Abstract
This article explores the impact of war-related traumatic experiences on political identities and political behavior by exploring different pieces of empirical evidence from the Spanish Civil War (1936–39), the Franco dictatorship (1939–75), and its aftermath. On one hand, the author analyzes semistructured interviews of survivors of the civil war and the dictatorship; on the other hand, she assesses data from a specialized survey implemented on a representative sample of the Spanish population. The analyses broadly suggest that, controlling for family leanings during the civil war, victimization experiences during the civil war and the subsequent dictatorship lead to the rejection of the perpetrators’ identities along the political cleavage that was salient during the war (i.e., left–right). The survey analysis also indicates that (1) although grudges related to severe wartime violations are transmitted through generations, moderate wartime violations do not have such a long-term political impact; (2) the political effects of victimization do not increase with proximity to the traumatic events (i.e., age); and (3) victimization experiences do not have a significant impact on identities along cleavages that were not salient during the war (i.e., center–periphery).
“How do civil wars intersect with elections? Is there a trademark breakdown dynamic? Are party identifications distinctively hardened by civil wars?” 1 David Mayhew suggests that academics have devoted insufficient time to addressing these analytical questions. In fact, very little research has been conducted on the consequences that events such as civil wars have on political identities. At the same time, in their comprehensive review of the literature, Blattman and Miguel argue, “[T]he social and institutional legacies of conflict are arguably the most important but least understood of all war impacts.” 2 This neglect overlooks the causal force of events such as civil wars. 3 This article aims to provide theoretical and empirical insights about patterns of continuity and change of political identities in countries that have undergone civil wars. The issue is addressed by exploring subnational variation in Spain, focusing on political identities of individuals before the Spanish Civil War and long after it (i.e., in the democratic period, post-1977) and taking into account the period in between these dates, in which there was a four-decade dictatorship under General Francisco Franco (1939–75). The article explores the identities of individuals who witnessed the civil war and endured victimization. It also examines identities of people who did not directly have these experiences but who learned about them from their parents and/or grandparents (resulting in an “imprint” of victimization that extends beyond one generation).
In this article, instead of presenting a fully developed theory, I undertake an empirical examination of these issues by drawing on original data. My interest lies in understanding the long-term effects (i.e., several years, or even decades, after) of victimization, as opposed to short-term effects (i.e., a few months or years after). Data availability has led researchers to focus on the short-term effects, when studying political attitudes, 4 attitudes toward violence, 5 intergroup trust, 6 or social risk and time preferences of victimized populations. 7 Hence, long-term effects have to some extent been neglected. The findings in these works have broadly pointed toward a “positive” effect of wartime victimization on political participation and intergroup trust: victimized people have been found to be more politically active and to display greater trust toward other citizens, as compared to nonvictimized people. These conclusions, which are counterintuitive, are not necessarily robust to the passage of time: we do not know how long these effects persist and therefore if they have any implications for political systems (i.e., beyond the postwar scenario). In addition, because these studies are based on countries that did not hold elections in the prewar period (e.g., Sierra Leone, Indonesia, Liberia, Burundi), most of this recent research can barely help us to understand the transformative effect of civil wars at the political level (i.e., we cannot compare postwar identities to prewar ones).
Following Costa and Kahn, who say that “by examining the past we can determine whether a phenomenon is transient or long-lasting,” 8 I argue that only by analyzing civil wars that have taken place in the distant past can we make a real attempt to answer the question of how wartime victimization affects political identities. At the same time, the case under analysis here is one in which democratic elections took place in the prewar period (only six months before the war), so there are results that we can compare with those in the postwar and postdictatorship periods. Furthermore, because of the specific features of the Spanish Civil War—in which some randomness existed in the initial zones controlled by each of the blocs—this case allows us to have some degree of exogeneity to the extent to which people fought on either side (each of the armies enforced routine conscription) as well as to the extent that people were victimized by the groups (groups could more easily victimize people living in their areas of control than otherwise). 9 This exogeneity is helpful when trying to study the impact of this wartime victimization on subsequent political identities.
This article therefore explores a historical civil war and its aftermath. It demonstrates that victimization does matter for political identities: it leads to the rejection of the perpetrators’ identities, and its impact lasts for more than one generation (i.e., beyond the individuals who experience the trauma). The article also shows how victimization matters: on one hand, only severe (and not moderate) victimization leaves an imprint; on the other hand, only the identities related to relevant wartime cleavages—in the case of Spain, the left–right cleavage—are those that have clear-cut political consequences.
Theoretical Framework: The Long-Term Effects of Violence
The idea that civil wars have long-term political effects is not new; at a purely anecdotal level, we have evidence of the endurance of political identities that were forged in conflicts that took place in the very distant past. For example, in the states of Missouri and Kansas, U.S. Civil War identities were given expression in a college football match more than 150 years later; 10 in Ireland, political families still identify with partisans of the Treaty of the Union and its enemies, who fought from 1922 to 1923. 11 Going beyond civil wars, evidence from Holocaust survivors and their offspring indicates that they have more moderate political attitudes and worldviews than people who did not experience the Holocaust. 12 Allinson argues that the experiences during the 1930s and 1940s in Japan affected how people responded to the political changes and economic opportunities that arose in the 1950s and afterward. 13 Horowitz similarly argues that people from postcommunist states that have undergone civil wars have less predisposition toward political and economic reforms because of their experiences. 14 In this article, I go beyond the anecdotal level and attempt to find systematic patterns in the long-term political effects of traumatic experiences concomitant to an armed conflict.
Civil wars can be defined as “armed combat within the boundaries of a recognized sovereign unit between parties subject to a common authority.” 15 We can think of civil wars as having four different (core) elements: (1) physical violence, which can be selective or indiscriminate, 16 direct or indirect; 17 (2) socialization and indoctrination; (3) nonphysical violence (e.g., displacement and migration, recruitment); and (4) social and economic change. Each of these components can have an effect on the political identities of individuals who have experienced a civil war. Among them, we can expect physical violence to have the most striking impact on survivors. 18
The psychological effects of different types of violence (e.g., being tortured, experiencing sexual violence, etc.) on individuals have been widely researched, for example with the study of the well-known posttraumatic stress disorder (PTSD). 19 Scholars have found that emotions and behavior are affected by traumatic shocks even when the experience is not firsthand. 20 In psychology, there is also a well-developed literature on trust, conflict resolution, and reconciliation, which focuses on the level of the group. 21 However, the specific effects of different types of traumatic experiences on individuals’ political identities and political behavior have been generally overlooked. 22 The exception has been research on attitudes toward justice and reconciliation, 23 and on specific attitudes or beliefs such as those regarding political regimes. 24 Nevertheless, insofar as civil wars have an intrinsic political dimension, 25 it is more than plausible to think that wartime experiences and related psychological distress will have political consequences; as Canetti-Nisim et al. state, “Personal exposure to political violence that results in psychological distress affects political worldviews.” 26 At the individual level, being a victim of or a witness to violence can lead to a number of feelings or psychological reactions toward the perpetrator (e.g., anger, moral outrage, resentment, terror, fear, sadness—to name but a few), which are likely to lead to a change in behavior 27 and attitudes 28 during war, but also toward the development of new political identities, or to the redefinition of previous political identities.
It is nonetheless difficult to conceptualize all the precise ways in which the aforementioned feelings influence subjects’ perception of political reality. Some would even argue that this is an idiosyncratic process that cannot be understood in a systematic way. In addition, these effects may occur only in specific circumstances. Canetti-Nisim et al. find, for example, that exclusionary attitudes from people who have been exposed to terrorist attacks show up only when they have experienced psychological distress and they feel threatened by the group associated with the source of this distress. 29 Pham et al., who analyze the impact of PTSD on attitudes toward reconciliation, write, “[O]penness to reconciliation is related to multiple other personal and environmental factors.” 30 At the same time, different feelings could lead to different reactions: a feeling like anger could lead to different reactions than a feeling such as fear. 31
For systematizing purposes, I argue that the effects of civil war victimization can be boiled down to three alternative attitudinal and behavioral responses vis-à-vis the perpetrators and their political identities. One effect is rejection of the identity represented by the armed group (e.g., due to revenge, resentment, or moral outrage). At the behavioral level, this leads to nonsupport for the political group(s) with the political identity (or label) in the elections, or to support for groups with rival identities. Rejection may also involve hostile feelings, attitudes, and behaviors toward social or ethnic groups associated with the perpetrator(s). 32 A second effect is acceptance of the identity represented by the armed group (e.g., due to terror or fear). At the behavioral level, acceptance leads to support for group(s) with the group’s political identity (and identification with it), and/or nonsupport for groups with a rival identity. A third response is demobilization or apathy, which leads to a rejection of the identities represented by all groups/parties (e.g., due to a combination of revenge, resentment, terror, and fear). 33 Demobilization leads to decreased political interest and, at the behavioral level, to political nonparticipation. 34 Finally, a fourth option, no effect, should also be considered. 35
Among all these effects, rejection is the most coherent with the psychological literature as well as with the existing literature on wartime attitudes and behavior. 36 Rejection can issue from all the aforementioned feelings (e.g., moral outrage, fear, discontentment, sadness), potentially in different degrees, and it can be argued that it will prevail over the other effects. Thus, the rejection effect will prevail over acceptance and demobilization effects (hypothesis 1).
Victimization in civil war can take a myriad of forms, as armed groups use different tools in a “repertoire of violence”; 37 some forms of victimization are more intense than others, and we can expect these to have differential political impacts. Following Wallace, we can distinguish between “moderate” and “severe” forms of victimization. 38 We can consider the former to include violations that encompass any form of physical harm, a life threat, or displacement (e.g., execution, being condemned to death, or being forced to leave a town/country). The latter includes violations that are degrading, but do not necessarily involve physical harm or a life threat for individuals (e.g., being imprisoned, fined, sacked from work). I hypothesize that these two forms of victimization—distinguished by their severity—have different political impacts. Specifically, moderate victimization is less likely than severe victimization to have an impact on political identities (hypothesis 2).
At the psychological level, victimization experiences are likely to have an impact not only on the individuals who suffer from them but also on their offspring. 39 Although the first generation is affected by the direct encounter, the second generation is affected through the process of socialization, which forms the background of political identities. 40 We can expect that the political effects of victimization will be transmitted across generations because victimization is a condition that is passed on to descendants through socialization processes. 41 At the same time, victimization is likely to be transmitted differently depending on its severity; more traumatic forms of victimization are more likely to be transmitted than less traumatic ones: the political effects of severe victimization are more likely to be transmitted through generations and therefore to prevail through time than the political effects of moderate victimization (hypothesis 3). At the same time, ceteris paribus, proximity with the victimizing events should lead to a stronger effect on political identities; memories are stronger when traumatic events have been lived firsthand, and the closer people are (in age) to the people who suffered them. The memory of events naturally tends to fade through time, with generational turnover: regarding victimizing events that have taken place in the distant past, the political effects of victimization will tend to increase together with the age of the individuals (hypothesis 4).
Empirical Exploration
I will now turn to the empirical test of these hypotheses, which will consist of a multimethod exploration of data from the Spanish Civil War and its aftermath. In the next subsection, I present a brief overview of the Spanish case. I then present evidence from semistructured interviews that I conducted on survivors of the civil war; subsequently, I analyze a specialized survey implemented on a representative sample of the Spanish population.
The Spanish Civil War and Its Aftermath
The Spanish Civil War began as a military coup against a legally constituted democratic government. It lasted for almost three years (July 18, 1936, to April 1, 1939) and caused nearly 800,000 deaths and more than 440,000 people to be externally displaced. The civil war took place between two main political blocs: (1) the army of the Republican government (or Loyalists), which also included militias of political parties, trade unions, and the International Brigades composed of foreign volunteers; and (2) the army of the rebels (Francoists or Nationalists), which also included factions of the regular army and various militias, in addition to aerial units of the fascist German and Italian armies, which sided with Franco. These blocs, which I will call left and right, respectively, mirrored the political divisions of the prewar period. Indeed, Spain was highly polarized along the left–right or class cleavage during the period of the Second Republic (1931–39) that preceded the outbreak of the civil war. 42 The class cleavage, which was strongly connected to a land distribution conflict, 43 was not the only cleavage in Spanish politics before the outbreak of the war: during the early decades of the twentieth century, peripheral nationalism had flourished politically, socially, and culturally. In the Basque Country, Galicia, and Catalonia, political forces representing regionalist interests not only had a hegemonic presence in their respective polities but also brought regional issues to the national level. The center–periphery cleavage was also significant during the civil war, when the Francoists heavily targeted members of the ethnic minorities. 44 Nonetheless, class was the main structuring cleavage in the civil war; as an indicator, at the time of the coup, the different regional political parties (and their constituencies) sided with the rebels or the government following their position on the left–right cleavage, and not following their positions on the center–periphery one. 45
Shortly after General Franco’s coup, Spanish territory became split between areas of Loyalist (Republican) and Rebel (Nationalist) control. The war was largely composed of pitched battles and aerial attacks, and in less than three years, the Nationalist army managed to conquer all the Loyalist territory and eventually win the war. During the war, both groups perpetrated executions in their own rearguards and conducted aerial bombings in their enemies’ rearguards. 46 Nationalist direct violence lasted several years after the war in the form of executions that had a protolegal nature. During and after the war, people were also victimized through displacement, torture, imprisonment, and being sacked from their jobs because of their ideologies. In this article, I explore the impact of all these forms of victimization on subsequent political identities of individuals.
Qualitative Research
Between April and August of 2007, I conducted a set of semistructured interviews with survivors of the Spanish Civil War. The target population was confined to individuals born any time before 1930 who had direct memory of the conflict. 47 I interviewed a total of thirty men and twenty-five women; the average age of the respondents was eighty-four years old. The interviewees were selected through “snowball techniques” in two different waves: a first sample of subjects (seventeen) was interviewed in a “pilot” process in April 2007; a second sample of subjects (thirty-eight) was interviewed from May to August 2007. 48 I employed the same interview protocol with all subjects: in particular, I asked them about their prewar identities, their wartime experiences (including victimization experiences), their dictatorship experiences, their postwar identities, their current identities, their current interest in politics, and the political identities of their offspring (if they had any). 49 I interviewed individuals in localities throughout the Spanish territory (mostly in the regions of Madrid, Catalonia, and Castile). 50 The interviewees had a wide range of social and economic statuses, and they reported very diverse wartime experiences: some of them lived in the Nationalist side during the war, others lived in the Republican side; some of them lived in battlefield zones, others lived in pure rearguard localities; some of them were combatants, others were refugees; some were directly victimized (e.g., through displacement, physical violence, etc.), others had no direct traumatic experience.
A summary of the main characteristics of the interviewees (gender, age, province during the war, victimization experiences, political identity pre- and postwar) is provided in Table 1. (Hereafter, I will refer to these respondents with their assigned identification number in this table.)
Testimonies of the Civil War. Summary of Semi-Structured Interviews
Although the semistructured interviews might involve some measurement problems (e.g., backward projection of current political preferences, report bias), the advantage of this method is that it allows the researcher to engage in deep conversations with the respondents and to access what Fuji calls “meta data” (information that goes beyond the interview itself). 51 In this particular case, the interviews put me in a key position to ask about sensitive issues such as political loyalties and wartime experiences, and they allowed me to assess feelings, sensations, and/or attitudes. 52 In fact, the interviews were accompanied by the expression of a myriad of feelings: some interviewees were initially reluctant to talk about that period, some expressed deep emotions when talking about their experiences (e.g., crying), and some did not let me record their testimony due to shame or fear of reprisals. In fact, fear was quite common among the interviewees. 53
Among my sample of interviewees, I observed mixed patterns regarding the effects of violence and other victimization experiences on political identities. Rejection was present in a considerable number of cases; the following testimony is a clear example of how it operated:
In my family [the Anarchists] killed two uncles of mine who were [Catholic] priests. And, as you will understand, then they all became Francoists. Nobody in my family would have liked Franco, but in those moments they all became Francoists because what was happening was terrible . . . it was impossible to live. In my family, who was religious and bourgeois, everyday we waited for someone to get killed. . . . My husband witnessed how [the Anarchists] killed his ten professors in school: “pum pum pum,” one after the other. Something like this has a huge impact on a fourteen-year-old child; obviously, after that, he became a Francoist. (testimony 10)
Or, for example, testimony 8, whose father was exiled to France and later killed in a Nazi concentration camp (Mathausen), told me that his political identity was clearly influenced by this experience; he exhibited resentment and sadness because of it, but he also felt a lot of anger. During our conversation he suggested that all the anger that he felt would be worse if the Francoists, rather than the Nazis, had killed his father.
Some of the individuals I interviewed identified strongly with one side of the conflict before the onset of the civil war, and their wartime experiences merely reinforced their ideological positions. For example, testimonies 11, 12, and 24 identified with the left to the degree that they volunteered to be combatants in the Republican army; they strongly identified with the left for the remainder of their lives. Likewise, testimony 26 came from a very conservative family, and, after being victimized by the left (one of her family members was assassinated), she remained highly conservative. Her political identification with the right seemed to have intensified. Testimony 44 argued that all her family was leftist and that the war “only made us more leftist”; Nationalists forcefully displaced several of her relatives because of their political ideology. Testimony 31 argued that “he has never traded shirt,” referring to the fact that he has always been a leftist; his father was a miner and a member of the Confederación Nacional del Trabajo (CNT) union. Testimony 36 gave a similar account. He identified as a leftist after fighting with the Republican army and for the remainder of his life. In some cases, continuity of political identities took place independently of wartime experiences: testimony 43, for example, told me, “I was a leftist before the war, and nothing of what happened during the war influenced my political leanings.” Testimony 9 was conservative before the war, he volunteered for the Nationalists, and he still displayed a strong right-wing ideology. He told me, “The civil war did not change me. I would be Francoist regardless because I believe that order is necessary.” While there were some cases of acceptance of the perpetrators’ political identity, acceptance was less frequent than rejection, and it happened mostly in cases where there were conflicting experiences—so that acceptance could be, in fact, the result of one rejection effect prevailing over another.
Among those who were combatants in the war, there is a strong coherence between their political identities and the side on which they fought, independent of their prewar identities. 54 This suggests that recruitment may be a powerful force that generates endogenous identities in the context of civil war. Testimony 47 explained that despite the fact that his father was a Republican—and that he also identified as a Republican in the early stages of the civil war—he decided to switch to the Nationalist side because he was motivated by the will to survive (when it was becoming clear that the Nationalists were winning the war); he stayed in the Francoist army thereafter and became a Franco supporter for the remainder of his life. Testimony 15 described a similar process. He was in Morocco doing military service when the war started, and even though he was a Republican, he enlisted with the Nationalist army to survive. This affected his political identity after the war: he was no longer a Republican.
Among the interviews conducted, I did not identify greater political involvement among those individuals who had suffered victimization during the civil war, nor did I notice greater intensity (or polarization) in their political loyalties. Some of the people I interviewed, although they had not been victimized during the war, were quite interested in politics; others who had been victimized displayed an extreme level of political apathy. 55 In fact, neutrality and political apathy were widespread among the interviewees. 56
Survey Analysis
In April 2008, the Centro de Investigaciones Sociológicas (CIS) undertook a survey of a representative sample of the Spanish population by asking a series of questions related to the civil war, the dictatorship, and the Law of Historical Memory that had been approved by the Spanish Parliament in December 2007. The survey targeted 2,936 respondents, older than eighteen, who underwent face-to-face thirty-minute interviews. The survey included a number of questions on the victimization experiences of the individuals (only for elders, who were defined as people older than sixty-five) and/or of the family (for the whole sample). People were also asked a myriad of questions about their past, including which side the family identified with during the civil war.
Table 2 depicts the correlation between family leanings with the sides of the civil war and current individual identification with these sides. We can observe that the correlation is quite high (the Pearson coefficient is significant at the 99 percent level; Kendall’s tau and Crammer’s V are also significant at the 95 percent level) but that they do not match perfectly; in other words, not everybody identifies with the same side as their family. 57
Individual and Family Identification with Sides in the Civil War
DK/NA = does not know/does not answer. Figures are column percentages.
I turn now to exploring the effects of civil war victimization on political identities. I first deal with victimization suffered by the individuals themselves (to test hypothesis 1) and then explore victimization of relatives or friends, in addition to individual victimization (to test hypotheses 1 to 4).
Individual Victimization
Individuals who were older than sixty-five years old (a total of 597 people in the sample) were asked specific questions about individual victimization experiences during the civil war. People between the ages of sixty-five and seventy-one could not possibly have any wartime experiences because they were born after 1936; thus, I code as “elders” only those older than seventy-one years old—those who were born the year of the outbreak of the war: 1936. 58 Table 3 shows the descriptive data on individual victimization experiences that were reported in the subsample of elders. We can see that those who had these experiences represent a very small share of this subsample. Only 54 out of 409 elders (13.2 percent) report having been victimized by one of the armed groups during the civil war. 59
Personal Victimization Experiences (elders)
Figures represent the percentage older than 71 in the sample.
Includes all people who suffered any type of victimization.
Because the subsample of elders includes very few cases, I cannot perform econometric analyses with it. However, if we analyze the profiles of each of these groups of victims at a descriptive level, we find a correlation between patterns of victimization during the civil war and voting behavior: those who were victimized by the Nationalist side vote for either leftist or Basque nationalist parties (e.g., PNV), whereas those who were victimized by the Republican side vote for the main right-wing party (i.e., the PP). Hence, among the various alternative effects suggested, rejection seems in Spain to be the most predominant. 60 Insofar as political participation is concerned, the survey data on elders suggest that voting in the elections is not significantly affected by victimization experiences. Table A3 of the appendix shows that the rate of participation in the March 2008 elections is not significantly different between subgroups of victimized and nonvictimized elders.
Family Victimization
In the survey, people were asked about the victimization experiences of their relatives and/or close friends during the civil war; Table 4 shows the distribution of responses across the different experiences they were asked about, in percentage levels. People were asked about different family members or close friends (up to three), and they were also asked about those responsible for the actions (Table 5).
Family Victimization during the Civil War, by Types (whole sample)
DK = does not know; NA = does not answer. Figures represent column percentages.
Side Reported to Be Responsible for Family Victimization (whole sample)
DK = does not know; NA = does not answer. Figures represent column percentages.
Although the rate of nonresponses to this victimization question was relatively greater among younger cohorts (see Figure A1 in the appendix for the rate of response to this question, by cohort), it is striking the degree to which people responded to it: only 22.2 percent of the sample did not answer concerning a first person (relative or friend). From this group, 52.58 percent reported victimization of this person, whereas 26.1 percent said that nothing happened to them. Imprisonment and death in combat constitute the most common form of reported victimization (10.7 percent); execution has a slightly lower percentage (8.6 percent), followed by having to leave Spain (3.9 percent) and having to hide (4.4 percent). The remaining forms are much less frequent. As can be seen in Table 5, of all those who report victimization of one family member/friend, 56 percent attribute responsibility to the Nationalist side, whereas 21.4 percent attribute responsibility to the Republican side; interestingly, this proportion is quite coherent with actual figures of violence in this civil war.
I proceeded to run a set of multivariate regressions to check the effect of wartime family victimization on the political identities of individuals. Again, the interest is in knowing whether (surviving) victimization influences political identity. First, I consider political identity broadly defined in terms of political blocs: left and right. These blocs have a rough correspondence with the sides fighting the civil war (left: Republican, right: Nationalist)—although the right has a democratic character nowadays, which differentiates it from the nondemocratic character of the Nationalist bloc during the civil war. I operationalize the dependent variable as a dummy variable, Leftist, which has a value of 1 if the individual is located to the left of the political spectrum (i.e., positions 1–5 on the ideological scale) and 0 if she or he is located to the right of the political spectrum (i.e., positions 6–10 on the ideological scale). 61 I believe this is the best way to operationalize political identities in the case of Spain, where the concept of Party Identification 62 is difficult to apply. 63
In a second set of analyses, I look at the determinants of the vote for peripheral nationalist parties in Catalonia and the Basque Country (i.e., Esquerra Republicana de Catalunya and Convergència i Unió in Catalonia; Partido Nacionalista Vasco, Aralar Na-Bai in the Basque Country) to see whether family victimization during the civil war has an effect on the vote for these parties and thus on identity around the center–periphery or nationalist cleavage. 64 This will allow us to control for the existence of a second (nationalist) dimension in specific areas of Spanish territory. Also, one could argue that this second dimension might have an influence on the effects of civil war violence on identities: as explained, ethnic minorities were victimized by the Nationalists (during the civil war) and by the Francoist regime (afterward), so rejection of the identity of these perpetrators may have led to a greater identification with the peripheral nationalist political parties—both to the left and to the right of the ideological spectrum.
Family victimization is the main independent variable in the regression analyses below. In a first operationalization, this is measured with a dummy variable with a value of 1 if the interviewee answers positively to any of the victimization items (for the first person; I will not focus on persons 2 and 3). 65 In addition to civil war victimization by the Nationalists (Nationalist Victimization) and/or the Republicans (Republican Victimization), I also include a measure of victimization during the Francoist dictatorship to account for events that occurred during the forty years of Francoist rule, which could affect people’s identities, and which could also have a counteractive effect on previous victimization. This is operationalized with a question similar to that measuring civil war victimization; in particular, people (if older than forty-five years old) were asked if themselves or a member of their family or a close friend were detained, incarcerated, sacked from work, fined, obliged to leave Spain, or executed. Table 6 shows the response distribution of this variable, for both individual and family victimization. The Francoist Victimization variable takes a value of 1 if the individual scores on any of these victimization items (i.e., regarding herself or himself or a close person) and 0 otherwise.
Individual and Family Victimization during the Francoist Dictatorship (1939–75)
DK = does not know; NA = does not answer. Percentages represent row percentages.
In a first set of analyses, I do not distinguish between types of victimization; in a second set of analyses, I include different measures of severe and moderate victimization, for both the civil war and the dictatorship. Specifically, I code the variables Severe Nationalist Victimization (for died in combat, died in bombing, was assassinated, was condemned to death, disappeared, and had to leave Spain, when perpetrated by the Nationalist army during the civil war), Severe Republican Victimization (for the same violations, when perpetrated by the Republicans during the civil war), Moderate Nationalist Victimization (for was imprisoned, had to hide, and was sacked from work, when perpetrated by the Nationalists during the civil war), Moderate Republican Victimization (for the same violations, when perpetrated by the Republicans during the civil war), Severe Francoist Victimization (for obliged to leave Spain or executed during the dictatorship), and Moderate Francoist Victimization (for detained, incarcerated, sacked from work, or fined during the dictatorship). 66 Table 7 depicts the sample distribution of each of these categories.
Moderate and Severe Victimization during the Civil War and the Dictatorship
A number of sociodemographic controls are also included in the regressions; their selection derives from the main set of variables identified and employed in the political behavior literature and is inspired by the empirical analyses performed in Aguilar et al. 67 I run regressions with different sets of variables introduced in subsequent stages to optimally capture their independent effects. 68 A first set of variables contains the sociodemographic controls: (1) Age (in years), (2) Gender (dummy with a value of 1 for men and 0 for women), (3) Size of the Municipality (a scale variable that takes values from 1 to 7), 69 (4) Interest in Politics, 70 (5) Education, 71 and (6) Religiosity. 72
A second set of variables relates to family socialization, which is crucial when analyzing the effects of particular events over generations who did not experience them directly: (7) Talked politics, which measures the extent to which the person was exposed to conversations about politics within the family; 73 (8) Family Nationalist, scored 1 if the family sympathized with the Nationalist side during the war, 0 otherwise; and (9) Family Republican, scored 1 if the family sympathized with the Republican side during the war, 0 otherwise. These two variables should allow us to control for socialization effects as well as for endogeneity issues associated with the fact that sympathizers from one side were more likely to be victimized by the enemy side, and vice versa.
A third set of explanatory variables refers to the victimization variables. A first set of models includes (10) Nationalist Victimization, (11) Republican Victimization, and (12) Francoist Victimization. A second set of models includes the nuanced victimization variables: (13) Severe Francoist Victimization, (14) Moderate Francoist Victimization, (15) Severe Nationalist Victimization, (16) Severe Republican Victimization, (17) Moderate Nationalist Victimization, and (18) Moderate Republican Victimization.
The results of a first set of logit regressions with the dummy dependent variable (Leftist) are presented in Table 8. 74
Logit Regressions for Leftist
Standard errors in parentheses.
p < .1. **p < .05. ***p < .001.
In Table 8, we observe that the sociodemographic control variables are generally very significant, with the exception of Age, which is not significant in any model. Being male increases the likelihood of being a leftist as well as being interested in politics and having higher levels of education. Living in bigger localities decreases the likelihood of being a leftist as well as being religious. Regarding the socialization variables, talking about politics has a slightly significant (positive) effect on the likelihood of being a leftist. As we would expect, having family from the Nationalist side in the civil war decreases the likelihood of being a leftist and having family from the Republican side increases it; these two variables are highly statistically significant. In model 3, the effects of victimization are significant in the hypothesized direction: reported victimization by the Nationalist side increases the odds of having a leftist political identity, and the opposite occurs with victimization by the Republican side. The effects of these variables are robust to the experiences of the dictatorship; in model 4, we observe that being a victim of Francoism also has a positive effect on leftist identity. None of the socialization and victimization variables change their substantive or statistical significance when we put them together in the same regression (i.e., in model 5). 75 These results are supportive of hypothesis 1. If we run ordinary least squares regressions with the dependent variable Leftscale, the results (Table A5 of the appendix) are consistent. 76
Following hypothesis 3, we would expect to find differences in the degree to which victimization experiences affect individuals depending on their age, namely, the effects being greater the older people are (because they are closer in time to the events). For this reason I run an additional regression, based on model 5 in Table 8, with a set of interaction terms between age and the three victimization variables. 77 The results (Table 9) demonstrate a counterintuitive finding: For Nationalist victimization, the rejection effect significantly diminishes with age. This implies that younger generations are more influenced by these experiences, which were closer to them in time, than are older generations. This result, which is contrary to hypothesis 4, might be associated with a greater freedom among younger generations that grew up in democracy to assume and react to these violations, as opposed to older generations that lived through a forty-year repressive dictatorship (the noneffect of the interaction of age with Republican victimization supports this mechanism). Or it could be associated with a bias related to the fact that people who endured the civil war (not necessarily victims) might see more complexity in it than those who just receive the narrative of it, often one-sided and partisan. 78 The rejection effect of Francoist victimization during the dictatorship does intensify with age, as expected. In this case, people might be less fearful of reporting victimization, as the democratic regime that followed the dictatorship has allowed for freedom of expression and open criticism toward the old regime. At the same time, the victimization in the dictatorship left less room for ambiguity for those who witnessed it, as the violations were clearly one-sided. 79
Logit Regressions for Leftist: Interactions of Victimization Variables with Age
Standard errors in parentheses.
p < .1. **p < .05. ***p < .001.
In Table 10, we can see the results with the dependent variable Peripheral Nationalist Vote, only for the subsamples of individuals in the Basque Country and Catalonia. These regressions include the same set of independent variables as Table 8. In this case, the results are striking, as none of the victimization variables show up as significant in explaining nationalism. This finding supports the idea that the major cleavage in the civil war was the left–right cleavage and that civil war victimization experiences did not have a major influence on identities articulated around the center–periphery cleavage.
Logit Regressions for Nationalist Vote, Basque Country and Catalonia
Standard errors in parentheses.
p < .1. **p < .05. ***p < .001.
Table 11 replicates the results of the analyses of Table 8 with the nuanced victimization variables. Although most of the variables in the models do not significantly change, we obtain some interesting results regarding the victimization variables. 80 In M5, both severe Republican and Nationalist victimization during the civil war demonstrate significance in explaining leftist ideology, and they have symmetric effects (negative and positive, respectively); in contrast, moderate victimization during the civil war is not significant for either of the two blocs. This is coherent with hypothesis 2. Yet when looking at Francoist victimization, the reverse happens: moderate (and not severe) victimization is statistically significant. 81 This result might be associated with the fact that moderate victimization during the dictatorship was a very institutionalized type of victimization, which could have penetrating and long-lasting effects for the lives of those affected (and their families). For example, several Republican interviewees reported extremely painful experiences of being ostracized in their localities after the war (not being allowed to work, being fined, being harassed, etc.); these events provoked a lot of anger against the regime. The marginal effects of each of these variables indicate that the family socialization variables are those that have a greater impact on leftist ideology (dy/dx is −0.15 and 0.27 for Nationalist and Republican family, respectively); at the same time, being a moderate victim of Francoism and being a severe victim of the Nationalists during the civil war have a similar positive impact on leftist identity: 0.07 and 0.067, respectively; being a severe victim of the Republicans during the civil war has the same effect in the opposite direction (−0.07). 82
Logit Regressions for Leftist, with Moderate and Severe Victimization
Standard errors in parentheses.
p < .1. **p < .05. ***p < .001.
To conclude, the multivariate regression analyses are supportive of the rejection hypothesis, which seems to have implications in the long term and to operate along the main war cleavage (i.e., left–right), and not along other currently relevant cleavages, such as the center–periphery one. The evidence is supportive of the existence of some sort of intergenerational transmission of victimization experiences. Regarding events in the distant past (e.g., the civil war), the impact of victimization is stronger for severe victimization than for moderate victimization. However, when looking at more recent events (i.e., the dictatorship), this impact is stronger for moderate forms of victimization as compared to severe ones, contrary to expected. Also contrary to what was expected, younger generations are more likely than older generations to be affected by familial victimization during the civil war; this does not happen, however, with familial victimization during the dictatorship, for which the impact does increase with age (and therefore proximity to the events).
The mechanisms underlying these results are not straightforward; they seem to reflect not only the emotional reactions to victimization but also other processes related to the sense of freedom, willingness to report violations, and similar. Regarding civil war victimization, the results suggest that older generations transmitted their victimization experiences (and their rejection of the political identities of the perpetrators) to younger generations through socialization. The fact that we are observing effects along the left–right cleavage (the main “macro cleavage” of the contest, along which most violations where committed) but not along the peripheral cleavage (less salient during the war but very important nowadays in Spanish politics) supports the view that this was the result of an intimate mechanism of transmission (i.e., through narratives within the family); in other words, this was not generated or motivated by the themes in the public sphere. Yet despite their influence on their children and grandchildren, these older generations, who endured strong ideological repression during the dictatorship, seem to be currently less affected by the victimization experiences than their offspring. It could be that the years under the dictatorship deactivated the older generations regarding the civil-war-related crimes. Or it could be that people who witnessed the civil war have a greater sense of its complexity than those who have been exposed only to narratives; this makes the former less reactive than the latter. At the same time, younger generations—and especially those who grew up under democracy—might feel freer to openly express their grievances about these past violent events, and even to ask for reparations. 83
Interestingly, we observe that the impact of moderate victimization during the dictatorship is stronger as individuals are closer in age to this period; this could be also related to the political freedom in the democratic period following the dictatorship. It is now possible for people to openly criticize the crimes of Francoism, and a number of transitional justice measures have been established. Furthermore, moderate violations during the dictatorship might be easier to acknowledge and report by generations that witnessed them, as these were part of an institutionalized system, which left little room for ambiguity. 84
Conclusions
This article has addressed the research question of what are the political consequences of civil wars by undertaking a multimethod analysis of the Spanish case. More specifically, I have presented a set of alternative hypotheses on the effects of war-related victimization on political identities, and I have explored them with various original pieces of empirical evidence. The results of the qualitative research and the survey analysis broadly indicate that violence—as well as other forms of victimization (e.g., imprisonment, displacement, forced labor, torture)—leads to the rejection of the political identity of the perpetrators. This rejection takes place not only among people who directly witnessed the civil conflict and were subject to these experiences but also among people from younger generations; in other words, familial victimization (not only individual victimization) matters. Also, the empirical results indicate that the effects of severe forms of victimization are stronger overall (as compared to moderate forms of victimization); these effects are also more resilient and therefore more prone to persist through time and generations. This is especially true for events in the distant past vis-à-vis more current events. We have also observed that the rejection of the identity of perpetrators operates across the main cleavage along which the Spanish Civil War was articulated (left–right) and not across other cleavages that might be currently relevant (center–periphery). Finally, in the analysis here, victimization does not seem to lead to increased political interest and political participation of individuals suffering from it; this contrasts with recent literature that has focused on the short-term impact of violence.
Methodological caveats have to be made regarding the results. First, the data might suffer from report bias and/or measurement error: in both the survey and the interviews, people could be projecting their current opinions back in time, or they might simply have bad memories and misreport victimization. Without having an experimental setting to deal with the question under consideration, and without having a good instrument for victimization, I believe that a research design that combines different methods and data (such as the one here) is the most appropriate. I have triangulated the evidence from the semistructured interviews and the survey to make the most of them both. Also, in the regression analyses, I have attempted to include all the control variables that were not only meaningful from a theoretical perspective but also helped counteract any possible omitted variable bias. The report bias, unfortunately, cannot be tackled with this type of data.
A second caveat has to do with the generalizability of the results. This is a study based on a single case, which is idiosyncratic because the winning side of the war went on to govern for forty years. This does not happen in many civil wars, to where the findings therefore cannot easily travel (although quite a number of cases exist where postwar political repression may severely restrict what people are free to say about the conflict). 85 It is worth mentioning that these characteristics of the Spanish Civil War, if anything, would make it harder to observe the aforementioned effects. Precisely because the Francoist regime sought to eliminate any trace of the violations perpetrated by the rebel side during the conflict (and, instead, it emphasized the violations committed by the loyalist side), we could expect the effect of violence to be one-sided (i.e., only regarding leftist violations) and not two-sided (as it is observed). Indeed, we observe that violations by the Nationalists, even if underrated during the forty-year dictatorship, have a similar impact on political identities to those perpetrated by the Republicans. This suggests that these are clear-cut effects of Nationalist victimization and that these would probably be stronger had there not been a dictatorship promoting one viewpoint, and two generations in between.
Overall, I would argue that the general results of the article are not necessarily affected by the idiosyncrasies of the Spanish case and that they should be applicable to other cases. These results broadly suggest that civil war victimization experiences are likely to influence the political identities of survivors and their offspring, that they are likely to do so in a way that is not favorable to the political identities of those associated with the perpetration of atrocities, and, finally, that these effects are likely to persist for at least one generation.
Footnotes
Appendix
Logit Regressions for Nationalist Vote, Basque Country and Catalonia, with Moderate and Severe Victimization
| Catalonia | Basque Country | |
|---|---|---|
| Age | 0.023*** | −0.016** |
| (0.01) | (0.01) | |
| Gender | 0.044 | 0.033 |
| (0.17) | (0.19) | |
| Town Size | −0.156*** | −0.322*** |
| (0.05) | (0.07) | |
| Interest Politics | −0.123 | 0.213 |
| (0.11) | (0.13) | |
| Education | 0.437*** | 0.147 |
| (0.15) | (0.16) | |
| Religion | 0.021 | 0.085 |
| (0.08) | (0.08) | |
| Talk Politics | 0.409*** | −0.021 |
| (0.12) | (0.15) | |
| Family Nationalist | 0.090 | −0.247 |
| (0.25) | (0.35) | |
| Family Republican | 0.445** | 0.079 |
| (0.20) | (0.21) | |
| Severe Franco Victim. | 0.710 | −0.431 |
| (0.66) | (0.36) | |
| Moderate Franco Victim. | 0.165 | 0.337 |
| (0.22) | (0.22) | |
| Severe Nationalist Victim. | −0.075 | 0.535* |
| (0.24) | (0.29) | |
| Severe Republican Victim. | −0.397 | −0.425 |
| (0.41) | (0.70) | |
| Moderate Nat Victim. | 0.360 | 0.977*** |
| (0.26) | (0.27) | |
| Moderate Rep Victim. | −0.370 | 0.582 |
| (0.41) | (0.61) | |
| Constant | −2.311*** | 0.277 |
| (0.51) | (0.62) | |
| Observations | 631 | 555 |
| χ2 | 61.558 | 59.195 |
Standard errors in parentheses.
p < .1. **p < .05. ***p < .001.
Acknowledgements
I am grateful for comments and suggestions on previous versions of this article to Gema García-Albacete, Stathis Kalyvas, Pedro Magalhães, Cory McCruden, Geoffrey Wallace, Elisabeth Wood, participants of the 2010 HiCN Workshop in Lisbon, and of the 2011 Fòrum de Recerca at Universitat Pompeu Fabra, Barcelona. I am also thankful for comments from the P&S editorial board. The usual caveat applies.
Declaration of Conflicting Interests
The author declared no conflict of interest with respect to the authorship and/or publication of this article.
Funding
The author(s) declares the following financial support for the research, authorship, and/or publication of this article: The Centro de Investigaciones Sociológicas (CIS), of the Spanish Ministry of the Presidency, has provided funding for the project with the Beca de Finalización de Tesis Doctorales and the Módulo de Encuesta grant. The research in this article has also benefited from funding from the MacMillan Center Dissertation Grant, the Leitner Program in International and Comparative Political Economy Research Grant, and the Leylan Fellowship in the Social Sciences, at Yale University.