religious institutions and resistance to repression: the

Religious Institutions and Resistance to Repression:

The Bishops Opposed to Argentina’s Dirty War

Pearce Edwards1

July 20, 2020

Word Count: 9,805

Abstract

Can religious leaders who oppose repression reduce its use in authoritarian regimes?Communal elites, such as religious leaders, may oppose such state violence which vio-lates basic rights. This paper argues that these leaders, part of institutions embeddedin local communities and with influence based on traditional power, reduce repres-sion when they oppose dictatorships. The argument’s main implication is tested inArgentina during the Dirty War of its 1976-1983 military dictatorship, using originalarchival data on the country’s Catholic bishops. Opposed bishops are associated withreduced disappearances and killings in areas under their jurisdiction. A variety ofempirical strategies, including an instrumental variables analysis leveraging the popewho appointed bishops, suggest a causal effect of opposed bishops. Evidence is pre-sented that opposed bishops reduced repression through an institutional mechanism,encouraging the actions of likeminded local agents and using public influence.

1PhD Candidate, Political Science, Emory University.

I thank Dan Brinks, Natalia Bueno, Courtenay Conrad, Michael Dodson, Jennifer Gandhi, HoracioLarreguy, Phil MacLeod, Miguel Rueda, Amy Erica Smith, Jeff Staton, Adam Scharpf, Chris Sullivan, AndyWedeman, Jason Wittenberg and the Emory Comparative Politics Reading Group for helpful feedback andsuggestions. The Archive of National Memory, the Center for Legal and Social Studies, the Haroldo ContiLibrary, the Pitts Theology Library, and the Benson Latin American Collection assisted with access toprimary sources. The Roper Center for Public Opinion Research provided access to survey data. Financialsupport came from the Halle Institute for Global Research and Emory Professional Development Support.

Religious Institutions and Resistance to Repression Edwards

Introduction

Religious institutions shape politics, even under dictatorships which restrict or co-opt their

activity. While scholars have shown religious institutions’ ability to sustain political iden-

tity (Wittenberg 2006), motivate participation (McClendon and Riedl 2019), nurture ac-

tivism (Amat 2019, Trejo 2012), encourage public goods provision (Tsai 2007), and stimulate

regime change (Woodberry 2012) in dictatorships, little work—with notable exceptions such

as Braun (2016)—studies how religious institutions affect a key aspect of politics in these

regimes: repression. This omission is striking, given that these institutions—particularly

their leaders—are among the “communal elite” on whom regimes depend to reinforce and

justify their use of coercive power (Slater 2010). If religious leaders oppose the use of coercive

power, they may be able to undercut it.

This paper argues that religious leaders’ opposition to repression decreases its use in

dictatorships. Opposed religious leaders draw on their institutions’ local embeddedness—

regular interaction with adherents in a given area (Carter and Hassan 2019, Harris 1998,

McClendon and Riedl 2019)—to gather information about repression with the assistance of

local agents. Opposed leaders also draw on public influence—derived from their institutions’

traditional power based on knowledge and sacred order handed down from the past (Condra,

Isaqzadeh and Linardi 2019, Nelson 1993, Pattin 2019, Weber 1947)—to increase their threat

to the regime. The argument implies repression is reduced in the areas under the jurisdiction

of opposed leaders. Leaders may reduce repression by interacting individually with the

regime, or by leveraging the institution’s embeddedness and influence with the public.

I examine the argument and its mechanisms in the Argentine military dictatorship of

1976-1983, in which the right-wing regime disappeared and killed thousands in a campaign

known as the Dirty War. I use original archival data on the country’s Catholic Church to

find those few among its leaders—bishops—who opposed repression in a religious institution

1


otherwise aligned with the military (Gill 1998, Morello 2015, Obregon 2005). Bishops could

leverage the resources and power of the Catholic Church, being the institution’s primary

executive authority in areas under their jurisdiction. Through the Church, bishops were

locally embedded: overseeing local religious life and service delivery. Likewise, bishops

had public influence derived from the Church: Argentines—more than 90% of whom were

Catholic adherents in the 1970s—agreed with the Catholic Church more than the military.

Because the Catholic pope appointed Argentine bishops without consideration for local

political, socioeconomic, and religious characteristics, I can describe the relationship between

opposed bishops and repression free of many potential confounding variables. I find a signifi-

cant, robust negative relationship: an average opposed bishop prevented 128 disappearances

and 32 killings in his jurisdiction during the dictatorship. However, identifying the relation-

ship between opposed bishops and repression in Argentina presents a challenge; opposition

and repression are often endogenous (Ritter and Conrad 2016).2 Demonstrating a causal

effect requires opposition neither caused by repression nor correlated with characteristics on

which the regime targets repression. I address endogeneity—and measurement error—using

an instrumental variables strategy which leverages the pope who appointed bishops. I also

rule out the possibility that bishops who supported the Dirty War account for the results.

I also present evidence that opposed bishops reduced repression through using institu-

tional resources, drawing on the local embeddedness and public influence of the Catholic

Church to resist state violence. First, opposed bishops protected and encouraged the actions

of likeminded local agents—priests—and were more effective in reducing repression when

a greater number of likeminded priests served in their jurisdictions. Second, a test using

novel micro-level data from the Argentine capital, Buenos Aires, underscores how priests

make bishops more effective: priests in opposed Catholic religious orders—organizations

2That is, repression affects opposition to repression or repression is targeted based on characteristicscorrelated with leaders’ opposition, such as demography.

2


with some independence from bishops—are associated with reduced repression in nearby

areas. Third, opposed bishops encouraged ecclesial base communities, progressive Catholic

networks which subverted the military regime. Fourth, I present evidence that opposed bish-

ops’ use of public influence derived from the Church had a lasting effect on public attitudes:

areas with opposed bishops had more Catholic adherents in 1988.

Existing repression research emphasizes states’ use of coercive power against domestic

threats (Blaydes 2018, Davenport 2015, Ritter and Conrad 2016). I consider a different

type of power—“traditional” or “symbolic” power (Condra, Isaqzadeh and Linardi 2019,

Nelson 1993, Weber 1947)—and describe how religious leaders use this power to command

local agents and exercise public influence (Brooke et al. 2020, Kubik 1994). The findings

connect literature on repression with research on religious leaders’ influence in dictatorships

(Braun 2016, Wittenberg 2006) and politics (Grzymala-Busse 2016, Spenkuch and Tillmann

2018). Incorporating religious leaders and other sources of traditional power into repres-

sion research can also extend theories about how subnational variation in regime opposition

shapes regimes’ strategies of repression (Blaydes 2018, Carter and Hassan 2019).

More generally, I draw on and further develop ideas from research on religious leaders,

civil society under dictatorship, and the role of the Catholic Church in 20th-century Latin

America (Gill 1998, Hale 2015, Mainwaring and Wilde 1989, Smith 2019, Trejo 2012). I show

that even a religious institution with a reputation for close alignment with dictatorship and

a national monopoly on religious adherence—such as the Catholic Church in Argentina—

may be divided: a minority of leaders can break with the institution’s prevailing views and

shape political outcomes in the ways and locations in which they have authority. This finding

suggests that looking within social or religious institutions otherwise reputed to be monolithic

or co-opted by regimes reveals rich variation in their subnational activities. Furthermore,

analysis within institutions reveals patterns, such as complementarities between leaders and

local agents, which could explain these institutions’ broader influence on politics.

3


The Effect of Religious Leaders on Repression

Social institutions, devised for structuring citizens’ social interactions, have “repeated, strong

interactions” with political regimes (Tilly 2006). I examine the interaction between one type

of regime—dictatorship—and leaders of one type of social institution—religious institutions.

Religious institutions engage their adherents with “ideational content, rituals and practices,

organizational resources, hierarchies, social networks, social identity categories, social insur-

ance, [and] service delivery” (McClendon and Riedl 2019, 26).

Religious institutions have acquiesced to dictatorships, standing by or lending influence

to repressive regimes (Slater 2010). In the Nazi regime, for example, Germany’s “rich as-

sociational life provided a critical training ground for eventual Nazi cadres and a base from

which the National Socialist German Workers’ Party (NSDAP) could launch its seizure of

power” (Berman 1997, 402). Spenkuch and Tillmann (2018) find that Catholic local religious

agents—priests—with connections to the NSDAP influenced adherents’ views, increasing vot-

ing for the fascist party in interwar German elections even as the Catholic Church initially

opposed the NSDAP. In Cold War-era Latin America, religious institutions often aligned

with military dictatorships, seeking the access to power these regimes offered (Gill 1998).

On the other hand, religious institutions have opposed dictatorships. Through education

and movements for social reform, religious opposition to colonial regimes made transitions to

democracy more likely (Woodberry 2012). In the Nazi-occupied Netherlands, local religious

communities helped Jews evade capture and deportation (Braun 2016). In Communist-era

Hungary, Catholic priests preserved citizens’ religious and political identities (Wittenberg

2006). In Sri Lanka, religious officials spoke against Tamil rebels’ “autocratic behavior”

(Gowrinathan and Mampilly 2019). In Mexico during the PRI regime, religious networks

in indigenous communities provided connections and resources necessary for mobilization

against economic reforms (Trejo 2012).

4


This paper focuses on leaders of religious institutions who oppose repression—the use

of state coercive power which violates basic rights (Davenport 2015)—in dictatorships. Op-

posed religious leaders are those who make statements and take actions which indicate dis-

approval of state violence.3 In so doing, their institutions become “reservoirs of....resistance

to arbitrary or tyrannical action” (Schmitter 1993, 15). Religious leaders who oppose repres-

sion amplify their resistance with two characteristics derived from their institutions: local

embeddedness and public influence.

First, religious leaders draw on institutions’ local embeddedness, aggregating local knowl-

edge through repeated interactions with adherents and agents (Carter and Hassan 2019,

McClendon and Riedl 2019, Wittenberg 2006). Leaders build “knowledge of the jurisdic-

tion” and “social and professional bonds” from these interactions (Carter and Hassan 2019,

6). Social and professional bonds facilitate the exchange of news, and thereby allow lead-

ers to gather information about local events (Harris 1998). Because news about repression

spreads by word of mouth (Blaydes 2018), leaders of locally embedded institutions learn

about state violence in their jurisdictions. In particular, leaders may learn about repression

from local agents who are most embedded with adherents (Braun 2016).

Second, religious leaders exercise public influence. The religious institution’s symbols and

trappings give leaders traditional power, “the sanctity of the order and the attendant powers

of control as they have been handed down from the past” (Nelson 1993, 656). Leaders’

traditional power lends influence distinct from the state’s legal power (Slater 2010, Weber

1947): religious adherents, members of the public, and even the regime and its agents give

religious leaders’ statements additional weight when forming attitudes and taking actions

(Condra, Isaqzadeh and Linardi 2019, McClendon and Riedl 2019). Thus, by acting on

information about state violence in their jurisdiction and opposing its use, religious leaders

become opinion leaders (Smith 2019). Leaders’ traditional power, used to oppose repression,

3In the paper, “repression” and “state violence” are used equivalently.

5


erodes the regime’s legal power and its ability to justify the use of coercion (Slater 2010).

Opposed religious leaders threaten the regime, as knowledge about repression, institu-

tional trust, and traditional power make them a “creator, repository, and propagator...of

values” parallel to the state (Kubik 1994, 119). The regime may control opposed religious

leaders with co-optation, further repression, or accommodation (Gandhi 2008). Co-opting

opposed leaders, attempting to purchase their loyalty, is one option. Yet leaders’ opposition

is often intransigent and rooted in moral conviction, making co-optation ineffective (Wit-

tenberg 2006). Repressing opposed leaders is another option. While repression may control

opposition from local agents or adherents, leaders’ institutional position gives them promi-

nence and influence with the public such that repression is counterproductive. Repressing

leaders angers adherents, leading to a net increase in opposition (Esberg 2020, Gautier 1998).

The risk of increased opposition limits repression: in Argentina during the Dirty War, for

example, the regime repressed less than two percent of opposed religious officials.4

The regime’s remaining option is to accommodate opposed leaders, conceding to them

and thereby reducing repression in areas under their jurisdiction. This produces an observ-

able implication: in areas of a regime’s territory where religious leaders oppose repression,

repression is reduced compared to areas where religious leaders do not oppose repression.

Furthermore, opposed leaders may reduce repression through two non-exclusive mechanisms:

leveraging institutional resources and taking individual actions.

Institutional Resources: In this explanation, religious leaders reduce repression through

drawing on the resources of the religious institution, particularly its local embeddedness and

public influence with adherents. First, leaders protect and encourage local agents who also

oppose repression. Among other actions, these agents help potential targets of repression

evade state violence (Braun 2016, Brooke et al. 2020). Religious institutions are a focal

point for evasion: networks of agents and adherents create a “free space” for these targets

4This figure is based on data from Pattin (2019) and CONADEP (1984).

6


(Gautier 1998). Second, religious leaders draw on the institution’s public influence to in-

crease opposition to repression: their political statements draw on religion’s traditional power

and influence public opinion (Condra, Isaqzadeh and Linardi 2019, Spenkuch and Tillmann

2018), while their actions include providing institutional resources to enable collective action

against repression (Amat 2019, Trejo 2012, Woodberry 2012).

Individual Actions: Under this mechanism, religious leaders appeal directly to regime

agents to reduce repression. Regime agents respond to these appeals, especially if the leaders

uplift the moral standing of agents who behave consistently with leaders’ values (Tsai 2007).

Agents who seek this approbation from religious leaders who oppose state violence may defy

orders to repress or even defect, thereby reducing repression.

The Argentine Dirty War and the Catholic Church

To test the argument’s main implication, I examine leaders of the Catholic Church—an im-

portant worldwide religious institution—in Argentina during the country’s 1976-1983 right-

wing military dictatorship. The lead-up to the dictatorship began when Juan Peron, pres-

ident of Argentina from 1946 to 1955, returned to power in 1973 via democratic elections

(O’Donnell 1988). However, Peron died in 1974 and his wife and vice president Isabel suc-

ceeded him. By 1975, labor conflict grew in response to inflation and fiscal austerity. The

military responded by repressing leftist groups in coordination with the paramilitary Argen-

tine Anticommunist Alliance. As violence escalated, the military staged a coup on March

24, 1976 and installed a dictatorship.

The new military regime implemented what became known as the Dirty War, deploying a

“systematic utilization of terror” against the population which extended far beyond its stated

targets (Obregon 2005, Sikkink 2004). The regime decentralized decision-making about

repression, empowering special repressive battalions and military officers assigned to various

7


geographic zones to kidnap, disappear, torture, and execute thousands of persons, taking

them from homes and workplaces to nearby clandestine detention centers (Scharpf and Glaßel

2019). Dismantling organized labor, the regime repressed unions, especially those in firms

with regime connections (Klor, Saiegh and Satyanath 2017). To prevent news of repression

from spreading, the regime blocked publication of victims’ obituaries in newspapers, forbade

public announcements of deaths, and obstructed human rights investigations (Morello 2015).

The regime thus kept the Argentine public uninformed about the scale of violence.5

The Argentine Catholic Church was known for its ideological alignment with the military

during the Dirty War (Gill 1998, Scharpf 2018). Many Catholic religious leaders—bishops—

“condoned the association that the military made in its public statements between state

terrorism and Christian virtues” (Romero and Brennan 2013, 238), and the military sought

the Church’s aid in justifying repression (Argentine Army 1975, Finchelstein 2014).6 To this

end, “the very day of the coup, March 24, 1976, members of the military junta met for a

long time with Archbishop Adolfo Tortolo of Parana...president of the Bishops’ Conference

of Argentina” (Mignone 1986, 19).

Substantively, the Argentine Church fits the scope of the argument: despite the pre-

vailing alignment of church and military, some bishops opposed the military’s campaign

of disappearances and killings. Admiral Emilio Massera, a leader of the coup, remarked

that “...bishops would influence military officials such that the military officials would carry

out less bloody repression” (Diario de Juicio 1985). For example, the “families of political

prisoners and disappeared persons in [the province of] Misiones....made public their apprecia-

tion...especially to the bishop of Posadas, Monsignor Jorge Kemerer...for his interventions in

favor of the freedom of the political prisoners of the province” (La Razon 1983). A prisoner

5Morello (2015) notes: “No previous military regime had resulted in such a bloodbath. Because of this,in the first months, no one believed the rumors” (76).

6National-level Catholic leadership shaped the ideology of the regime, posing the fight against leftistopponents as a defense of “Western Christian civilization” (Romero and Brennan 2013).

8


recounted that “Monsignor Marengo, Bishop of Azul, was worried about my fate and spoke

with General Alfredo Saint Jean, asking that they not kill me, which worked” (CONADEP

1984). Juan Tome, Bishop of Mercedes, “intervened before the authorities to liberate” six

persons (Excelsior 1979). I later evaluate possible mechanisms to explain this influence.

Furthermore, the Argentine Church was locally embedded and possessed public influence.

Bishops are the executive of a Catholic jurisdiction known as a diocese. Within each diocese,

bishops control the employment of agents—priests—who are embedded in local Catholic

parishes. Bishops also supervise a bureaucracy which administers religious life: in 1971 in

the Argentine capital of Buenos Aires, for example, the bishop oversaw 155 parish churches,

116 community religious organizations, and 262 schools (Archdiocese of Buenos Aires 1971).

For influence, opinion data suggests the Argentine public—up to 90% of whom were Catholic

adherents in the 1970s—agreed with the Catholic Church more than the military. In a 1971

survey asking the institutions with which respondents most agreed, 78% expressed as much

or more agreement with the Catholic Church as they did with the military (Turner 1971).

There is also evidence Catholic bishops were appointed independently of regime politics

and socioeconomic conditions.7 A Concordat between the Vatican and the 1966-1973 military

dictatorship stipulated the “Argentine state recognizes and guarantees the Roman Catholic

apostolic Church...jurisdiction in the sphere of its own competence” (Mignone 1986, 76). This

Concordat ended a system of state-controlled appointments, and Catholic popes thereafter

appointed bishops. Accordingly, Argentine regimes treated appointments with “kid gloves”

for fear of angering the pope (Gill 1998, 160). For example, though Admiral Massera wanted

several opposed bishops to resign, he pressed the issue only lightly (Morello 2015). Popes

appointed bishops according to personal preferences, tending to overlook bishops’ political

activities: “bishops [in Argentina] are nominated in gratitude for favors...because they had

money, or they had done something good as the church perceived it” (Marchak 1999, 252).

7Trejo (2012), Osorio, Schubiger and Weintraub (2020), and Tunon (2018) make similar arguments.

9


Research Design

Measuring Bishop Opposition

To test the effect of bishop opposition, I first determine which Argentine Catholic bishops op-

posed the regime’s repression. I create a biographical sketch for each of the 82 bishops during

the Dirty War period using sources drawn from three Argentine human rights archives, three

contemporaneous Catholic periodicals, and declassified United States intelligence. From this

sketch, I code a bishop as opposed to repression if he made pre-regime statements and ac-

tions supportive of the leftist groups the regime repressed, as well as if he made pre-regime

statements and actions opposing state violence and endorsing human rights.8 Bishops are

then linked to departments under their jurisdiction, Argentina’s second-level administrative

division and the geographic unit of analysis.9

An example of an opposed bishop is Antonio Brasca, of the diocese of Rafaela. From

1968 to 1970 Brasca participated in meetings of the progressive Movement of Priests for the

Third World (MSTM)—a clerical movement linked with the political left which spoke against

perceived socioeconomic injustice, was labeled a “red bishop” in 1974 for his progressive

leanings, and refused the military’s request that he hold a memorial for an officer. An

example of an unopposed bishop is Desiderio Collino, of the diocese of Lomas de Zamora.

In 1974 he denounced the MSTM, and during the regime he opposed the investigations of

human rights organizations into repression. In all, twenty-five bishops made statements or

took actions indicating opposition to repression. Fifty-five bishops ranged from statements

and actions indicating no position on repression to statements and actions indicating support

for repression. Two bishops lacked sufficient information to code.

I create the main explanatory variable Bishop Opposed i,t and assign it the value of 1 if de-

8The appendix reports the full coding procedure, as well as a list of sources used in coding.9In the Buenos Aires province, this division is a partido. For simplicity, all units are labeled departments.

10


Figure 1: Bishop Opposition to the Dirty War, 1976

Note: Map depicts Argentine bishop attitudes towards the military regime’s campaign of repression in 1976.Each geographic division is a department. Dioceses are composed of departments.

partment i in year t had an opposed bishop, and 0 otherwise. Because some bishops changed

during the dictatorship, opposition varies both between years and across departments.10

Opposition is linked to geography because information about repression in department i is

most likely to come to the attention of the bishop whose diocese contains department i.11

Figure 1 depicts a cross-section of bishop opposition from 1976. It is possible there is mea-

surement error in bishop opposition which could bias estimates with this variable. After the

10Retirement and natural death induce bishop changes (Tunon 2018). Sixteen dioceses (27%) had morethan one bishop between 1976 and 1983.

11Bishops must live in their dioceses and acquire local knowledge per Catholic Canon 396—“the Bishopis bound to visit his diocese in whole or in part each year”—and Canon 383—“the diocesan Bishop is to besolicitous for all Christ’s faithful entrusted to his care.”

11


presentation of the main results, I address this form of error.

Measuring Repression

For the dependent variables of interest, I create Disappearances i,t, the count of disappear-

ances in department i in year t, and Killings i,t, the count of killings in a given department-

year.12 These variables record the location in which the repression occurred: typically in

or near the victim’s place of residence, rather than—for victims of killings—the location

in which the victim was found. Data are from the Argentine National Commission on the

Disappearances of Persons (CONADEP) report, which records 8,961 victims of repression

during the Dirty War and specifies the location and date of each disappearance and killing.

These data are “highly reliable” sources (Sikkink 2004, 96).

Nonrandom measurement error in the dependent variables could, if present, bias an

estimate of the relationship between bishops’ opposition and repression. Bishops unopposed

to repression could have obstructed CONADEP as it gathered data on disappearances and

killings after the collapse of the regime. This would lead to fewer reports of disappearances

and killings in departments under these bishops’ jurisdiction.13 Conversely, bishops opposed

to repression helped CONADEP gather data on disappearances and killings (Brysk 1994).

However, such systematic measurement error would bias estimates against the expectations

of the argument, meaning I likely underestimate the effect of opposed bishops on repression.

Other Covariates

I include covariates which measure departments’ socioeconomic characteristics before the

dictatorship. From the 1970 Argentine census, I include literacy rates, the log of total

population, the proportion of young men, and the proportion of foreign-born residents. Each

12These were the modal types of repression used by the regime during the dictatorship.13The Argentine human rights community maintains CONADEP’s tally underestimates the count of dis-

appearances and killings by a factor of three or more.

12


covariate corresponds with potential predictors of repression: it often occurred in urban areas,

targeted organized labor, and—through collusion with other regimes in what was known as

Plan Condor—citizens of other countries suspected of leftist activism. Furthermore, opposed

bishops may have sought out areas with characteristics reflecting their social preferences.

I account for departments’ political characteristics with two covariates. First, I include

pre-dictatorship voting behavior with a covariate for the vote share of the Peronist Justicialist

party in September 1973 presidential elections (Ministry of the Interior 1973).14 Second, I

include the pre-dictatorship presence of left-wing militants in a department based on military

intelligence reports (Argentine Army 1975). I also account for religious characteristics which

could affect bishop opposition and repression: whether the bishop served an archdiocese, the

bishop’s tenure in office, and the count of priests affiliated with the MSTM in 1967.15

How Are Bishops Appointed?

Bishops’ appointments were independent of regime politics for two main substantive rea-

sons. First, these appointments fill vacancies which opened when a bishop reached a church-

mandated retirement age of 75 or died in office, rather than at the discretion of a domestic

political actor (Tunon 2018). Second, appointments to vacancies were consistent with the

1966 Concordat, which recognized the “independence of the church in its spiritual domain,

the naming of bishops, and the creation of ecclesiastical boundaries” (Villa 2000, 253). Not

only did the Catholic Church have autonomy in appointing bishops, but the boundaries of

bishops’ authority were outside state control.16

Even with evidence of the church’s autonomy, prospective bishops nevertheless could

14These were the last democratic elections held before the dictatorship.15The bishop tenure covariate also accounts for a potential SUTVA violation in which a unit’s exposure to

treatment is a function of the bishop’s time in office. With this covariate, I am able to compare department-years with similarly-tenured bishops.

16Half of provinces contained multiple dioceses, and three dioceses spanned provinces. Five-sixths of thedictatorship’s military subzones contained multiple dioceses, and twelve dioceses spanned subzones.

13


have sought appointments to areas with their preferred political or demographic character-

istics. For example, bishops who opposed repression could have requested appointments in

dioceses with MSTM presence. To rule out this selection process, I compare pre-dictatorship

characteristics of department-years with bishops who opposed repression with characteristics

of department-years with bishops unopposed to repression. I regress the bishop opposition

variable on political and demographic characteristics and the count of priests who in 1967

affiliated with MSTM. Department-years are balanced on pre-dictatorship characteristics be-

tween bishops who opposed and bishops who did not oppose repression: none of the estimates

are significantly different from zero. Figure 2 depicts results from the test.17

Figure 2: Balance of Department Characteristics by Bishop Opposition

Note: Figure depicts estimated differences in department-year characteristics between opposed and unop-posed bishops. Variables are standardized through z-transformation, with 95% confidence intervals displayed.

Finally, bishop appointments during the dictatorship could have become endogenous.

17Full results are in the appendix, including a diocese-level balance test which remains robust, a test forspatial clustering of opposition which indicates spatial independence, and evidence that appointments madeduring the 1966-1973 dictatorship, after the Concordat, were likewise balanced.

14


That is, once repression began, the Vatican could have conditioned the appointment of

opposed bishops on the level of repression in departments under their potential jurisdiction.

Opposed bishops may have replaced unopposed bishops where repression was greater as an

effort by the Vatican or the bishop to obstruct the regime. Opposed bishops also may have

replaced unopposed bishops where repression was lesser if the Vatican sought to prevent

opposed bishops from disrupting a larger amount of state violence.

To address the possibility of strategic appointments, I conduct a test with an explanatory

variable for a future change to an opposed bishop, ∆Bishop Opposedi,t+1. The variable

measures whether department i changed from an unopposed bishop to an opposed bishop

between year t and year t + 1. If opposed bishops were appointed based on current levels

of repression, then this variable would have a significant relationship with the dependent

variable of disappearances and killings in year t. Results from this test indicate no consistent

relationship between future appointment of opposed bishops and current repression.18

Statistical Estimation

I use linear regression to estimate opposed bishops’ effect on repression, including specifica-

tions with time and geographic-unit fixed effects. The estimating equation with the different

measures of repression, a vector of department-level covariates Xi, department fixed effects

γi, year fixed effects µt, and standard errors clustered at the diocese level εi is:

Repressioni,t = α + β · Bishop Opposedi,t + δ ·Xi + γi + µt + εi

The coefficient estimate for bishop opposition is identified if opposition is independent

of confounding variables conditional on included covariates. I include year fixed effects to

account for events affecting repression across Argentina in particular years, such as the regime

18The appendix reports full results from this test.

15


scaling back repression before a 1979 visit from the Interamerican Commission on Human

Rights. Department fixed effects account for unobserved, time-invariant geographic unit

characteristics. Some specifications replace department fixed effects with military subzone

fixed effects to account for unobserved commanding officer characteristics—the dictatorship

gave considerable autonomy to the commanding officers in these subzones (Scharpf 2018)—

which could affect bishop opposition and repression. I also report a specification with no

geographic unit fixed effects, given concerns about the ability of two-way fixed effects model

to adjust for both temporal and unit confounding (Imai and Kim 2020).

Analysis

The Effect of Opposed Bishops

Opposed bishops reduced repression in areas under their jurisdiction: the estimated change in

disappearances and killings which corresponds to an opposed bishop in a given department-

year is consistently negative and statistically significant. There were between one and two

fewer disappearances and between one-third and one-half fewer killings per year, on average,

in department-years with an opposed bishop than those with an unopposed bishop. Given

the average diocese has eight departments and the dictatorship lasted almost eight years,

an average bishop who opposed repression could have reduced disappearances by 128 and

killings by 32 had he been a bishop throughout the dictatorship. Furthermore, this reduction

is underestimated if opposed bishops assisted in reporting repression.

I report coefficient estimates for opposed bishops with greater detail in Table 1. The

first four columns report estimates from specifications with the first dependent variable,

Disappearances, while the second four columns report estimates from specifications with the

second dependent variable, Killings. These estimates are consistently negative.

16


Table 1: Estimated Effects of Opposed Bishops on Repression

Disappearances Killings(1) (2) (3) (4) (5) (6) (7) (8)

Bishop Opposed −1.34∗ −2.15† −1.04† −2.19† −0.49∗ −0.49∗ −0.36∗ −0.58∗

(0.61)1 (1.12) (0.58) (1.20) (0.21) (0.23) (0.18) (0.28)

Adj. R2 0.27 0.07 0.29 0.24 0.13 0.10 0.16 0.16Observations 3851 3763 3731 4003 3851 3763 3731 4003Covariates N Y Y N N Y Y NFixed Effects Year, Sub.2 Year Year, Sub. Year, Dpt.3 Year, Sub. Year Year, Sub. Year, Dpt.

∗∗p < 0.01, ∗p < 0.05, †p < 0.101 All standard errors clustered at the diocese level.2 Military subzone fixed effects.3 Department fixed effects.

Robustness Tests

I first address the possibility of measurement error in the explanatory variable first by using

stricter coding criteria: only bishops who took repeated public actions against repression

are coded as opposed. The estimates, reported in the appendix, are generally larger in

magnitude and increased in significance. I next recode each bishop’s opposition one at a

time and re-estimate the main results. The procedure determines whether the main results

are sensitive to small changes in measuring bishop opposition. The results, reported in the

appendix, suggest almost complete robustness to measurement error for any single bishop.

The distribution of the dependent variables—repression concentrated in urban areas and

in the first years of the dictatorship—presents several possible concerns. First, outliers could

drive the results: repression in the capital city of Buenos Aires was about sixteen standard

deviations higher than any other department. The main specifications are re-estimated with

data dropping Buenos Aires. I continue this procedure, dropping each diocese in turn and

re-estimating the main specifications.19 Second, excess zeros in the dependent variables

could lead to a violation of the normality assumption for the regression models’ error terms.

To address this problem, I re-estimate the main specifications with data only from 1976 to

19This also addresses a concern that the death of the Bishop of La Rioja in a suspicious automobile accidentcould introduce endogeneity if the regime sought to manipulate the appointment of his replacement.

17


1978, during which more than 90% of disappearances and killings occurred. I also aggregate

the data to the diocese-year level and re-estimate the main specifications. Finally, negative

binomial regression models are specified. Results remain consistent across these tests.

Spillovers also present a threat to inference: opposed bishops could have affected repres-

sion in nearby departments without an opposed bishop. This would be a concern if, for exam-

ple, a military officer with authority over departments with and without an opposed bishop

redirected repression to the departments without an opposed bishop. To test for spillovers,

I create two new variables, Bishop Opposed, 50kmi,t and Bishop Opposed, 100kmi,t, which

count the departments under the jurisdiction of an opposed bishop within 50 and 100-

kilometer radii of department i in year t. These variables are then interacted with Bishop

Opposed in a linear regression model (Ichino and Schundeln 2012). If there are spillovers,

the baseline terms for Bishop Opposed, 50km and Bishop Opposed, 100km in the interaction

would be positive and significant. That is, repression in department-years with an unop-

posed bishop would increase in the count of nearby departments with an opposed bishop.

The baseline terms approach zero and are not significant.

There are also potential problems in how I account for temporal effects. First, includ-

ing time and geographic unit fixed effects in the model could fail to adjust for confounding

from these sources (Imai and Kim 2020). I address this potential problem by replacing year

fixed effects with a linear trend, subtracting time-dependent patterns in repression while

also allowing for valid comparisons in specifications which also include unit fixed effects.

Accounting for such patterns may also be preferable to year effects, given repression strictly

decreased from 1976 to 1983. Results remain consistent with a linear trend. Second, covari-

ates could have time-varying effects, whereas the included covariates are time-invariant. To

address this problem, I interact covariates with year fixed effects and find consistent results.

Finally, I subject the main results to a sensitivity analysis. The procedure derives a

robustness value—the level of a potential unobserved confounding variable’s association with

18


bishop opposition and repression needed to render the main results insignificant—measured

by partial R2 (Cinelli and Hazlett 2020). This procedure is preferable to other sensitivity

analyses because it makes no assumptions about the functional form of bishop opposition

assignment or the unobserved variable’s distribution. The results, reported in the appendix,

indicate such a variable must explain more variation in repression than population to render

the main results for killings insignificant, and explain about as much variation as population

to render the main results for disappearances insignificant.

An Instrumental Variables Strategy

There remain two challenges to interpreting the main results as causal estimates of bishop

opposition to repression. First, bishops’ opposition could be the effect, rather than the cause,

of less repression. If this is the case, only bishops in areas with less repression could have

felt secure enough to oppose repression. Second, there could still be measurement error in

the explanatory variable. Bishops who did not oppose repression either in their statements

or actions may have feigned a lack of opposition to gain more influence with the regime in

reducing repression (Gautier 1998).

To address these challenges, I build on the work of Tunon (2018) and employ an instru-

mental variables strategy which uses as an instrument the pope who appointed each bishop.

John Paul II, who became pope in 1978, preferred bishops who were less theologically pro-

gressive than bishops appointed by other popes: he had “a conservative, orthodox and

somewhat authoritarian worldview” early in his tenure (Kirk 1985). Because popes appoint

bishops according to personal preferences, bishops appointed by John Paul II are more likely

conservative and orthodox and thus less likely to oppose the right-wing military dictator-

ship. The exogenous change—popes take power upon natural death of their predecessors—in

which pope appointed Argentine bishops is a potential instrument for bishop opposition.20

20Paul VI, John Paul I, and John Paul II were each pope during the regime. Tunon, who employs a similar

19


The instrumental variables procedure is as follows: the first stage regresses bishop oppo-

sition on Pope Liberal, an indicator of whether John Paul II did not appoint the bishop. This

is a strong instrument, with a first-stage F -statistic of 14. The second stage uses predicted

values from the first stage regression to form the instrumented variable Bishop Opposition

and makes it the right-hand-side predictor for the dependent variables, Disappearances and

Killings. Because John Paul II became pope later in the regime and patterns of repression

varied over time, I include a time trend to achieve conditional independence.

Results from the two-stage procedure address concerns about reverse causality: the esti-

mated local average treatment effect—the effect of bishop opposition among those bishops

whose opposition was a function of being appointed by a different pope than John Paul II—is

3.39 fewer disappearances and 1.07 fewer killings in a given department-year. These esti-

mates are significant at the p = 0.1 level.21 Given the instrumental variables estimates are

larger in magnitude than coefficient estimates in the main results, the explanatory variable

likely contains some random measurement error.

Alternative Explanation: Supporting Bishops

An alternative explanation is that the results reflect supporting bishops increasing repression

rather than opposed bishops reducing it. While both processes could occur, the effects

of opposed bishops may not be robust to accounting for supporting bishops. To address

this possibility, I first code unopposed bishops as either supporting—making statements or

taking actions consistent with support for repression—or neutral.22 I include this bishop

support variable in model specifications alongside bishop opposition, with neutral bishops as

the reference category. If the alternative explanation is correct, the coefficient estimate for

strategy with Brazilian bishops, instruments appointing pope with exogenous vacancies given bishops maystrategically resign. However, such a strategy is unnecessary here: no bishops in the sample resigned.

21The appendix reports full results, including validation that appointing pope affects repression onlythrough bishop appointments: popes have no relationship with repression in areas with unopposed bishops.

22This procedure is similar to coding bishop opposition, which is reported in greater detail in the appendix.

20


opposed bishops would be zero, while the coefficient estimate for supporting bishops would be

positive. Results, reported in the appendix, suggest the opposite: opposed bishops’ effects

are negative and significant even in comparison to neutral bishops. Supporting bishops,

meanwhile, have no less repression in areas under their jurisdiction than neutral bishops.23

Testing the Institutional Mechanism

How did opposed bishops reduce repression during Argentina’s Dirty War? There are two

non-exclusive mechanisms through which these religious leaders could have effects: insti-

tutional resources and individual actions. In this section, I provide evidence that opposed

bishops’ effects are consistent with an institutional mechanism in two ways. First, opposed

bishops protected and encouraged the actions of likeminded local agents—priests—who were

also embedded among adherents. Second, opposed bishops’ use of their institution’s public

influence affected Argentine attitudes beyond the dictatorship. While individual actions may

also have been at work, systematic data is difficult to obtain given the Church’s withholding

of Dirty War-era records.24

Qualitative evidence suggests bishops’ use of institutional resources—the Church’s local

embeddedness and public influence—were a means of reducing repression: Morello (2015),

evaluating Catholic resistance, argues the regime “responded better to the pressure of civil

society than to actions taken in private” (186). Using institutional resources involved opposed

bishops protected and encouraged likeminded local religious agents, such as priests. For

example, opposed bishops may have worked with these agents to shelter potential targets

of repression. Potential targets disclosed the existence of “a network of safe houses” with

Church support which concealed “those who would have disappeared” (Duzdevich 2019).

Bishops and priests also visited prisoners to offer the Church’s aegis (Mignone 1986).

23I conduct an additional test to assess whether supporting bishops increased repression in the appendix24In the appendix I test proxies for individual actions which suggest they may have been ineffective.

21


Bishops and the Movement of Priests for the Third World

If opposed bishops protected and encouraged opposed priests’ actions, these bishops should

be more effective at reducing repression with a larger number of likeminded priests in their

diocese. Priests opposed to repression, particularly those in the Movement of Priests for

the Third World (MSTM), were activists in Argentina’s poor communities and criticized the

Argentine military from the late 1960s onward (Bresci 1994). Bishops, in turn, could create

conditions favorable for MSTM priests to act with their authority to appoint priests, to

make rules for the diocese, and to issue specific precepts which obligate priests’ behavior, To

that end, opposed bishops gave MSTM priests a “wide range within which to operate” and

“protected them” (Dodson 1974, 66). For example, Bishop Jaime de Nevares of Neuquen

defended MSTM actions before regime officials in 1977, and at least seven opposed bishops

affirmed the MSTM in meetings as early as 1969 (Pattin 2019).

Unopposed bishops, rather than protecting and encouraging opposed priests’ actions,

could have used their authority to appoint, make rules, and obligate opposed priests in order

to constrain them. Indeed, some of these bishops imposed constraints as soon as the MSTM

formed. In 1969, Archbishop Guillermo Bolatti of Rosario forbade priests from activism in

the poor communities in his diocese. In the early 1970s, Bishop Juan Laise of San Luis

expelled a priest who sympathized with leftist movements from his diocese. Bishop Leon

Kruk of San Rafael pushed for priests’ education to take a more “traditional” role which

emphasized obedience to political authorities, and also banned leftist Catholic publications

in his diocese. In 1974, a group of unopposed bishops denounced the MSTM in a right-wing

Catholic periodical.25

To test the implication that opposed bishops were more effective at reducing repression

with opposed priests, I interact a measure of priest opposition to repression—the count of

priests in a diocese affiliated with the MSTM in 1967—with the opposed bishop variable

25These examples are drawn from various primary sources which are detailed in full in the appendix.

22


and estimate models otherwise identical to the main specifications. If opposed bishops

reduced repression by encouraging likeminded priests’ actions, the interaction term should

be negative. The results, reported in the appendix, are consistent with this expectation.

Furthermore, the baseline term for opposed bishops—the effect of these bishops on repression

with no MSTM priests in their diocese—is attenuated. These findings suggest bishops’ and

priests’ effects on repression are complementary.

Confirming Priests’ Influence: Micro-Level Evidence

I next validate the finding above that repression is reduced when a bishop imposes fewer

constraints on opposed priests’ actions. Under an opposed bishop, these priests are less at

risk of being forbidden from activism, expelled from their diocese, or mandated to preach

a message unopposed to repression. Opposed priests thus reduce repression when their

bishop, at minimum, chooses not to constrain their actions, or protects them from attempts

originating outside the Church to impose constraints. There are also situations in which

bishops are less able—even if they are willing—to constrain opposed priests’ actions. In

these situations, opposed priests should likewise reduce repression.

Priests in Catholic religious orders—international organizations whose priests were given

charge of some local Catholic jurisdictions known as parishes—are one such situation. Priests

of an order are under the dual authority of their order’s superior and the area’s bishop, mak-

ing their parishes less constrained by the bishop than “secular” parishes under the bishop’s

sole authority (McDermott 2004).26 Thus a bishop unopposed to repression can impose

fewer constraints on the parishes of religious orders than he can on secular ones. As a result,

priests in opposed religious orders should reduce repression similarly to opposed priests in

the employ of an opposed bishop.

To test whether opposed orders reduced repression, I document the religious orders which

26In Argentina, orders received parishes before, and thus free from the interference of, the military regime.

23


spoke out against the Dirty War (Catoggio 2010). Based on these accounts, I determine which

orders had a record of opposition to repression and match those orders to parishes in their

charge in the Archdiocese of Buenos Aires, whose bishop was unopposed to repression.27

With data from the Archdiocese of Buenos Aires (1971), I geolocate each Catholic parish

in the diocese which existed in 1971. For disappearances and killings, measurements are

imprecise below the department level. For example, the precise last location of disappearance

victims were frequently unknown. As a proxy measure, I geolocate the physical sites of

repression—clandestine detention centers (CCDs)—in Buenos Aires (CONADEP 1984).28

The regime selected CCD locations endogenously, converting only 34 of 50 Buenos Aires

police stations into CCDs. Figure 3 maps Catholic parishes and CCDs in Buenos Aires.

Figure 3: Catholic Parishes and Detention Centers in Buenos Aires, 1970s

Note: The left panel depicts the geographic location of parish churches in Diocese of Buenos Aires in 1971,with parishes in the charge of opposed religious orders shaded. The right panel depicts the geographiclocation of clandestine detention centers in the same area during the dictatorship.

If priests in opposed religious orders reduced repression, CCDs—as a proxy for repression—

should be more distant from parishes with opposed religious orders than other parishes. To

27The archbishop concealed information about repression from the public (Finchelstein 2014). An enu-meration of these orders and corresponding parishes is in the appendix.

28A test in the appendix shows CCDs strongly predict disappearances and killings at the national level.

24


test the implication, I regress the variable Distance to Nearest CCD (km)p, which measures

the distance in kilometers from a parish church p to the nearest detention center, on the

explanatory variable, Opposed Order p, which indicates whether parish p was in the charge

of an opposing order. I also include zone and neighborhood fixed effects, accounting for

geographic characteristics which could explain both the presence of opposed religious orders

and the availability or selection of CCD locations.

Table 2 reports the results: parishes in the charge of opposed religious orders were, on

average, about 0.4 kilometers farther from the nearest CCD than other parishes. These

estimates are substantively large—about one standard deviation of the dependent variable

among parishes without opposed orders.29 The findings suggest that priests can reduce

repression when they have fewer constraints on their actions imposed by a bishop.

Table 2: Religious Orders and Repression in Buenos Aires

Parish Distance to Nearest CCD (km)(1) (2) (3)

Opposed Order 0.45† 0.51∗ 0.38∗

(0.25) (0.23) (0.18)

Adj. R2 0.06 0.21 0.38Observations 154 154 154Fixed Effects Zone Neighborhood∗∗p < 0.01, ∗p < 0.05, †p < 0.10

Bishops and Ecclesial Base Communities

I next consider how opposed bishops protected and encouraged another likeminded part of

the Catholic Church: ecclesial base communities. These were progressive local networks

composed of priests, other agents, and adherents which connected with marginalized social

groups and engaged in “semi-clandestine” activities which subverted the regime (Obregon

2007). Base communities were part of a regional movement which connected the Catholic

29A robustness test validating the opposed order coding scheme is included in the appendix.

25


Church with the poor and emphasized religious commitment to social justice (Mainwaring

and Wilde 1989). Unlike the MSTM or religious orders, however, ecclesial base communities

did not emerge in Argentina until after the dictatorship began.

If opposed bishops’ used institutional resources to encourage local opposition in the

Church, such as ecclesial base communities, these communities should be more likely in

areas under the jurisdiction of opposed bishops than in areas under the jurisdiction of unop-

posed bishops. To test this implication, I gather data on which dioceses supported ecclesial

base communities and when ecclesial base communities received support in these areas (Ma-

rina 2014). Using similar models to those for the main results, except with support for

base communities as an outcome, I find a positive, significant relationship between opposed

bishops and support for ecclesial base communities.30 Taken together with evidence that

opposed bishops were more effective with a larger number of opposed priests, this result

suggests such bishops reduced repression through an institutional mechanism.

Exercising Public Influence

Opposed bishops also took actions which leveraged the Catholic Church’s public influence

to pressure the regime (Hale 2015, Smith 2019, Trejo 2012). Archbishop Vicente Zazpe of

Santa Fe—leader of about four million Catholics—preached a message pressing the “urgency

of justice” with respect to human rights issues (U.S. State Department 1979). Other times

bishops mobilized directly through domestic human rights organizations such as the Mothers

of the Plaza de Mayo and the Ecumenical Movement for Human Rights. Bishop Justo

Laguna of Moron, a friend of Argentine human rights activist and Nobel Peace Prize winner

Adolfo Perez Esquivel, publicly denounced the regime’s repression (U.S. State Department

1979). Bishop Jorge Novak of Quilmes was among the leaders of a 15,000-person march

which presented “over 210,000 signatures calling for information about missing persons to

30Full results are reported in the appendix.

26


an official of the executive branch” (U.S. State Department 1983).

If bishops used public influence to pressure the regime to reduce repression, they would

also have affected public attitudes in areas under their jurisdiction. By resisting the regime’s

repression, opposed bishops may have won support among the public as such resistance

became more favored upon the regime losing support and collapsing in 1983. Therefore, I

test whether opposed bishops increased adherence to the Catholic Church after the regime

(Wittenberg 2006). I aggregate the data to a department-level cross section and estimate

the relationship between bishop opposition—measured as the proportion of years during the

dictatorship a department had an opposed bishop—and the proportion of Catholic adherents

in 1988. Bishop opposition corresponds to significantly more post-regime Catholics.

Tests of the institutional mechanism suggest opposed bishops used the local embedded-

ness and public influence of the Catholic Church to resist state violence. Opposed bishops

were more effective with likeminded priests, as such priests with fewer constraints on their

actions could reduce repression. Opposed bishops also supported progressive ecclesial base

communities, and affected public attitudes after the regime collapsed. Supplementary tests

in the appendix suggest opposed bishops may also have affected the Argentine human rights

movement by lending their resources to its mobilization activities.

Discussion

Religious leaders can reduce repression in dictatorships. This study of Catholic bishops in

Argentina during the 1976-1983 military regime shows that bishops opposed to repression

reduced disappearances and killings in areas under their jurisdiction. An average bishop

who opposed repression could have prevented 128 disappearances and 32 killings over the

course of the dictatorship. I offer evidence that bishops’ appointments were independent of

diocese characteristics, and leverage exogenous variation in the appointing pope to suggest a

27


causal effect of opposed bishops. These findings affirm the influence of communal elites such

as religious leaders in dictatorships, and show religious institutions’ stances toward regimes

are not monolithic. In particular, I document the opposed leaders in an Argentine Catholic

Church which otherwise had a reputation for alignment with the regime.

Religious leaders’ influence likely generalizes beyond the case studied here. First, Ar-

gentina’s regime-aligned Church makes this a hard case in which to estimate the effects of

opposed leaders. When the Catholic Church is less regime-aligned, such as in Chile under

Pinochet, religious resistance is even more likely to be effective (Amat 2019). Second, reli-

gious institutions are ubiquitous in dictatorships: in 2000, 87% of citizens in dictatorships

were adherents of a major world religion—an increase of 8% over the prior three decades.31

Drawing on their embeddedness and influence, religious leaders take center stage in opposing

(and supporting) these regimes. Considerable evidence shows religious leaders’ influence in

Southeast Asian (Slater 2010), Eastern European (Wittenberg 2006), Sub-Saharan African

(McClendon and Riedl 2019), and Latin American (Gill 1998, Smith 2019) regimes. This in-

fluence also extends to eroding democracies, such as the role of Catholic parishes in curtailing

violence in the Philippine drug war (Brooke et al. 2020).

Despite the breadth of their influence in dictatorships, religious leaders’ effects on re-

pression could break down in several ways. First, in religiously fragmented states—where

public adherence is divided among several institutional religions or no religious affiliation at

all—leaders will be less embedded and less credible. A smaller share of adherents limits in-

formation gathering and public recognition of leaders’ traditional power. This may constrain

leaders’ effectiveness in Sub-Saharan Africa, for example. Second, if a regime can prevent an

increase in opposition when targeting leaders—because state violence is pervasive enough to

deter expression of any such opposition or because foreign powers offer support for repression

(Chyzh and Labzina 2018)—then repression of opposed leaders, rather than accommodation,

31The appendix contains more detailed information about scope conditions.

28


would be a best response. For example, United States support for the El Salvadoran military

regime helped it survive backlash to the death of Archbishop Oscar Romero in 1980. Third,

the regime may co-opt entire religious institutions, integrating them with the state and pre-

cluding the appointment of opposed leaders (Wainscott 2017). Fourth, religious leaders may

oppose a regime without also opposing repression. The Argentine regime was synonymous

with the Dirty War, merging the two objects of opposition. Yet not all dictatorships are

predicated on campaigns of state violence.

This paper’s focus on religious leaders who command institutional resources such as

organizational capacity, networks, and service provision comports with a larger literature

which shows these institutions can check the arbitrary use of state power (Braun 2016,

Brooke et al. 2020, Hale 2015). By showing opposed religious leaders reduce state violence,

I advance this scholarship. Studying leaders’ use of institutional resources also follows the

call of Davenport (2015) for a “rigorous attempt at documenting the internal workings of

those who challenge political authorities” (308). Religious leaders’ internal workings involve

using the resources and public influence of their institution to oppose repression and protect

the actions of likeminded agents. Furthermore, showing the interaction between religious

leaders and their agents suggests their actions are complementary. Future research should

similarly explore the internal workings of other challenger groups, studying how interactions

among leaders and agents shape their effectiveness in resisting state violence.

In this paper, I have also considered the effects of a different type of power than the

coercion typically studied in repression research. The symbolic, or traditional, power of reli-

gious leaders differs from state coercive power, affecting official and public behavior through

appeal to morality (Jung 2020) and sacredness (Condra, Isaqzadeh and Linardi 2019, Kubik

1994). While literature has considered the role of religion in politics and qualitative accounts

document the interaction of dictatorships and religious institutions, the study of traditional

power has made few inroads into the quantitative study of dictatorships. The use of tradi-

29


tional power could explain outcomes such as regime durability, military loyalty, and public

support for regimes (Slater 2010), reinforcing or substituting for co-optation or repression.

Finally, the paper has revealed an Argentine Catholic Church more divided than in other

accounts of the institution during this era (Morello 2015, Obregon 2005). While on aggregate

the Church aligned with the right-wing regime, a minority of bishops broke from the insti-

tution’s prevailing views. This finding suggests two paths for future research. First, scholars

may discover variation in other religious institutions and explore how such variation may bet-

ter explain these institutions’ influence. Second, the origins of opposed bishops’ preferences

present a puzzle. Whereas many accounts predict Protestant competition prompts Catholic

leaders’ concern for human rights (Smith 2019, Trejo 2012), the lack of such competition in

1970s Argentina (Gill 1998) indicates scholars should consider other explanations for these

bishops’ opposition.

References

Amat, Consuelo. 2019. “The Hydra Effect: When Repression Creates New Opposition

against Authoritarianism.” Working paper .

Archdiocese of Buenos Aires. 1971. “Guıa Eclesiastica.” Ediciones AICA .

Argentine Army. 1975. “A la directiva del comandante general del ejercito Nro. 404/75

(Lucha contra la subversion).” Buenos Aires, Argentina.

Berman, Sheri. 1997. “Civil Society and the Collapse of the Weimar Republic.” World

Politics 49(3):401–429.

Blaydes, Lisa. 2018. State of Repression: Iraq under Saddam Hussein. Princeton: Princeton

University Press.

30


Braun, Robert. 2016. “Religious Minorities and Resistance to Genocide: The Collective

Rescue of Jews in the Netherlands during the Holocaust.” American Political Science

Review 110(1):127–147.

Bresci, Domingo A. 1994. Movimiento de Sacerdotes para el Tercer Mundo: Documentos

para la memoria historica. Buenos Aires: Comision de Estudios de Historia de la Iglesia

en Latinoamerica.

Brooke, Steven, David Buckley, Clarissa David and Ronald Mendoza. 2020. “Populist Vio-

lence and Social Resistance: The Catholic Church and the Philippine Drug War.” Working

paper .

Brysk, Alison. 1994. “The politics of measurement: The contested count of the disappeared

in Argentina.” Human Rights Quarterly 16:676–692.

Carter, Brett L. and Mai Hassan. 2019. “Regional Governance in Divided Societies: Evidence

from the Republic of Congo and Kenya.” Journal of Politics Forthcoming.

Catoggio, Marıa Soledad. 2010. “Cambio de Habito: Trayectorias de religiosas durante la

ultima dictadura militar argentina.” Latin American Research Review 45(2):27–48.

Chyzh, Olga and Elena Labzina. 2018. “Bankrolling Repression? Modeling Third-Party

Influence on Protests and Repression.” American Journal of Political Science 62(2):312–

324.

Cinelli, Carlos and Chad Hazlett. 2020. “Making Sense of Sensitivity: Extending Omitted

Variable Bias.” Journal of the Royal Statistical Society: Series B (Statistical Methodology)

82(1):39–67.

CONADEP. 1984. Legajos de comision sobre los desaparecidos. Accessed July 5, 2019.

31


Condra, Luke N, Mohammad Isaqzadeh and Sera Linardi. 2019. “Clerics and scriptures:

Experimentally disentangling the influence of religious authority in Afghanistan.” British

Journal of Political Science 49(2):401–419.

Davenport, Christian. 2015. How Social Movements Die: Repression and Demobilization of

the Republic of New Africa. Cambridge: Cambridge University Press.

Diario de Juicio. 1985. “2 de agosto de 1985.” Editorial Perfil S.A. .

Dodson, Michael. 1974. “Priests and Peronism: Radical Clergy and Argentine Politics.”

Latin American Perspectives 1(3):58–72.

Duzdevich, Aldo. 2019. Salvados por Francisco. Madrid: Ediciones B.

Esberg, Jane. 2020. “Anticipating Dissent: The Repression of Politicians in Pinochet’s

Chile.” Journal of Politics Forthcoming.

Excelsior. 1979. “La Embajada de Mexico Asilo a Campora.” Accessed July 5, 2019 .

Finchelstein, Federico. 2014. The Ideological Origins of the Dirty War: Fascism, Populism,

and Dictatorship in Twentieth Century Argentina. Oxford: Oxford University Press.

Gandhi, Jennifer. 2008. Political Institutions under Dictatorship. Cambridge: Cambridge

University Press.

Gautier, Mary L. 1998. “Church Elites and the Restoration of Civil Society in the Communist

Societies of Central Europe.” Journal of Church and State 40(2):289–317.

Gill, Anthony. 1998. Rendering Unto Caesar: The Catholic Church and the State in Latin

America. Chicago: University of Chicago Press.

32


Gowrinathan, Nimmi and Zachariah Mampilly. 2019. “Resistance and Repression under the

Rule of Rebels: Women, Clergy, and Civilian Agency in LTTE Governed Sri Lanka.”

Comparative Politics 52(1):1–20.

Grzymala-Busse, Anna. 2016. “Weapons of the Meek: How Churches Influence Public Pol-

icy.” World Politics 68(1):1–36.

Hale, Christopher W. 2015. “Religious Institutions and Civic Engagement: A Test of Reli-

gion’s Impact on Political Activism in Mexico.” Comparative Politics 47(2):211–230.

Harris, Margaret. 1998. “A Special Case of Voluntary Associations? Towards a Theory of

Congregational Organization.” British Journal of Sociology 49(4):602–618.

Ichino, Nahomi and Matthias Schundeln. 2012. “Deterring or Displacing Electoral Irreg-

ularities? Spillover Effects of Observers in a Randomized Field Experiment in Ghana.”

Journal of Politics 74(1):292–307.

Imai, Kosuke and In Song Kim. 2020. On the Use of Two-Way Fixed Effects Regression

Models for Causal Inference with Panel Data. Technical report Harvard University IQSS

Working Paper. Cambridge, MA.

Jung, Jae-Hee. 2020. “The Mobilizing Effect of Parties’ Moral Rhetoric.” American Journal

of Political Science 64(2):341–355.

Kirk, John. 1985. “John Paul II and the Exorcism of Liberation Theology: A Retrospective

Look at the Pope in Nicaragua.” Bulletin of Latin American Research 4(1):33–47.

Klor, Esteban F., Sebastian M. Saiegh and Shanker Satyanath. 2017. “Croynism in State

Violence: Evidence from Labor Repression During Argentina’s Last Dictatorship.” Work-

ing paper .

URL: https://papers.ssrn.com/sol3/papers.cfm?abstractid=2851822

33


Kubik, Jan. 1994. Power of Symbols Against the Symbols of Power: The Rise of Solidarity

and the Fall of State Socialism in Poland. Penn State Press.

La Razon. 1983. “Liberan a 7 Detenidos Polıticos en Misiones, Dos de Ellos Enfermos.”

Accessed July 11, 2019 .

Mainwaring, Scott and Alexander Wilde. 1989. The Progressive Church in Latin America.

Notre Dame, IN: University of Notre Dame Press.

Marchak, Patricia. 1999. God’s Assassins: State Terrorism in Argentina in the 1970s. Mon-

treal: McGill-Queen’s University Press.

Marina, Alejandro. 2014. “Comunidades Eclesiales de Base – CEBs in Argentina.” Small

Christian Communities .

URL: https://smallchristiancommunities.org/

McClendon, Gwyneth and Rachel Beatty Riedl. 2019. From Pews to Politics: Religious

Sermons and Political Participation in Africa. Cambridge: Cambridge University Press.

McDermott, Rose. 2004. “Ecclesiastical Authority and Religious Autonomy: Canon 679

Under Glass.” Studia Canonica 38(2):461.

Mignone, Emilio F. 1986. Witness to the Truth: The Complicity of Church and Dictatorship

in Argentina, 1976-1983. New York: Orbis Books.

Ministry of the Interior. 1973. “Resultados electorales: elecciones generales, 23 Septiembre

1973.” Direccion Nacional Electoral .

Morello, Gustavo. 2015. The Catholic Church and Argentina’s Dirty War. Oxford: Oxford

University Press.

34


Nelson, Reed E. 1993. “Authority, Organization, and Societal Context in Multinational

Churches.” Administrative Science Quarterly 38(4):653–682.

Obregon, Martın. 2005. “La iglesia argentina durante el ‘Proceso’ (1976-1983).” Prismas,

Revista de historia intelectual 9:259–270.

Obregon, Martın. 2007. La iglesia argentina durante la ultima dictadura militar: El terror

desplegado sobre el campo catolico (1976-1983). In Historizar el pasado vivo en America

Latina, ed. Anne Perotin-Dumon. Universidad Alberto Hurtado pp. 1–43.

O’Donnell, Guillermo. 1988. Bureaucratic Authoritarianism: Argentina 1966-1973, in Com-

parative Perspective. University of California Press.

Osorio, Javier, Livia Schubiger and Michael Weintraub. 2020. “Legacies of Resistance: Mobi-

lization Against Organized Crime in Mexico.” Comparative Political Studies Forthcoming.

Pattin, Sebastian. 2019. Entre Pedro y el pueblo de Dios: Las concepciones de autoridad en

el catolicismo Argentino. Rosario: Prohistoria ediciones.

Ritter, Emily Hencken and Courtenay R. Conrad. 2016. “Preventing and Responding to

Dissent: The Observational Challenges of Explaining Strategic Repression.” American

Political Science Review 110(1):85–99.

Romero, Luis Alberto and James P. Brennan. 2013. A History of Argentina in the Twentieth

Century. University Park: Penn State University Press.

Scharpf, Adam. 2018. “Ideology and state terror: How officer beliefs shaped repression

during Argentina’s ’Dirty War’.” Journal of Peace Research 55(2):206–221.

Scharpf, Adam and Christian Glaßel. 2019. “Why Underachievers Dominate Secret Po-

lice Organizations: Evidence from Autocratic Argentina.” American Journal of Political

Science Forthcoming:1–16.

35


Schmitter, Philippe C. 1993. “Some Propositions About Civil Society and the Consolidation

of Democracy.” Institut fur Hohere Studien .

Sikkink, Kathryn. 2004. Mixed Signals: U.S. Human Rights Policy in Latin America. Ithaca:

Cornell University Press.

Slater, Dan. 2010. Ordering Power: Contentious Politics and Authoritarian Leviathans in

Southeast Asia. Cambridge University Press.

Smith, Amy Erica. 2019. Religion and Brazilian Democracy: Mobilizing the People of God.

New York: Cambridge University Press.

Spenkuch, Jorg L. and Philipp Tillmann. 2018. “Elite Influence? Religion and the Electoral

Success of the Nazis.” American Journal of Political Science 62(1):19–36.

Tilly, Charles. 2006. Regimes and Repertoires. Chicago: University of Chicago Press.

Trejo, Guillermo. 2012. Popular Movements in Autocracies: Religion, Repression, and In-

digenous Collective Action in Mexico. Cambridge: Cambridge University Press.

Tsai, Lily L. 2007. “Solidary Groups, Informal Accountability, and Local Public Goods

Provision in Rural China.” American Political Science Review 101(2):109–122.

Tunon, Guadalupe. 2018. “When the Church Votes Left: How Progressive Religion Hurts

Gender Equality.” Working paper.

Turner, Frederick C. 1971. “ARMISC1971-Turner: Various Argentina Adults.” University

of Connecticut - Roper Center for Public Opinion Research .

U.S. State Department. 1979. “Secret: Political Reporting (Buenos Aires).” Accessed June

14, 2019 .

36


U.S. State Department. 1983. “Confidential Memo: Human Rights Developments.” Accessed

June 14, 2019 .

Villa, Nestor Daniel. 2000. Manual de Derecho Eclesiastico. Buenos Aires: Universidad

Libros.

Wainscott, Ann Marie. 2017. Bureaucratizing Islam: Morocco and the War on Terror. Cam-

bridge University Press.

Weber, Max. 1947. The Theory of Social and Economic Organization. New York: Free Press.

Wittenberg, Jason. 2006. Crucibles of Political Loyalty: Church Institutions and Electoral

Continuity in Hungary. Cambridge: Cambridge University Press.

Woodberry, Robert. 2012. “The Missionary Roots of Liberal Democracy.” American Political

Science Review 106(2):244–274.

37


Religious Institutions and Resistance to Repression

Appendices

Contents

Appendix A: Descriptive Statistics 2

Appendix B: Bishop Appointments 7

Appendix C: Robustness Tests 16

Appendix D: Instrumental Variables Analysis 34

Appendix E: Alternative Explanation 38

Appendix F: Mechanism Tests 42

Institutional Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

Religious Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Individual Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Appendix G: Archival Evidence 57

Appendix H: Scope Conditions 61

1


Appendix A: Descriptive Statistics

1. Two-sample t-test for Argentines’ agreement with the Catholic Church and with the

armed forces, Table A1. Data are from a survey of 890 Argentine adults, conducted by

Turner (1971). It is also likely that respondents’ agreement with the armed forces is

overestimated: the survey was conducted over the telephone during a military dicta-

torship and respondents may have had an incentive to falsify their preferences toward

the dictatorship (Kuran 1991).

2. Descriptive statistics, department level, Table A2.

3. Descriptive statistics, department-year level, Table A3.

4. Map detailing Catholic diocese boundaries in 1976, Figure A1.

5. List of bishops opposed to the regime, Table A4.

6. Coding criteria for bishop opposition.

Table A1: Two-Sample t-Test, Agreement with Argentine Institutions

Church Mean Armed Forces Mean t df p-value CI Lower CI Upper

35.18 24.56 7.46 1590.3 1.37 · 10−13 7.83 13.41

2


Table A2: Descriptive Statistics, Department Level

Variable Minimum Maximum Mean Median Std. Dev Missing

Population, 1970 427 2972453 47603.82 15439 158650.83 27Literate Share, 1970 0.32 0.9 0.72 0.74 0.11 28Foreign Residents, 1970 0 0.43 0.06 0.05 0.05 28Child Mortality Rate, 1970 0.13 1.49 0.47 0.42 0.25 28Peronist Vote, 1973 28.5 94.3 58.66 57 11.5 20Militant Presence, 1975 0 1 0.3 0 0.46 0

Table A3: Descriptive Statistics, Department-Year Level

Variable Minimum Maximum Mean Median Std. Dev Missing

Disappeared 0 796 1.69 0 20.93 48Killed 0 129 0.43 0 4.17 48Bishop Opposed 0 1 0.27 0 0.44 69Bishop Age 41 76 61.88 62 6.5 4

3


Figure A1: Argentine Catholic Diocese Boundaries, 1976

Note: Figure depicts dioceses (colors) overlaid on department boundaries (lines) based on the Catholicdioceses existing in 1976 in Argentina.

4


Table A4: Argentine Bishops Opposed to the Regime, 1976-1983

Bishop Diocese

Agustin Adolfo Herrera San FranciscoAlberto Devoto GoyaAlcides Jorge Pedro Casaretto RafaelaAlfredo Guillermo Disandro Villa MariaAntonio Alfredo Brasca RafaelaAntonio Maria Aguirre San IsidroArcenio Raul Casado JujuyCarlos Horacio Ponce de Leon San NicolasEnrique A Angelelli La RiojaGerardo Eusebio Sueldo OranJaime Francisco De Nevares NeuquenJorge Kemerer PosadasJorge Novak QuilmesJose Agustin Marozzi ResistenciaJuan Carlos Ferro ConcepcionJuan Jose Iriarte ReconquistaJusto Oscar Laguna MoronLuis Juan Tome Mercedes-LujanManuel Marengo AzulMiguel Esteban Hesayne ViedmaMiguel Raspanti MoronMoises Julio Blanchoud Rio CuartoPedro Boxler GualeguaychuRaul Marcelo Scozzina FormosaVicente Faustino Zazpe Santa Fe

5


Coding Criteria for Bishop Opposition

The following criteria were used to code a bishop as opposed to the regime.

• Did the bishop have a connection or involvement with the Movement of Priests for the

Third World before the dictatorship? If yes, label as potential opposed bishop.

• Did the bishop have a connection or involvement with other sectors of society which

were targets of the dictatorship—such as student activists and organized labor—before

the dictatorship? If yes, label as potential opposed bishop.

• Did the bishop take actions or make statements indicating opposition to repression

before the coup on March 24, 1976? If so, label as potential opposed bishop.

• Was the bishop mentioned in Mignone (1986) as either a public or private opponent of

the regime? If so, label as potential opposed bishop.

• Did the bishop have frequent associations and issue statements or take actions jointly

with other bishops who are labeled as potential opposition bishops? If so, label as

potential opposed bishop.

• Did the bishop have at least two pieces of data which fulfill one or more of the above

criteria? If so, validate and confirm coding as opposed bishop.

See Appendix H for more details on the archival sources used.

6


Appendix B: Bishop Appointments

1. Test of endogenous bishop appointments using change in bishop opposition from cur-

rent year to the following year, Table B1. If the presence of regime-opposed bishops

is exogenous as has been asserted and empirically tested in the body of the paper,

then it should also be the case that dioceses to which regime-opposed bishops would

be appointed in the future had similar levels of repression to dioceses to which these

bishops would not be appointed. This means there should be no “effects” of an opposed

bishop the year before his appointment to a particular diocese. Results from the test

demonstrate no consistent relationship between next-year opposed bishops (∆ Bishop

Opposed, t+ 1) and current-year (t) repression – both disappearances and killings.

2. Results of balance tests for opposed bishops with data at the department-year level,

Table B2. Test indicates opposed bishops were not likely to be placed in dioceses which

varied on observed social, economic, and political characteristics.

3. Results of balance tests for opposed bishops with data aggregated to the diocese level,

Table B3. This could be a more theoretically grounded test if the bishop or pope

considered diocese characteristics on aggregate when making decisions to request or

give an appointment, respectively.

4. Results of balance tests for opposed bishops at the diocese level depicted graphically,

Figure B1.

5. Results of procedure testing for spatial clustering of bishop opposition, Figure B2.

The procedure is analogous to randomization inference. First, geographic clusters

of departments are generated using k-means clustering with four, six, eight, and ten

7


clusters. Second, a χ2 test of independence is conducted and a test statistic from the

true assignment of bishop opposition obtained. Third, I simulate five hundred random

assignments of bishop opposition—clustered by diocese—based on bishop opposition at

the start of the dictatorship. Fourth, I generate a null distribution of χ2 test statistics

from these five hundred assignments. Fifth, I compare the test statistic from the true

assignment against the null distribution of test statistics. In all four cases, the test

statistic is not significant at the p = 0.05 level, suggesting there is not geographic

clustering on the opposed bishop variable.

6. Results of balance tests for opposed bishops at the diocese level based on whether bish-

ops were appointed during the 1966-1973 Argentine Revolution military dictatorship.

There could be a concern if that dictatorship did not abide by the terms of the 1966

Concordat and only allowed bishops who were pro-military to be appointed. However,

there is no evidence that bishops appointed during this period were less likely to be

opposed to repression.

7. Graphical depiction of simple difference in means for opposed and unopposed bishops

across the two dependent variables: disappearances and killings, Figure B3.

8


Table B1: Test of Endogenous Appointments of Regime-Opposed Bishop

Disappearances, t Killings, t(1) (2) (3) (4) (5) (6) (7) (8)

∆ Bishop Opp., t + 1 −0.05 −1.51 0.68 1.40† 0.25 −0.22 0.52 0.52∗∗∗

(0.05) (1.51) (1.31) (0.76) (0.15) (0.26) (0.42) (0.15)Log(1970 Pop.) 3.96∗ 2.59∗∗∗ 0.81∗∗ 0.75∗∗∗

(1.83) (0.68) (0.24) (0.21)Young Men Share −67.58 −11.94 −13.17∗∗ −3.38

(43.45) (7.50) (4.92) (2.22)Literate Share −27.55 −11.20∗∗ −5.06∗ −3.17∗

(17.83) (3.62) (2.22) (1.26)Child Mortality 5.93 4.05∗ 0.67 1.77

(4.85) (1.73) (0.61) (0.94)Foreign Residents 46.44∗ 10.79∗ 9.57∗∗∗ 4.64∗

(22.37) (5.01) (2.73) (2.14)Archdiocese 5.16 3.63∗∗ 1.06∗ 1.28∗∗

(3.11) (1.19) (0.44) (0.38)Tenure −0.09 −0.01 −0.01 −0.00

(0.11) (0.05) (0.02) (0.01)Peronist Vote −0.05 −0.03 0.00 −0.00

(0.09) (0.03) (0.01) (0.01)Militant Presence 4.49∗ 5.00∗ 0.97∗ 2.07

(2.11) (2.14) (0.41) (1.51)(Intercept) 6.74∗∗ 4.00 −6.74 277.98∗∗∗ 4.54∗∗∗ 0.29 −3.63 32.80∗∗∗

(2.09) (13.86) (7.77) (1.47) (0.64) (2.32) (2.87) (0.42)Adj. R2 0.01 0.08 0.32 0.27 0.14 0.10 0.18 0.18Observations 3496 3286 3258 3496 3363 3286 3258 3496Fixed Effects Year, Sub. Year Year, Sub. Year, Dpt. Year, Sub. Year Year, Sub. Year, Dpt.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

9


Table B2: Results for Bishop Opposition Balance Test, by Department-Year

Bishop Opposed(1) (2) (3) (4) (5) (6) (7)

Log(1970 Pop.) 0.06(0.04)

Young Men Share 0.03(0.06)

Literate Share −0.05(0.06)

Child Mortality −0.02(0.04)

Foreign Residents 0.03(0.04)

Peronist Vote −0.04(0.04)

MSTM Priests 0.04(0.06)

(Intercept) 0.27∗∗∗ 0.27∗∗∗ 0.27∗∗∗ 0.27∗∗∗ 0.27∗∗∗ 0.27∗∗∗ 0.27∗∗∗

(0.06) (0.06) (0.06) (0.06) (0.06) (0.06) (0.06)

Adj. R2 0.02 0.00 0.01 0.00 0.00 0.01 0.01Observations 3835 3819 3819 3819 3819 3875 4051∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

10


Table B3: Results for Bishop Opposition Balance Test, by Diocese

Bishop Opposed(1) (2) (3) (4) (5) (6) (7)

Log(1970 Pop.) −0.01(0.05)

Young Men Share 0.03(0.05)

Literate Share −0.03(0.05)

Child Mortality 0.01(0.05)

Foreign Residents 0.06(0.06)

Peronist Vote −0.03(0.05)

MSTM Priests 0.01(0.05)

(Intercept) 0.32∗∗∗ 0.32∗∗∗ 0.32∗∗∗ 0.32∗∗∗ 0.32∗∗∗ 0.32∗∗∗ 0.32∗∗∗

(0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05)

Adj. R2 -0.01 -0.01 -0.01 -0.01 0.00 -0.01 -0.01Observations 78 78 78 78 78 78 79∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

11


Figure B1: Balance Test for Bishop Opposition at the Diocese Level

Note: Figure depicts coefficient estimates and confidence intervals for the relationship between potentiallypredictive covariates and bishop opposition. Variables are standardized via z-transformation.

12


Figure B2: Test of Spatial Clustering by Bishop Opposition

Note: Figure depicts null distribution of χ2 statistics for bishop opposition across four clustering schemes.Vertical line indicates position of the test statistic for the true assignment of opposition. In all four cases,the true assignment is not significant at the p = 0.05 level, suggesting no clustering.

13


Table B4: Results for Bishop Opposition Balance Test, 1966-1973 Regime, by Diocese

Bishop Opposed1966-1973 Appointment −0.05

(0.12)(Intercept) 0.33∗∗∗

(0.06)Adj. R2 -0.01Num. obs. 79∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

14


Figure B3: Simple Difference in Means for Opposed and Unopposed Bishops

Note: Figure depicts group means for disappearances and killings between opposed and un-opposed bishopsat the department-year level.

15


Appendix C: Robustness Tests

1. Main results including covariates, Table C1.

2. Main results with more strict coding criteria, Table C2.

3. Main results recoding bishops one at a time, Figure C1.

4. Main results dropping City of Buenos Aires, Table C3.

5. Main results dropping diocese of La Rioja, Table C4.

6. Main results dropping dioceses one at a time, Figure C2.

7. Main results with only data from 1976-1978, Table C5.

8. Consideration of localized spillovers, Tables C6 and C7. This test considers the possi-

bility that repression was redirected away from areas with anti-regime bishops and into

areas with bishops who did not resist the regime. Therefore, the mechanism would not

be deterrence of repression through the threat of publicity but rather the displacement

of repression to other areas in Argentina. The presence of a regime-opposed bishop’s

territory within 50 and 100 kilometers is considered, and the coefficient of interest is

the baseline term (meaning no opposed bishop) for the number of departments within

50 and 100 kilometers with an opposed bishop. A positive and significant coefficient

signifies positive spillovers in these areas, which would suggest a SUTVA violation.

9. Localized spillovers with autocovariate, Tables C8 and C9. This test adds an auto-

covariate term which takes the average bishop opposition of surrounding departments

and thus accounts for a lack of independence in the number of departments with an

16


opposed bishop within a certain radius (Braun 2016). Autocorrelation is likely to be

a problem in the tests for localized spillovers since opposed bishops’ jurisdiction spans

the many adjacent departments which compose a diocese. The latter test yields several

negative and significant results for the baseline term on number of departments within

a 100km radius, the opposite sign as would be expected if spillovers were a problem.

10. Main results with linear time trend instead of year fixed effects, Table C10. Given

the well-documented temporal pattern of repression in Argentina, with the greatest

frequency of repression occurring at the beginning of the regime and dropping sharply

thereafter, the linear time trend may be a an appropriate specification.

11. Main results with random effects instead of fixed effects, Table C11. The procedure

estimates a separate intercept for each group in the grouping variable, instead of esti-

mating a separate coefficient for each group.

12. Main results with year fixed effects interacted with department level covariates, Table

C12. Such an interaction allows for the potentially confounding covariates to have a

time-varying influence on repression.

13. Main results with data aggregated to the diocese-year level, Table C13. Estimated

effects remain negative and increase in significance. However, this estimation strat-

egy loses the ability to hold constant military subzone characteristics, as these span

dioceses.

14. Negative binomial results for the main hypothesis, Table C14.

15. Sensitivity analysis, Figures C3 and C4. The procedure suggests how much variation

an omitted variable would need to explain relative to the log of 1970 population on

repression to invalidate significant results for opposed bishops. Plots are based on the

partial R2 procedure in Cinelli and Hazlett (2020).

17


Table C1: Estimated Effects of Opposed Bishops on Repression, Showing Covariates


Bishop Opposed −1.34∗ −2.15† −1.04† −2.19† −0.49∗ −0.49∗ −0.36∗ −0.58∗

(0.61) (1.12) (0.58) (1.20) (0.21) (0.23) (0.18) (0.28)Log(1970 Pop.) 3.60∗ 2.29∗∗∗ 0.74∗∗ 0.67∗∗∗

(1.64) (0.61) (0.22) (0.19)Young Men Share −58.89 −10.64 −11.45∗∗ −3.05

(37.97) (6.61) (4.25) (1.94)Literate Share −26.57 −10.22∗∗ −4.97∗ −2.93∗

(16.32) (3.23) (2.03) (1.13)Child Mortality 4.81 3.19∗ 0.51 1.42


(20.16) (4.34) (2.45) (1.85)Archdiocese 3.87 2.77∗∗ 0.79∗ 0.97∗∗

(2.55) (0.99) (0.36) (0.31)Tenure −0.02 0.01 0.01 0.01

(0.08) (0.05) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.00


(1.76) (1.93) (0.35) (1.31)(Intercept) 16.18∗∗∗ 6.11 −3.98 244.34∗∗∗ 4.51∗∗∗ 0.89 −2.50 28.99∗∗∗

(3.41) (12.52) (6.68) (1.60) (0.74) (2.00) (2.42) (0.45)

Adj. R2 0.27 0.07 0.29 0.24 0.13 0.10 0.16 0.16Observations 3851 3763 3731 4003 3851 3763 3731 4003Fixed Effects Year, Sub. Year Year, Sub. Year, Dpt. Year, Sub. Year Year, Sub. Year, Dpt.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

18


Table C2: More Strict Bishop Opposition Coding


Bishop Opposed (Strict) −1.48∗ −1.76 −1.28† −2.71∗ −0.49∗ −0.41† −0.40∗ −0.72∗

(0.70) (1.10) (0.67) (1.24) (0.21) (0.22) (0.19) (0.30)Log(1970 Pop.) 3.51∗ 2.29∗∗∗ 0.72∗∗ 0.66∗∗∗

(1.61) (0.60) (0.21) (0.19)Young Men Share −58.14 −9.98 −11.27∗ −2.84

(37.83) (6.68) (4.25) (1.95)Literate Share −25.77 −10.09∗∗ −4.80∗ −2.88∗

(16.06) (3.27) (2.00) (1.14)Child Mortality 5.07 3.32∗ 0.56 1.47


(20.31) (4.35) (2.50) (1.86)Archdiocese 4.21 2.93∗∗ 0.86∗ 1.03∗∗

(2.64) (1.00) (0.37) (0.32)Tenure −0.04 0.01 0.00 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.00


(1.84) (1.95) (0.36) (1.32)(Intercept) 16.23∗∗∗ 5.59 −4.23 244.33∗∗∗ 4.50∗∗∗ 0.78 −2.61 28.99∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

19


Figure C1: Coefficient Estimates Recoding Bishops One at a Time

Note: Figure depicts coefficient estimates and confidence intervals for the effect of opposed bishops ondisappearances from specifications including subzone and year fixed effects, when recoding each bishop’sopposition in turn. Bishop order is artificial.

20


Table C3: Effects of Bishop Opposition, Dropping City of Buenos Aires


Bishop Opposed −1.18∗ −1.50∗ −1.06† −1.45† −0.47∗ −0.42† −0.48∗ −0.48∗

(0.56) (0.65) (0.56) (0.77) (0.21) (0.21) (0.19) (0.24)Log(1970 Pop.) 1.96∗∗∗ 2.29∗∗∗ 0.57∗∗∗ 0.66∗∗

(0.49) (0.60) (0.15) (0.19)Young Men Share −20.82∗∗ −10.93 −7.60∗∗∗ −4.13∗

(6.90) (6.64) (1.94) (1.76)Literate Share −10.44∗∗ −10.50∗∗ −3.33∗∗ −2.48∗

(3.56) (3.21) (1.18) (0.94)Child Mortality 0.91 3.21∗ 0.11 0.67

(0.97) (1.42) (0.30) (0.35)Foreign Residents 21.70∗∗∗ 9.97∗ 6.59∗∗∗ 2.92∗

(4.64) (4.31) (1.45) (1.24)Archdiocese 1.51∗ 2.74∗∗ 0.55∗ 0.87∗∗

(0.68) (0.94) (0.25) (0.29)Tenure 0.05 0.04 0.01 0.01

(0.03) (0.04) (0.01) (0.01)Peronist Vote 0.01 −0.03 0.01 0.00

(0.02) (0.03) (0.01) (0.01)Militant Presence 2.53∗∗ 4.08∗ 0.71∗

(0.83) (1.94) (0.29)(Intercept) 14.95∗∗∗ −3.71 −5.23 4.06∗∗∗ 4.29∗∗∗ −0.20 −1.23 1.41∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

21


Table C4: Effects of Bishop Opposition, Dropping Diocese of La Rioja


Bishop Opposed −1.23∗ −2.21† −1.01 −1.78 −0.47∗ −0.49∗ −0.36† −0.47†

(0.59) (1.19) (0.61) (1.13) (0.21) (0.23) (0.18) (0.25)Log(1970 Pop.) 3.75∗ 2.43∗∗∗ 0.77∗∗ 0.71∗∗∗

(1.72) (0.64) (0.23) (0.20)Young Men Share −58.24 −9.85 −11.38∗ −2.83

(37.81) (6.79) (4.28) (2.01)Literate Share −28.30 −10.69∗∗ −5.35∗ −3.06∗

(17.15) (3.47) (2.14) (1.16)Child Mortality 4.84 3.42∗ 0.48 1.49


(20.37) (4.37) (2.45) (1.81)Archdiocese 4.06 2.83∗∗ 0.83∗ 0.99∗∗

(2.61) (0.99) (0.37) (0.32)Tenure 0.00 0.03 0.01 0.01

(0.08) (0.05) (0.01) (0.01)Peronist Vote −0.08 −0.04 −0.00 −0.01


(1.91) (1.94) (0.36) (1.31)(Intercept) 16.26∗∗∗ 6.17 −5.36 244.42∗∗∗ 4.53∗∗∗ 0.94 −2.91 29.01∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

22


Figure C2: Coefficient Estimates Dropping Dioceses One at a Time

Note: Figure depicts coefficient estimates and confidence intervals for the effect of opposed bishops ondisappearances from specifications including subzone and year fixed effects, when dropping each diocese inturn. Diocese order is artificial.

23


Table C5: Effects of Bishop Opposition, 1976-1978 Only


Bishop Opposed −2.95† −5.03† −2.15 −3.56∗∗ −1.17∗ −1.17∗ −0.82† −0.96∗

(1.48) (2.79) (1.36) (1.27) (0.54) (0.56) (0.43) (0.40)Log(1970 Pop.) 9.19∗ 5.87∗∗∗ 1.91∗∗ 1.74∗∗∗

(4.18) (1.55) (0.55) (0.50)Young Men Share −151.11 −29.13 −30.30∗∗ −8.65

(95.38) (17.00) (10.77) (5.13)Literate Share −67.86 −27.52∗∗ −12.92∗ −8.08∗

(41.08) (8.82) (5.21) (3.10)Child Mortality 12.11 8.49∗ 1.29 3.75


(47.62) (10.91) (6.09) (4.95)Archdiocese 9.78 7.46∗∗ 2.03∗ 2.61∗∗

(6.31) (2.50) (0.91) (0.81)Tenure −0.05 0.13 0.02 0.03

(0.26) (0.13) (0.04) (0.03)Peronist Vote −0.15 −0.09 −0.00 −0.01


(4.83) (5.01) (0.92) (3.41)(Intercept) 30.73∗∗∗ 4.32 −21.70 618.11∗∗∗ 8.05∗∗∗ −0.94 −10.26 71.57∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

24


Table C6: Localized Spillovers, 50 km Radius

Disappearances Killings(1) (2) (3) (4) (5) (6)

Bishop Opposed −1.64∗ −1.80 −1.35† −0.58∗ −0.46∗ −0.42†

(0.78) (1.13) (0.75) (0.27) (0.21) (0.21)Opp. Depts. in 50km 0.04 0.17 0.02 0.01 0.03† 0.01

(0.03) (0.14) (0.03) (0.01) (0.02) (0.01)Bishop Opposed*Opp. Depts −0.01 −0.12 0.00 −0.00 −0.02 −0.00

(0.04) (0.13) (0.03) (0.01) (0.01) (0.01)Log(1970 Pop.) 2.93∗ 2.29∗∗∗ 0.64∗∗ 0.66∗∗

(1.17) (0.61) (0.19) (0.19)Young Men Share −48.81 −10.21 −9.91∗∗ −2.91

(29.75) (6.73) (3.41) (1.92)Literate Share −19.57 −10.27∗∗ −3.90∗ −2.93∗

(11.41) (3.16) (1.69) (1.11)Child Mortality 5.32 3.24∗ 0.59 1.44

(4.03) (1.41) (0.49) (0.80)Foreign Residents 26.96∗ 9.04∗ 6.33∗∗ 3.87∗

(10.42) (4.20) (1.89) (1.91)Archdiocese 4.11 2.80∗∗ 0.82∗ 0.98∗∗

(2.60) (0.98) (0.36) (0.31)Tenure −0.03 0.01 0.00 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.09 −0.03 −0.01 −0.01

(0.10) (0.02) (0.01) (0.01)Militant Presence 0.86 3.86 0.38 1.62

(1.38) (1.97) (0.38) (1.33)(Intercept) 13.14∗∗ 6.80 −5.21 3.49∗∗∗ 1.01 −2.86

(3.80) (12.13) (6.49) (0.94) (1.91) (2.30)R2 0.28 0.09 0.29 0.14 0.11 0.17Adj. R2 0.27 0.08 0.29 0.13 0.10 0.16Observations 3859 3771 3739 3859 3771 3739Fixed Effects Year, Sub. Year Year, Sub. Year, Sub. Year Year, Sub.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

25


Table C7: Localized Spillovers Test, 100 km Radius


Bishop Opposed −0.54 −1.71† −0.60 −0.24 −0.39† −0.19(0.90) (1.03) (0.91) (0.31) (0.22) (0.27)

Opp. Depts. in 100km 0.00 0.04 −0.00 0.00 0.01 −0.00(0.01) (0.03) (0.01) (0.00) (0.00) (0.00)

Bishop Opposed*Opp. Depts. −0.01 −0.03 −0.01 −0.00 −0.00 −0.00(0.02) (0.03) (0.01) (0.01) (0.00) (0.00)

Log(1970 Pop.) 3.40∗ 2.27∗∗∗ 0.70∗∗ 0.66∗∗

(1.51) (0.60) (0.21) (0.19)Young Men Share −55.16 −11.02 −10.76∗∗ −3.17

(35.63) (6.84) (4.00) (1.97)Literate Share −23.80 −10.06∗∗ −4.46∗ −2.88∗

(14.62) (3.15) (1.91) (1.13)Child Mortality 4.95 3.12∗ 0.53 1.39


(18.21) (4.26) (2.29) (1.83)Archdiocese 3.82 2.77∗∗ 0.78∗ 0.97∗∗

(2.51) (0.98) (0.35) (0.31)Tenure −0.03 0.02 0.00 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.00

(0.08) (0.03) (0.01) (0.01)Militant Presence 2.55 4.12∗ 0.60 1.71

(1.31) (1.94) (0.35) (1.30)(Intercept) 16.33∗∗∗ 4.58 −3.39 4.61∗∗∗ 0.61 −2.27

(3.36) (11.72) (6.92) (0.85) (1.94) (2.51)Adj. R2 0.27 0.07 0.29 0.13 0.10 0.16Observations 3859 3771 3739 3859 3771 3739Fixed Effects Year, Sub. Year Year, Sub. Year, Sub. Year Year, Sub.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

26


Table C8: Localized Spillovers Test, 50 km with Autocovariate


Bishop Opposed −1.39† −1.66 −1.18 −0.51† −0.42† −0.37†

(0.74) (1.15) (0.74) (0.26) (0.23) (0.21)Opp. Depts. in 50 km −0.02 0.15 −0.01 −0.00 0.02 −0.00

(0.04) (0.14) (0.03) (0.01) (0.02) (0.01)Bishop Opposed*Opp. Depts. −0.02 −0.12 0.00 −0.01 −0.02 −0.00

(0.03) (0.13) (0.02) (0.01) (0.01) (0.01)Log(1970 Pop.) 2.93∗ 2.25∗∗∗ 0.64∗∗ 0.65∗∗

(1.19) (0.63) (0.20) (0.19)Young Men Share −50.27 −10.74 −9.82∗∗ −3.12

(31.82) (6.52) (3.60) (1.80)Literate Share −19.87 −10.06∗∗ −3.81∗ −2.88∗

(11.99) (3.20) (1.77) (1.11)Child Mortality 5.63 3.38∗ 0.67 1.45


(10.79) (4.12) (1.99) (1.92)Archdiocese 3.99 2.71∗∗ 0.78∗ 0.93∗∗

(2.57) (1.00) (0.35) (0.32)Tenure −0.03 0.01 0.00 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.09 −0.03 −0.01 −0.00


(1.51) (1.98) (0.40) (1.34)(Intercept) 7.49 7.11 −7.52 1.75 0.83 −3.62

(3.83) (12.91) (6.07) (1.06) (1.99) (2.28)Adj. R2 0.28 0.08 0.29 0.14 0.10 0.16Observations 3715 3715 3715 3715 3715 3715Autocovariate Y Y Y Y Y YFixed Effects Year, Sub. Year Year, Sub. Year, Sub. Year Year, Sub.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

27


Table C9: Localized Spillovers Test, 100km with Autocovariate


Bishop Opposed −0.34 0.14 −0.39 −0.20 −0.12 −0.14(0.89) (1.81) (0.91) (0.30) (0.29) (0.27)

Opp. Depts. in 100km −0.03∗∗ −0.05 −0.02 −0.01∗∗ −0.01 −0.01∗

(0.01) (0.04) (0.01) (0.00) (0.01) (0.00)Bishop Opposed*Opp. Depts. −0.02 −0.04 −0.01 −0.01 −0.01 −0.00

(0.01) (0.04) (0.01) (0.00) (0.01) (0.00)Log(1970 Pop.) 2.78∗ 2.16∗∗ 0.61∗∗ 0.62∗∗

(1.15) (0.64) (0.20) (0.20)Young Men Share −48.41 −10.87 −9.58∗∗ −3.14

(31.05) (6.48) (3.54) (1.78)Literate Share −18.99 −9.36∗∗ −3.68∗ −2.65∗

(11.98) (3.28) (1.80) (1.16)Child Mortality 5.17 3.24∗ 0.59 1.41

(3.96) (1.43) (0.48) (0.81)Foreign Residents 26.71∗ 7.71 6.31∗∗ 3.48

(10.16) (3.91) (1.89) (1.86)Archdiocese 3.76 2.74∗∗ 0.76∗ 0.94∗∗

(2.46) (1.00) (0.34) (0.32)Tenure −0.04 0.01 0.00 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.07 −0.04 −0.00 −0.01


(1.42) (1.98) (0.38) (1.32)(Intercept) 7.33 6.17 −6.94 1.70 0.77 −3.44

(3.73) (12.58) (6.20) (1.11) (1.99) (2.26)R2 0.28 0.09 0.29 0.14 0.11 0.17Adj. R2 0.28 0.08 0.29 0.14 0.10 0.16Observations 3715 3715 3715 3715 3715 3715Autocovariate Y Y Y Y Y YFixed Effects Year, Sub. Year Year, Sub. Year, Sub. Year Year, Sub.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

28


Table C10: Effect of Bishop Opposition with Linear Time Trend


Bishop Opposed −1.27∗ −1.94† −0.79 −1.89† −0.47∗ −0.42† −0.28 −0.47†

(0.60) (1.11) (0.60) (1.08) (0.21) (0.22) (0.18) (0.26)Time Trend −0.96∗∗ −0.97∗∗ −0.97∗∗ −0.95∗∗ −0.26∗∗∗ −0.26∗∗∗ −0.26∗∗∗ −0.25∗∗∗

(0.31) (0.33) (0.32) (0.31) (0.07) (0.07) (0.07) (0.07)Log(1970 Pop.) 3.60∗ 2.29∗∗∗ 0.74∗∗ 0.66∗∗∗

(1.64) (0.60) (0.22) (0.19)Young Men Share −58.70 −10.40 −11.40∗∗ −2.98

(37.90) (6.58) (4.24) (1.94)Literate Share −26.45 −9.99∗∗ −4.93∗ −2.86∗

(16.30) (3.22) (2.03) (1.12)Child Mortality 4.82 3.22∗ 0.51 1.43


(20.09) (4.37) (2.44) (1.85)Archdiocese 3.93 2.89∗∗ 0.80∗ 1.01∗∗

(2.56) (1.03) (0.36) (0.33)Tenure −0.04 −0.01 0.00 0.00

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.00


(1.77) (1.91) (0.35) (1.31)(Intercept) 1908.69∗∗ 1930.44∗∗ 1912.70∗∗ 2124.65∗∗ 513.46∗∗∗ 515.70∗∗∗ 514.67∗∗∗ 525.86∗∗∗

(622.60) (650.59) (636.53) (621.83) (131.58) (136.22) (134.18) (129.60)Adj. R2 0.27 0.07 0.28 0.23 0.12 0.09 0.15 0.15Observations 3851 3763 3731 4003 3851 3763 3731 4003Fixed Effects Subzone Subzone Dpt. Subzone Subzone Dpt.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

29


Table C11: Random Effects Models for Main Results


Bishop Opposed −1.30 −2.06∗ −0.95 −1.34 −0.49∗∗ −0.48∗∗ −0.35† −0.26(0.80) (0.93) (0.95) (1.04) (0.18) (0.18) (0.21) (0.19)

Log(1970 Pop.) 3.59∗∗∗ 2.32∗∗∗ 0.74∗∗∗ 0.68∗∗∗

(0.32) (0.35) (0.06) (0.08)Young Men Share −58.88∗∗∗ −11.22 −11.45∗∗∗ −3.36

(12.90) (12.66) (2.54) (2.74)Literate Share −26.50∗∗∗ −10.45 −4.96∗∗∗ −3.06∗

(5.89) (6.22) (1.16) (1.34)Child Mortality 4.82∗ 3.36 0.51 1.49∗∗

(2.09) (2.38) (0.41) (0.51)Foreign Residents 41.98∗∗∗ 10.02 8.64∗∗∗ 4.31∗

(7.31) (7.88) (1.44) (1.70)Archdiocese 3.90∗∗∗ 2.88∗∗ 0.79∗∗∗ 1.01∗∗∗

(0.93) (1.06) (0.18) (0.23)Tenure −0.02 0.01 0.01 0.01

(0.05) (0.05) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.01

(0.04) (0.04) (0.01) (0.01)Militant Share 3.87∗∗∗ 4.23∗ 0.84∗∗∗ 1.75∗∗∗

(1.00) (2.04) (0.20) (0.44)(Intercept) 15.71 0.76 −0.08 2.06 2.24 −0.80 −3.26 0.50

(13.05) (7.29) (14.43) (1.10) (1.46) (1.45) (2.05) (0.28)Log Likelihood -16730.80 -16782.46 -16227.72 -17654.52 -10859.21 -10684.72 -10518.61 -11298.43Observations 3851 3763 3731 4003 3851 3763 3731 4003Groups: Subzone 18 18 18 18Groups: Year 8 8 8 8 8 8 8 8Groups: Department 503 503∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

30


Table C12: Effect of Bishop Opposition, Time-Department Level Covariate Interactions

Disappearances Killings(1) (2) (3) (4)

Bishop Opposed −3.33† −1.77∗ −0.65∗ −0.58∗

(1.94) (0.84) (0.29) (0.24)Log(1970 Pop) 13.73∗ 12.16∗∗ 3.45∗∗ 3.31∗∗

(5.82) (4.41) (1.12) (1.04)Young Men Share −168.53 −122.75 −42.44∗ −34.12∗

(119.30) (86.83) (19.73) (16.60)Literate Share −66.48 −50.82 −14.89 −12.81

(45.99) (31.88) (7.99) (6.69)Child Mortality 15.68 13.33 2.42 2.97

(13.09) (9.80) (2.31) (2.14)Foreign Residents 112.34 85.32 27.64∗ 24.16∗

(60.02) (44.16) (10.98) (9.85)Peronist Vote −0.14 −0.13 −0.01 −0.02


(5.56) (5.47) (1.59) (2.15)(Intercept) −36.95 −42.61 −10.69 −12.96

(39.29) (36.09) (9.92) (10.12)Adj. R2 0.17 0.39 0.24 0.31Observations 3763 3731 3763 3731Fixed Effects Year Year, Sub. Year Year, Sub.Interactions Y Y Y Y

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

31


Table C13: Effects of Bishop Opposition, Aggregated to Diocese-Year Level

Disappearances, t Killings, t(1) (2) (3) (4)

Bishop Opposed −12.45∗ −16.38∗ −2.50∗ −1.73(5.28) (8.23) (1.21) (1.49)

Log(1970 Pop.) 27.96∗ 3.67(12.67) (2.31)

Young Men Share −749.62 −64.38(535.42) (85.98)

Literate Share −212.29 −10.50(200.14) (38.93)

Child Mortality 89.87∗ 6.11(44.45) (7.38)

Foreign Residents 285.07 26.02(164.38) (24.68)

Archdiocese 8.20 4.99(8.14) (2.63)

Tenure −0.20 −0.01(0.34) (0.08)

Peronist Vote −1.91 −0.08(1.07) (0.15)

Militant Presence −1.81 3.01(12.44) (3.03)

(Intercept) 67.50∗∗∗ 138.51 19.55∗∗∗ −2.48(18.85) (237.90) (4.71) (59.43)

Adj. R2 0.12 0.27 0.18 0.28Observations 445 397 445 397Fixed Effects Year Year Year Year∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

Table C14: Negative Binomial Estimates, Bishop Opposition


Bishop Opposed −0.57∗∗∗ 0.15 −0.69∗∗∗ -0.02(0.07) (0.10) (0.01) (0.28)

(Intercept) 1790.56∗∗∗ 1752.26∗∗∗ 2051.36∗∗∗ 2024.6∗∗∗

(17.93) (65.05) (60.77) (162.70)

Observations 4003 3763 4003 3763Controls No Yes No YesTime Trend Yes Yes Yes Yes

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

32


Figure C3: Graphical Representation of Sensitivity Analysis, Disappearances Models

Note: Figure depicts the correlation between an omitted variable and the dependent variable necessary torender the main results insignificant.

Figure C4: Graphical Representation of Sensitivity Analysis, Killings Models

Note: Figure depicts the correlation between an omitted variable and the dependent variable necessary torender the main results insignificant.

33


Appendix D: Instrumental Variables Analysis

1. Results for instrumental variables for bishops by papal appointment, Table D1.

2. Diagnostics for instrumental variables analysis, Table D2. The results suggest a strong

instrument, with an F-statistic of 14.

3. Validation of exclusion restriction for IV estimation: relationship of Pope with repres-

sion in dioceses without regime-opposed bishops, Table D3. The IV assumes that the

appointing Pope affects repression only through his appointment of regime-opposed

bishops. Therefore, there should be no direct relationship between Pope and repres-

sion in areas in whose bishops are unopposed to the regime. This is borne out in the

findings, in which there is a consistently null relationship between a liberal Pope (Paul

VI) and repression in the sample of unopposed bishops.

4. Sensitivity analysis for instrumental variables model, Figure D1. The procedure, from

Wang et al. (2018), estimates the proportion of the instrumental variable results that

would need to be explained by structural error (violation of the exclusion restriction and

endogeneity of the instrument and outcome with unobserved confounders) in order to

invalidate the results. The black horizontal line in the figure represents this threshold.

The procedure reveals that up to 5% of the results would need to be explained by

structural error. In other words, despite the evidence that appointing pope does not

have a direct effect on repression there is some margin for such a direct effect which

could otherwise be undetected.

34


Table D1: Instrumented Bishop Opposition by Papal Appointment

Disappearances Killings(1) (2)

Bishop Opposed −3.39† −1.07†

(2.00) (0.64)Year −0.86∗∗∗ −0.26∗∗∗

(0.24) (0.07)(Intercept) 1706.33∗∗∗ 512.30∗∗∗

(476.29) (146.87)

Observations 4003 4003

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

Table D2: Instrumental Variables Diagnostics

Df1 Df2 Statistic P-value

Weak Instruments 1 3787 13.73 0.00Wu-Hausman 1 3786 0.11 0.73

35


Table D3: Validation of Exclusion Restriction: Papal Effects on Non-Opposed Dioceses


Pope Liberal −0.01 0.90 −0.42 −3.13 0.01 −0.12 0.06 −0.64(0.03) (2.31) (1.30) (1.88) (0.00) (0.40) (0.31) (0.34)

Log(1970 Pop.) 4.28∗ 2.57∗∗∗ 0.84∗∗ 0.74∗∗

(2.10) (0.72) (0.27) (0.23)Young Men Share −77.16 −10.32 −14.49∗ −1.37

(51.96) (10.24) (5.76) (3.16)Literate Share −22.55 −8.99 −4.27∗ −2.58∗

(14.88) (4.50) (1.96) (1.14)Child Mortality 6.53 4.38∗ 0.76 1.85

(5.50) (1.71) (0.65) (0.98)Foreign Residents 49.21 12.76∗∗ 10.26∗∗ 4.81∗∗

(26.49) (4.55) (3.47) (1.66)Archdiocese 4.38 3.11∗ 0.76 0.97∗∗

(3.05) (1.15) (0.39) (0.34)Tenure −0.14 −0.04 0.00 −0.01

(0.21) (0.07) (0.03) (0.02)Peronist Vote −0.06 −0.06 −0.00 −0.02


(2.01) (2.55) (0.45) (2.20)(Intercept) 8.31∗∗ 1.87 −9.64 248.88∗∗∗ 2.27∗∗ 0.56 −4.74 29.88∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

36


Figure D1: Instrumental Variable Sensitivity Analysis

37


Appendix E: Alternative Explanation

1. Test of alternative explanation in which supporting bishops could increase repression

in their dioceses, Table E1. It could be the case that the results do not indicate

the effects of regime-opposed bishops in reducing repression but rather the effect of

regime-supporting bishops in increasing repression. The argument has asserted that

the actions of regime-supporting bishops and bishops who are simply unopposed to

the regime will have the same effects. To test this claim and address the alternative

explanation, additional coding is conducted among unopposed bishops. A variable,

Bishop Support is constructed which takes the value of 1 if the bishop was unopposed

and made statements and took actions consistent with supporting the military and

justifying violence against targets of repressions. It takes the value of zero if the

bishop was unopposed and did not make such statements or take such actions.

2. Test of alternative explanation with opposed bishops dropped, so comparisons are only

made among unopposed bishops, Table E2. Estimates of the effect of bishop support

on repression are most not significant and, if anything, slightly negative.

3. Verification that the results for the test of the alternative explanation are not an artifact

of the reduced sample size (2961 department-year observations), Figure E1. To do so, I

re-estimate the main results for bishop opposition on 500 random samples of the same

size as the test of the alternative explanation. Every estimate is negative, and more

than half are significant at the p = 0.05 level.

38


Table E1: Comparing Effects of Opposed and Supporting Bishops


Bishop Opposed −1.81† −3.46 −1.71∗ −3.57∗ −0.57† −0.60† −0.50∗ −0.84∗

(0.91) (2.17) (0.79) (1.71) (0.29) (0.35) (0.24) (0.37)Bishop Support −0.71 −1.80 −0.84 −2.29∗ −0.12 −0.15 −0.18 −0.43†

(0.86) (2.07) (0.66) (1.13) (0.27) (0.34) (0.21) (0.24)Log(1970 Pop.) 3.56∗ 2.28∗∗∗ 0.74∗∗ 0.66∗∗∗

(1.61) (0.59) (0.21) (0.19)Young Men Share −59.80 −10.65 −11.53∗ −3.05

(38.69) (6.62) (4.33) (1.94)Literate Share −26.38 −10.42∗∗ −4.96∗ −2.98∗

(16.02) (3.28) (2.00) (1.14)Child Mortality 4.55 2.96∗ 0.48 1.37


(20.07) (4.30) (2.47) (1.84)Archdiocese 3.93 2.74∗∗ 0.79∗ 0.96∗∗

(2.61) (0.97) (0.36) (0.31)Tenure 0.02 0.03 0.01 0.01

(0.07) (0.05) (0.01) (0.01)Peronist Vote −0.05 −0.03 −0.00 −0.00


(1.89) (1.95) (0.35) (1.31)(Intercept) 16.67∗∗∗ 7.02 −3.46 244.45∗∗∗ 4.59∗∗∗ 0.96 −2.39 29.01∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

39


Table E2: Supporting Bishops and Repression


Bishop Support −1.05 −1.37 −0.39 −3.71∗ −0.23 −0.12 −0.09 −0.60†

(0.92) (1.94) (0.63) (1.65) (0.28) (0.35) (0.19) (0.34)Log(1970 Pop.) 4.22∗ 2.56∗∗∗ 0.83∗∗ 0.73∗∗

(2.05) (0.72) (0.26) (0.22)Young Men Share −79.14 −9.76 −14.42∗ −1.41

(54.28) (10.33) (5.98) (3.22)Literate Share −23.17 −9.03 −4.23∗ −2.61∗

(15.44) (4.53) (1.97) (1.16)Child Mortality 6.13 4.22∗ 0.71 1.82

(5.24) (1.74) (0.62) (0.98)Foreign Residents 49.94 12.31∗ 10.03∗∗ 4.86∗∗

(28.10) (4.79) (3.51) (1.70)Archdiocese 4.32 3.12∗ 0.78 0.95∗∗

(2.98) (1.18) (0.39) (0.35)Tenure −0.08 −0.05 0.00 −0.00

(0.14) (0.08) (0.02) (0.02)Peronist Vote −0.05 −0.06 −0.00 −0.02


(2.15) (2.56) (0.47) (2.19)(Intercept) 15.47∗∗∗ 4.53 −9.95 245.60∗∗∗ 4.27∗∗∗ 0.46 −4.59 29.20∗∗∗

(4.11) (14.02) (8.64) (2.34) (0.81) (2.31) (3.60) (0.62)Adj. R2 0.28 0.08 0.29 0.23 0.14 0.10 0.16 0.15Obseravations 2791 2717 2697 2915 2791 2717 2697 2915Fixed Effects Year, Sub. Year Year, Sub. Year, Dpt. Year, Sub. Year Year, Sub. Year, Dpt.

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

40


Figure E1: Coefficient Estimates with 500 Random Samples of 2961 Observations

Note: Figure depicts coefficient estimates and confidence intervals for the effect of opposed bishops ondisappearances from specifications including subzone and year fixed effects, when taking random samples ofequivalent size to the test of the alternative explanation. Estimates are ordered by coefficient size.

41


Appendix F: Mechanism Tests

Institutional Actions

1. Test for bishop-priest complementarities, Table F1. This test uses the number of

priests aligned with the Movement of Priests for the Third World at the only time

such a comprehensive tally was taken (1967) as an indicator of a diocese’s overall level

of priest opposition to the regime. Given that priests are the agents of bishops, an

implication of the protecting targets mechanism is that more priests aligned with this

progressive movement would be willing to assist opposed bishops in their efforts to help

targets evade repression.

2. Test for bishop-ecclesial base community complementarities, Table F2. This test in-

teracts the indicator of whether a diocese had ecclesial base communities in 1976.

Ecclesial base communities were under the authority of diocesan bishops, and were

hubs of activity for the progressive church during the dictatorship. An implication of

the protecting targets mechanism is that the base communities, like opposed priests,

would assist opposed bishops in helping targets evade repression.

3. Outcome test for bishop opposition’s relationship with post-dictatorship Catholicity

in each diocese, Table F3. Dioceses are averaged across different bishops’ opposition

status if there were multiple bishops during the dictatorship. The results suggest

bishops opposed to the regime had dioceses which had 1.3% more Catholic adherents

in 1988 than bishops who did not oppose the regime.

4. Test for bishop opposition’s relationship with post-dictatorships sites of memory which

the human rights movement pushed to create for remembering repression, Table F4.

42


Dioceses’ bishop opposition is assigned according to the attitudes of the bishop who

presided for the majority of the dictatorship. Bishop opposition is interacted with

militant presence—a strong predictor of repression based on the main results. Controls

are included which would plausibly relate to additional number of post-dictatorship

human rights activity: population and pre-dictatorship political preferences.

5. Test for bishop opposition’s relationship with post-dictatorship trials for crimes against

humanity in which former regime repressors were defendants, Table F5. Trials are the

cumulative sum of trials from 2005-2016 in the province corresponding to the bishop’s

diocese. Dioceses’ bishop opposition is assigned according to the attitudes of the bishop

who presided for the majority of the dictatorship.

6. Finally, Rupflin (2015) validates this mechanism in the case of Jaime de Nevares of

the Diocese of Neuquen: “during the dictatorship, the Diocese of Neuquen provided

important material and symbolic resources...for the relatives of the disappeared and

the people who acted in defense of human rights” (66).

43


Table F1: Priest-Bishop Interaction Effects


Bishop Opposed −0.66 0.15 −1.06 −1.86 −0.19 −0.21 −0.26 −0.51(0.69) (1.58) (0.76) (1.26) (0.20) (0.24) (0.21) (0.30)

MSTM Priests 0.32∗∗∗ 1.22 0.06 −0.50∗∗ 0.11∗∗∗ 0.16 0.04 0.02(0.06) (0.94) (0.11) (0.17) (0.02) (0.09) (0.03) (0.03)

MSTM*Bishop Opp. −0.25∗∗∗ −1.13 −0.06 −0.09 −0.10∗∗∗ −0.14† −0.05∗ −0.02(0.07) (0.87) (0.11) (0.09) (0.03) (0.08) (0.03) (0.02)

Log(1970 Pop.) 2.64∗∗ 2.29∗∗∗ 0.62∗∗∗ 0.65∗∗

(0.83) (0.64) (0.16) (0.20)Young Men Share −24.01 −9.86 −6.92∗ −2.18

(17.34) (7.88) (2.60) (2.09)Literate Share −18.34∗ −10.58∗ −4.17∗∗ −2.86∗

(8.68) (4.12) (1.47) (1.33)Child Mortality 5.66 3.28∗ 0.60 1.50

(4.19) (1.52) (0.52) (0.87)Foreign Residents 31.89∗∗ 8.66 7.38∗∗∗ 3.49

(11.61) (4.46) (1.79) (1.91)Archdiocese −1.16 3.11 0.18 0.96

(3.38) (1.70) (0.52) (0.48)Tenure −0.00 −0.00 0.01 0.01

(0.10) (0.05) (0.02) (0.01)Peronist Vote −0.05 −0.03 −0.00 −0.01


(2.09) (2.04) (0.42) (1.40)(Intercept) 15.36∗∗∗ −2.56 −3.00 268.05∗∗∗ 4.31∗∗∗ 0.05 −2.28 28.00∗∗∗


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

44


Table F2: Opposed Bishops and Ecclesial Base Communities

Base Community Support(1) (2) (3) (4)

Bishop Opposed 0.18† 0.16† 0.19∗ −0.07(0.09) (0.08) (0.08) (0.10)

Log(1970 Pop.) 0.01 0.04∗∗

(0.02) (0.01)Young Men Share 0.50 0.32

(0.42) (0.39)Literate Share −0.17 −0.16

(0.43) (0.21)Child Mortality 0.09 0.15∗

(0.08) (0.07)Foreign Residents −0.19 0.35

(0.44) (0.33)Archdiocese −0.02 0.04

(0.04) (0.07)Tenure 0.00 −0.00

(0.00) (0.00)Peronist Vote −0.00 −0.00

(0.00) (0.00)Militant Presence 0.08 −0.03

(0.09) (0.05)(Intercept) −0.14 0.04 −0.55 −0.04∗

(0.07) (0.28) (0.31) (0.02)Adj. R2 0.26 0.14 0.29 0.70Observations 3899 3795 3763 4051Fixed Effects Year, Sub. Year Year, Sub. Year, Dpt.∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

45


Table F3: Relationship of Bishop Opposition with Post-Regime Catholicity

Post-Regime Catholicity(1) (2) (3)

Bishop Opposed 2.24∗ 3.01∗ 1.95†

(1.04) (1.44) (1.07)Log(1970 Pop.) 0.52 −0.27

(0.54) (0.22)Literate Share 12.72∗ 4.51

(5.79) (4.27)Foreign Residents −16.84∗∗ −9.70

(5.98) (4.97)Child Mortality 7.15∗∗ 2.21∗

(2.08) (1.06)Young Men Share 0.21 −0.85

(11.06) (6.65)Archdiocese 0.14 −1.77

(1.50) (1.75)Peronist Vote 0.07 −0.02

(0.04) (0.03)MSTM Priests −0.12 −0.16

(0.10) (0.10)(Intercept) 90.97∗∗∗ 70.35∗∗∗ 93.70∗∗∗

(2.06) (9.87) (4.32)

Adj. R2 0.49 0.21 0.57Observations 486 473 469Fixed Effects Subzone Subzone∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

46


Table F4: Relationship of Bishop Opposition with Post-Regime Sites of Memory

Post-Regime Sites of Memory(1) (2) (3)

Bishop Opposed 0.04 0.02 0.04(0.03) (0.03) (0.03)

Log(1970 Pop.) 0.06∗∗∗ 0.06∗∗∗

(0.01) (0.01)Foreign Residents 0.65∗∗ 0.45∗

(0.21) (0.21)Literate Share −0.08 −0.03

(0.14) (0.18)Young Men Share −0.52 −0.63

(0.29) (0.32)Archdiocese 0.00 0.01

(0.03) (0.03)Bishop Tenure −0.00 −0.00

(0.00) (0.00)Peronist Vote 0.00 0.00

(0.00) (0.00)MSTM Priests 0.00 0.01∗

(0.00) (0.00)(Intercept) 0.21 −0.39∗ −0.46∗

(0.14) (0.16) (0.22)


47


Table F5: Relationship of Bishop Opposition with Post-Regime Trials

Post-Regime Human Rights Trials(1) (2) (3)

Bishop Opposed 0.01 −0.00 −0.00(0.03) (0.05) (0.06)

Log(1970 Pop.) 0.10∗∗∗ 0.12∗∗∗

(0.02) (0.02)Foreign Residents 0.66∗ 0.07

(0.30) (0.30)Literate Share −0.04 0.18

(0.21) (0.25)Young Men Share −0.50 −0.48

(0.45) (0.51)Archdiocese 0.02 0.03

(0.04) (0.05)Bishop Tenure −0.00 0.00

(0.00) (0.00)MSTM Priests 0.00 0.00

(0.00) (0.00)Peronist Vote 0.00 0.00

(0.00) (0.00)(Intercept) 0.77∗∗∗ −0.85∗∗ −0.72∗

(0.15) (0.25) (0.35)


48


Religious Orders

1. Enumeration of parishes in Buenos Aires with opposed religious orders based on de-

scription of order activities in Catoggio (2010), Table F6.

2. Test indicating CCDs are a valid proxy for repression, Table F7. With a national-

level test using CCD locations, the relationship between CCDs and each measure of

repression is positive and significant.

3. Test of an alternative explanation in which religious orders more generally drive the

results, Table F8.

Table F6: Opposed Religious Orders in Buenos Aires

Parish Order

San Pedro Apostol SalesianosSan Juan Evangelista SalesianosSan Carlos Borromeo SalesianosSan Juan Bosco SalesianosNrta. Sra. de los Remedios SalesianosSan Patricio PalotinosSanta Isabel de Hungrıa PalotinosNtra Sra. de las Mercedes AsuncionistaSan Martın de Tours AsuncionistaSanto Cristo Sagrados CorazonesJesus Salvador Sagrados CorazonesNtra. Sra. de Fatima Sagrados CorazonesNtra. Sra. de los Dolores Sagrados CorazonesSanta Maria Magdelena de Betania Sagrados Corazones

49


Table F7: CCDs As a Valid Proxy for Repression


CCDs 5.87∗∗ 5.13∗∗ 1.00∗∗∗ 0.97∗∗∗

(1.97) (1.71) (0.15) (0.16)Log(1970 Pop.) −0.58 −1.11 0.02 0.01

(0.55) (1.08) (0.06) (0.08)Young Men Share −23.08∗ 7.37 −5.33∗∗ 0.45

(11.02) (8.73) (1.91) (1.69)Literate Share −14.01∗ −9.88∗∗ −2.70∗∗ −2.73∗∗

(5.40) (3.63) (0.78) (0.95)Child Mortality 1.90 0.18 0.04 0.84

(1.37) (1.36) (0.20) (0.52)Foreign Residents 6.74 −8.25 2.53∗ 0.59

(5.08) (7.20) (1.14) (1.35)Archdiocese 1.89∗ 1.25 0.48∗ 0.73∗∗

(0.92) (0.95) (0.23) (0.24)Bishop Tenure −0.01 0.02 0.00 0.00

(0.04) (0.03) (0.01) (0.01)Peronist Vote −0.05 −0.05 0.00 −0.01

(0.05) (0.03) (0.01) (0.01)Militant Presence 3.00∗∗ 0.15 0.70∗ 0.94

(1.08) (1.73) (0.28) (0.95)(Intercept) 21.09 25.14∗ 2.55∗ 2.28

(11.93) (12.46) (1.07) (1.40)

Adj. R2 0.50 0.57 0.39 0.40Observations 3828 3796 3828 3796Fixed Effects Subzone Subzone∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

50


Table F8: Religious Orders and Nearest CCD Distance

Distance to Nearest CCD (km)(1) (2) (3)

Any Order 0.13 0.15 0.08(0.10) (0.09) (0.08)

(Intercept) 0.78∗∗∗ 0.66∗∗∗ 0.50∗∗∗

(0.04) (0.08) (0.11)

Adj. R2 0.01 0.15 0.34Observations 154 154 154Fixed Effects Zone Neighborhood∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

51


Individual Actions

Background: Opposed bishops may also have reduced repression through individual ac-

tions, attempting to persuade regime officials. For example, Bishop of Viedma Miguel

Hesayne appealed privately to Minister of the Interior Albano Harguindeguy regarding re-

pression in his diocese. Hesayne asserted that “torture is immoral whoever employs it. It is

violence and violence is un-human and un-Christian” (Diario de Juicio 1985). However, bish-

ops’ private influence is difficult to falsify: there are few accounts of bishops’ interactions

with regime officials, and the Argentine Catholic Church forbids release of Dirty War-era

records. Nonetheless, I present the following tests using proxies for the individual actions

mechanism which suggest such actions may have been ineffective.

1. Test for bishop-episcopal conference complementarities, Table F9. This test interacts

an episcopal conference indicator—whether a bishop was an active member of the

Argentine episcopal conference—with the main indicator for a resisting bishop. In other

words, under this mechanism the members of the episcopal conference who resisted

the regime would be more effective at reducing repression than ordinary bishops who

opposed the regime, due to the authority of these conferences (Manzanares 1988).

2. Test for persuadable regime officers, Table F10. This test interacts an indicator for

whether a given department was under the jurisdiction of the Argentine cavalry during

the regime with the opposed bishop indicator. Cavalry officers were considered the

most reluctant to carry out the Dirty War, and thus could be more willing to reduce

repression under persuasion from an opposed bishop (Scharpf 2018).

3. Test for persuadable regime officers accounting for possible ceiling effects, Table F11.

This test interacts an indicator for whether a given department was under the juris-

52


diction of an Argentine officer in a known nationalist branch of the military (Infantry,

Artillery, and Communications) per Scharpf (2018) with the opposed bishop indica-

tor. If cavalry officers reduced repression to its minimum extent to maintain “law and

order”, opposed bishops would be no more effective in persuading them. Nationalist

officers, on the other hand, would have more repression that could be reduced from

opposed bishops’ appeals. However, there is no evidence that opposed bishops were

any more effective in reducing repression with these officers.

53


Table F9: Opposed Bishop Heterogeneous Effects by Episcopal Conference Service


Bishop Opposed −0.66 −2.45∗ −1.20† −0.26 −0.63∗ −0.37†

(0.69) (1.02) (0.71) (0.21) (0.24) (0.20)Episcopal Conference 1.57 −0.11 −1.08 0.56 −0.23 −0.17

(1.01) (1.98) (1.39) (0.32) (0.38) (0.37)Bishop Opposed*Conference −2.49† 1.20 −0.17 −0.84 0.58 −0.15

(1.33) (2.44) (1.39) (0.55) (0.43) (0.44)Log(1970 Pop.) 3.62∗ 2.28∗∗∗ 0.75∗∗∗ 0.66∗∗∗

(1.63) (0.61) (0.22) (0.19)Young Men Share −58.76 −11.57 −11.42∗∗ −3.18

(37.81) (7.00) (4.25) (2.01)Literate Share −26.52 −10.69∗∗ −5.02∗ −3.03∗

(15.84) (3.49) (1.98) (1.16)Child Mortality 4.92 2.84∗ 0.55 1.35


(19.70) (4.23) (2.44) (1.82)Archdiocese 3.88 3.39∗ 0.91∗ 1.07∗∗

(2.28) (1.41) (0.43) (0.39)Tenure −0.02 0.03 0.01 0.01

(0.10) (0.06) (0.01) (0.01)Peronist Vote −0.06 −0.03 −0.00 −0.01


(1.76) (1.87) (0.35) (1.30)(Intercept) 15.53∗∗∗ 5.90 −3.12 4.28∗∗∗ 0.84 −2.34


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

54


Table F10: Opposed Bishop Heterogeneous Effects by Cavalry Officers


Bishop Opposed −1.59∗ −2.21† −1.10† −0.57∗ −0.49∗ −0.37†

(0.70) (1.21) (0.63) (0.24) (0.24) (0.19)Cavalry Officer −1.96∗ −0.98 −2.06 −0.62∗∗ −0.29 −0.65∗

(0.96) (0.92) (1.13) (0.22) (0.19) (0.26)Bishop Opp.*Cavalry 1.32 −0.17 −0.29 0.39 −0.26 −0.19

(0.92) (1.72) (1.04) (0.24) (0.36) (0.27)Log(1970 Pop.) 3.54∗ 2.31∗∗∗ 0.72∗∗ 0.67∗∗∗

(1.64) (0.61) (0.22) (0.19)Young Men Share −58.63 −9.86 −11.34∗ −2.81

(38.05) (6.51) (4.26) (1.91)Literate Share −26.43 −10.48∗∗ −4.88∗ −3.01∗

(16.47) (3.26) (2.02) (1.14)Child Mortality 4.68 3.17∗ 0.48 1.42


(20.33) (4.40) (2.48) (1.86)Archdiocese 3.86 2.85∗∗ 0.78∗ 1.00∗∗

(2.53) (1.00) (0.35) (0.32)Tenure −0.01 0.01 0.01 0.01

(0.09) (0.05) (0.01) (0.01)Peronist Vote −0.07 −0.03 −0.00 −0.00


(1.79) (1.92) (0.36) (1.31)(Intercept) 16.61∗∗∗ 7.05 −3.84 4.64∗∗∗ 1.15 −2.48


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, †p < 0.10

55


Table F11: Opposed Bishop Heterogeneous Effects by Nationalist Officers


Bishop Opposed −1.68 −4.01∗∗ −2.34∗∗ −0.48 −1.03∗∗ −0.69∗∗

(1.09) (1.48) (0.86) (0.26) (0.35) (0.26)Nationalist Officer 1.82 −0.64 0.83 0.57 −0.02 0.33

(0.93) (1.21) (0.81) (0.29) (0.22) (0.21)Bishop Opp*Nationalist 0.15 2.18 1.50 −0.10 0.61∗ 0.37

(1.20) (1.27) (0.98) (0.30) (0.30) (0.27)Log(1970 Pop.) 3.60∗ 2.27∗∗∗ 0.73∗∗ 0.66∗∗∗

(1.68) (0.60) (0.22) (0.19)Young Men Share −58.69 −9.55 −11.39∗∗ −2.64

(37.99) (6.78) (4.25) (1.95)Literate Share −26.33 −9.96∗∗ −4.78∗ −2.84∗

(16.67) (3.20) (2.03) (1.12)Child Mortality 5.12 3.33∗ 0.55 1.45


(20.82) (4.41) (2.50) (1.88)Archdiocese 4.03 2.86∗∗ 0.79∗ 0.99∗∗

(2.70) (0.96) (0.37) (0.31)Tenure −0.01 0.01 0.01 0.01

(0.08) (0.05) (0.01) (0.01)Peronist Vote −0.07 −0.02 −0.00 −0.00


(1.82) (1.91) (0.36) (1.31)(Intercept) 15.17∗∗∗ 6.37 −5.29 4.18∗∗∗ 0.85 −2.98


∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

56


Appendix G: Archival Evidence

1. Enumeration of archival data, Table G1. The table indicates each archival source,

the number of documents used, the type of documents, and their general contents.

Archival data was used to construct the measure of bishop opposition used as the

main independent variable, and also to develop the qualitative discussion of the mech-

anisms. Documents used come from a larger universe of documents from which keyword

searches were used to identify documents related to relevant topics. This is presented

in the interest of transparency (Balcells and Sullivan 2018).

2. Discussion of archival evidence for individual action mechanism. The discussion rein-

forces the lack of support for this mechanism in the quantitative evidence.

57


Table G1: Summary of Archival Documents Used in Coding

Source No. Docs Useda Type Topics

ANMb 137 Press, reports, Bishop activities,testimonials process of repression

CELSc 158 Press, military memos, Bishop activities,academic studies repression, human rights

Conti Library6 Books Ecclesiastical law,

MSTM, military strategy

CyRd 31 Magazines Bishop activities,MSTM activities

Criterioe 12 Magazines Bishop activities

Cabildof 23 Magazine Bishop activities,regime strategy

U.S. govt.g232 Memos, publications Bishop activities,

intel. briefs human rights, repression

TOTAL589

aThis represents not total documents consulted but documents from which coding decisions were made.bArchive of National Memory.cCenter for Legal and Social Studies.dCristianismo y Revolucion, a progressive Catholic magazine published in Argentina 1966-1971.eA mainstream Catholic magazine.fA conservative Catholic magazine, using editions published from 1973-1982.gFrom the Argentine Declassification Project; consulted using keyword searches.

58


Archival, Primary, and Secondary Evidence for Individual Ac-

tion Mechanism

Bishops’ individual actions toward agents of repression is a mechanism which links bish-

ops’ opposition to the regime and the reduction of repression. For example, an anti-regime

bishop could plead with military officers in his diocese to consider the effects of repression

on human life and dignity. In turn, these officers would lessen the frequency or severity of

the repression they carried out as a result of shame. While anti-regime bishops did take

such actions, including withholding Catholic sacraments from repressive agents in several

dioceses (Mignone 1986, Morello 2015), there are two reasons to rule out this mechanism.

First, military officers had strong preferences toward repression which developed through

the combination of (1) selection into military service (Wallace 2014), (2) long-term accul-

turation to the ideology of the Argentine armed forces (Scharpf 2018), and (3) incentives

to repress harshly for career advancement (Arendt 1963, Gregory 2009, Scharpf and Glaßel

2019). Second, repressive agents’ interactions with the Argentine church occurred through

the Military Vicariate, a parallel Catholic institution embedded in the armed forces and

stacked with officials who saw the Dirty War as just. The Vicariate had 270 chaplains for

500,000 soldiers, a greater density of clergy than most civilian dioceses (Ruderer 2015, 14).

With the Military Vicariate so prominent in the life of the Argentine armed forces, it is

unlikely that repressive agents would be exposed in a churchgoing or confessorial capacity to

anti-regime bishops. This is consistent with the argument of Tsai (2007), who finds that the

co-embeddedness of the public, social leaders, and regime agents in the same institutions is a

necessary condition for the effectiveness of moral suasion. The primary interaction between

the military and bishops would have been through bishops’ communications with the regime

about the disappeared. The Military Vicariate encouraged officers’ strong preferences for

59


repression by presenting the Dirty War as a just war between true religion and atheism, a

frame which decreases altruistic behavior and encourages hostility toward out-groups (Hoff-

mann et al. 2019). For example, Commander of Naval Operations Luis Maria Mendıa told his

unit that “he had consulted with ecclesial authorities and they had approved the method of

[killing people by dropping them out of airplanes] with the consideration that it was a Chris-

tian and humanitarian death” (Museo Sitio Memoria 2019). Furthermore, military officers

continued to espouse the necessity of repression after the dictatorship (Marchak 1999).

60


Appendix H: Scope Conditions

1. Evidence of the Catholic Church’s public influence in Argentina, from a survey in which

respondents were asked to indicate their agreement with domestic institutions on a

100-point scale (Turner 1971). Responses are presented in Table H1. On aggregate, a

two-sample t-test indicates greater agreement with the Church than with the military

is significant. Individually, 78% of respondents expressed as much or more agreement

with the Church as they did the military.

2. Additional evidence for application of theory to the Argentine case, Table H2. The

table shows responses to a 1970 survey question about the institutions or groups re-

spondents believed were least indifferent to “those that are more at the margin” in

the country (CIMS 1970). Respondents rated the Catholic Church least indifferent. A

quote from Morello (2015) is also relevant to this point:

“The rest of the social space that was unoccupied by the state was the Churchitself. Since being ‘Catholic’ was identical to being a citizen, it was believed thatall social groups were represented by the Church; the Church identified itself asthe ‘public sphere’ ”(187).

3. Religious adherence to major institutional world religions (Buddhism, Christianity,

Hinduism, and Islam) by regime type, Figure H1. Regime type is defined according to

the criteria in Cheibub, Gandhi and Vreeland (2010). Between 1970 and 2000, religious

adherence declined in democracies (from 90% to 87%) while increasing in dictatorships

(from 79% to 87%). The increase in adherence in dictatorships is likely due to a shift in

which countries were dictatorships. For example, many more secular Eastern European

countries became democracies between 1970 and 2000. We would expect that religious

leaders are more likely to have influence in a given country as the share of adherents

61


increases: there are more members of the local community who would be congregants,

and more members of the public who would find religious leaders’ statements credible.

Data are from the World Religion Database (2020).

4. Religious adherence to major institutional world religions in dictatorships by region,

Figure H2. The high levels of adherence across continents—exceeding 80% in 2000—

suggests generalizability across cases in these regions, as suggested by the list of ex-

amples in the main text discussion. Particularly noteworthy are the rise of major

institutional religions in Africa and Asia, contrasting with the decline in the Americas.

5. Religious adherence to a country’s dominant religion, defined by religion with maximum

share of adherents in a given year, by regime type – Figure H3. It could be the case

that religious adherence is divided among many religious groups, therefore diluting

the influence of leaders who only have a small constituency to which to appeal. This

figure, compared with Figure H1 demonstrates that in democracies and dictatorships

alike, the dominant religion composes a large share of the religious adherents in a given

country.

6. Religious adherence to a country’s dominant religion: in dictatorships by region, Figure

H4. Dominant religions retain, across regions, a large share of total religious adherence

in their respective countries. We might expect the effects of religious leaders to be

greater when their respective religion is dominant.

7. Religious adherence in twenty Latin American countries, 2004-2018, Figure H5. Data

are from the Latin American Public Opinion Project (2020). This share includes mem-

bers of major world religions as well as practitioners of native religions and smaller

sects. With the exception of Uruguay (2006-2018), Chile (2016-2018), and Argentina

(2018), religious adherence exceeds 80%. Consistent with the discussion of global pat-

62


terns, we would expect religious leaders to exercise greater influence as the share of

adherents increases.

8. Catholic share in twenty Latin America countries, 2004-2018, Figure H6. Data are

from the Latin American Public Opinion Project (2020). Similarly to the discussion

of global patters, Catholicism is the dominant religion in Latin American countries

in the sample. However, Catholic adherence is more variable across country and is

generally declining as it faces competition from Protestant and Protestant-Evangelical

traditions (Gill 1998, Smith 2019). However, we might expect that even as the influence

of Catholic leaders declines as their share of adherents declines, there would be more

opposition to dictatorial behavior among those leaders because of the well-established

progressive turn Catholic leaders take when facing Protestant competition (Gill 1998,

Trejo 2012). This increase in opposition could offset the decrease in adherence in terms

of opposition’s aggregate political effects, though it remain an open empirical question.

63


Table H1: Public Agreement with Argentine Institutions, 1971

Institution Mean Agreement

Unions 41.4

Church 35.2

Industrial Firms 32.8

Politicians 30.5

Military 24.6

Table H2: Attitudes Toward Argentine Social Groups and Institutions, 1970

Group % Calling Indifferent

Catholic Church 7.3%

Working Class 7.7%

Military 8.5%

Politicians 12.2%

Agricultural Elite 18.5%

Urban Middle/Upper Class 24.4%

Not Sure/No Answer 21.3%

64


Figure H1: Institutional Religious Adherence by Regime Type

65


Figure H2: Institutional Religious Adherence in Dictatorships, by Region

66


Figure H3: Dominant Religion Adherence by Regime Type

67


Figure H4: Dominant Religion Adherence in Dictatorships, by Region

68


Figure H5: Religious Share in Latin America, 2004-2018

69


Figure H6: Catholic Share in Latin America, 2004-2018

70


References

Arendt, Hannah. 1963. Eichmann in Jerusalem: A Report on the Banality of Evil. New

York: Viking Press.

Balcells, Laia and Christopher M. Sullivan. 2018. “New findings from conflict archives: An

introduction and methodological framework.” Journal of Peace Research 55(2):137–146.

Braun, Robert. 2016. “Religious Minorities and Resistance to Genocide: The Collective

Rescue of Jews in the Netherlands during the Holocaust.” American Political Science

Review 110(1):127–147.

Catoggio, Marıa Soledad. 2010. “Cambio de Habito: Trayectorias de religiosas durante la

ultima dictadura militar argentina.” Latin American Research Review 45(2):27–48.

Cheibub, Jose Antonio, Jennifer Gandhi and James Raymond Vreeland. 2010. “Democracy

and dictatorship revisited.” Public Choice 143:67–101.

CIMS. 1970. “CIMS Poll1970-BUSINESS: Attitudes Toward Business.” Centro de Investi-

gaciones Motivacionales y Sociales .

Cinelli, Carlos and Chad Hazlett. 2020. “Making Sense of Sensitivity: Extending Omitted

Variable Bias.” Journal of the Royal Statistical Society: Series B (Statistical Methodology)

82(1):39–67.

Diario de Juicio. 1985. “2 de agosto de 1985.” Editorial Perfil S.A. .

Gill, Anthony. 1998. Rendering Unto Caesar: The Catholic Church and the State in Latin

America. Chicago: University of Chicago Press.

71


Gregory, Paul. 2009. Terror by Quota: State Security from Lenin to Stalin (an Archival

Study). New Haven: Yale University Press.

Hoffmann, Lisa, Matthias Basedau, Simone Gobien and Sebastian Prediger. 2019. “Uni-

versal Love or One True Religion? Experimental Evidence of the Ambivalent Effect of

Religious Ideas on Altruism and Discrimination.” American Journal of Political Science

Forthcoming:1–18.

Kuran, Timur. 1991. “Now Out of Never: The Element of Surprise in the East European

Revolution of 1989.” World Politics 44(1):7-48.

Latin American Public Opinion Project. 2020. “Data - All Merge 2004-2018.” Vanderbilt

University .

Manzanares, Julio. 1988. “The Teaching Authority of Episcopal Conferences.” Jurist 48:234–

263.

Marchak, Patricia. 1999. God’s Assassins: State Terrorism in Argentina in the 1970s. Mon-

treal: McGill-Queen’s University Press.

Mignone, Emilio F. 1986. Witness to the Truth: The Complicity of Church and Dictatorship

in Argentina, 1976-1983. New York: Orbis Books.

Morello, Gustavo. 2015. The Catholic Church and Argentina’s Dirty War. Oxford: Oxford

University Press.

Museo Sitio Memoria. 2019. Secretaria de Derechos Humanos y Pluralismo Cultural. Ac-

cessed July 13, 2019.

Ruderer, Stephan. 2015. “Between Religion and Politics: The Military Clergy during the

Late Twentieth-Century Dictatorships in Argentina and Chile.” Journal of Latin American

Studies 47(3):463–489.

72


Rupflin, Barbara. 2015. ““Somos ovejas de su rebano”. El papel de la diocesis de Neuquen

para las vıctimas de la dictadura militar argentina (1976-1983).” Archives de sciences

sociales des religions (170):61–77.

Scharpf, Adam. 2018. “Ideology and state terror: How officer beliefs shaped repression

during Argentina’s ’Dirty War’.” Journal of Peace Research 55(2):206–221.

Scharpf, Adam and Christian Glaßel. 2019. “Why Underachievers Dominate Secret Po-

lice Organizations: Evidence from Autocratic Argentina.” American Journal of Political

Science Forthcoming:1–16.

Smith, Amy Erica. 2019. Religion and Brazilian Democracy: Mobilizing the People of God.

New York: Cambridge University Press.

Trejo, Guillermo. 2012. Popular Movements in Autocracies: Religion, Repression, and In-

digenous Collective Action in Mexico. Cambridge: Cambridge University Press.

Tsai, Lily L. 2007. “Solidary Groups, Informal Accountability, and Local Public Goods

Provision in Rural China.” American Political Science Review 101(2):109–122.

Turner, Frederick C. 1971. “ARMISC1971-Turner: Various Argentina Adults.” University

of Connecticut - Roper Center for Public Opinion Research .

Wallace, Geoffrey P.R. 2014. “Martial Law? Military Experience, International Law and

Support for Torture.” International Studies Quarterly 58:501–514.

Wang, Xuran, Yang Jiang, Nancy R Zhang and Dylan S Small. 2018. “Sensitivity analysis

and power for instrumental variable studies.” Biometrics 74(4):1150–1160.

World Religion Database. 2020. “All religions by country.” Institute on Culture, Religion,

and World Affairs at Boston University .

73