth.In the latter, vote for Carlist candidates positively affects, and the incidence of desertions negatively affects HB vote.Another way to look at the effects of population growth on the persistence of Carlist legacies is by using as an indicator of the inflow of immigrants to a community the percentage of people in each village born in the Basque Country.We have estimated the models with this indicator, and the results (shown in the Appendix, Table A1) essentially hold.Distance to the coast is another indicator of a community's isolation.Coast villages are more likely to interact with people from other regions and countries through commerce.The industrialisation of the Basque Country led to increased commerce in iron ore and steel with Britain.Therefore, we could expect that distance to the coast as an indicator of exposure to commerce should affect the likelihood of the persistence of the traditions, beliefs and preferences associated with the Carlist wars.Models ( 4) and ( 5) of Table 5 show that this is the case.The interaction terms between distance to the coast and two indicators of support for Carlism in the nineteenth century (vote for Carlist candidates and subscription to the Carlist official journal) are both positive and significant.In those villages close to the coast, the legacies of Carlism were considerably more diluted, and their impact on modern political violence was much diminished.However, this is not the case for our third indicator of Carlist support, deserters from the Carlist Army (model 6).In this case, it seems that the influence of deserters on HB vote is stronger in villages close to the coast, contrary to our expectations.
Finally, models ( 7), ( 8) and ( 9) show interactions with another proxy of isolation, the size of the community.The results show that the impact of the legacy of Carlism on support for terrorism is fainter in larger cities, while in small communities, the traditions of Carlism are more strongly felt.
All four indicators of the degree of isolation of local communities-population growth, the share of the population born in the Basque Country, distance to the coast and village size-have the expected effects on the influence of Carlist legacies on HB vote (except for distance to the coast for deserters of the Carlist Army).According to these four indicators, villages that are more isolated are those where Carlist traditions had a larger effect on support for terrorism in 1979-80.By contrast, the effects of Carlism seem to be much weaker in those villages that were more exposed to external influences in the odd century between the Second War and the 1970s.
In the theoretical section, we have discussed not just the mechanisms that explain the subsistence across time of legacies of the past but the possible content of those legacies.The black box of the explanation is usually left unopened in most studies about the current impact of traumatic past events.We have tried to go further by providing a possible content of the beliefs and norms transmitted through generations.We have argued that Carlist Wars instilled extreme distrust and deep disaffection towards the central state.To see whether this is the case, we test the influence of Carlist support in the nineteenth century on distrust towards the central state.We measure political distrust as the percentage of people in each village that reject the Spanish Constitution.
Table 6 shows the support for the 1978 Spanish Constitution if it were to be approved nowadays, according to Euskobarometer.Those who say that they would vote "no" in a referendum on the Constitution show rejection of the rules and territorial definition that govern the democratic Spanish state.This captures a historical grievance of Basque nationalism: during the Spanish transition to democracy, moderate Basque nationalists asked for the abstention in the referendum on the Spanish constitution, while radical nationalists asked for its rejection.The three indicators of support for Carlism are significantly associated with a negative vote in a possible referendum on the Spanish Constitution.A negative vote is more likely in those villages with a stronger Carlist legacy.
Therefore, we have seen that the legacies of political violence in the nineteenth century are associated with supporting terrorism in the 1970s, even controlling by contemporary causes of terrorism.We have further seen that Carlism left a legacy of distrust and disaffection towards the central government, which is coherent with a wealth of anecdotic evidence on distrust towards the centralising policies of the Madrid governments during the Carlist wars.And finally, we have found that these legacies of past political violence are especially persistent in those communities that have remained more isolated in the odd century that separates the Carlist wars from the 1970s terrorism.Now we will perform some robustness checks.First, we will check whether the results hold with a new dependent variable, the number of ETA activists per county.Second, we have included four variables from the interwar period between the two Carlist wars to test whether pre-existent sociocultural variables-and not Carlist legacies-explain our result.Third, we check whether Carlist legacies explain not just terrorism but support for moderate nationalism.

Carlism and ETA activists
We have measured support for terrorism as HB vote.Now, we will test the correlation between our variables of support for Carlism in the nineteenth century and another indicator of support for terrorism: the number of ETA activists born in each Basque county. 59As an indicator of ETA's support, it is highly correlated to HB vote (r 2 =0.66); unfortunately, we have this measure at the county level, reducing the number of cases to just twenty.However, we have checked for robustness whether our two indicators of support for Carlism in the nineteenth century correlate with the county of birth of ETA's activists arrested by the security forces.
There is indeed a strong correlation between the three variables.Carlist vote in 1890-92 has a correlation of 0.45 with ETA's activists across counties, as can be graphically seen in Figure 10.There is a 0.36 (non-significant) correlation between subscriptions to the Gaceta de Oñati and ETA's activists.

Interwar variables and support for terrorism
As another robustness check to the long-term relationship between Carlist civil wars and support for terrorism in the 1970s, we have included four variables of the interwar period.We are checking whether support for terrorism in the 1970s is explained not by the civil wars of the nineteenth century but by other contemporary variables to the civil wars.These are Basque speakers in 1867, the percentage of industrial workers per village in 1860, per capita wealth in 1845 and literacy in 1860. 62Fractionalisation based on linguistic cleavages is sometimes seen as a predictor of conflict. 63ccording to this argument, what lies behind support for terrorism is not Carlist long-term legacies but the persistence of differentiated Basque linguistic identity, whose origins go back to a previous period.The distribution of industrial workers across villages, literacy and per capita wealth is, in turn, indicators of the economic structure and modernisation.Unfortunately, per capita wealth is not available for one of the provinces (Alava).Some of the models of Table 7 show a positive connection between Basque speakers and HB vote in some of the models.There is a positive connection between Basque speakers and HB vote in some of the models.However, the results show an independent effect of two of the indicators of Carlism-Carlist vote and deserters of the Carlist Army.Subscriptions to the Carlist journal show a positive if insignificant, association.If we include the literacy variable exclusively, excluding the other three interwar variables (not shown in Table 7), subscription to the Gaceta de Oñati remains positive and significant, suggesting that this indicator is not simply a proxy for literacy.Early industrialisation (industrial workers per capita in 1860) is positively associated with HB vote, although the indicators of Carlist support remain significant.Literacy in 1860 reduces support for HB in two of the models, but two of the indicators of Carlist support (Carlist vote and deserters of the Carlist army) remain significant, as is the case for the models that include per capita wealth.There is, therefore, an independent effect of Carlist legacies on HB vote, even after including these four interwar variables.

Carlism and nationalism
A third robustness check concerns the association between Carlism and Basque nationalism, not just the violent type.There is scholarly literature that connects the Carlist's defence of Basque autonomy and privileges with the development of Basque nationalism, especially the moderate Basque Nationalist Party (PNV), 64 although some authors have put into question this association. 65If there were indeed a connection between Carlism and nationalism, then it would be harder to argue that Carlist legacies could partly account for the support for terrorism in the 1970s; instead, Carlism would be the basis of all forms of national movements in the Basque Country, both radical and moderate.We have estimated three models, shown in Table 8, with PNV vote in 1979-80 as the dependent variable and our three indicators of support for Carlism in the nineteenth century as exogenous variables. 66Two out of three of our indicators-Carlist vote and subscriptions to the Carlist official journal-are non-significant.The third one, deserters of the Carlist Army, show that support for moderate nationalism was higher in villages where support for the Carlist Army was weakest.These results clearly show that Carlism is not related to support for moderate nationalism.

Conclusion
The Basque Country was the scene of two civil wars in the ninteenth century.One of the warring factions-the Carlistas-was mainly based in the Basque Lands.We have shown in this article that those communities that supported terrorism in the Basque Country one hundred years after the end of the Second Carlist War were, to a significant extent, the ones who supported Carlism in the ninteenth century.Those places where Carlist vote was especially strong, where the subscriptions to the official Carlist journal during the first war were numerous and where desertion to the Carlist Army was a rare event, were the same where HB (the political party attached to ETA) had its main strongholds.Crucially, this connection between Carlism and terrorism does not hold for moderate nationalism.The relationship between HB vote and Carlism is robust to the introduction of relevant control variables.
Furthermore, we have demonstrated that the persistence of these historical legacies was especially pronounced in villages that mainly remained isolated in the century, separating the two Carlist wars from terrorism in the 1970s.Two villages with the same history of support for Carlism, one of them the recipient of strong flows of immigrants, the other with a stagnant or declining population, will show different effects of Carlism on terrorist support in the 1970s.In the first one, Carlist legacies will mostly fade, while in the second, the connection between Carlism and vote for HB will be pretty robust.
However, it has be acknowledged that these results can be affected by unobservables that could simultaneously affect support for Carlism in the nineteenth and support for terrorism in the 1970s, such as the prevalence of illiberal values.We have tried to minimise this problem with matching techniques, as well as considering as many control variables as we thought that could be theoretically linked to support for nineteenth-century Carlism and twentieth-century terrorism (all of them subjected to the limitation of the availability of historical data).This, and the coherence of the results linked to the theoretical mechanisms of transmission of beliefs and preferences over time, lend some plausibility to the idea that there is a causal connection between legacies of violence in the past and political violence in the present, although the data only allow us to claim for sure that support for Carlism in the nineteenth century and support for terrorism more than one hundred years later are strongly associated.There are, besides, potential alternative interpretations of the data.It may be the case that political entrepreneurs just exploited the legacies of the civil wars of the nineteenth century in the twentieth century to increase support for terrorism (as the abovementioned reference to how ETA considered Carlist heroes of the civil wars as forerunners of its cause).This may be the case, but still, the fact that these arguments found fertile ground in areas of strong Carlist support in the nineteenth century while failing flat in places where Carlist traditions were weaker probably tells us something about the influence of Carlism on support for terrorism in the twentieth century.
Regarding the mechanisms for the transmission of beliefs, an alternative explanation to the one we have provided in the article is that people self-select into isolated or non-isolated communities, both in the past and present.This is undoubtedly a concern; there could be beliefs and preferences explaining both selection of rural versus urban settings, for example, and political beliefs and preferences.Although this could be the case, it is also true that isolated places have not, by definition, experienced inflows of population that prefer rural over urban communities.Another possibility is that people do not move out of their isolated communities because they like this lifestyle due to unobserved variables that also determine their political preferences (in our case, towards political violence).It is difficult with our data to ascertain if this is the case.However, our explanation that the association between Carlist legacies and support for terrorism is stronger in more isolated places because, in these places, there are fewer exogenous influences is at least theoretically plausible and not disconfirmed by the data.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 3 .
Figure 3. Subscriptions to the Gaceta de Oñati (1836).Source: Biblioteca de la Real Academia de la Historia

Figure 7 .
Figure 7. Support for Carlism during the 1833-40 War and vote for HB.

Figure 8 .
Figure 8. Carlism and HB vote for villages with high and low population growth.

Figure 9 .
Figure 9. Deserters of the Carlist Army and HB vote for villages with high and low population growth.

Table 1 .
Summary statistics

Table 3 .
Carlism in the nineteenth century and support for terrorism OLS regressions.Standard errors clustered at the village level.***p < .01**p < .05*p < .1

Table 4 .
Carlism in the nineteenth century and support for terrorism

Table 6 .
Carlism in the nineteenth century and rejection of the 1978 Spanish constitution OLS regressions.Standard errors clustered at the village level.***p < .01**p < .05*p < .1

Table A2 .
Carlist vote during the second republic and vote for HB OLS regressions.Standard errors clustered at the village level.***p < .01**p < .05*p < .1