Politics by two means: direct and indirect violence in civil war

Laia Balcells
Yale University
laia.balcells@yale.edu; laia.balcells@gmail.com

“Prepared for delivery at the Annual Meeting of the Political Science Association Meeting,
Toronto (CA), September 3rd – 6th 2009”

Abstract:
This paper distinguishes between ‘direct’ and ‘indirect’ violence during conventional civil wars.
These two types differ in their forms of production: while indirect violence is unilaterally
perpetrated by the armed group, direct violence is jointly produced by the armed group and
civilians, and it hinges on local collaboration. These differences have consequences for the
spatial variation of each of these types: indirect violence is hypothesized to be positively
associated with levels of prewar support for the enemy group; in contrast, direct violence is
hypothesized to increase together with the level of political parity between factions in a locality.
The predictions are tested with self-collected data from 1,710 municipalities in Catalonia and
Aragon during the Spanish civil war (1936-1939).

1
 INTRODUCTION*

This paper seeks to explain intentional violence against noncombatants in conventional civil

wars. Direct (or ‘face-to-face’) violence is distinguished from indirect violence,1 and two

different logics, corresponding to each of these categories, are provided. Political factors are

placed at the core of the explanation, thereby challenging recent research on the dynamics of

violence during conflict that has given primacy to economic, organizational or military factors

(e.g. Azam 2006, Kalyvas 2006, Humphreys and Weinstein 2006). I argue that political

alignments are key in explaining the perpetration of violence during civil war, and that they do

so in different ways depending on the nature of violence -as defined by its form of production.

This is the case because, regarding the perpetration of violence, political identities intervene at

two different levels: 1) in the decision to target (by the armed groups), 2) in the decision to

collaborate, i.e. acquiescing in the perpetration of violence and/or promoting it (by the civilians).

In this paper, I focus on a civil conflict that primarily takes the form of a conventional

contest, the Spanish civil war (1936-1939) (hereafter, also SCW). Conventional civil wars are

those that “have clear frontlines, in which attacks take place mostly from barricades and stable

positions, and in which there are big major battles that are usually determinants for the war

*
Earlier versions of this paper have benefited from feedback received at several venues: the HiCN
Workshop 2008, and the ISA and ASN Conventions 2009, the Midwest Political Science Association Annual
Meeting 2009, and the Elections and Violence conference at Yale University. I am especially grateful for comments
and suggestions from: Stathis Kalyvas, Macartan Humphreys, Thad Dunning, Elisabeth Wood, David Carter,
Alexandra Hartman, James Fearon, José Luis Ledesma, Matthew Kocher, Carles Boix, Ana Arjona, Keith Darden,
Abbey Steele, Victor Lapuente, Gema Garcia-Albacete, Juan G. Bertomeu, Juan Linz and Vivek Sharma. I would
also like to thank Albert Sesé, Anna Palacios, Josep Ventura and Mercè Tafalla for assisting in the arduous process
of digitalization of the data, Nacho Campanero for his help in mapping it, and Paul Rigg and Núria Ventura for
editing suggestions. Special thanks are due to the John Enders Fellowship; the Macmillan Center; the Leitner
Program in Political Economy at Yale University and the Juan March Institute, who have all provided financial
support for the project.
1
In the typology here, direct and indirect violence are both intentional; the classification thus differs from
that in Valentino et al (2004), who also include non-intentional deaths (i.e. as a consequence of famine) as indirect
violence. The typology also differs from Arreguin-Toft’s (2001), whose categories refer to warfare strategies:
‘direct’ stands for approaches that target an adversary's armed forces in order to destroy that adversary's capacity to
fight; ‘indirect’ stands for approaches that seek to destroy an adversary's will to fight.
2
outcomes” (Kalyvas 2005). One of the main differences between them and irregular or guerrilla

wars is that -except for zones that are extremely close to the frontline (i.e. ‘battlefield areas’) -

the control of the armed groups over the population under their dominion is overwhelming; this

is not the case in irregular civil wars, in which areas of total control coexist with areas of

fragmented control where this must be shared with the rival. This means that while in guerrilla

wars violence is largely the result of warfare and competition to gain territory, in conventional

civil wars violence against civilians is much less connected to military competition. Indeed,

because warfare has implications for civilian victimization, we need different theories in order to

understand violence across categories of wars as distinguished by the type of warfare (Kalyvas

and Balcells 2008, Balcells 2010).

The hypotheses in this paper are tested with data from municipalities in the regions of

Catalonia and Aragon (Spain) during the SCW, which have been collected from primary and

secondary historical sources. The particular focus on these regions derives not only from the

availability of fine-grained historical data,2 but also from the fact that these territories show a

rich variation in the dependent variable(s) and the independent variable(s) that will be taken into

consideration. Very importantly, these two regions allow for the comparison of violence that

took place during the same period of time (1936-1939) in areas controlled by two very different

armed groups. On the one hand, what I label as ‘Nationalist Aragon’ was controlled by the well-

organized and disciplined Nationalist army, supported by irregular Falangist and other right-wing

militias; on the other hand, what I label as ‘Republican Catalonia’ was controlled by the loosely

organized and less disciplined Republican army, which during a long time was under the

2
Reliable microlevel data on violence during the Spanish civil war is still not available for a large number of
provinces.
3
influence of the anarcho-syndicalist and communist militias.3 This variation allows testing for

the role of organizational factors vis-à-vis local level political variables at explaining lethal

violence against civilians.

The organization of the paper is as follows: the next section introduces the theoretical

framework and hypotheses. Section 3 includes a comprehensive test of the hypotheses, using

multivariate regression techniques and a set of self-built datasets (one for Catalonia, one for

Aragon, and a third one for both regions). Section 4 concludes the article with a summary of the

main findings of the paper.

2. THEORETICAL FRAMEWORK

This paper tackles the determinants of violence against civilians in the rear territories of a

conventional civil war; that is, it excludes battlefield violence, which predominantly involves the

assassination of combatants.4 In this section, I first define direct and indirect violence,

identifying the main dimensions over which these two types diverge. I then present a set of

hypotheses on the determinants of each of these types of violence, which I argue are connected –

although in diverging ways- to the distribution of political loyalties at the local level.

Direct violence can be defined as violence that is perpetrated with light weaponry (e.g.

guns, knives, shotguns, machetes) in a ‘face-to-face’ type of interaction (e.g. individual or mass

executions). The production of direct violence entails the interaction of armed groups and

civilians living in the localities where this violence takes place: in order to perpetrate direct

3
This was the case until May 1937, when the Communists imposed a much more centralized structure.
However, by that time, a large proportion of the executions had already taken place on the Republican side (Solé i
Sabaté and Villarroya 1989, Balcells 2007).
4
Unlike in irregular wars, violence against civilians and combatants in conventional civil wars takes place in
clearly delineated spaces. For a more developed comparison of the ‘logics’ of violence in conventional vis-à-vis
irregular civil wars, see Balcells (2010).
4
violence, armed groups have to take different actions that either require the collaboration of local

citizens, or that are potentially hindered by a lack of collaboration (i.e. ‘veto’) of the civilians.

These include actions such as the localization of suspects, arrest, transportation to locations of

execution, and/or the execution itself. Local civilians may denounce their neighbors, help

identify them, or they can even arrest them (this enhances an armed group’s capacity to

assassinate the civilian supporters of its rival). Conversely, civilians can hide potential victims,

they can help them flee to other places, or they can give false indications to the groups (this

constrains the lethal capacity of groups).5

Indirect violence, by contrast, is perpetrated with heavy weaponry (e.g. tanks, fighter

planes), and it does not require face-to-face interaction with the victims. Because its technology

of production allows it, indirect violence is unilateral from an armed actor’s perspective, giving

very limited agency (if any) to civilians. Also, indirect violence can be perpetrated in territories

where the armed group has no territorial control (e.g. through aerial strikes). All this makes the

logic of direct and indirect violence necessarily different.

At the theoretical level, direct violence against civilians during conventional civil wars is

puzzling (Balcells 2007, 2010): from a perspective that emphasizes the strategic motivations of

armed groups and civilians (Kalyvas 2006), this violence is counterintuitive because the armed

group does not attain obvious utility from killing civilians who live in zones they fully control.

And, insofar as the internal organization of groups does not vary across the rear territories, as is

often the case in conventional wars, violence cannot be explained based on the greater or lesser

degree of internal cohesiveness of the groups (Weinstein 2006, Humphreys and Weinstein

5
Civilians can also presumably be neutral to the actions of the groups (Wood 2003). Yet, remaining neutral
does not seem to be easy in wartime contexts (Kalyvas 2006; Petersen 2001).
5
2006).6 The same applies to explanations based on greed motivations of groups (e.g. Azam

2006): since direct violence against civilians is disconnected from dynamics of military

conquest, spatial variation in victimization cannot be explained by attempts to occupy more

‘desirable’ (i.e. wealthier) territories.

Indirect violence in the enemy’s controlled territories is somewhat less puzzling than

direct violence, as it is likely to be driven by the ‘requirements’ of the military struggle. The

logic in this case should be the same as the one explaining strategic bombings in international

(Pape 1996, Downes 2008, Belkin et al. 2002, Allen 2007) or internal (Arreguin-Toft 2001,

Kalyvas 2006, Downes 2008, Kocher et al. 2008, Lyall 2009) wars, in which violence is aimed at

eliminating the military capacities of the enemy and/or coercing the enemy to change their

policies.7 However, even if we assume that armed groups sometimes choose to attack civilian

locations that have no particular geostrategic or military interest in order to gain bargaining

power, we still need to explain what leads them to target some locations and not others.

This paper relies on the assumption that variation in intentional violence against civilians

in conventional civil wars can be explained from a framework in which groups are driven by

strategic motivations, and in which this is aimed at the elimination of supporters of the enemy

who –due to their political militancy- may eventually become a military threat (Balcells 2010).

From this perspective, political identities are crucial for the groups’ recognition of potential

6
In conventional wars, armed groups often rely on routinized conscription mechanisms, and consist of
disciplined and strong organizations (Kalyvas and Balcells 2009). And, even if this is not the case (i.e. the
Republican army in the early stages of the SCW) there are no reasons to think that the internal structure of the group
will vary across the rearguard territory.
7
“In conventional conflicts, belligerents employ indiscriminate violence against an enemy’s civilian
population –to the extent they are able and the circumstances warrant –to gain coercive leverage over the
adversary’s government. Belligerents also use indiscriminate violence in conventional conflicts to eliminate
unwanted populations from conquered territory” (Downes 2008: 9).
6
threats behind the frontline; as a consequence, at the local level, the targeting of noncombatants

has a very close connection with prewar political alignments.

From this starting point, the following pages develop an explanation of direct and indirect

violence during conventional civil wars.

Direct Violence 8

Let’s imagine a hypothetical country where a civil war erupted after a period of intense

political confrontation between political parties A and B, whose platforms are now championed

by respective armed groups A and B. The citizenry of this country has been mobilized along the

A-B cleavage. Now imagine a hypothetical armed group A that is patrolling a territory that has

been newly conquered from group B. The two groups, which fight a conventional war with

relatively stable frontlines, enjoy exclusive military control of relative large areas from which

they have excluded the rival group. Relevant interactions in the territory controlled by A involve

combatants of this group and all civilians living in it.

In addition to fighting B on the battlefield in order to increase the share of territory under

its control, A is interested in getting rid of strong supporters of B (hereafter, also BSS), who are

perceived as a potential threat. The crucial interactions leading to direct violence take place at the

local level, where the degree to which A targets civilians depends on two factors: (a) the number

of BSS living in it and (b) the behavior of civilians in the locality, who can choose to back the

killings or to constrain them. More specifically:

(a) In each locality there are political activists or individuals who are highly mobilized

and who identify strongly with one of the groups. As a general norm we can expect that the

8
This sub-section section draws heavily on Balcells (2010).
7
presence of BSS will be proportional to the existence of supporters of B in a locality (the same

should hold for A and ASS).9

(b) Collaboration with the armed group in power (in our example, A) is subject to

constraints. In this regard, local civilians who identify with group A are likely to take into

consideration the effects of violence for the future of their locality. I argue that these

considerations are shaped by the distribution of local power between groups, as expressed

electorally: when electoral power approaches parity (i.e. the margin of victory is small), violence

can decisively alter the local political balance; in this context, A supporters are likely to

opportunistically push the armed group toward violence against B supporters. However, where A

supporters are either a distinct majority or a distinct minority (i.e. the margin of victory is large),

they are likely to restrain their respective armed groups. Where they are a majority, they do not

need to use violence in order to change the status quo; where they are a minority, only genocidal

levels of violence would help reverse the balance - short of that, they would endanger themselves

without altering their position vis-à-vis B supporters.10 Translating this conjecture into

operational terms, we would expect A’s supporters to promote armed group violence where the

prewar electoral balance approaches parity (i.e. a 50-50 percent distribution) and to constrain

armed group violence where this balance moves away from parity.

In a nutshell, as political competition approaches parity, both the presence of strong

supporters of the enemy and civilian behavior combine to generate greater levels of violence.

9
The number of Bss will also be determined by the presence of particular institutions mobilizing along the
war cleavage lines (e.g. trade union, churches and professional organizations). Also, we can also expect more Bss in
places with a history of social unrest and political confrontation between A and B.
10
Following Valentino (2004), interest in genocide usually comes from strategically-oriented national
leaders, with long-term views (i.e. “radical communization of their societies”). These motivations are therefore less
likely to exist among local level leaders, who will be interested in maximizing their chances of survival (first and
foremost), of holding power in the near future, and of eliminating their private adversaries (Kalyvas 2006).
8
Hypothesis 1: The greater the degree of prewar electoral parity between groups, the higher the
level of direct violence perpetrated by the armed group controlling that locality.

This hypothesis assumes a one-shot, static setting; elsewhere (Balcells 2010), this is extended to

subsequent time periods.

Indirect Violence

Now imagine the same country, civil war and territory. B is a well-equipped armed

group, which has heavy artillery that permits it to shell from land (in places close to the

frontline), sea (in places close to the seashore), and air (presumably, anywhere).11 In addition to

using this technology to attack A on the frontlines and/or militarily strategic enclaves, B can

choose to deploy some of its resources to attack civilian locations living in A’s rearguard.12

These attacks, while barbaric by definition, can be perpetrated on a selective basis: that is, the

group can decide to assassinate civilians in a particular locality and not in another.13 This choice

will be in some ways necessary for armed groups because their resources are limited.14

Military-strategic factors normally play a crucial role in the decision to bomb a location.

For this reason, industrial and other infrastructure locations are likely to be targeted. The same

happens with strategic communication enclaves, such as harbors, nodal train stations or roads.

11
Thus, I conceptualize indirect violence as aerial, artillery or maritime bombardments. The use of nuclear
weapons can also be included in this category, but given its rarity (especially in civil conflicts), I will not take this
technology into account.
12
It does not make sense to think that B will deploy resources to assassinate civilians indirectly in its own
rearguard.
13
Indirect violence is not necessarily indiscriminate, although it surely is following Kalyvas’s (2006)
definition. Yet, this definition does not take into account the fact that there can be some degree of selectivity even in
attacks against localities/collectivities. Pape (1996) explains, for instance, that aerial power has a lot of comparative
advantage (over land power and naval power) with regard to selectivity: “Unlike sea power, bombing can focus on
specific categories of targets, attacking either political, economic, population, or military targets in isolation or
combination” (45). Also, Steele (2009) argues that when groups target civilians due to a shared characteristic they
are not perpetrating indiscriminate violence: she calls this ‘collective targeting’.
14
Furthermore, genocide may not always be in the strategic interest of the military leaders (Valentino 2004).
9
Obviously, all these variables have been widely scrutinized in the security studies literature (e.g.

Pape 1996, Horowitz and Reiter 2001, Downes 2008). Yet, political factors may play a relevant

role too: everything else being equal, B is likely to attack places with a greater number of strong

supporters of the enemy group, i.e. number of ASS. This is because B is ultimately interested in

eliminating those people who are providing greater support to A during the civil war (highly

mobilized people are those who are likely offering greater logistical and material support to A),

or who will constitute a potential threat in a future period (when the group will potentially exert

control over this territory). Given the degree of imprecision of indirect type of attacks such as

bombings, the armed groups can only make sure that they are targeting strong supporters of the

enemy by attacking locations with a relatively large number of these supporters. In these

locations the effects of indirect violence are more likely to be ‘positive’ for the aims of the group

than elsewhere.

If we conceptualize the degree of support towards the enemy group in a locality –and

therefore the relative number of strong supporters - with the degree of prewar electoral support

for the enemy group, it follows that:

Hypothesis 2: The greater the electoral support towards a group during the prewar period, the
greater the likelihood that a locality will be the target of indirect violence by the enemy group.

3. EMPIRICAL TEST

In this section, hypotheses 1 and 2 are tested with self-collected data from the Spanish regions of

Aragon and Catalonia. During the SCW, with very few exceptions, most of the Catalan territory

10
was under Republican control until the beginning of 1939, with very few exceptions.15 This

means that direct violence took place in two stages: a first period (from July 1936 to 1938/39) in

which violence was perpetrated by leftist militias and the Republican army; and a second period

(during and after its occupation of the territories) in which violence was perpetrated by the

Nationalist army and right-wing militias. The focus here will only be on the violence that took

place in the first stage, that is, on violence perpetrated by the Republican army and the anarchist

and communist militias (the left). The distribution of this violence across municipalities of

Catalonia is depicted in Map 1.16

[Map 1 about here]

Indirect violence took place in Catalonia in the form of aerial strikes perpetrated not only by the

Nationalist army, but also by the Italian and German armies –whose actions were however

supervised by Nationalist commanders. This violence took place throughout the whole conflict:

in 1936 and 1937 they mostly focused on strategic places such as big industries, harbors or

crossroads locations, but they increased in intensity and dispersion in 1938 and until the end of

the conquest of Catalonia in February 1939 (Solé i Sabaté and Villarroya 1986). Map 2 depicts

the spatial distribution of the totality of aerial strikes perpetrated by the right in Catalonia (1936-

1939).

[Map 2 about here]

In Aragon, a total of 366 municipalities were under Nationalist control during the entire civil

war; the remaining municipalities in the region (582) were under leftist control at some point

15
A few localities close to the Ebro frontline were conquered by the Nationalists in mid-1938. The first
Catalan town to be occupied by the Nationalist army was Lleida (3rd of April of 1938). The total occupation of
Catalonia ended on the 12th of February of 1939 (Solé i Sabaté 2000).
16
Focusing on this period will allow a comparison with violence that was perpetrated by the other armed
group (the Nationalists) during the same time, in another region (Aragon).
11
during the war –they are labeled as Republican (vis-à-vis Nationalist). The distribution of

municipalities by control zones (and provinces) of this region is shown in Table A4 of the

Appendix.17 The focus here is on rightist executions that took place in the first subset of

localities (i.e. Nationalist), which spatial distribution is depicted in Map 3.

[Map 3 about here]

Why focus only on this area of Aragon? On the one hand, in locations of the Republican zone of

Aragon (in light brown, on the map), rightist violence took place in a second phase of the

conflict, once leftist violence had already taken place; thus, violence is likely to be (at least

partially) influenced by factors such as revenge, resentment or retaliation (in connection to the

events in the previous phase) - these types of factors are not contemplated in the theoretical

framework above.18 On the other hand, I do not analyze leftist violence in Republican Aragon

because the theoretical framework does not account for violence taking place in a ‘battlefield’

zone such as this one.19 Violence against noncombatants in ‘battlefield’ areas is likely to be

determined by military factors (e.g. the location of the troops or militias, the location of the

battles, etc.), and not by political variables.20

The hypotheses will be tested by means of multivariate linear regression techniques using

three nested self-built cross-sectional datasets: 1) a dataset of all 1,060 municipalities in

Catalonia; 2) a dataset of the 366 ‘Nationalist’ municipalities in Aragon; 3) a dataset of all

municipalities in Aragon and Catalonia (a total of 2,008). With regard to direct violence I will

17
I thank José Luis Ledesma for helping me develop this classification.
18
Violence in non-initial phases of a civil war are addressed at a theoretical level elsewhere (Balcells 2009,
2010).
19
The characterization of this area as a ‘battlefied zone’ draws on evidence from numerous sources, including
Casanova (1985), Ledesma (2003, 2006a, 2006b, 2009), Martínez de Baños et al. (2004), Maldonado (2007).
20
To prove this, I run the same regressions that will be run for the ‘rearguard ‘territories, and for the subset of
localities in this ‘battlefield’ zone (Table A6 of the Appendix). As expected, the coefficient of the main explanatory
variable in the model (competition) do not show as significant in these estimations.
12
estimate negative binomial II (NB) and zero inflated negative binomial (ZINB) regressions,

which are count models appropriate for the nature of the dependent variable(s) (i.e. number of

people executed by the armed group). NB permits to control for overdispersion; ZINB allows

controlling both for overdispersion and for the excess of zeros in the dependent variable (Long

1997). This model generates two set of estimates: a first set of estimates explaining the

probability of the non-occurrence of violence (that is, that the DV is always 0) –through a logit

regression; and a second set of estimates explaining the number of counts –through a NB

regression- for all those cases that are ‘not always zero’. Although the ZINB regression will

prove to be more suitable to the data,21 the NB results are also included so as to supply greater

robustness.22 With regard to indirect violence, I will estimate a binomial regression model with a

dummy dependent variable for bombed locations.23

The datasets have been built from primary and secondary sources on the SCW, which are

detailed in Tables 1 and 2, for Catalonia and Aragon respectively. They include history books,

local histories, official censuses, trade union bulletins and official military records.

[Tables 1 and 2 about here]

The dependent variables in the different models are: ‘Executed Left’ (for Catalonia) and

‘Executed Right’ (for Aragon), which measure the total number of victims of direct violence (for

21
In all the regressions below, the Vuong test will indicate that the zero inflated specification is necessary,
that is that ZINB results should be more reliable than NB results. Also, if we test the different count regression
models potentially applicable to these data, and we check graphically the way they fit to the real data (following
Long 1997, 247-248), the ZINB model is seen to be the most appropriate.
22
In addition to the count models, I run a set of OLS regressions with the dependent variable normalized on
the size of the locality, i.e. executed per thousand inhabitants. The results are robust, and they are available upon
request.
23
In this case, count models (with number of bombings as the dependent variable) will be used as robustness
checks. They are not included here due to length constraints, but they are available upon request.
13
each of the localities in the datasets). For Catalonia, ‘Bombings’ is a dummy variable with value

1 if the locality received any aerial strike during the civil war, and 0 if not.24

The independent variables in the models are: ‘Support Left’, which measures the

percentage of support towards leftist political parties in the Spanish general elections of 16th

February 1936 (i.e. the parties in the Popular Front coalition); ‘Support Right’, which measures

the percentage of support towards rightist political parties (i.e. those in the Nationalist Front

coalition); ‘Competition’, which is the main independent variable, and consists of an index

created from the returns in the 1936 elections, capturing the extent to which there is a balance of

power or parity between two factions: 1-[(%VoteLeft36-%VoteRight/100)]2. This index has

value 0 when one of the groups received all of the votes in the elections (i.e. a 100% share), and

it has value 1 when both groups received 50% of the vote in the elections;25 following the

theoretical framework, we expect this variable to have a positive effect on the number of

executions. An alternative parity index, ‘Compabs’ will be used in a set robustness checks: 1-

|(%VoteLeft36-%VoteRight/100)|.26 An additional independent variable, which will be further

explained below, is ‘Non-volatity’; this variable measures over time stability of the vote in a

locality. In the regressions for Aragon, I will include an additional independent variable:

‘previous violence’.27 This variable, which is not available for Catalonia, measures political

24
Unfortunately, fine-grained data on bombings is unavailable for Aragon. The different primary and
secondary sources I have consulted provide data on selected locations, not for the whole territory.
25
See figure A2 in the appendix for the Kernel density estimate for this variable.
26
This index has a greater variance than ‘Competition’. See figure A3 in the Appendix for the Kernel density
estimate for this variable.
27
This variable takes value 1 if there was any violent conflict in the period January- July 1936 and 0 if not.
The data is obtained from Casanova (1985: 52). I put together all his categories of conflict, which are: strikes;
occupation of private properties including those ending with expulsion; order alterations or clashes between groups;
governmental intervention to solve conflicts; violent aggressions against peasants; occupation of communal lands
and situations of tension solved through negotiations. None of these categories include either crimes of passion or
normal delinquency.
14
violence in a locality during the prewar period; I expect it to have a positive incidence on

wartime violence, as it is an indicator of the presence of ‘strong supporters’ of the groups.

The control variables in the models are: ‘Catholic center’, which is a dummy for places

that had a catholic archbishop, and that therefore had relatively large numbers of religious

population -we expect it to have a positive effect on leftist violence and a negative effect on

rightist violence; ‘Affiliation CNT’ and ‘Affiliation UGT’, which additional proxies for political

and social conflict in a locality, as well as for number of strong supporters –I expect these to take

a positive sign for violence perpetrated by both armed groups.

A set of geographical variables that have potential effects on violence will also be

included in the regressions as controls: ‘frontline’ captures the uncertainty that is likely to take

place in zones close to the war frontline(s), where levels of victimization can be expected to be

boosted by each of the groups. For Catalonia, I use a dummy variable; for Aragon, I use a

somewhat more refined measure, ‘latitude’ of the locality: since the frontline was to the East of

all these territories, the greater the latitude, the greater the proximity to the frontline. ‘Sea’ is a

dummy for localities in the seashore; this should capture the effect of a potential escape route on

the number of killings taking place in a particular area (we expect that this will reduce them).28

This variable only applies to the Catalonia dataset. Proximity to the French ‘border’ should also

capture the effect of proximity to an escape route; as before, I use a dummy in the Catalonia

dataset, and a more refined measure, ‘longitude’, for Aragon: the greater the longitude, the

further south the locality and therefore the bigger the distance from France.29 Altitude is a

measure for ‘rough terrain’, and it should capture the effect that knowledge of local terrain and

28
During the civil war, many refugees left Catalonia by sea (Doll-Petit 2004).
29
I will run the regressions for Catalonia also with ‘longitude’ and ‘latitude’, and the results do not change. I
present here the results with the dummy variables ‘frontline’ and ‘border’ because they are more intuitive.
15
access difficulties has on violence against civilians: in rough terrain locations, people can hide in

the mountains or forests in order to avoid being killed to a greater extent than in other places;

thus, we can expect this variable to take a negative sign.30 Finally, I include inhabitants of the

village in 1936 (‘population’) in order to control for size of the locality.

Table 3 depicts the results of the NB and ZINB models for leftist executions in

Catalonia;31 Models 1 and 2 include ‘competition’ as the main independent variable. Models 3

and 4 replicate Models 1 and 2, but they include ‘support left’ as the main independent variable,

instead of ‘competition’. I include this variable in order to check for the competing hypothesis

that support for the group in a locality reduces levels of direct violence. I call this the ‘non-

domination’ hypothesis, implying that the group kills more the less it dominates a locality,

because it has a relative larger number of non-supporters. At the methodological level, the

inclusion of this variable permits us to check for the possibility that the results obtained with the

variable ‘competition’ are not driven by the imposition of a particular functional form (i.e. non-

linearity).

[Table 3 about here]

The estimates in table 3 indicate that, as hypothesized, ‘competition’ is substantively and

statistically significant for explaining direct violence: the greater the level of parity between

political factions in a locality, the greater the number of leftist executions. Yet, the coefficient is

not significant in the second part of the ZINB equation, this variable is not relevant to explain the

occurrence of violence. These results are coherent with the fact that, at the empirical level, there

30
In alternative specifications, I use ‘altitude range’ of a locality to proxy for rough terrain, which is a
measure somewhat more consistent with Fearon and Laitin’s (2003). The results are robust.
31
Note that the ZINB models have two columns each, one for the logit regression estimating non-occurrence
of violence, another for the negative binomial regression estimating number of deaths in places with non-negative
number of deaths.
16
are some factors that affect the occurrence of violence and that are independent of the degree of

competition in a locality: for example, in many localities of the Republican side, the priest is the

only victim of leftist violence.32

‘Support Left’ is not significant to explain violence, so the alternative ‘non-domination’

hypothesis can be ruled out. With regard to the control variables, ‘CNT Affiliation’ also has a

significant effect, and it indicates that the presence of affiliates increases the number of

assassinations.33 ‘UGT affiliation’ is also positive and significant to explain levels of violence.

With regard to the geographical variables, ‘Frontline’ has a positive effect on violence, as

predicted; proximity to the French ‘border’ and to the ‘sea’ take the expected negative signs,

indicating that the possibility of fleeing reduces the degree of victimization at the local level.

‘Altitude’ also takes a negative sign, capturing the negative effect of rough terrain over

executions. Finally, ‘Catholic center’ has a very strong positive effect on level of executions,

indicating that the presence of large number of strong supporters of the right (i.e. religious

people) leads to greater levels of violence by the left.34

32
As is explained by Delgado (1992) killing the priest became a sort of a revolutionary obligation that could
not easily be avoided by militiamen (even in those cases where they spared the lives of other citizens, e.g.
landowners). Figure A1 displays the distribution of ‘number of priests executed’ across the sample (not including
the city of Barcelona, which is an outlier -with 429 cases). If we run a logit regression on the execution of at least
one priest in a locality (1=one priest or more executed in the locality, 0=no priest execution) on the same
explanatory variables above, the variable competition is not significant. Table A5 of the appendix display the results
of this regression.
33
In fact, this variable cannot be included in the logit piece of the ZINB model because it overpredicts
violence; in other words, there are no places with positive levels of CNT affiliation that did not observe leftist
violence.
34
As in the case of ‘CNT affiliation’, ‘Catholic center’ is a variable that also over-determines the occurrence
of violence, and that cannot be included in the logit part of the ZINB models.
17
 The results of the NB regressions for rightist violence in Aragon are depicted in Table 4;

the ZINB models are summarized in table 5.35

[Tables 4 and 5 about here]

The results for Aragon are also supportive of hypothesis 1, and they are coherent with what we

observed for Catalonia. Levels of direct violence increase together with the degree of political

parity between groups; yet, again, ‘competition’ cannot explain the occurrence of violence.

In models 3 and 4 of Table 4, support for the left shows as significant in explaining levels

of violence by the Nationalists. I suspected that this variable could be capturing a non-linear

effect, and this is why I estimated a fifth model that included a quadratic term of support left

(SupportLeft2): it took a negative and significant value –proving that the effect is indeed non-

linear.36

As expected, in models 2 and 4 of both tables 4 and 5, ‘previous violence’ is significant

to explain levels of direct violence. This variable accounts for a greater intensity in targeting (by

the armed group) in localities with a greater density of strong supporters of the enemy.

Overall, the results in tables 3 to 5 reveal that the same mechanism accounting for

variation in levels of violence perpetrated by the left in one territory (i.e. Catalonia) also explains

the variation in levels of violence perpetrated by the right in another territory (i.e. Aragon). This

is the case despite the fact that the armed groups had clearly divergent organizational

35 In the ZINB models, ‘previous violence’, ‘UGT Affiliation’ and ‘catholic center’ cannot be introduced into
the logit equation due to overdetermination issues. For some unknown reason, model 3 could not be optimized in the
ZINB regressions.
36
This finding is in fact redundant of the significance of the variable competition. Yet, the non-linear effect of
‘support left’ is also reflected in the logit regression of the ZINB models (in Model 5 of Table 5). This is
contradictory with the results of the variable competition (Models 1 and 2), which, as we have seen, does not explain
non-occurrence of violence.
18
characteristics and ideological platforms.37 Further evidence along these lines is provided by

Figure 1, which depicts the predicted number of killings in Catalonia and Aragon, by levels of

political competition. This variable displays a very similar substantive impact on levels of

violence perpetrated by both armed groups.

[Figure 1 about here]

We can better test for the effect of local political vis-à-vis organizational factors by running a set

of regressions with a dataset pooling together all municipalities in Catalonia and Aragon.38 The

dependent variable in these regressions is ‘total number of deaths’ (with no distinction of

perpetrator): ‘killed’; a dummy for ‘control zone’ (1 if Republican, 0 if Nationalist) will allow

measuring control by one or the other armed group; competition will capture, again, political

parity at the local level.

Table 6 includes the results of three alternative regression models: Model 1, with the

usual measure of competition; Model 2, with an alternative measure of competition (index with

absolute values); Model 3, with the variable ‘support left’. Geo-referencing variables are not

included in this regression because these would have contradictory effects (the frontline was to

the West for the localities in Catalonia, and to the East for Aragon), and they could not be

interpreted properly. Altitude is included here for localities in both regions. 39

[Table 6 about here]

The results in table 6 are confirmatory of the hypothesis that political competition, measured in

both quadratic and absolute terms, is explanatory of levels of violence in localities for both the

37
In the particular case of Spain, the thesis that violence perpetrated on each side followed ‘different logics’
(Herreros and Criado 2009) is challenged by these results.
38
The descriptive statistics of the variables in this dataset are listed in Table A3 of the Appendix.
39
It must be noted that localities in ‘battlefield zone’ are included in this regression. These are the localities in
Republican Aragon.
19
Republican and the Nationalist side. The coefficient for ‘ZoneRep’ indicates that armed group is

not explanatory of levels of violence; yet, being in the Republican zone increases the likelihood

of a locality experiencing violence; this may be indicative that there are differences between with

regard to the spatial distribution of violence. If we look at actual numbers, we can see that

violence was more widespread in Republican Catalonia than in Nationalist Aragon. While

42.53% of municipalities in Nationalist Aragon had at least one executed, 61.77% of localities in

Republican Catalonia had at least one executed. Thus, in Republican Catalonia violence was

more widespread across in Nationalist Aragon.40

To test the causes of indirect violence at the local level, I explore the determinants of a

locality being targeted anytime during the conflict. As said, I use data only on Catalan localities,

as there is no systematic data on bombings available for Aragon. The converse of hypothesis 2 is

that a group will likely perpetrate less indirect attacks in places where it has greater electoral

support; to facilitate the interpretation of the results, the hypothesis is operationalized in this

way. In other words, the main independent variable in the model is ‘support for the right’ in the

elections, and this is expected have a negative effect on bombings perpetrated by Nationalist

forces.41

CNT affiliation and UGT affiliation are included in the regression model in order to

control for the presence of strong supporters of the left –thus, we can expect these two variables

to have a positive effect on bombings. Other control variables are latitude and longitude of the

locality (in degrees); since they allow for a precise measurement of the geographical location,

these variables serve as a corrector for the omitted variable bias driven by the non-inclusion of

40
The total number of executed by the right in the 366 localities of Nationalist Aragon were 6,035; the total
executed by the left in the 1,062 localities in Republican Catalonia were 8,009.
41
Figure A2 in the Appendix shows the Kernel density estimate of this variable. The distribution is close to
‘normal’ although it is slightly skewed: the median is at a 46.5% and not at the mean level, which is 47.72%.
20
military and geostrategic variables in the regression model. Altitude of the locality also allows us

to control for geostrategic factors– we can expect locations in more mountainous (and therefore

higher) places to be less relevant from a militaristic perspective, and therefore less prone to be

targeted by aerial strikes.42

[Table 7 about here]

The results in Table 7 show that, controlling for all other variables, support for the right has a

negative effect on the likelihood of a locality being bombed, as predicted. ‘CNT affiliation’ and

‘UGT affiliation’ have both positive and significant signs, thus indicating that the right was more

likely to indirectly attack enclaves of the left. Also, as expected, altitude has a negative effect on

likelihood of bombings. Catholic center is not statistically significant. Latitude has a negative

significant effect, indicating that eastern locations were more likely targeted (this is consistent

with Map 1, which shows that the localities on the seashore were those more heavily affected by

indirect attacks); longitude, in contrast, has no significant effect on bombings.

Figure 2 depicts a post-estimation graph of the predicted likelihood of bombing in a

locality by levels of support for the right in the elections (all other variables in the regression are

set at their mean level). This permits us to better observe the substantive impact of this variable,

which is non-negligible.43

[Figure 2 about here]

42
As before, the inclusion of the variable ‘altitude’ leads us to lose a significant number of cases. If we run
this same regression without this variable (and a total of 1,052 cases) the results do not change.
43
In additional to the logit regression, I have performed additional tests with alternative specifications, which
provide robustness to these findings: 1) the same logit regression above, but with ‘competition’ as the independent
variable; 2) NB and ZINB regressions with the total number of aerial strikes in a locality; 3) NB and ZINB
regressions with number of people killed by aerial strikes; 4) logit and count regressions with % support to the right
in the 1933 elections (instead of the 1936 elections). The results are consistent with the ones in Table 3: they show
that support for the right in the elections is a crucial variable explaining indirect violence at the local level.
Importantly, the variable competition does not prove to be relevant for bombings, and this allows us to rule out the
hypothesis that indirect violence is explained by the same factors that account for direct violence.
21
One implication of the theoretical framework above is that direct violence will reach greater

levels in places with deeper prewar political cleavages, where supporters of the groups are likely

to be perceived as more ‘committed’ to political platforms, and therefore as more threatening

both for the armed group and for their local political enemies, who will thereby have greater

incentives to eliminate them.

I test this implication by proxying depth of the political cleavages with a measure of

electoral stability of a locality; what I call ‘Non-Volatility’. If patterns of support for the political

blocs are stable, and therefore volatility is low, we can think that cleavages are deeper than if

patterns of support for the blocs are unstable, and therefore volatility is high. I measure non-
44
volatility with electoral returns of both 1933 and 1936, and This variable (a dummy) is

included as an additional explanatory variable to the ZINB regressions for Catalonia and Aragon

above.

[Tables 8 and 9 about here]

This conjecture is confirmed for Catalonia (Table 8): violence is greater in places with stability

of voter patterns; non-volatility does not explain the occurrence of violence, though. In Aragon

(Table 9), by contrast, non-volatility does explain neither occurrence nor levels of violence.

Hypothetically, ‘competition’ and ‘non-volatility’ could have an interactive effect; we

test for this by adding an interaction term (competition*non-volatility) to the regressions above.

We do not present the regression coefficients here because in non-additive models these are not

44
These data are returns, at the local level, in the Spanish Courts/Parliament elections (General elections),
which took place in November 13th 1933 and February 16th 1936, respectively. I collected these data both for
Catalonia and Aragon. Volatility is a variable that has value 1 if the difference in % support for the left between
these two elections is greater than 10%, and 0 otherwise. Non-volatility takes the reverse values.
22
effects, and their interpretation, as well as that of their standard errors, is not straightforward.45

Following Kam and Franzese (2007:32), we compare predicted number of executions at varying

meaningful levels of competition and non-volatility, holding any other variables in the model at

the sample mean.

[Figures 3 and 4 about here]

I have graphed the predicted values in order to facilitate their interpretation (Figures 3 and 4); in

figure 3 we observe that stability has an interactive effect with competition: the marginal effect

of competition on violence is rather modest in ‘volatile’ places, and it is more substantive in

electorally ‘stable’ locations. Figure 4 -for Aragon- depicts a similar pattern.46 In brief, these

tables indicate that political parity influences levels of violence, but that it does so in a greater

extent in places with deeper cleavages than elsewhere. The mechanism driving this is, again,

local level collaboration with the armed groups.

To sum up, the results of the empirical test are coherent with the idea that civilian

targeting in conventional civil wars is very much coupled with political identities and to prewar

local level political dynamics. This is consistent across armed groups and control zones, except

for ‘battlefield’ areas, where violence is mostly related to military dynamics. The results also

show that, as hypothesized, political cleavages and identities affect levels of direct and indirect

violence in different ways. Regarding direct violence, the distribution of political loyalties

determines the extent to which groups find collaboration among the local population that enables

45
“Mere presentation of regression coefficients and their standard errors is inadequate for the interpretation of
intearactive effects. As we have seen, the estimated effects of variables involved in interactive terms and the
standard errors of these estimated effects vary depending on the values of the conditioning variables” (Kam and
Franzese 2007: 60).
46
The distribution of volatile and non-volatile localities, as we have classified them, is a follows: in
Catalonia, 37.38% classified as non-volatile, and 62.62% classified as volatile. In Aragon, 11.68% classified as non-
volatile, and 88.32% classified as volatile.
23
a more ‘precise’ elimination of political enemies. In the case of indirect violence, political

loyalties serves as an indicator for the armed groups of the extent to which the perpetration of

imprecise attacks is going to be functional in eliminating political enemies.

4. CONCLUSIONS

This paper has sought to explain the dynamics of violence against civilians in a civil war context,

by extending the analytical focus to a civil war that was fought conventionally—namely the

Spanish civil war. I have analyzed sub-national variation in one single civil war in a way that

follows current practice in the field and provides significant empirical leverage. Data collection

at the micro-level (i.e. at the level of the locality), even though it is time-consuming, allows for a

better exploration of the mechanisms than cross-national datasets (e.g. ACLED), or regional

datasets (i.e. Herreros and Criado 2009), which not only have greater measurement issues,47 but

also do not permit precise operationalization for theoretically crucial variables (e.g. geographic

location, political competition). The focus has been on two distinct types of violence that take

place in conventional conflicts: direct violence (in the armed group’s rear territory) and indirect

violence (in the enemy’s rear territory). Obviously, the distinction between these types of

violence can also be made for non-conventional wars, although it may not be as theoretically

relevant.

Several implications follow: because civilian agency matters for direct violence and it

does not matter for indirect violence, political alignments are associated with each of these types

in different ways. On the one hand, direct violence is not linearly associated with the number of

47
In the case of the SCW, the micro-level focus has allowed me to concentrate on regions for which data on
violence is extremely reliable. Indeed, historical research on the conflict has been more developed –and, in fact, has
been pioneering - for the rest of Spain in Aragon and Catalonia. For most of the remaining regions in Spain, there
are still many lacunae and imprecision in the civil war casualties figures (on this topic, see Juliá 2004). Using them
is methodologically ‘risky’.
24
supporters of the enemy group (i.e. we do not observe greater leftist violence in places with a

greater proportion of rightist supporters) –as one could naively assume (i.e. the ‘non-domination

hypothesis’); instead, the relationship is curvilinear, as indicated by the robust significance of the

variable ‘competition’. We observe greater violence in places with a greater degree of parity

between political blocs. I have argued that this is the case because, in these contexts, the armed

group has enemies to kill, and local civilians have an interest in eliminating political enemies (in

order to change the balance of power). On the other hand, we do observe a negative linear

relationship between levels of support for a group in a locality and indirect violence. I have

argued that this is the case because, ceteris paribus, with these indirect attacks the groups are

trying to maximize the elimination of strong supporters of the enemy.

Second, the converging results regarding violence that took place in two different regions

during the same time period, and perpetrated by armed groups with very different characteristics,

indicate that local level political dynamics are particularly robust to explain levels of violence.

While organizational factors seem to be capturing the degree of diffusion of violence across a

rear territory, they are not explanatory of levels of violence. Thus, the findings here are in partial

contradiction with Weinstein (2006) and Humphreys and Weinstein (2006), although they also

indicate that organizational variables can have a complex effect on violence, i.e. operating at

different levels. Their explanatory power may be conditional on the type of conflict; this is

something to be explored in further comparative research. The results also partially challenge

purely strategic approaches such as Kalyvas’s (2006), which overlook the role of local political

factors, and cannot account for violence taking place in civil wars fought conventionally.

25
 In conclusion, the contribution of this paper has been at adding further complexity and

insight into previous theoretical explanations of the phenomenon of wartime violence against

civilians, which now need to be refined in the light of new theoretical and empirical explorations.

26
REFERENCES

Allen, Susan H. 2007. “Time Bombs: Estimating the Duration of Coercive Bombings
Campaigns.” Journal of Conflict Resolution 51(1): 112-133.

Arreguín-Toft. 2001. "How the Weak Win War: A Theory of Asymmetric Conflict."
International Security 26(1): 93-128.

Azam, Jean-Paul. 2006. On Thugs and Heroes. Economics of Governance 7 (1):53-73.

Balcells, Laia. 2009. “Behind the Lines: Causes and Consequences of Violence against Civilians
in Conventional Civil Wars” Ph.D Dissertation. Yale University.

Balcells, Laia. 2010. “Rivalry and Revenge: Violence against Civilians in Conventional Civil
Wars.” International Studies Quarterly 54 (2) (Forthcoming).

Balcells, Laia. 2007. “Rivalry and Revenge. Killing Civilians in the Spanish Civil War.”Working
Paper 2007/233, Juan March Institute, Madrid.

Belkin Aaron et al. 2002. “When is Strategic Bombing Effective? Domestic Legitimacy and
Aerial Denial” Security Studies 11(4): 51-88.

Casanova, Julián. 1985. Anarquismo y revolución en la sociedad rural aragonesa. 1936-1938

(Anarchism and Revolution in the Rural Aragonian Society. 1936-1938). Madrid: Siglo
Veintiuno de Espana Editores.

Casanova, Julián, ed. 2001. El pasado oculto. Fascismo y violencia en Aragón (1936-1939) (The
Hidden Past: Fascism and Violence in Aragón (1936-1939)), 3rd ed. Zaragoza: Mira Editores.

Collier, Paul, and Anke Hoeffler. 2004. “Greed and Grievance in Civil War”. Oxford Economic
Papers 56: 563-595.

Cucó Giner, Antonio. 1970. “Contribución a un estudio cuantitativo de la C.N.T.” (Contribution

to a quantitative study of the CNT). Saitabi XX: 181-202.

Downes, Alexander. 2008. Targeting Civilians in War. Ithaca: Cornell University Press.

Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency and Civil War.” American
Political Science Review 97 (1): 75-90.

Germán Zubero, Luis Gonzalo. 1982. “Elecciones y partidos políticos en Aragón durante la
Segunda República: estructura económica y comportamiento político” (Elections and political
parties in Aragon during the Second Republic: economic structure and political behavior). Ph.D
Dissertation. Departamento de Historia Moderna y Contemporánea, Universidad de Zaragoza,
Zaragoza.
27
Herreros, Francisco and Henar Criado. 2009. “Pre-emptive or Arbitrary. Two forms of Lethal
Violence in a Civil War.” Journal of Conflict Resolution 53 (3): 419-445.

Horowitz, Michael and Dan Reiter. 2001. “When Does Aerial Bombing Work? Quantitative
Empirical Tests, 1917-1999.” Journal of Conflict Resolution 45 (2): 147-73.

Humphreys, Macartan and Jeremy Weinstein. 2006. “Handling and Manhandling Civilians in
Civil War.” American Political Science Review 100(3): 429-447.

Juliá, Santos, ed. 2004. Víctimas de la guerra civil (Victims of the Civil War). Madrid: Temas de
Hoy.

Kalyvas, Stathis N. 2005. “Warfare in Civil Wars.” In J. Angstrom, ed. Rethinking the Nature of
War. Abingdton: Frank Cass.

Kalyvas, Stathis N. 2006. The Logic of Violence in Civil War. New York: Cambridge University
Press.

Kalyvas, Stathis N. and Laia Balcells. 2008. “International System and Technologies of
Rebellion: How the Cold War Shaped Internal Conflict.” Under Review.

Kam, Cindy D. and Robert J. Franzese Jr. 2007. Modeling and Interpreting Interactive
Hypotheses in Regression Analysis. Ann Arbor: The University of Michigan Press.

Ledesma, José Luis. 2003. Los días de llamas de la revolución. Violencia y política en la
retaguardia republicana de Zaragoza durante la Guerra Civil (The burning days of the
revolution. Violence and politics in the republican rearguard of Zaragoza during the Civil War).
Zaragoza: Institución Fernando el Catolico (C.S.I.C).

Lyall, Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks?” Journal of Conflict
Resolution 53 (3): 331-362.
Long, Scott J. 1997. Regression Models for Categorical and Limited Dependent Variables.
London: Sage Publications.

Maldonado, José María. 2007. El frente de Aragón. La guerra civil en Aragón (1936-1938) (The
frontline of Aragon. The civil war in Aragon (1936-1938). Zaragoza: Mira Editores.

Martínez de Baños, Fernando (Ed). 2004. Guerra Civil Aragón (Civil War Aragon). Cuarte de
Huerva: Delsan.

Pape, Robert A. 1996. Bombing to Win. Air Power and Coercion in War. Ithaca: Cornell
University Press.

Petersen, Roger. 2001. Resistance and Rebellion: Lessons from Eastern Europe. New
York: Cambridge University Press.
28
Solé i Sabaté, Josep M. and Joan Villarroya. 1986. Catalunya sota les bombes (Catalonia under
the bombs). Barcelona: Publicacions de l'Abadia de Montserrat.

Solé i Sabaté, Josep M. and Joan Villarroya. 1989. La Repressió a la Reraguarda de Catalunya
(1936-1939) (The Repression in the Rearguard of Catalonia (1936-1939)). Barcelona:
Publicacions de l'Abadia de Montserrat.

Steele, Abbey. 2009. “Seeking Safety: Avoiding Displacement and Choosing Destinations in
Civil Wars.” Journal of Peace Research 46(3): 419-430.

Valentino, Benjamin. 2004. Final Solutions. Mass Killing and Genocide in the 20th Century.
Ithaca & London: Cornell University Press.

Valentino, Benjamin A., Paul Huth and Dylan Balch-Lindsay. 2004. “Draining the sea: Mass
Killing, Genocide, and Guerrilla Warfare.” International Organization 58 (2): 375-407.

Vilanova, Mercè. 2005. Atles Electoral de la Segona República a Catalunya (Electoral atlas of
the Second Republic in Catalonia). Barcelona: Enciclopèdia Catalana.

Weinstein, Jeremy M. 2006. Inside Rebellion: the Political Economy of Rebel Organization. NY:
Cambridge University Press.

Wood, Elisabeth J. 2003. Insurgent Collective Action and Civil War in El Salvador. New York:
Cambridge University Press.

Periodicals

UGT. Boletin de la Unión General de Trabajadores de España (Bulletin of the Workers’s

General Union of Spain). Noviembre 1931. Año III. Núm. 35.

UGT. Boletin de la Unión General de Trabajadores de España (Bulletin of the Workers’s

General Union of Spain). Diciembre 1931. Año III. Núm. 36.

CNT. El Congreso Confederal de Zaragoza. Mayo de 1936 (General confederal Congress of

Zaragoza. May 1936)

Electronic Sources

Conferencia Episcopal Española: http://www.conferenciaepiscopal.es/

Global Ganzeeter Version 2.1: http://www.fallingrain.com/world

Instituto Nacional de Estadística (National Statistics Institute): www.ine.es

29
Institut Cartogràfic de Catalunya (Cartographic Institute of Catalonia), "Coordenades de Caps de
Municipi de Catalunya": http://www.icc.cat/

Tables, Figures and Maps

Map 1. Leftist Direct Violence in Catalonia

30
Map 2. Rightist Indirect Violence in Catalonia

31
Map 3. Rightist Direct Violence in Aragon

32
 Table 1. Dependent and Independent Variables in the Models (Catalonia)

Name of the
Characteristics Data Sources
Variable
Total number of people executed by Solé i Sabaté and Villarroya
Executed Left
the left in a locality (1989)

% support for the Popular Front in the

Support Left Vilanova (2005)
1936 general elections
1 if locality bombed by the
Bombings Nationalists at any time during the Solé i Sabaté (1986)
war, 0 if not
Index from 0 (minimum parity) to 1
Competition Vilanova (2005)
(maximum parity)
% inhabitants affiliated to the CNT in CNT (1936), Cucó i Giner
CNT Affiliation
a locality (1970)
% inhabitants affiliated to the UGT
UGT Affiliation UGT (1931)
in a locality

Inhabitants of the municipality in Instituto Nacional de

Population
1936 Estadística
Dummy variable, 1 if the Conferencia Episcopal
Catholic center municipality had an archbishop in Española
1936; 0 otherwise
Dummy variable, 1 if the
municipality is in a county that shares Solé i Sabaté and Villarroya
Frontline
the military frontline at any time (2005)
during the war, 0 if not
Dummy variable, 1 if the Institut Cartogràfic de
Border municipality is in a county that shares Catalunya
the French border, 0 if not
Dummy variable, 1 if the
Institut Cartogràfic de
Sea municipality is in a county with
Catalunya
seashore, 0 if not
Altitude of the municipality, in Institut Cartogràfic de
Altitude
meters Catalunya
Institut Cartogràfic de
Latitude Degrees (UTM, fus 31, datum ED50)
Catalunya
Institut Cartogràfic de
Longitude Degrees (UTM, fus 31, datum ED50) Catalunya

1 if %left 1936-%left 1933 <=10%

Non-Volatility Vilanova (2005)
0 if %left 1936-%left 1933 > 10%
33
 Table 2. Dependent and Independent Variables in the Models (Aragon)

Name of the
Characteristics Data Sources
Variable
Total number of people executed by
Executed Right Casanova (2001)
the right in a locality

Index from 0 (minimum parity) to 1

Competition Germán (1982)
(maximum parity)

% support for the Popular Front in the

Support Left Germán (1982)
1936 general elections

% inhabitants affiliated to the CNT in CNT (1936), Cucó i Giner

CNT Affiliation
a locality (1970)

% inhabitants affiliated to the UGT in

UGT Affiliation UGT (1931)
a locality

Inhabitants of the municipality in Instituto Nacional de

Population
1936 Estadística
Dummy variable, 1 if the municipality Conferencia Episcopal
Catholic center had an archbishop in 1936; 0 Española
otherwise
Latitude
Latitude degrees Global Gazetteer 2.1
(Frontline, Sea)

Longitude
Longitude degrees Global Gazetteer 2.1
(Border)

Altitude Altitude of the municipality, in feet Global Gazetteer 2.1

1 if violent events in the prewar

Previous Conflict Casanova (1985)
period, 0 if not

34
 Table 3. Executed by the Left in Catalonia. NB and ZINB Models

M1 M2 M2 M3 M4 M4
NB: Number ZINB: ZINB: NB: ZINB: ZINB:
of executed Number of Non- Number of Number of Non-
executed violence executed executed violence
(NB) (Logit) (NB) (Logit)

Competition 1.16*** 1.85*** 0.85 ----------- ------------ --------------

(0.318) (0.15) (0.56)

Support Left ----------- 0.00021 -0.004 0.0055

----------
---------- (0.0049) (0.003) (0.013)
0.265** 0.042 -0.063
Frontline 0.271 0.238*** 1.44*
(0.136) (0.035) (0.227)
(0.152) (0.152) (0.855)
Population 0.402*** 0.114*** -2.1***
(*1000) (0.0002) (0.0002) 0.08 0.35*** -0.008***
(0.0087) (0.12) (0.04) (0.002)
CNT 0.016***
0.046* --------------- 0.126** 0.05*** -------
Affiliation (0.024)
(0.024) (0.062) (0.017)
UGT 0.052*** 0.1
0.0611 0.113 0.041 0.163
Affiliation (0.072) (0.108)
(0.068) (0.092) (0.039) (0.59)
-0.214*** -0.011 -0.44** -2.54** -0.478**
Border -0.282** (0.055) (0.28) (0.18) (0.153) (0.564)
(0.141)
-0.23***
Sea -0.335*** 0.408* -0.373** -0.08** 2.52**
(0.036)
(0.13) (0.14) (0.134) (1.25)
Altitude -0.224*** 0.78** -1.3*** -0.2 -2.7**
(meter*1000) -0.7*** (0.2) (0.33)
(0.2) (0.39) (0.134) (0.39)

Catholic 1.48***
0.991*** --------------- 2.22** 0.88* ---------
Center (0.043)
(0.695) (0.86) (0.468)

0.00091*** 0.153 -0.56

Constant 1.7 1.17*** -1.35
(0.306) (0.147) (0.547)
(0.29) (0.24) (1.17)
-0.423 0.16 0.16 0.67 0.24 0.24
Lnalpha (0.0944) (0.0835) (0.0835)
(0.074) (0.09) (0.09)
1.527 1.17 1.17 1.955 1.27 1.27
Alpha (0.184) (0.106) (0.106)
(0.114) (0.113) (0.113)
Observations 870 582 869
582 869 582
Standard Errors in Brackets
Sig Level: *.1, **.05, *** .001

35
 Table 4. Executions in ‘Nationalist’ Aragon. NB Models

M1 M2 M3 M4 M5

Competition 1.694*** 1.624*** ------------ ----------- -------------

(0.48) (0.47)
Population 0.000 0.000 0.000 0.000 0.000
(0.00) (0.00) (0.00) (0.00) (0.00)
CNT Affiliation 0.016 -0.053 0.003 -0.071 0.024
(0.07) (0.06) (0.07) (0.05) (0.08)
UGT Affiliation 0.116** 0.092*** 0.093*** 0.077*** 0.126**
(0.05) (0.03) (0.03) (0.03) (0.06)
Latitude -0.554* -0.647* -0.419 -0.496 -0.566*
(0.30) (0.33) (0.32) (0.36) (0.30)
Longitude 0.522 0.461 0.443 0.351 0.528
(0.41) (0.44) (0.42) (0.44) (0.41)
Altitude (Feet) -0.001*** -0.001*** -0.001*** -0.001*** -0.001***
(0.00) (0.00) (0.00) (0.00) (0.00)
Catholic Center 2.911*** 2.325*** 2.928*** 2.347*** 2.921***
(1.10) (0.83) (1.10) (0.85) (1.12)
Prewar Conflict --------- 1.287** ---------- 1.334** ----------
(0.51) (0.64)
Supp Left ----------- ------------ 0.023*** 0.022*** 0.074***
(0.01) (0.01) (0.03)
SuppLeft2 --------------- ------------- ------------- ------------- -0.001**
(0.00)
Constant 25.720** 29.626** 20.495 23.688 26.192**
(12.86) (14.09) (13.75) (15.34) (12.78)
LnAlpha
Constant 0.739*** 0.665*** 0.774*** 0.699*** 0.738***
(0.13) (0.13) (0.14) (0.14) (0.13)
Observations 251 251 251 251 251
chi2 140.058 176.454 125.699 167.692 140.883
Standard Errors in Brackets
Sig Level: *.1, **.05, *** .001

36
 Table 5. Executions in ‘Nationalist’ Aragon. ZINB Models
Variable M1 M2 M4 M5
(NB) (NB) (NB) (NB)

Competition 1.360*** 1.177*** ----------- -------------

(0.43) (0.42)
Population 0.000 0.000 0.000* 0.000
(0.00) (0.00) (0.00) (0.00)
CNT Affiliation 0.034 -0.039 -0.073 0.013
(0.05) (0.05) (0.05) (0.05)
UGT Affiliation 0.086** 0.072* 0.052 0.067
(0.04) (0.04) (0.04) (0.04)
Catholic Center 2.021*** 1.661** 1.602** 1.991***
(0.76) (0.74) (0.73) (0.76)
Latitude -0.072 -0.196 -0.141 -0.002
(0.29) (0.28) (0.27) (0.28)
Longitude 0.952*** 0.810** 0.818** 0.912***
(0.35) (0.35) (0.32) (0.33)
Altitude (Feet) -0.001*** -0.001*** -0.001*** -0.001***
(0.00) (0.00) (0.00) (0.00)
Prewar Conflict ---------------- 1.062*** 1.115*** -------------
(0.33) (0.34)
Support Left --------------- --------------- 0.023*** 0.031
(0.01) (0.02)
SuppLeft2 ---------------- ----------------- -0.000
(0.00)
Constant 6.289 11.400 9.320 3.528
(12.15) (12.04) (11.60) (11.98)
INFLATE (Logit) (Logit) (Logit) (Logit)
Competition -1.328 -1.410 ------------ --------------
(1.04) (0.97)
Population -0.002** -0.002** -0.002** -0.002**
(0.00) (0.00) (0.00) (0.00)
CNT Affiliation -0.131 -0.155 -0.201 -0.112
(0.17) (0.20) (0.26) (0.17)
Latitude 0.789 0.741 0.459 1.109
(0.65) (0.64) (0.58) (0.67)
Longitude 2.095 1.803 1.954* 1.501
(1.32) (1.19) (1.08) (1.08)
Altitude (Feet) 0.001*** 0.001*** 0.001*** 0.001***
(0.00) (0.00) (0.00) (0.00)
Support Left -------------- --------------- 0.002 -0.119**
(0.02) (0.05)
SuppLeft2 -------------- ---------------- 0.002**
(0.00)
Constant -31.674 -29.813 -18.779 -45.379
(28.40) (28.10) (25.44) (29.13)
Lnalpha
Constant 0.238 0.124 0.117 0.225
(0.15) (0.15) (0.16) (0.15)
Observations 251 251 251 251
chi2 207.368 218.257 220.324 210.119
Standard Errors in Brackets. Sig Level: *.1, **.05, *** .00

37
Figure 1. Predicted Executions on Level of Prewar Competition (ZINB models)

CATALONIA (Leftist Violence) ARAGON (Rightist Violence)

8

8
7

7
6

Predicted Executions
Predicted Executions

6
5

5
4

4
3

3
2

2
1
1

0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
Competition Competition

38
 Table 6. Executions in Catalonia and Aragon. ZINB Models

Variable M1 M2 M3

Population 0.000*** 0.000*** 0.000***

(0.00) (0.00) (0.00)
ZoneRep 0.038 0.038 0.057
(0.11) (0.11) (0.11)
CNT Affiliation 0.069** 0.069** 0.075**
(0.02) (0.02) (0.02)
UGT Affiliation 0.063** 0.064** 0.068**
(0.03) (0.03) (0.03)
Catholic Center 1.637*** 1.634*** 1.654***
(0.37) (0.37) (0.37)
Border -0.526*** -0.521*** -0.505***
(0.12) (0.12) (0.12)
Sea -0.158 -0.152 -0.081
(0.10) (0.10) (0.10)
Competition 0.511*** -------------- -------------
(0.17)
Compabs --------------- 0.391** --------------
(0.15)
Support Left ---------------- --------------- -0.002
(0.00)
Constant 1.301*** 1.458*** 1.752***
(0.17) (0.14) (0.13)

INFLATE (Logit) (Logit) (Logit)

ZoneRep -2.306*** -2.328*** -2.400***
(0.41) (0.41) (0.41)
Population -0.008*** -0.008*** -0.008***
(0.00) (0.00) (0.00)
Border 0.801*** 0.815*** 0.782***
(0.31) (0.31) (0.32)
Sea 0.385 0.396 0.433
(0.38) (0.38) (0.38)
Competition 0.326 -------------- -----------------
(0.58)
Compabs ---------------- 0.249 ------------------
(0.53)
Support Left ----------------- -------------- 0.002
(0.01)
Constant 2.657*** 2.762*** 2.863***
(0.63) (0.57) (0.54)
Lnalpha
Constant 0.479*** 0.482*** 0.486***
(0.05) (0.05) (0.05)
Observations 1709 1709 1709
chi2 894.482 892.110 886.257
Standard Errors in Brackets. Sig Level: *.1, **.05, *** .001
39
 Table 7. Logit on Bombings (Catalonia)

DV: Bombings (Dummy)

Support Right % -0.015**

(.007)
-0.552**
Latitude (*1000)
(0.22)

0.38
Longitude (*1000)
(0.35)

Altitude -1.4***
(meters*1000) (.5)
Population 0.366***
(*1000) (.079)
0.05*
CNT Affiliation
(.026)
0.16*
UGT Affiliation
(.085)
2.21
Catholic Center
(1.37)

-16.32
Constant
(15.73)

Observations 870

Lr Chi2 (8) = 124.51

Prob>Chi2 = 0.0000
Pseudo R2 = 0.1673
Standard Errors in Brackets
Sig Level: *.1, **.05, *** .001

40
Figure 2. Predicted Likelihood of Bombings on Level of Prewar Rightist Support

.4
.3
.2
.1
0

0 20 40 60 80 100
%Support Right 1936

[Key: grey lines depict 95% confidence interval]

41
Table 8. ZINB Regression with Non-Volatility (Catalonia)

M1
NB (Number of deaths)
Competition 1.333***
(0.36)
Non-Volatility 0.217**
(0.10)
CNT Affiliation 0.044***
(0.02)
UGT Affiliation 0.030
(0.04)
Frontline 0.310**
(0.13)
Border -0.288*
(0.15)
Sea -0.079
(0.14)
Altitude -0.000
(0.00)
Population 0.000***
(0.00)
Catholic Center 0.831
(0.45)
Constant -0.322
(0.34)

Inflate Logit (Non-Violence)

Competition 1.380
(1.40)
Non-Volatility 0.342
(0.43)
CNT Affiliation ---------------

UGT Affiliation 0.030

(0.50)
Frontline 0.804
(0.80)
Border -0.357
(0.51)
Sea 1.622
(1.15)
Altitude 0.002*
(0.00)
Population -0.006***
(0.00)
Constant -1.955*
1.17
Lnalpha Constant 0.147
(0.11)
N 869
chi2 450.572
Standard Errors in Brackets. Sig Level: *.1, **.05, *** .001
42
Table 9. ZINB regressions with Non-Volatility (Aragon)

M1
NB (Number of deaths)

Competition 1.399***
0.44
Non-Volatility 0.213
0.24
CNT Affiliation 0.044
0.05
UGT Affiliation 0.081*
0.04
Latitude -0.122
0.30
Longitude 0.967***
0.37
Altitude (feet) -0.001***
0.00
Population 0.000
0.00
Catholic Center 2.168***
0.78
Constant 8.298
12.64

Inflate Logit (Non-Violence)

Competition -1.480
1.05
Non-Volatility -0.113
0.71
Latitude 0.901
0.69
Longitude 1.825
1.39
Altitude (feet) 0.001***
0.00
Population -0.002**
0.00
Constant -36.718
30.11
Lnalpha Constant 0.232
0.15
N 251
chi2 207.451
Standard Errors in Brackets. Sig Level: *.1, **.05, *** .001

43
Figure 3. Predicted Number of Executions by the Left in Catalonia. Interactive Model

Figure 4. Predicted Number of Executions by the Right in Aragon. Interactive Model

44
 Appendix

Descriptive Statistics of Independent and Dependent Variables

Table A1. Catalonia Dataset

Variable Obs Mean Std. Dev. Min Max

Population 1058 1647.56 19726.11 50 637841

Executed Left 1062 7.54 73.65 0 2328

Support Left 1058 52.27 16.94 2.2 100
Competition 1058 0.88 0.16 0 1

Affiliation CNT % 1062 0.98 4.49 0 49.61

Affiliation UGT% 1058 0.088 1.023 0 20.36

Frontline 1060 0.21 0.40 0 1
Border 1060 0.22 0.41 0 1
Sea 1060 0.28 0.45 0 1
Altitude (Meters) 875 368.22 317.3 0 1539
Catholic center 1062 0.008 0.0865 0 1
Bombings 1062 0.13 0.34 0 1
Non-Volatility 1062 0.626 0.484 0 1
Novolatcomp 1058 0.55 0.4443 0 1

45
 Table A2. Aragon Dataset

Variable Obs Mean Std. Dev. Min Max

Executed Right 366 16.48907 186.1591 0 3543

Support Left 255 27.33264 17.46444 .2503129 81.92336

Competition 255 .672952 .283288 .0099875 .9997024

Population 361 1446.961 8583.183 139 162121

CNT Affiliation 361 .4877972 2.223111 0 26.02069

UGT Affiliation 361 .2685374 1.612073 0 13.22931

Latitude 359 41.75901 .5840134 40.46667 42.76667

Longitude 359 -1.212163 .4631542 -2.15 .15

Altitude (Feet) 359 2404.563 988.3771 567 6676

Catholic Center 366 .0081967 .0902874 0 1

Previous Violence 366 .0519126 .2221542 0 1

46
 Table A3. Catalonia and Aragon Dataset

Variable Obs Mean Std. Dev. Min Max

Executed 2010 11.62055 103.9019 0 3567

Population 1996 1399.688 14833.55 50 637841

Zone Rep. 2010 .818408 .3856039 0 1

CNT Affiliation 2000 .6175011 3.460033 0 77.22008

UGT Affiliation 2000 .1051037 1.047569 0 20.36492

Catholic Center 2010 .0069652 .0831872 0 1

Border 2008 .1792829 .3836846 0 1

Sea 2008 .1469124 .3541067 0 1

Altitude (Feet) 1976 1682.697 1339.706 0 6676

Competition 1716 .8068128 .2277111 0 1

Compabs 1716 .6391341 .2509973 0 1

1716 42.74381 20.7501 .2503129 100
Support Left

Table A4. Aragon Municipalities by Control Zones

Province Nationalist Republican Total

Huesca 77 282 359

Teruel 34 248 282
Zaragoza 255 52 307

Total 366 582 948

47
Figure A1. Number of Priests Executed in a Locality. Histogram

Table A5. Logit on ‘Priest Executed’ (Catalonia)

DV: Priest Executed

Competition 0.489
(0.50)
CNT
Affiliation 0.082
(0.06)
UGT
Affiliation 0.057
(0.08)
Frontline 0.350*
(0.19)
Border -0.090
(0.20)
Sea -0.448**
(0.20)
Altitude -0.001***
(0.00)
Population 0.001***
(0.00)
Constant -0.633
(0.48)
N 862
chi2 66.166
Standard Errors in Brackets Sig Level: *.1, **.05, *** .001

48
 Table A6. Leftist Executions in ‘Republican’ Aragon. NB Models

M1 M2 M3 M4 M5
Competition -0.413 -0.423 -------- --------- ----------
(0.29) (0.29)
Population 0.001*** 0.001*** 0.001*** 0.001*** 0.001***
(0.00) (0.00) (0.00) (0.00) (0.00)
CNT Affiliation 0.021* 0.022* 0.026** 0.026** 0.025**
(0.01) (0.01) (0.01) (0.01) (0.01)
UGT Affiliation -0.088 -0.090 -0.036 -0.037 0.042
(0.06) (0.06) (0.04) (0.04) (0.07)
Latitude -0.179 -0.176 -0.180 -0.176 -0.183
(0.14) (0.14) (0.13) (0.13) (0.14)
Longitude 0.378** 0.371** 0.416** 0.408** 0.368**
(0.17) (0.17) (0.17) (0.17) (0.16)
Altitude -0.000** -0.000** -0.000** -0.000** -0.000**
(0.00) (0.00) (0.00) (0.00) (0.00)
Catholic Center -1.668* -1.695** -1.697** -1.730** -1.668**
(0.86) (0.86) (0.85) (0.85) (0.84)
Prewar Conflict ------------ -0.140 ---------- -0.183 ----------
(0.35) (0.35)
Support Left ------------ ----------- -0.016*** -0.016*** 0.017
(0.00) (0.00) (0.01)
Support Left2 ------------ ------------ ---------- ------------ -0.000**
(0.00)
Constant 8.932 8.807 9.156 9.009 8.869
(5.79) (5.81) (5.68) (5.70) (5.75)
Lnalpha
Constant 0.267 0.266 0.233 0.233 0.209
(0.10) (0.10) (0.10) (0.10) (0.10)
N 395 395 395 395 395
chi2 138.633 138.554 180.567 180.990 161.022
aic 2142.030 2143.886 2132.331 2134.080 2128.024
Standard Errors in Brackets Sig Level: *.1, **.05, *** .001

49
Figure A2. Kernel Density Estimate for Support Right 1936

Kernel density estimate

.025
.02
.015
Density
.01 .005
0

0 20 40 60 80 100
support_right36
kernel = epanechnikov, bandwidth = 3.7

Figure A3. Kernel Density Estimate for Competition

Kernel density estimate

8
6
Density
4 2
0

0 .2 .4 .6 .8 1
competition
kernel = epanechnikov, bandwidth = .02

50
 Figure A4. Kernel Density Estimate for Compabs

Kernel density estimate

2
1.5
Density
1 .5
0

0 .2 .4 .6 .8 1
compabs
kernel = epanechnikov, bandwidth = .05

51