Why Underachievers Dominate Secret Police

Organizations: Evidence from Autocratic

Argentina
Adam Scharpf GIGA German Institute of Global and Area Studies
Christian Gläßel University of Mannheim

Abstract: Autocrats depend on a capable secret police. Anecdotal evidence, however, often characterizes agents as surprisingly
mediocre in skill and intellect. To explain this puzzle, this article focuses on the career incentives underachieving individuals
face in the regular security apparatus. Low-performing officials in hierarchical organizations have little chance of being
promoted or filling lucrative positions. To salvage their careers, these officials are willing to undertake burdensome secret
police work. Using data on all 4,287 officers who served in autocratic Argentina (1975–83), we study biographic differences
between secret police agents and the entire recruitment pool. We find that low-achieving officers were stuck within the
regime hierarchy, threatened with discharge, and thus more likely to join the secret police for future benefits. The study
demonstrates how state bureaucracies breed mundane career concerns that produce willing enforcers and cement violent
regimes. This has implications for the understanding of autocratic consolidation and democratic breakdown.

Replication Materials: The data and materials required to verify the computational reproducibility of the results,
procedures and analyses in this article are available on the American Journal of Political Science Dataverse within the
Harvard Dataverse Network, at: https://doi.org/10.7910/DVN/PGFOXW.

“We don’t want clever people. We want medio- police forces run extensive spying networks to surveil and
crities.” detect conspiring enemies within the elite, the security
—Officer, Greek Military Police1 apparatus, and society at large. Acting on the suspicion
of subversion, units such as Czar Nicholas’s Okhrana,

A
utocracies are notorious for their systematic
Hitler’s Gestapo, or Assad’s Air Force Intelligence Direc-
violation of civil liberties and human rights
torate detained, tortured, and even killed their victims.
(e.g., Davenport 2007; Frantz and Kendall-Taylor
The relentless persecution of opponents makes secret
2014; Valentino 2004). At the core of autocratic repres-
police forces indispensable for the survival of autocratic
sion stand secret police forces (Arendt 2017; Friedrich
regimes. All the more surprising is that historical stud-
and Brzezinski 1965). For centuries, regimes have made
ies describe secret police agents as remarkably mediocre
extensive use of and largely depended on the loyalty of
in skill and intellect. Many members of Stalin’s NKVD,
such organizations (Baldwin 1934; Greitens 2016; Plate
for example, appear to have been poorly educated, lacking
and Darvi 1982). Charged with the responsibility to pro-
the capability of other Soviet bureaucrats (Gregory 2009).
tect the regime from internal and external threats, secret
This resonates with piecemeal information hinting at the

Adam Scharpf is Postdoctoral Research Fellow, GIGA German Institute of Global and Area Studies, Neuer Jungfernstieg 21, 20354 Ham-
burg, Germany ([email protected]). Christian Gläßel is PhD Candidate, Department of Political Science, and Researcher,
Collaborative Research Center SFB 884: Political Economy of Reforms, University of Mannheim, Parkring 47, 68159 Mannheim, Germany
([email protected]).
We are grateful to Belén González for her continued support in this project. We thank Sabine Carey, Kristine Eck, Alexander De Juan,
Felix Haass, Kristen Harkness, Anna-Lena Hönig, Nikolay Marinov, Cosima Meyer, Neil Mitchell, Felix Olsowski, Katrin Paula, Tom
Pepinsky, Jan Pierskalla, Constantin Ruhe, Gerald Schneider, Lucia Tiscornia, Christian von Soest, and Claudius Wagemann, as well as the
participants of the 2018 PEDD, NEPS, ECPR, and 2019 ISA conferences for helpful feedback. We also thank the three anonymous reviewers
and the editors for insightful comments and suggestions. This work was supported by the University of Mannheim’s Graduate School of
Economic and Social Sciences funded by the German Research Foundation.
1
The quote refers to the recruitment practice of the Greek secret police (Amnesty International 1977, 31).
American Journal of Political Science, Vol. 00, No. 00, xxxx 2019, Pp. 1–16

C 2019, Midwest Political Science Association DOI: 10.1111/ajps.12475

1
2 ADAM SCHARPF AND CHRISTIAN GLÄßEL

poor quality of secret police officers in other countries To test our theory, we draw on Argentina’s last dicta-
(Browder 1996; Persak and Kamiński 2005; Plate and torship. The case allows us to study secret police recruit-
Darvi 1982).2 However, whether underqualified agents ment in a country where sectarian cleavages are absent
systematically dominated the secret police in any of these and respective social classes too large to offer a mean-
cases is still unknown. Notwithstanding this open empir- ingful rationale for the selection of agents. Our analysis
ical question, theoretical work rationalizes why dictators offers systematic evidence on the composition of Intelli-
might staff key security organizations with incompetent gence Battalion 601—the most notorious and powerful
agents (Egorov and Sonin 2011; Zakharov 2016). They secret police unit in the history of Argentina. Between
suggest that leaders strategically recruit less-skilled of- 1975 and 1983, Battalion 601 was the operative brain
ficials, whom they deem to be more loyal. We advance behind the surveillance, detention, and forced disappear-
these arguments by dissecting the supply side in the re- ance of thousands of Argentine citizens, while assuming
cruitment of agents to explain why underachievers want a leading role in Operation Condor—a secret rendition
to join the secret police in the first place and why leaders network among right-wing dictatorships that targeted in-
accept them. dividuals across the Western Hemisphere (Dinges 2004;
We argue that the organizational structure of the reg- Ministerio de Justicia y Derechos Humanos [Ministry of
ular security apparatus3 produces career incentives for Justice and Human Rights; MJyDH] 2015).
less-skilled officials to work in the secret police. In hierar- We collect and analyze original microlevel data on
chical organizations, officials with only mediocre achieve- the profiles of officials to find out why individuals want
ments in the early stages of their career have little chance to serve in units designed to oppress societies. To the best
of being promoted and gaining lucrative positions either of our knowledge, this is the first empirical study on the
inside or outside the coercive bureaucracy. This generates enrollment of secret police agents. Combining informa-
strong career pressures for underachieving but aspira- tion from various archival sources, we compile a data set
tional officials to demonstrate commitment. The ardu- of more than 4,000 military officers who constituted the
ous nature of secret police work offers underachievers the entire recruitment pool for the Argentine secret police.
opportunity to signal their value to the regime and get These officers represent all individuals who entered the
ahead of competitors for higher positions. Leaders can officer corps between 1947 and 1975. We link this infor-
exploit these incentives and staff their secret police with mation with recently released data on all agents working
zealous officials. in Battalion 601 to study biographic differences between
Our theory makes sense of recruitment in regimes secret police members and their peers in all other parts
with and without salient social cleavages. Studies sug- of the army. The results from our analysis offer strong
gest that in countries with politicized divisions between support for our theoretical expectation and corroborate
classes, religions, or ethnicities, autocrats recruit secu- each part of our theoretical mechanism. Overall, we find
rity personnel from favored social groups (Albrecht and that officers who were under severe pressure due to poor
Ohl 2016; Greitens 2016; Hassan 2017; Sassoon 2016). performance in their early career stages were more likely
Leaders expect these individuals to follow repressive or- to join Battalion 601.
ders out of their shared interest in the regime’s survival. Adopting an organizational perspective, this study
Our organizational perspective explains which individu- offers unique insights into the institutional anatomy of
als within these societal segments are most likely to join authoritarian regimes. Despite the importance and long-
the secret police. In many other countries, sectarian or lasting consequences of secret police forces for politi-
social cleavages are too weak or the favored groups too cal regimes and domestic societies, the systematic study
large and heterogeneous to predict individual loyalty. By of such organizations has been hampered by sparse and
focusing on the career incentives universal to coercive bu- unreliable information (Barros 2016; Policzer 2009). We
reaucracies, our argument also explains which agents are expose the clandestine organization at the center of a dic-
willing to join secret police forces in states without such tatorship and scrutinize the individual agents who serve
divisions. in it. Other than psychological predispositions, deep ide-
ological convictions, or strong ethnic identities, we iden-
tify mundane but universal career concerns as a prime
2 motivator for officials to engage in arduous secret po-
Cases include Czechoslovakia, Germany, Greece, Iran, Paraguay,
Poland, Romania, South Africa, and Uruguay. lice work. Leaders exploit these incentives to maximize
3 compliance within organizations that require high lev-
Security apparatus and coercive bureaucracy refer to the state’s body
of formal security organizations, such as military, police, and in- els of operational autonomy to carry out their tasks but
telligence agencies. that are otherwise difficult to control. Career pressures
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 3

therefore serve regimes as the lubricant of their repressive the selection of agents is highly important. Discrim-
machines. inating among security personnel, regimes commonly
fill units with officials deemed most suitable (Bellin
2004; Harkness 2016). One might therefore expect that
Research on Dictators and Coercive leaders select the brightest and most capable bureaucrats
Agents to serve in the secret police (Baldwin 1934; McMahon and
Slantchev 2015). Anecdotal evidence, however, suggests
Studies on coercive institutions usually assume the otherwise. For example, in communist Poland, members
perspective of governments that are concerned about of the Bezpieka secret police have been described as “ex-
disobedience within the security apparatus (Bellin 2004; tremely undereducated,” with “no political or social expe-
Gregory 2009; Greitens 2016; Hassan 2017; Quinlivan rience” (Dudek and Paczkowski 2005, 242–43). Accounts
1999). For autocratic governments, disobedient secret of the czarist Okhrana, Lenin’s Cheka, Hitler’s Security
police forces are particularly dangerous, as their refusal to Service, the State Security of Czechoslovakia, and other
carry out orders commonly marks the end of the leader’s organizations hint at similar patterns (Blažek and Žáček
rule (Dragu and Przeworski 2019). Secret police agents 2005; Browder 1996; Zuckerman 1996). In view of the fact
uncover conspiracies and protect the regime against that regimes depend on an effective secret police, these
internal and external enemies, which makes leaders historical descriptions are puzzling.
vulnerable to resistance from within those organizations Two strands of research speak to the recruitment of
(Greitens 2016; Plate and Darvi 1982). Noncompliance incompetent officials. The literature on nepotist regimes
in the core repressive unit reveals the government’s weak- conceives the selection of low-skilled agents as an unin-
ness and enables opposition groups, rival regime elites, tended side effect of the leader’s recruitment decisions.
and other security forces to turn against it (Albrecht and Regarding the assignment of top officials, studies empha-
Ohl 2016; DeMeritt 2015; Dragu and Lupu 2018). size the centrality of personal connections to the ruler
Leaders seek to prevent these risks with utmost vigor. (Slater 2003; Taylor 2011). Sassoon (2016, 124–25), for
To deter disobedience and increase the compliance of example, describes how Saddam Hussein appointed his
agents, dictators set up monitoring systems and impose family members to lead the intelligence services irrespec-
draconian punishments (Gregory 2009; Quinlivan 1999; tive of their expertise. As it is impossible to fill entire
Sudduth 2017). Yet, control mechanisms are known to be organizations with friends and relatives, dictators have
largely ineffective (Brehm and Gates 1999). Monitoring been argued to staff important units with members of the
does not alter the preferences of subordinates, and many social groups deemed most loyal. Both General Qaddafi
governments lack the resources to identify and sanction and President Assad reportedly drew on aligned religious
rogue officials (Cole 2015; Conrad and Moore 2010).4 minorities and ethnic tribes to maximize loyalty among
Such control problems are particularly pronounced with their security forces (Quinlivan 1999).
regard to secret police forces (Greitens 2016; Policzer However, ethnicity and socioeconomic class are
2009). By design, these units operate in secrecy and broad categories that offer only limited information on
with great autonomy to be effective—by necessity, they the personal interests and future behavior of individu-
are experts in espionage and counterespionage (Plate als. It is unknown which markers leaders use to select
and Darvi 1982, 85–101). As a result, even parallel agents from within favored segments of society. Accord-
security forces often struggle to control the secret units ing to Sassoon (2016, 124), even in countries with highly
ultimately responsible for the regime’s protection (Dragu politicized social cleavages, such as Iraq and Syria, the
and Przeworski 2019; Gregory 2009). This makes it composition of security organizations is more heteroge-
inherently difficult to keep the secret police in check, neous than is commonly assumed—a fact that current
and it often allowed these organizations to accumulate studies do not explain. Furthermore, in many countries,
tremendous powers—turning them into what some have highly salient ethnic cleavages are absent or leaders lack
termed a “state within the state” (Arendt 2017, 556) or allied socioeconomic strata. In the absence of politically
a bureaucratic “monster” (Stepan 1988, 16). exploitable cleavages, it is unknown how regimes staff
Because of these control problems and the devastat- their core repressive units.
ing consequences from secret police failure, for rulers The second strand of research explicitly discusses po-
tential trade-offs between competence and loyalty to ex-
4 plain why leaders might deliberately employ low-skilled
Training and indoctrination do not guarantee obedience either,
as they imperfectly replace personal preferences, which exacerbates agents. According to Egorov and Sonin (2011), leaders
control problems if agents enjoy large autonomy (Scharpf 2018). fear that smart officials use their competence to oust them
4 ADAM SCHARPF AND CHRISTIAN GLÄßEL

in the future.5 Zakharov (2016), in turn, expects skilled Second, the organizations are pyramid shaped, with
agents to remain idle once their ruler is in danger, be- an abundance of positions at the bottom but only few
cause they consider themselves indispensable under any lucrative posts at the top (Blau and Scott 2003, 32). Of-
successor regime. Together, this suggests that leaders have ficials striving for promotion face a declining number
a strategic incentive to select less competent secret police of available positions at each career step, and only few
agents for the sake of greater loyalty. Our theory comple- individuals can ascend to the highest ranks. These built-
ments and refines these arguments. It dissects the supply in bottlenecks generate competition among junior and
side of organizational recruitment to explain why under- midlevel officials for senior billets (Kung and Chen 2011,
achieving officials want to join the secret police in the first 29–31). The career pressures are especially pronounced
place. We show that career pressures inherent to regular in security organizations with “up-or-out” promotion
security organizations drive the very individuals preferred systems, in which individuals who have been repeatedly
by dictators into the secret police. passed over for promotion are threatened with forced re-
tirement to make way for the next cohort (Baker, Jensen,
and Murphy 1988, 604–5).
Third, in pyramid-shaped organizations, officials
A Theory of Careers within Coercive climb up the hierarchy until they reach a position where
Institutions they are no longer competitive against their peers of equal
rank (Lazear 2004, 160). Leaders value both competence
We first examine inherent characteristics of the state’s and loyalty; both traits can be bases for promotion. How-
coercive bureaucracy, which typically constitutes the re- ever, as competence is an observable trait, whereas loyalty
cruitment pool for the secret police, in order to explain is not, the competitiveness of an individual is largely de-
how organizational bottlenecks create career pressures termined by her past performance (Lazear 2004, 159).
for officials with a hitherto weak track record. We then Officials typically lay the foundation for future advance-
describe why burdensome secret police service offers un- ment at an early stage (Becker and Strauss 1956, 256). The
derachieving officials the opportunity to advance their importance of early achievements is particularly well doc-
career and why leaders are likely to accept them into the umented for careers within the military.7 Cadets who per-
secret police. form well at the academy have better chances of gaining
Historically, many secret police forces have recruited access to higher education facilities, where they acquire
personnel from the regime’s larger security apparatus— management and leadership skills necessary to ascend to
including the military, gendarmeries, police departments, the upper echelons. Without such training, officials are
and conventional intelligence agencies (Browder 1996; likely to be stuck at midlevel ranks and lack the skills
Gregory 2009; Greitens 2016; Persak and Kamiński 2005; for lucrative employment outside the security apparatus
Sassoon 2016; Zuckerman 1996). These security organi- (Biderman and Sharp 1968, 388).8
zations share three general characteristics that shape the Due to these inherent characteristics of security or-
career trajectories and incentives of officials. First, the ganizations, low-performing officials face grim career
organizations are hierarchically structured. Members of prospects. This pressures them into pursuing unortho-
such institutions generally strive for promotions because dox routes by which they can climb up to higher ranks
rank within the hierarchy determines the amount of per- (Kung and Chen 2011, 28). One way to improve their own
sonal power, pay, and social prestige (Baker, Jensen, and promotion prospects in such circumstances is to apply for
Murphy 1988, 599; Svolik 2012, 168–69).6 unpopular assignments where underachievers can stand
out from their peers (Becker and Strauss 1956, 257). We
argue that the secret police offers exactly such posts.
5
In contrast to Egorov and Sonin (2011), McMahon and Slantchev
(2015) argue that leaders always have an incentive to hire skilled
guardians, while limiting their resources. Rulers would hire incom-
7
petent agents only if they mistakenly considered them competent. In prototypical military organizations, “promotions are consis-
As the authors point out, constituent components of the studies’ tently based on achievement criteria, which include relative stand-
proposed mechanisms are barely quantifiable and have thus not ing in one’s graduating class from the military academy, [and]
been systematically tested (Egorov and Sonin 2011, 906; McMahon attendance at advanced training centers” (Nordlinger 1977, 43).
and Slantchev 2015, 307–11). Early performance is even important in nepotist regimes, as per-
6 sonal networks have less impact on promotions at lower ranks
Top positions in coercive bureaucracies often entail perks, includ- (Moore and Trout 1978, 460–61; Sassoon 2016, 106–8).
ing priority health care and luxurious housing (Bellin 2004; Brooks
8
1998), or grant access to exclusive networks of corruption or illegal Civilian employers value managerial and leadership skills rather
activities (Droz-Vincent 2007). than purely security-related proficiencies (Janowitz 1988, 64).
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 5

In the name of protecting the regime, secret police status” otherwise (Browder 1996, 22, 83).10 Driven by the
agents must spy on, intimidate, torture, and even kill hope of future reward and promotion, underachieving
people. In contrast to regular police officers or soldiers, officials are therefore likely to put themselves forward for
agents do not use these measures against ordinary the secret police.
criminals or foreign armies but “more or less arbitrarily Regimes, on the other hand, have good reasons to ad-
selected classes of the population” (Friedrich and mit underperformers to their key coercive unit and to not
Brzezinski 1965, 22). This requires agents to regularly counteract the self-selection of such officials. From the
perpetrate repression that defies moral norms, incurs so- perspective of leaders, underachievers are beneficial be-
cial stigma, and entails high psychological burdens even cause they must diligently work toward the autocrat’s in-
for trained specialists (Grossman 1996, 222–26; Huggins, terests to retain their career chances. For example, Alexan-
Haritos-Fatouros, and Zimbardo 2002, 214–31).9 dru Drăghici, the former interior minister of Romania,
Brazilian secret police officers, for example, “manifested was well aware that his secret police agents “had a fairly
such stress-related symptoms as insomnia, hypertension, low level of training and general knowledge, but that these
fear, and depression [...] exacerbated by an inability to shortcomings were compensated for by their powerful
talk about their work” (Huggins, Haritos-Fatouros, and revolutionary enthusiasm” (Deletant 2005, 304). By con-
Zimbardo 2002, 15). Compared to service in the regular trast, leaders who assign high-performing officials to serve
security apparatus, secret police work is emotionally in the secret police must fear that their subordinates will
much more “difficult, arduous, and exhausting” (Plate not execute unpleasant orders. Having lucrative career
and Darvi 1982, 128). High-performing officials, who options, competent officials make unreliable secret police
do not face career pressure, therefore have little incentive agents, as they are likely to stay passive when the regime
to engage in burdensome secret police tasks (Browning is under attack (Zakharov 2016).
1998, 169). For underachievers, this reduces the competi- In the extreme case, leaders have to worry that ca-
tion from peers with hitherto superior achievements and pable agents even use their competence to actively work
improves their prospects for promotion within the secret against the government or seize an all-dominant position
police. in the regime (Egorov and Sonin 2011; Geddes, Wright,
Furthermore, pressured officials with poor past per- and Frantz 2018). Such risks are lower when it is not the
formance have an incentive to join the secret police be- most capable officials but rather loyal mediocrities who
cause they can expect that those who carry out undesir- are entrusted with the leaders’ protection. In this way,
able secret police tasks send a signal of commitment to the regime can fully benefit from top performers working
the regime. As secret police agents, they have the oppor- within the regular security apparatus or the larger regime
tunity to build a reputation of being loyal and zealous— bureaucracy while exploiting the career pressures of un-
characteristics that are highly valued by both superiors derachievers to forge a loyal secret police. Finally, direct
and leaders. In the words of a Uruguayan officer, re- superiors within the secret police, who joined the force
pressive zeal was “rewarded by the authorities either in because of their own weak performances, should also fa-
promotion or in assignment” (Plate and Darvi 1982, vor less-skilled subordinates. Acting upon their own ca-
141). While officials have difficulties proving such re- reer concerns, superiors hope that underachieving sub-
liability in the regular security apparatus (Moore and ordinates will not compete with or supersede them. This
Trout 1978, 455–56), secret police service allows agents to strengthens the career prospects for agents with weak per-
manifest their loyalty. Underachievers may therefore hope formances. Given the overlapping incentives of the regime
that their service in the secret police will not only improve and its agents, we expect that officials with inferior past
their prospects for promotion within the secret police it- achievements are more likely to serve in the secret police.
self but also open doors to higher ranks back in the regular
security apparatus. Anecdotal evidence suggests that un-
derachieving agents do indeed see their secret police ser-
vice as an opportunity to salvage their faltering careers. Empirical Case
In Nazi Germany, for example, police officers “blocked by
the ‘bottleneck’ of seniors ahead of them” sought to join We focus on Argentina’s last dictatorship to unearth, tri-
the Gestapo—anticipating that they “lacked any hope of angulate, and analyze data on officials in the secret police
10
Similarly, Polish functionaries saw the secret police as a vehi-
9
Agents may also worry about future repercussions, as dictators’ cle for social advancement (Dudek and Paczkowski 2005, 243).
henchmen are often the first to be held accountable (DeMeritt Paraguayan officials who “could not make it up the ranks” (Plate
2015). and Darvi 1982, 134) took over the unpopular task of torture.
6 ADAM SCHARPF AND CHRISTIAN GLÄßEL

and the surrounding security apparatus. Sparse infor- FIGURE 1 Graduate Performance and Career
mation has constrained research on authoritarian insti- Achievement
tutions (Barros 2016; Geddes, Wright, and Frantz 2018;
Policzer 2009). Past regimes thus offer the best oppor-
tunity to study secret police forces and their internal
composition (Balcells and Sullivan 2018). To the best
of our knowledge, the case of Argentina is unique in
that it permits us to systematically test whether and why
underperformers are more likely to serve in the secret
police.

Argentina’s Coercive System

The coercive system of 1970s Argentina comprised
regular law enforcement agencies and the military, as well
as a number of civil and military intelligence services.
The latter included the state intelligence secretariat, the
superintendence of the federal police, and the regular
intelligence agencies of the army, navy, and air force
(MJyDH 2015). Throughout Argentina’s modern history,
the army held a key position inside this system, repeatedly
intervened in politics, and ousted several democratically
elected governments (Potash 1996). Despite these
Note: The plot shows the relationship between offi-
interventions, the army’s recruitment, training, and
cers’ performances at the military academy and sub-
promotion system resembled a prototypical security
sequent career achievements. Graduation rank gives
organization with a hierarchical and meritocratic struc-
each officer’s placement relative to her respective co-
ture (Atkins and Thompson 1972).11 With an up-or-out
hort at the academy. The rank of zero identifies the
system in place, military officers faced considerable career
best-performing officers. Junior and midlevel officers
pressures from the outset. Figure 1 depicts the strong
and general officers denote ranks at retirement. Gray
relationship between early performance at the military
areas denote probability densities, dark areas indicate
academy and subsequent advancement to the highest
interquartile ranges, and black lines give medians.
positions. Cadets who graduated at the top of their
class were much more likely to rise to the general ranks,
whereas most graduates at the bottom of their class had in March 1976. Political power now was in the hands of
to leave the organization as midlevel officers. a junta called the “National Process of Reorganization”
(Lewis 2002, 131–32).12
For the military, the insurgent uprisings constituted
Intelligence Battalion 601 clear signs of a Soviet-supported attempt to take over Ar-
gentina and indeed the Western Hemisphere. The junta
Between 1975 and 1983, Argentina’s security forces faced was convinced that the insurgent groups were only the
violent uprisings by two insurgent groups: the People’s visible manifestations of a clandestine enemy that oper-
Revolutionary Army and the left-wing Peronist Mon- ated from within the Argentine society (Lewis 2002, 137–
toneros. In response, the military pressured the demo- 43). Intelligence therefore turned into the key resource in
cratic government of Isabel Perón for far-reaching pow- uncovering and destroying subversive networks. In this
ers; by 1975, the army was in charge of all matters of context, Intelligence Battalion 601 became the secret po-
internal security (Heinz 1999, 621–33, 680–85). Despite lice unit centrally responsible for the nationwide targeting
their authority, generals led by Jorge Videla ousted Perón of thousands of alleged subversives (MJyDH 2015).
To streamline the state’s repressive capacities, the mil-
11
Successful military careers typically featured good academy per- itary leadership established Battalion 601 as the single
formances, first assignments as junior officers, and advanced train-
ing at the Higher War School. High-performing midlevel officers
12
later managed larger units or acquired staff positions before be- Our results are not affected by changes in junta personnel or rebel
coming generals. capacity (see supporting information [SI], Section SI.4, 21–22).
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 7

most powerful organization “to which all other security FIGURE 2 Rank Distribution within Battalion
units were subordinate” (Dinges 2004, 112). Directly re- 601
porting to the junta and the General Army Command,
the battalion possessed what Greitens calls “coordinat-
ing authority” (2016, 25).13 All conventional intelligence Colonel

agencies and regular army units had to pass their intelli-

Midlevel
Lieutenant Colonel
gence on to 601 (MJyDH 2015, 11–29). Battalion agents
processed all incoming information before feeding it back
Major
into the coercive system, as well as up or down the chain of
command. At the same time, 601 collected its own intelli-
Captain
gence and conducted a huge number of repressive opera-

Junior level
tions. In short, “disappearance [was] 601 work” (National
First Lieutenant
Security Archive [NSA] 2002a, 1).
Internally, Battalion 601 consisted of three depart-
ments.14 Central de Contrainteligencia, in charge of Second Lieutenant

counterespionage, and Central de Apoyo, responsible for 0 10 20 30 40

Percentages
logistics, were overseen by the battalion’s vice chief. The
third department, called Central Reunión, constituted
the operational heart of Battalion 601 and was in charge
of ground-level policing and capture-or-kill missions ion.17 In general, every year soldiers could formally state
(MJyDH 2015, 24–29). It enjoyed great autonomy and their preferences for serving at specific locations or join-
reported to the chief of 601 only. To repress coarsely ing particular army units. Prospective secret police mem-
defined target groups, Central Reunión harbored seven bers could voice their desire to serve in 601, which was
task forces.15 The agents infiltrated organizations, spied then decided on by the general staff. Figure 2 shows that,
on citizens, and disappeared alleged subversives. Victims in particular, junior officers switched over to the secret
were typically kidnapped and deported to one of the police unit.18 Upon admission, 601 commanders unilat-
secret detention and interrogation centers, where most erally assigned agents to one of the departments within
of them were tortured and eventually killed (Lewis 2002, the battalion. On average, agents served in the unit for al-
147–59).16 most 3 years. Many of them reentered the regular security
apparatus after their secret police service.

Transitioning to the Battalion

Internal Validity
As Battalion 601 originated from the army, military of-
ficers could join the battalion and return to the regular The case of Argentina’s last dictatorship and its coercive
security apparatus with relative ease. Before entering 601, system allows us to test crucial components of our argu-
officers had worked in conventional army units, on mil- ment and to rule out alternative explanations. Argentina is
itary bases, or at army headquarters. In these positions, an ethnically and religiously homogeneous society. From
most officers trained enlisted ranks and assisted superiors 1890 to 1940, Prussian military advisors professionalized
with the administration of the army. Some of them also the Argentine army, turning it into a merit-based organi-
commanded units in regular military operations against zation (Atkins and Thompson 1972). Explanations based
insurgents (MJyDH 2015, 131–62). According to a former on favoritism are therefore unlikely to confound our anal-
601 agent, there were no specific entry requirements or ysis.
special processes governing the transition to the battal- Furthermore, 601 exclusively recruited members of
the Argentine officer corps, which enables us to compare
early career performances and isolate their effects. As part
13 of the regular armed forces, the battalion employed the
See SI.2 (4) for visualizations of Argentina’s coercive system and
601’s central position.
17
14
SI.2 (5) provides an organizational chart. Information is based on the authors’ interview with a former
Battalion 601 agent (August 2018).
15
SI.2 (5) offers information on the task forces. 18
Colonels served as Battalion chiefs, and lieutenant colonels
16
SI.2 (6) provides a mission report on a capture-or-kill operation headed individual departments. Regression results do not change
by 601 task force officers. when we exclude midlevel officers (SI.4, 20).
8 ADAM SCHARPF AND CHRISTIAN GLÄßEL

same military rank structure and promotion schemes— who graduated with officer i in cohort j. Graduation rank
which facilitated transitions to 601 and a smooth reinte- ranges from 0 to 100, with larger values indicating worse
gration into regular units afterward. Finally, formal entry relative performances. Officers with excellent grades who
requirements and selection procedures did not change graduated at the top of their class receive a value of 0,
during the junta’s rule. The case is therefore well suited whereas underachieving officers who graduated at the
to empirically test our theory. bottom of their class receive a value of 100. Higher val-
ues of Graduation rank should therefore predict greater
probabilities of serving in Battalion 601.
We include several pretreatment variables in the
Data and Method statistical analysis to account for potential confounders.
Officers may have held strong ideological convictions or
We compile an original data set to assess whether un- come from influential social classes, which could have
derachieving officers were more likely to serve in Ar- affected their graduation performance and motivated
gentina’s secret police. Using recently published infor- them to serve in the secret police. In Argentina, social
mation by the Argentine Ministry of Justice and Human class and ideological convictions clustered in army
Rights, we identify all officers who served in Battalion 601 branches (Scharpf 2018). Nationalist officers of low so-
(MJyDH 2015). We link this list to data on each officer’s cioeconomic status mostly served in the infantry branch,
professional career record (Figueroa 2008). These data whereas cavalry officers usually came from influential
stem from more than 20 army and government archives, military families of the comparatively liberal upper class.
offering unique information on all 4,287 army officers— The binary variables Cavalry officer and Infantry officer
including all 152 Battalion 601 members who graduated indicate whether an individual joined the respective
from the Colegio Militar de la Nación, the army’s offi- branch.
cer academy. Our final data set comprises all 30 officer Performances at the academy may have also been
cohorts who graduated between 1947 and 1975. In each affected by the quality of primary education. Educational
cohort, at least one officer served in 601. Officers who did quality is likely to influence professional achievements,
not serve in the battalion but worked in any other army shape upward mobility, and correlate with willingness to
unit function as counterfactuals in the empirical analysis. serve in 601. Since fine-grained socioeconomic data are
The unit of observation in our data set is the individ- unavailable before 1975, we calculate the variable Home
ual officer. Our dependent variable is binary, identifying literacy rate, measuring the literacy rate in each officer’s
each officer who served in Battalion 601 at any point be- home province (Lupu and Stokes 2009).22 The variable
tween 1975 and 1983.19 Officers who never joined the Cadet age controls for an officer’s age upon entering the
battalion are coded as 0. Given the type of our dependent academy. Young, ambitious cadets may have shown little
variable, we employ logistic regressions with standard er- hesitation about serving in 601 to boost their careers,
rors clustered on officer cohorts in the main analysis.20 while lacking the experience or skills needed to do well at
Our key independent variable is Graduation rank, the academy.
which measures each officer’s performance at the officer We further account for the content of military train-
academy relative to her cohort (Figueroa 2008).21 The ing, as it may affect both performance and willingness to
variable captures the relative educational performance at join the secret police. Training under military government
the beginning of the officer’s career with is a proxy for exposure to repressive counterrevolutionary
 
Ranki − 1 training. It measures the share of training (percentage)
Graduation ranki, j = ∗ 100, that cadets completed under former military rule
Cohort size j − 1
(Geddes, Wright, and Frantz 2014). Finally, graduation
where Rank is the absolute position of an officer i among
results may hinge on the amount of time officers spent at
her peers and Cohort size is the total number of cadets
the academy. Officers may have taken more time because
they had problems completing the coursework, or as
19
SI.1 (3) shows summary statistics. a way to strategically reduce course load and improve
20
We replicate all results using matched samples and ordinary least their grades. We therefore control for Training length,
squares with birth province and cohort fixed effects (SI.4, 19, 23– measuring the number of months an officer spent at the
24).
academy.
21
As in other officer academies, such as West Point, academic
achievements and military leadership performance accounted for
22
roughly 80% of the final grade, and the rest was a combination of Data are available only for election years. We code the literacy
athletic performance and disciplinary record. rate of the first election year after the officer’s birth year.
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 9

FIGURE 3 Battalion 601 Members vs. Other itary academy were more likely to serve in the regime’s
Army Members by Academy secret police unit. The finding is stable across different
Performance model specifications. Results for the control variables
show that the officers’ backgrounds also mattered. The
0 statistically significant, positive estimate for Infantry of-
ficer and the nonsignificant coefficient for Cavalry officer
show that infantry officers were more likely to serve in
601. This indicates that officers from lower classes with-
20 out extensive personal connections were more willing to
join the battalion. The positive, statistically significant
estimate for the variable Training under military govern-
ment indicates that officers who underwent training dur-
Graduation rank

40 ing previous military dictatorships were more likely to

serve in the battalion. Cadets’ age, quality of primary ed-
ucation, and training length at the military academy are
not significantly correlated with battalion membership.
60
To gauge the effect size of our key finding, we calcu-
late predicted probabilities for an average infantry officer
across all values of Graduation rank, holding other con-
trol variables at their means (King, Tomz, and Wittenberg
80
2000). Figure 4 shows that an infantry officer who grad-
uated at the top of her class had an average probability of
0.025 (0.02–0.04) of serving in the battalion. For the same
officer graduating at the bottom of her class, the probabil-
100
ity of joining Battalion 601 was 0.09 (0.05–0.14)—a 360%
–.008 –.004 0 .004 .008
increase. This result shows that officers with weak perfor-
DensityBattalion − DensityArmy
mance at an early career stage were more likely to serve in
the secret police. Next, we substantiate the core aspects of
Note: For each graduation rank, a black dot shows the
our theoretical mechanism with empirical evidence.
difference in densities between Battalion 601 members
and officers in any other unit. Positive x-values indi-
cate that a given graduation rank was overrepresented
among 601 members. Negative x-values indicate that a
given graduation rank was underrepresented among 601
Probing the Mechanism
members.
Our argument consists of several steps. In security or-
ganizations with an up-or-out system, underperforming
Empirical Results officials face early retirement. Such officials have difficul-
ties in reaching higher ranks, as poor past performances
Were underachieving officers more likely to serve in Bat- impair access to advanced training. In view of these grim
talion 601? Anecdotal evidence suggests that members did career prospects, officials are eager to join the secret po-
indeed lack ability. In 1979, a 601 member who served as lice, as it offers a promising alternative way to stay in
an informant to the U.S. embassy acknowledged that the the security apparatus and reach higher ranks. Regimes
unit did not consist of the “best men” (NSA 2002b, 7). can exploit the incentives of underachievers by accepting
Descriptive evidence corroborates this statement. Figure 3 them into the secret police. We offer evidence for each
shows that, relative to other army units, in Battalion 601 step.
low performers were overrepresented. This lends initial
support to our argument that underachieving officers
were more likely to serve in the secret police. Early Performance and the Risk
Turning to the results of the statistical analysis, of Retirement
Table 1 shows that Graduation rank is statistically sig-
nificant and positively correlated with membership in The first component of our mechanism postulates
Battalion 601. Officers who underperformed at the mil- that up-or-out promotion systems produce great career
10 ADAM SCHARPF AND CHRISTIAN GLÄßEL

TABLE 1 Logistic Regressions for Battalion 601 Membership, 1975–83

(1) (2) (3) (4)

Graduation rank 0.014∗∗∗ 0.014∗∗∗ 0.014∗∗∗ 0.013∗∗
(0.004) (0.004) (0.004) (0.004)
Cavalry officer −0.780∗∗ −0.551 −0.563
(0.298) (0.308) (0.310)
Infantry officer 0.517∗∗ 0.507∗∗
(0.171) (0.173)
Home literacy rate 1.291 0.992
(0.694) (0.777)
Cadet age 0.001 0.006
(0.006) (0.005)
Training under military gov. 0.007∗
(0.004)
Training length 0.008
(0.010)
Constant −4.058∗∗∗ −3.968∗∗∗ −5.513∗∗∗ −7.176∗∗∗
(0.189) (0.190) (1.535) (1.469)
AIC 1,295.09 1,289.00 1,272.73 1,263.34
Wald ␹ 2 14.18∗∗∗ 20.86∗∗∗ 28.86∗∗∗ 44.34∗∗∗
Pseudo R2 0.02 0.02 0.03 0.04
Observations 4,287 4,287 4,216 4,216
Note: Coefficients with robust standard errors are clustered on cohorts.
∗
p < .05, ∗∗ p < .01, ∗∗∗ p < .001.

pressures for underperforming officials. If correct, weak graduation rank should be negatively correlated with the
achievements at the academy should increase both the of- likelihood of such training. To test this, we code a binary
ficers’ risk of (early) retirement and likelihood of serving variable indicating whether officers attended advanced
in 601. We test this implication with a refined coding of training centers.24 Results offer strong support for this
our dependent variable and assess how graduation rank part of our mechanism. Predicted probabilities in Figure 6
influences the probability of serving in any unit other than show that officers with low early-career performance were
Battalion 601, retiring from the army, or working in the unlikely to undergo advanced training, and therefore
battalion.23 Results support our expectation. Predicted they were disadvantaged in the competition for lucrative
probabilities, shown in Figure 5, indicate that strong positions within and outside the coercive bureaucracy.
academy performance is positively correlated with longer
service in the army (a). In contrast, weak achievements
increase the likelihood of both retirement (b) and service Rewards for Secret Police Service
in 601 (c). Officers threatened by early retirement turned
to the battalion in the hope of salvaging their careers. Our mechanism also suggests that underachieving offi-
cials join the secret police in the hope of improving their
bleak career prospects. Qualitative evidence indicates that
Limited Access to Advanced Training many 601 agents gained access to advanced training at the
Higher War School over the course of their secret police
We have argued that officials with weak graduation service. Upon completion, officers returned to regular
results are stuck in the security apparatus because they army units or the general staff. Some later transitioned to
are less likely to receive advanced training—a prerequisite Argentina’s federal and provincial police, or they pursued
for future promotions and a successful career. If correct, careers in civil intelligence agencies. Few officers even left

23
The coding for each officer is mutually exclusive due to the cross-
24
sectional data structure. See SI.3 (7, 9) for coding description of We code graduations at the Higher War School, Higher Technical
the dependent variable and multinomial regression results. School, or Army Information School.
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 11

FIGURE 4 Graduation Rank and Likelihood of Serving in Battalion 601

Note: The plot shows predicted probabilities for 601 membership with 95% confi-
dence intervals for an average infantry officer, who was 18 years old when entering
the academy, grew up in a province with an 83% literacy rate, and underwent about 4
years of training, half of it under military governments (based on Model 4 in Table 1).

FIGURE 5 Early Achievements and Career Outcomes

Note: The plots show predicted probabilities for career outcomes (a: continued service in army; b: retirement
from army; c: service in 601) at different graduation ranks with 95% confidence intervals (based on Model 2 in
Table SI.3.1, 9).

the security apparatus for positions in the state adminis- their peers. Using coarsened exact matching, we test
tration (MJyDH 2015, 131–62). whether 601 membership increased service time and rank
Quantitatively, we analyze two reward types: service at retirement.25 Figure 7 shows that officers of Battalion
time and rank at retirement. If secret police agents are 25
Officers are matched on graduation rank and cohorts to compare
rewarded for their service, they should remain longer in individuals with similar track records. SI.3 (10–11) provides im-
the security apparatus and ascend to higher ranks than balance statistics and regression results for both outcome variables.
12 ADAM SCHARPF AND CHRISTIAN GLÄßEL

FIGURE 6 Early Career Performance and as gains in income and pension rights.26 The Argentine
Access to Advanced Training regime rewarded agents for secret police service.

Internal Placement of Secret Police Agents

Drawing on theoretical literature (Egorov and Sonin
2011; McMahon and Slantchev 2015; Zakharov 2016),
our argument suggests that leaders can exploit the ca-
reer pressures of underachieving officials. This part of
our mechanism is most difficult to test. We therefore rely
on an observable implication inspired by anecdotal ev-
idence from Nazi Germany: The leadership was careful
to assign well-educated members of the secret police to
administrative positions, whereas agents with poor track
records were tasked with hands-on repression (Browder
1996, 187–88). For Argentina, this suggests that the junta
Note: The plot shows predicted probabilities for ad- should have placed low-performing members in Cen-
vanced military training, with 95% confidence intervals tral Reunión of Battalion 601, which required the highest
(based on Model 2 in Table SI.3.2, 9). loyalty and was most directly involved in ground-level
policing. The implication rests on the assumption that
incoming officers hoped for a positive impact from secret
FIGURE 7 Battalion 601 Membership and police service on their career, but had little knowledge re-
Service Time in Army garding which department within Battalion 601 would be
best suited to help them achieve this goal. We believe that
.15 Army members
601 members
this is a plausible assumption given the secrecy surround-
ing the battalion’s internal structure and the information
by a former 601 member that leaders alone decided over
.1 placements within the unit. To further rule out the pos-
Density estimate

sibility that agents knew about the respective purposes of

the different departments, the test exclusively focuses on
the initial positions of incoming members right after the
.05
creation of Battalion 601 as the secret police.27
The refined dependent variable identifies the first
position that officers held within Battalion 601. It distin-
0
guishes between officers who served in Central Reunión
310 320 330 340 350 and officers who worked in another 601 unit. Graduation
Months rank should be positively correlated with the proba-
bility of serving in Central Reunión. Results from lo-
Note: Plot shows density estimates for the total service gistic regressions and Heckman selection models show
time in the army comparing matched officers with and that low-performing officers were indeed more likely to
without service in 601; gray areas denote 95% confidence serve in Central Reunión when entering Battalion 601.28
intervals (based on Model 4 in Table SI.3.4, 10). The regime seems to have strategically exploited the ca-
reer pressures of underachieving individuals by assigning
them to specific branches of the secret police.
601 indeed served longer in the army than officers who
never joined the unit. On average, 601 membership 26
SI.3 (11, 15–16) offers additional tests and visualizations, showing
granted agents 16 additional months in the military. This that service in Battalion 601 increased the chances of reaching
is a substantive but conservative effect estimate, since higher ranks, particularly for officers with weak graduation results.
27
several officers implicated in repression faced discharge We analyze internal placements in the first 5 years of the battal-
after the junta’s demise in 1983. The additional service ion’s existence.
28
time came with higher military ranks at retirement as well See SI.3 (12–13).
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 13

Internal Promotions pyramids” (Sassoon 2016, 38). Young party members of-
ten have to zealously carry out dirty work like surveillance
Finally, we have argued that career concerns of direct su- and intimidation to make it to the top (Svolik 2012, 168–
periors might reinforce the dominance of underachievers 69), again closely resembling the career pressures of junior
in the secret police. Superiors who have joined the secret Argentine army officers (Kung and Chen 2011, 28). Our
police due to their own career pressure should favor un- theory predicts that ambitious cadres with weak prospects
derachieving subordinates, as these pose less competition for advancement in the party hierarchy are likely to pro-
and are unlikely to supersede them. Since direct superiors vide the human resources for secret police units within
have little say in recruitment questions, but can influence such regimes.
internal promotion decisions, we scrutinize promotion This study might also speak to intelligence services
patterns within Battalion 601. If superiors valued under- in democracies. Contrasting with autocratic secret police
achievers, graduation rank should have had less effect on forces that uphold the iron rule of despots, democratic
internal promotions than in the rest of the army. We code intelligence agencies are supposed to protect the general
a binary variable indicating whether agents advanced in public. Even though such service is surely less arduous,
rank during their service in 601. The results from our sta- threats to national security sometimes also require agents
tistical analysis suggest that performance at the military in democracies to intervene in citizens’ rights through
academy is indeed an insignificant predictor of promo- nationwide spying or enhanced interrogation programs.
tions within the battalion (SI.3, 13, 16–17). The incentives Such contingencies might deter high performers, whereas
of direct superiors in the secret police seem to have mit- officials under strong career pressure are not in the posi-
igated the otherwise strong relationship between early tion to decline assignments. Career concerns could there-
career performance and promotion prospects. This un- fore also shape the composition of democratic intelli-
derscores that secret police service provides underachiev- gence units facing trade-offs between national security
ing officials with an exceptional opportunity to salvage and civil liberties.
their career. Apart from that, several factors might moderate the
general applicability of our mechanism. First, senior of-
ficials of competing intelligence organizations might put
External Validity more emphasis on competence to outdo rival appara-
tuses. Such competition is common in coup-proofed per-
Our findings can offer insights into numerous cases. sonalist dictatorships. Second, over time, exceptionally
The organization of Argentina’s security apparatus durable dictatorships might be better able to nurture
and the close association between the army and the both competent and loyal secret police forces through
secret police are most representative of military regimes lifelong indoctrination efforts.29 Third, in stable and less
(Geddes, Wright, and Frantz 2014). Similar historical violent regimes, secret police work may be less arduous
cases include the dictatorships of Brazil and Chile, whose and thus less well suited to demonstrate individual loy-
secret police forces also originated from the military alty. This may increase the willingness of high performers
(Greitens 2016; Policzer 2009; Stepan 1988). The logic and reduce the incentives for underachievers to join the
of our theoretical argument should, however, also travel secret police. Fourth, in societies with deep ethnic or so-
to autocracies that have a strategic incentive to limit cial cleavages, some leaders may hope to ensure the loyalty
institutional and personnel overlaps between the secret of secret police agents solely by recruiting from favored
police and the military. Hybrid and personalist regimes groups. By contrast, even in highly polarized societies,
often rely on security organizations like the regular police leaders who fear coups by competent individuals should
or paramilitary forces (Geddes, Wright, and Frantz have an incentive to employ underachievers from within
2018; Hassan 2017). Although these organizations may favored groups and exploit their career pressures to max-
not employ strict up-or-out systems, their hierarchical imize loyalty.
structure with a limited number of top posts generates
competition among junior officials. Resulting career
pressures are likely to produce similar incentives for
underachievers to join the secret police.
29
Our theoretical argument may also inform the com- The Argentine regime lasted for 8 years, which is marginally
position of secret police forces in party-based regimes longer than the most common regime spell and slightly shorter
than the median duration of all autocracies since World War I
where security and party apparatuses form “interlocking (SI.2, 6).
14 ADAM SCHARPF AND CHRISTIAN GLÄßEL

Conclusion with established career trajectories and merit-based pro-

motion systems produce the incentives for unconditional
Why do underqualified agents dominate the secret police? loyalty, which can be exploited by regimes. Even in such
We highlight officials’ motivation for serving in secret an institutionalized apparatus, individuals may be willing
police organizations to explain this puzzle. In bureaucra- to serve their leaders and ruthlessly target societal oppo-
cies, officials with limited past achievements are forced to sition groups. This might be bad news for democracies,
demonstrate their commitment in order to improve their where performance-based careers are the cornerstone of
chances of gaining future promotions and lucrative posi- the state apparatus. Our study explains how governments
tions. Under pressure to outperform their better-qualified can accomplish swift autocratic turns without major bu-
peers, such underachievers have large incentives to dili- reaucratic resistance. Officials facing career pressures are
gently work in the key coercive unit and show their loy- likely to serve as willing executioners, while their well-
alty to the regime. Underachieving but ambitious officials placed peers remain silent bystanders.
therefore fulfill the single most important requirement of Finally, this study also points to unintended conse-
secret police agents—a strong personal interest in com- quences of transitional justice measures. Our argument
pliance. Leaders who depend on a loyal secret police force implies that future punishment makes secret police ser-
can expect individuals motivated by self-preservation to vice even more costly. However, the looming risks of
loyally execute violence. discharge and incarceration are unlikely to fully drain
The quantitative analysis of Battalion 601, one of the the pool of willing executioners, as the least competent
most notorious intelligence units in the Western Hemi- agents with the lowest career prospects will still gamble
sphere, provides empirical evidence for our argument. on gains from loyal regime service. Imminent sanctions
Drawing on original individual-level data on more than may therefore entail increased suffering for civilians, as
4,000 officers, we show that those who underperformed poorly qualified agents lack the skills and incentives to
early on in their careers were indeed more likely to serve in target regime enemies selectively.
the battalion. Our results also demonstrate why this was
the case. Top performers at the academy were more likely
to attain advanced training and high-ranking regime po-
sitions, whereas most low performers were stuck at mid- References
dling ranks or even had to leave the military altogether.
Underperforming officials were under considerable ca- Albrecht, Holger, and Dorothy Ohl. 2016. “Exit, Resistance,
reer pressure and therefore had an incentive to join the Loyalty: Military Behavior during Unrest in Authoritarian
Regimes.” Perspectives on Politics 14(1): 38–52.
secret police.
Amnesty International. 1977. Torture in Greece: The First Tor-
One key implication of this study pertains to the se-
turers’ Trial 1975. London: Amnesty International.
curity apparatus within nepotist regimes, where loyalty
Arendt, Hannah. 2017. The Origins of Totalitarianism. New
results from favored groups’ access to goods and posts. York: Penguin.
Our argument suggests that individuals without influen- Atkins, George P., and Larry V. Thompson. 1972. “German
tial networks or career patrons face constraints in reaching Military Influence in Argentina, 1921–1940.” Journal of Latin
the most lucrative regime positions. The lack of personal American Studies 4(2): 257–74.
connections can therefore generate the same pressures Baker, George P., Michael C. Jensen, and Kevin J. Murphy. 1988.
as the shortage of merit in other regimes. As such, this “Compensation and Incentives: Practice vs. Theory.” Journal
of Finance 43(3): 593–616.
study may also explain puzzling cases where individuals
Balcells, Laia, and Christopher M. Sullivan. 2018. “New Find-
from excluded segments of society have served in key co-
ings from Conflict Archives: An Introduction and Method-
ercive units. Similar to merit-based systems, sidelined or ological Framework.” Journal of Peace Research 55(2): 137–
even excluded agents should perceive secret police ser- 46.
vice as a way to advance their careers within nepotist Baldwin, Roger N. 1934. “Political Police.” In Encyclopedia of
regimes. the Social Sciences, Volume XII, ed. Edwin R. A. Seligman and
A second key implication concerns the institutional Alvin Johnson. New York: MacMillan, 203–7.
roots of authoritarian stability. Conventional wisdom sees Barros, Robert. 2016. “On the Outside Looking In: Secrecy and
the Study of Authoritarian Regimes.” Social Science Quar-
patrimonial relations as key elements of a durable auto- terly 97(4): 953–73.
cratic architecture. By contrast, our study shows that an Becker, Howard S., and Anselm L. Strauss. 1956. “Careers, Per-
institutionalized, meritocratic bureaucracy does not con- sonality, and Adult Socialization.” American Journal of Soci-
tradict autocratic longevity. Hierarchical organizations ology 62(3): 253–63.
WHY UNDERACHIEVERS DOMINATE THE SECRET POLICE 15

Bellin, Eva. 2004. “The Robustness of Authoritarianism in the Kamiński. Warsaw: Institute of National Remembrance,
Middle East: Exceptionalism in Comparative Perspective.” 221–83.
Comparative Politics 36(2): 139–57. Egorov, Georgy, and Konstantin Sonin. 2011. “Dictators and
Biderman, Albert D., and Laure M. Sharp. 1968. “The Con- Their Viziers: Endogenizing the Loyalty-Competence Trade-
vergence of Military and Civilian Occupational Structures: off.” Journal of the European Economic Association 9(5): 903–
Evidence from Studies of Military Retired Employment.” 30.
American Journal of Sociology 73(4): 381–99. Figueroa, Abelardo Martı́n. 2008. Promociones Egresadas del
Blau, Peter M., and W. Richard Scott. 2003. Formal Organi- Colegio Militar de la Nación (1873–2007). Buenos Aires: So-
zations: A Comparative Approach. Stanford, CA: Stanford ciedad Militar Seguro de Vida Institucion Mutualista.
University Press. Frantz, Erica, and Andrea Kendall-Taylor. 2014. “A Dictator’s
Blažek, Petr, and Pavel Žáček. 2005. “Czechoslovakia.” In A Toolkit: Understanding How Co-optation Affects Repres-
Handbook of the Communist Security Apparatus in East Cen- sion in Autocracies.” Journal of Peace Research 51(3): 332–
tral Europe, 1944–1989, ed. Krzysztof Persak and Lukasz 46.
Kamiński. Warsaw: Institute of National Remembrance, 87– Friedrich, Carl J., and Zbigniew K. Brzezinski. 1965. Totalitar-
161. ian Dictatorship and Autocracy. Cambridge, MA: Harvard
Brehm, John, and Scott Gates. 1999. Working, Shirking, and University Press.
Sabotage: Bureaucratic Response to a Democratic Public. Ann Geddes, Barbara, Joseph Wright, and Erica Frantz. 2014. “Auto-
Arbor: University of Michigan Press. cratic Breakdown and Regime Transitions: A New Data Set.”
Brooks, Risa. 1998. Political-Military Relations and the Stability Perspectives on Politics 12(2): 313–31.
of Arab Regimes. Oxford: Oxford University Press. Geddes, Barbara, Joseph Wright, and Erica Frantz. 2018. How
Browder, George C. 1996. Hitler’s Enforcers: The Gestapo and the Dictatorships Work: Power, Personalization, and Collapse.
SS Security Service in the Nazi Revolution. Oxford: Oxford Cambridge: Cambridge University Press.
University Press. Gregory, Paul R. 2009. Terror by Quota: State Security from
Browning, Christopher R. 1998. Ordinary Men: Reserve Police Lenin to Stalin (An Archival Study). New Haven, CT: Yale
Battalion 101 and the Final Solution in Poland. New York: University Press.
Harper Perennial. Greitens, Sheena Chestnut. 2016. Dictators and Their Secret
Cole, Wade M. 2015. “Mind the Gap: State Capacity and the Police: Coercive Institutions and State Violence. Cambridge:
Implementation of Human Rights Treaties.” International Cambridge University Press.
Organization 69(2): 405–41. Grossman, Dave. 1996. On Killing: The Psychological Cost of
Conrad, Courtenay R., and Will H. Moore. 2010. “What Stops Learning to Kill in War and Society. Boston: Back Bay
the Torture?” American Journal of Political Science 54(2): Books.
459–76. Harkness, Kristen A. 2016. “The Ethnic Army and the State: Ex-
Davenport, Christian. 2007. “State Repression and the Tyran- plaining Coup Traps and the Difficulties of Democratization
nical Peace.” Journal of Peace Research 44(4): 485–504. in Africa.” Journal of Conflict Resolution 60(4): 587–616.
Deletant, Dennis. 2005. “Romania.” In A Handbook of the Com- Hassan, Mai. 2017. “The Strategic Shuffle: Ethnic Geography,
munist Security Apparatus in East Central Europe, 1944–1989, the Internal Security Apparatus, and Elections in Kenya.”
ed. Krzysztof Persak and Lukasz Kamiński. Warsaw: Institute American Journal of Political Science 61(2): 382–95.
of National Remembrance, 285–328. Heinz, Wolfgang S. 1999. “Determinants of Gross Human
DeMeritt, Jacqueline H. R. 2015. “Delegating Death: Military Rights Violations by State and State-Sponsored Actors in
Intervention and Government Killing.” Journal of Conflict Argentina, 1976–1983.” In Determinants of Gross Human
Resolution 59(3): 428–54. Rights Violations by State and State-Sponsored Actors in Brazil,
Dinges, John. 2004. The Condor Years: How Pinochet and His Uruguay, Chile, and Argentina, 1960–1990, ed. Wolfgang S.
Allies Brought Terrorism to Three Continents. New York: New Heinz and Hugo Frühling. Leiden, The Netherlands: Marti-
Press. nus Nijhoff, 593–845.
Dragu, Tiberiu, and Adam Przeworski. 2019. “Preventive Re- Huggins, Martha K., Mika Haritos-Fatouros, and Philip G. Zim-
pression: Two Types of Moral Hazard.” American Political bardo. 2002. Violence Workers: Police Tortures and Murderers
Science Review 113(1): 77–87. Reconstruct Brazilian Atrocities. Berkeley: University of Cal-
ifornia Press.
Dragu, Tiberiu, and Yonatan Lupu. 2018. “Collective Action and
Constraints on Repression at the Endgame.” Comparative Janowitz, Morris. 1988. The Professional Soldier: A Social and
Political Studies 51(8): 1042–73. Political Portrait. New York: Free Press.
Droz-Vincent, Philippe. 2007. “From Political to Economic Ac- King, Gary, Michael Tomz, and Jason Wittenberg. 2000. “Mak-
tors: The Changing Role of Middle Eastern Armies.” In De- ing the Most of Statistical Analyses: Improving Interpreta-
bating Arab Authoritarianism: Dynamics and Durability in tion and Presentation.” American Journal of Political Science
Nondemocratic Regimes, ed. Oliver Schlumberger. Stanford, 44(2): 341–55.
CA: Stanford University Press, 195–211. Kung, James Kai-sing, and Shuo Chen. 2011. “The Tragedy of
Dudek, Antoni, and Andrzej Paczkowski. 2005. “Poland.” In the Nomenklatura: Career Incentives and Political Radical-
A Handbook of the Communist Security Apparatus in East ism during China’s Great Leap Famine.” American Political
Central Europe, 1944–1989, ed. Krzysztof Persak and Lukasz Science Review 105(1): 27–45.
16 ADAM SCHARPF AND CHRISTIAN GLÄßEL

Lazear, Edward. 2004. “The Peter Principle: A Theory of De- Sassoon, Joseph. 2016. Anatomy of Authoritarianism in the Arab
cline.” Journal of Political Economy 112(S1): S141–63. Republics. Cambridge: Cambridge University Press.
Lewis, Paul H. 2002. Guerrillas and Generals. Westport, CT: Scharpf, Adam. 2018. “Ideology and State Terror: How Officer
Preager. Beliefs Shaped Repression during Argentina’s ‘Dirty War.’”
Lupu, Noam, and Susan C. Stokes. 2009. “The Social Bases of Journal of Peace Research 55(2): 206–21.
Political Parties in Argentina, 1912–2003.” Latin American Slater, Dan. 2003. “Iron Cage in an Iron Fist: Authoritarian
Research Review 44(1): 58–87. Institutions and the Personalization of Power in Malaysia.”
McMahon, R. Blake, and Branislav L. Slantchev. 2015. “The Comparative Politics 36(1): 81–101.
Guardianship Dilemma: Regime Security through and from Stepan, Alfred. 1988. Rethinking Military Politics: Brazil and the
the Armed Forces.” American Political Science Review 109(2): Southern Cone. Princeton, NJ: Princeton University Press.
297–313. Sudduth, Jun Koga. 2017. “Strategic Logic of Elite Purges in
Ministerio de Justicia y Derechos Humanos (MJyDH). 2015. Dictatorships.” Comparative Political Studies 50(13): 1768–
“El Batallón de Inteligencia 601.” http://www.saij.gob.ar/do 1801.
cs-f/ediciones/libros/Batallon inteligencia 601.pdf. Svolik, Milan W. 2012. The Politics of Authoritarian Rule. New
Moore, David W., and B. Thomas Trout. 1978. “Military Ad- York: Cambridge University Press.
vancement: The Visibility Theory of Promotion.” American Taylor, Brian D. 2011. State Building in Putin’s Russia: Polic-
Political Science Review 72(2): 452–68. ing and Coercion after Communism. New York: New York
National Security Archive. 2002a. “Document 13: Sub- University Press.
ject: Hypothesis—The GOA as Prisoner of Army In- Valentino, Benjamin A. 2004. Final Solutions: Mass Killing and
telligence, August 18, 1980.” https://nsarchive2.gwu.edu/ Genocide in the Twentieth Century. Ithaca, NY: Cornell Uni-
NSAEBB/NSAEBB73/800818dos.pdf. versity Press.
National Security Archive. 2002b. “Document 9: Subject: Nuts Zakharov, Alexei V. 2016. “The Loyalty-Competence Trade-off
and Bolts of the Government’s Repression of Terrorism- in Dictatorships and Outside Options for Subordinates.”
Subversion, August 7, 1979.” https://nsarchive2.gwu.edu/ Journal of Politics 78(2): 457–66.
NSAEBB/NSAEBB73/790807dos.pdf.
Zuckerman, Fredric S. 1996. The Tsarist Secret Police and Russian
Nordlinger, Eric A. 1977. Soldiers in Politics: Military Coups and Society, 1880–1917. New York: New York University Press.
Governments. Englewood Cliffs, NJ: Prentice-Hall.
Persak, Krzysztof, and Lukasz Kamiński. 2005. A Handbook of
the Communist Security Apparatus in East Central Europe,
1944–1989. Warsaw: Institute of National Remembrance. Supporting Information
Plate, Thomas, and Andrea Darvi. 1982. Secret Police: The Inside
Story of a Network of Terror. London: Robert Hale.
Additional supporting information may be found online
Policzer, Pablo. 2009. The Rise and Fall of Repression in Chile.
in the Supporting Information section at the end of the
Notre Dame, IN: University of Notre Dame Press.
article.
Potash, Robert A. 1996. The Army and Politics in Argentina,
1962–1973: From Frondizi’s Fall to the Peronist Restoration. SI.1: Summary statistics
Stanford, CA: Stanford University Press.
SI.2: Case information
Quinlivan, James T. 1999. “Coup-Proofing: Its Practice and SI.3: Probing the mechanism
Consequences in the Middle East.” International Security
24(2): 131–65. SI.4: Additional robustness checks