You are on page 1of 16

Research Articles

Journal of Peace Research


2015, Vol. 52(3) 269–284
A spatial analysis of the impact of West ª The Author(s) 2015
Reprints and permission:
German television on protest mobilization sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0022343314554245
during the East German revolution jpr.sagepub.com

Charles Crabtree
Department of Political Science, Pennsylvania State University
David Darmofal
Department of Political Science, University of South Carolina
Holger L Kern
Department of Political Science, Florida State University

Abstract
Formal models of revolutionary collective action suggest that ‘informational cascades’ play a crucial role in overcom-
ing collective action problems. These models highlight how information about the aggregate level of participation in
collective action conveys information about others’ political preferences, and how such informational cues allow
potential participants to update their beliefs about the value of participating in antiregime collective action. In
authoritarian regimes, foreign mass media are often the only credible source of information about antiregime pro-
tests. However, limited robust evidence exists on whether foreign media can indeed serve as a coordination device
for collective action. This article makes use of a detailed dataset on protest events during the 1989 East German rev-
olution and exploits the fact that West German television broadcasts could be received in most but not all parts of
East Germany. Across a wide range of Cox proportional hazards models and conditional on a rich set of observables,
it finds that the availability of West German television did not affect the probability of protest events occurring. The
evidence presented here does not support the widely accepted ‘fact’ that West German television served as a coordi-
nation device for antiregime protests during the East German revolution. More broadly, it also calls into question
strong claims about the effects of communication technology on revolutionary collective action.

Keywords
authoritarian regimes, collective action, democratization, East Germany, mass media, natural experiment

Introduction to democracy, in particular during the third wave of


democratization (Huntington, 1991: 100–106; Schmit-
Overthrowing an authoritarian regime is incredibly diffi-
ter, 1996: 34–35; Whitehead, 1996: 5–8). However, the
cult. It requires ordinary citizens to overcome collective
state of empirical research in this area can still be
action problems and to coordinate their antiregime beha-
described as inchoate. Only a handful of studies have
vior (Lichbach, 1995). Such coordination is complicated
empirically examined the question of whether foreign
both by the prevalence of preference falsification (Kuran,
mass media can indeed facilitate revolutionary collective
1989, 1991), which makes it difficult to accurately gauge
action. Our article contributes to this literature, and the
the strength of antiregime sentiments among the citi-
zenry, and the absence of independent domestic media
in authoritarian regimes. The theoretical literature thus Corresponding author:
stresses the importance of foreign media in transitions hkern@fsu.edu
270 journal of PEACE RESEARCH 52(3)

literature on media effects in authoritarian regimes more communist regimes in Eastern Europe. Western observ-
generally, by examining the role played by West German ers were impressed with East Germany’s economic per-
television (WGTV) during the East German revolution.1 formance and regarded it as a socialist success story
Because of its historical significance and the unparalleled (Kopstein, 1997: 1–13). The Socialist Unity Party of
wealth of primary sources that has become available after Germany (Sozialistische Einheitspartei Deutschlands,
German reunification, the East German revolution is a SED) ruled East Germany through an implicit social
prominent case in the literature on social movements contract that rewarded political acquiescence with stea-
and revolutionary collective action (e.g. Hirschman, dily improving living standards (Pollack, 2000: 35–40;
1993; Lohmann, 1994; Pfaff & Kim, 2003; Pfaff, Dale, 2005: 59–81). As the 1980s progressed, however,
2006). According to this literature, WGTV played a key the East German regime became increasingly incapable
role during the East German revolution by broadcasting of buying off its people. During the 1970s, East German
news about the escalating political crisis directly into East economic growth had been spurred by easy access to
German living rooms. The literature suggests that by international credit and economic aid from the Soviet
spreading knowledge of successful protests and the unex- Union. The 1970s oil shocks, rising interest rates, and
pected vulnerability of the East German dictatorship, reduced Soviet subsidies revealed the structural weak-
WGTV was able to alter East Germans’ perceptions of nesses of East Germany’s centrally planned economy.
political opportunity and to facilitate the activation and With its focus on heavy industry and the extensive use
diffusion of protest (Kuran, 1991: 37; Opp, Voss & of factor inputs, the East German economy was ill
Gern, 1993: 254–255; Opp & Gern, 1993: 675–676; equipped to participate in the technological revolution
Hirschman, 1993: 198; Jarausch, 1994: 44; Grix, that was propelling growth in capitalist economies (Sti-
2000: 32–33). glitz, 1994). By the late 1980s, stagnation had become
In this article, we evaluate this widely held claim that evident (Kopstein, 1997). Even though nominal wages
WGTV served as a coordination device for protest activ- continued to rise, inflation, frequent consumer goods
ities during the East German revolution. Our research shortages, and the lack of adequate housing left many
design is based on a natural experiment, as we make East Germans with the impression that their living stan-
use of the fact that WGTV broadcasts could be dards were stagnating, if not declining (Schneider,
received in most but not all parts of East Germany 1996). The gap between West and East Germany had
(Kern & Hainmueller, 2009: 380–382). Relying on never been as plainly visible as in the late 1980s, as the
a detailed micro-level dataset of more than 2,700 pro- availability of WGTV and visits to and from the West
test events that took place between September 1989 facilitated direct comparisons of living standards (Dale,
and March 1990, we use Cox proportional hazards 2005: 82–97).
models to model the risk of protest events at the Economic discontent was not the only reason for the
county-level conditional on the availability of WGTV communist regime’s lack of popular support. The appal-
and an unusually rich set of covariates. In the next ling state of the environment also undermined its legiti-
section, we begin by providing some historical back- macy. The East German economy was geared towards
ground on the East German revolution. We then maximizing output regardless of the ecological conse-
review the literature on mass media and revolutionary quences. East Germany’s large chemical industry and the
collective action. Following that, we introduce our use of brown coal for electrical power production, com-
research design and statistical model. We then present pounded by the lack of effective pollution abatement
and discuss our results. The last section concludes. technology, led to dramatic ecological damages. East
Germany’s emission rates of pollutants such as particu-
late matter, sulfur dioxide, and nitrogen oxides were
The 1989 East German revolution among the highest in Europe. Indeed, per capita sulfur
Until its collapse in 1989, the German Democratic dioxide and particulate matter emissions were more than
Republic (GDR) was seen as one of the most stable 15 times as high as in West Germany (Kuhrt, Buck &
Holzweissig, 1996).
1
As in other communist countries, political dissent was
WGTV refers to ARD and ZDF, the two major West German not tolerated in East Germany. Lacking the opportunity
public broadcasting stations. Commercial television had been
introduced in West Germany in the mid-1980s, but it did not broad-
to voice their concerns, many dissatisfied East Germans
cast to East Germany and could only be received in some areas near left their country, either legally after they had received
West Berlin. exit visas (after a waiting period that could last up to ten
Crabtree et al. 271

years) or illegally. Dissidents who remained could only German regime eventually relented and allowed all East
find support within the confines of the Protestant Germans occupying West German embassies to leave for
church. A number of clergymen supported small groups West Germany (Zelikow & Rice, 1995: 63–101).
of political activists concerned with issues of peace, sus- The regime’s demand that the sealed trains carrying
tainable development, the environment, and human them to West Germany had to cross East German terri-
rights by providing them with access to modest resources tory so that they could be formally stripped off their citi-
such as telephones, copying machines, and meeting zenship and expelled from East Germany backfired,
space. Some activists dismissed from their state jobs also however. Many East Germans learned about this com-
found employment with the church. In the late 1980s, a promise solution from WGTV. Along the train route,
tiny East German samizdat press emerged that attempted large-scale riots broke out between the police and desper-
to create a public sphere independent of the regime ate East Germans who attempted to board the trains.
(Neubert, 1998; Pollack, 2000: 197–200; Dale, 2005: The East German leadership, focused on the celebration
103–107). of the 40th anniversary of the GDR on 7 October, had
The majority of East Germans, however, found nei- seriously misread the public mood. While East German
ther the political program nor the lifestyle of these polit- media remained silent about the riots, WGTV broadcast
ical activists attractive. Most East Germans valued higher dramatic footage of the arrival of thousands of East Ger-
living standards and the freedom to travel to the West man emigrees in West Germany. This mass exodus and
more than the ‘socialism with a human face’ that dissi- the way it was mismanaged by the East German regime
dents attempted to bring about (Pollack, 2000: 205– significantly fueled public protests (Hirschman, 1993:
208; Dale, 2005: 98–119). Opposition groups remained 186–193; Pfaff & Kim, 2003: 415–416).
on the fringes of popular political activity, neither orga-
nizing nor playing a key role in the protests that erupted
in September and October 1989 (Opp, Voss & Gern, Mass media and revolutionary collective action
1993; Opp & Gern, 1993). The first ‘Monday demon- According to relative deprivation theory (Gurr, 1970),
strations’ in Leipzig were predominantly led by East Ger- people rebel when they perceive a discrepancy between
mans who had applied for exit visas and were hoping that their material expectations and the regime’s ability to
public protest would make them enough of a nuisance to ensure the living standards to which they feel entitled.
the regime to be allowed to leave the country (Pollack, Such a gap undoubtedly existed for most East Germans,
2000: 209–234). who coveted Western consumer goods and wished for
In contrast to Hungary and Poland, no significant the freedom to travel to the exotic places shown on
impetus for reform arose from within the ruling party. WGTV.2 Yet relative deprivation can hardly account for
Following Mikhail Gorbachev’s accession to the leader- the outbreak of the East German revolution (or any of
ship of the Soviet Communist Party in 1985, relations the East European revolutions). After all, living condi-
between the GDR and the Soviet Union swiftly soured, tions in West Germany had been significantly better
with East Germany distancing itself from Gorbachev’s than in East Germany throughout the existence of the
reform program. Most East Germans in contrast wel- two German states, yet mass protests did not break out
comed Gorbachev’s reforms and hoped for similar until summer 1989 (with the exception of the June
changes in East Germany (Süss, 1996). 1953 uprising, which was quickly suppressed by Soviet
The East German regime’s situation became increas- troops and tanks).3
ingly difficult when Hungary and Poland both effectively The theory of political opportunity structure (Tarrow,
left the Soviet bloc at the beginning of 1989. In January, 1998) is based on the notion that people do not rebel
non-communist parties were legalized in Hungary. In when they are most discontent, but when a formerly
February, round-table talks began in Poland and in June, closed system of political opportunities for change opens
the Polish Communist Party relinquished its hold on up. This was undoubtedly the case in East Germany.
power. In May, Hungary started to dismantle its border
fortifications with Austria. Throughout the summer,
2
thousands of East Germans vacationing in Hungary WGTV was extremely popular in East Germany. Almost all East
made use of what seemed like a once-in-a-lifetime oppor- Germans with access to WGTV watched it regularly not only
because of its political programming but also for entertainment
tunity to escape to the West. Others occupied West Ger- (Kern & Hainmueller, 2009).
man embassies in Prague, Warsaw, and Budapest to 3
Tilly, Tilly & Tilly (1975) is a classic criticism of relative
force their emigration to West Germany. The East deprivation theory.
272 journal of PEACE RESEARCH 52(3)

Gorbachev’s domestic reforms sowed discontent among stemming from loss of personal autonomy and sacrifice
East Germans when the East German regime refused to of personal integrity. The larger the divergence between
implement similar reforms in East Germany. More an individual’s public and private preferences, the larger
importantly, the reforms also changed East Germans’ the psychological costs of ‘living a lie’ (Havel, 1985: 20).
perceptions of the Soviet commitment to guarantee the Now assume that someone is opposed to the regime.
survival of the East German regime against domestic Holding his public and private preferences constant, as
challenges. The failed 1953 uprising had taught East the opposition movement grows, there comes a point
Germans that regime change would not be possible at which the psychological costs of publicly supporting
against the wishes of the Soviet Union. Yet Gorbachev’s a secretly despised status quo outweigh the costs of join-
reforms and his remark during a WGTV interview in ing the opposition, and so the individual joins the oppo-
East Berlin on 6 October that ‘Those who are late will sition movement. This point is called his revolutionary
be punished by life itself’ were interpreted by East Ger- threshold: the opposition size required to change his
mans as at least tacit approval of political reform in East behavior from feigned support for the status quo to open
Germany.4 Yet while political opportunity theory cor- support for the opposition. A ‘revolutionary bandwagon’
rectly highlights the (external) conditions that were nec- ensues when someone joins the opposition, and that
essary for the overthrow of the East German regime, it leads others to join the opposition as well. Because of
does not explain the process of micro-level mobilization such revolutionary bandwagons, authoritarian regimes
of antiregime protest. that once appeared unshakeable can see their support
The literature on bandwagon processes and behavioral crumble in no time. Whether such a cascade occurs
cascades explicitly addresses the micro-level problem of depends on the unobservable distribution of revolution-
protest participation by acknowledging that the decision ary thresholds, which is why revolutions are easy to
of one individual to join collective action depends upon explain ex post but impossible to predict ex ante (Kuran,
the decisions of others (Granovetter, 1978; Schelling, 1989, 1991).
1978). Kuran (1989, 1991) develops a model for partic- Lohmann (1993, 1994) offers a slightly different
ipation in revolutionary collective action. He posits that model in which turnout at protest events among political
every individual has two preferences over regime type: a moderates provides information about others’ political
private preference and a public preference. The private preferences. Both Kuran (1991) and Lohmann (1994)
preference corresponds to one’s true preference, whereas use their respective models to shed light on protest mobi-
the public preference is the preference one chooses to lization during the East German revolution, stressing
publicly reveal. Preference falsification denotes the situa- how protest participation depended on credible informa-
tion in which private and public preferences diverge. tion about others’ political preferences. Neither their for-
Whether an individual joins the opposition depends mal models nor empirical analyses, however, contain
on the trade-off between two payoffs, one external and detailed discussions of the channels through which infor-
the other one internal. The external payoff captures the mation about the size of the opposition movement
incentives (e.g. future personal rewards if the opposition (Kuran, 1989, 1991) or aggregate turnout of moderates
is successful) and disincentives (e.g. risking one’s life) for in protest events (Lohmann, 1993, 1994) is transmitted.
publicly declaring one’s support for the opposition The widely cited article by Lohmann (1994), for exam-
movement. The larger the existing opposition move- ple, succeeds admirably in describing how the Monday
ment, the smaller are the disincentives for participation, demonstrations in Leipzig helped to change public per-
since the likelihood of punishment for siding with the ceptions of the vulnerability of the East German regime.
opposition is decreasing in the size of the opposition. It also notes that mass demonstrations in Leipzig ‘trig-
If many people participate in antiregime protests, the gered a wave of political protest throughout the GDR’
regime will be unable to punish most of them. If only (Lohmann, 1994: 42). Her account, however, does not
a few protest, in contrast, the regime will be able to explain how East Germans learned about protest events
severely punish all of them. The internal payoff is rooted that the vast majority of them could not directly observe.
in the psychological costs of preference falsification This omission is particularly critical in the context of
regimes with heavily censored, state-owned mass media
such as East Germany.
4
In fact, Gorbachev’s exact words were less pithy. The sentence
A number of authors have stressed WGTV’s role dur-
quoted above seems to originate with Gorbachev’s press secretary, ing the East German revolution. Kuran (1991: 37) for
Gennadi Gerassimow (Plog, 2004). example writes with respect to the demonstrations that
Crabtree et al. 273

took place in East Berlin during the celebrations of the lowering the transaction costs of cooperating with the
40th anniversary of the GDR that WGTV ‘immediately government. Pierskalla & Hollenbach (2013) in con-
played these events back to the rest of East Germany. trast present cross-national evidence for 27 African
The scenes alerted disgruntled citizens in every corner countries that suggests that the availability of cell
of the country to the pervasiveness of discontent, while phone technology facilitates violent collective action.
the government’s weak response revealed its vulnerabil- The recent literature on the Arab Spring and the
ity.’ Opp & Gern (1993: 675–676) note that ‘the [Leip- Color Revolutions likewise discusses the impact of social
zig] Monday prayers, together with the demonstrations, communication technologies, although studies often fall
contributed to the emergence of protest in other East short of offering credible causal identification strategies.
German cities. People were informed, primarily on Some scholars claim that social media either cause
WGTV, about the events in Leipzig, and the expectation (Shirky, 2011) or play a role in causing revolutions and
formed that citizens in each city would meet sponta- transitions to democracy (Howard, 2010; Hussain &
neously on the city square for Monday demonstrations.’ Howard, 2013). Others avoid explicit causal claims but
Jarausch (1994: 44) similarly notes that ‘Western televi- argue that social media are powerful tools that the oppo-
sion coverage enabled acts of symbolic defiance to reach a sition can use to disseminate information, circumvent
wider audience, spreading unrest.’ With regard to the censorship, and organize and coordinate collective action
exiting crisis, Hirschman (1993: 198) writes that ‘pic- (Chebib & Sohail, 2011; Eltantawy & Wiest, 2011;
tures of the exodus soon flooded the TV screens, with Khamis & Vaughn, 2011; Goldstein, 2007; Lynch,
the result of not just causing established critics [ . . . ] 2011, 2012: 10–11).
to sharpen their criticism but also of making activists out Most directly related to our research are articles by
of long-passive average citizens’. Opp, Voss & Gern Kern (2011) and Grdĕsić (2014) that both look at
(1993: 254–255, 260) and Grix (2000: 32–33) similarly WGTV’s role as a coordination device for protest activ-
stress WGTV’s role in disseminating political informa- ities during the East German revolution. Kern (2011)
tion not available from the state-controlled East German compares counties without WGTV to a matched
media. comparison group of counties with WGTV, but does
In the social movements literature, empirical research not find any evidence that WGTV affected the speed
on mass media as a facilitator of collective action has or depth of protest diffusion. Grdĕsić (2014) argues,
largely been limited to the impact of domestic mass based on vector autoregressions and Granger causality
media in democratic societies (e.g. Andrews & Biggs, tests, that WGTV news reports about East German
2006; Roscigno & Danaher, 2001; Myers, 2000). Work protests ‘Granger caused’ protests in East Germany
on the impact of domestic mass media in authoritarian the following day.
settings in contrast is in its infancy. Yanagizawa-Drott
(2014) finds that radio broadcasts served as a coordina-
tion device for collective violence during the Rwandan Research design
genocide. Enikolopov, Petrova & Zhuravskaya (2011), Our research design takes advantage of the fact that
in an article on media effects in Russia, demonstrate the WGTV broadcasts could be received in most but not
strong impact of the only national television channel all parts of East Germany. Specifically the northeastern
independent of the government on voting for opposition part of East Germany and the Dresden district in the
parties during the 1999 parliamentary elections. Adena southeast were by and large cut off from WGTV broad-
et al. (2013) document the role that radio propaganda casts due to East Germany’s topography and their dis-
played in undermining the democratic institutions of the tance from West German broadcast transmission
Weimar Republic and the rise of nazism in pre-WWII towers. Historical maps of WGTV’s over-the-air signal
Germany. strength used in previous research (Kern & Hainmuel-
Recent research has also examined the role of hor- ler, 2009: Figures 1 and 3) allow us to distinguish
izontal, ‘social’ communication technologies such as between broad areas of East Germany with and without
cell phones, Facebook, and Twitter as opposed to WGTV. However, these maps have a serious drawback
older vertical communication technologies (i.e. news- in that they are not detailed enough to allow us to reli-
papers, radio, and television) (Warren, 2014, 2015). ably determine the availability of WGTV at the county
Shapiro & Weidmann (2015), for example, show that level. We therefore use the Longley-Rice electromag-
the availability of cell phone networks reduced collec- netic signal propagation model in conjunction with ter-
tive violence in Iraq between 2004 and 2009 by rain data and data on the location and technical
274 journal of PEACE RESEARCH 52(3)

Figure 1. Signal strength of WGTV in East Germany as predicted by Longley-Rice model

characteristics of WGTV broadcast transmitters to average signal strength in the center of the city of Dres-
model WGTV’s signal strength across East Germany den. For a county to have access to WGTV, we require
(see Figure 1).5 We then discretize this continuous that at least 50% of the county receives a signal equal to
measure of WGTV signal strength to distinguish or above that threshold (see Figure 2). While the map in
between East German counties with and without Figure 2 closely reproduces the overall pattern of the
WGTV. We set the threshold value as the modeled historical maps shown in Kern & Hainmueller
(2009), our approach classifies a number of counties
5
differently than Kern (2011). In the online appen-
The same approach has been used in research on media effects in
dix, we show that our results are unaffected when
economics to model the availability of radio and television signals
(Olken, 2009; Enikopolov, Petrova & Zhuravskaya, 2011; we use Kern’s (2011) classification instead. More-
DellaVigna et al., 2014). See the online appendix for a detailed over, there we also show that our results are entirely
discussion. unaffected by the exact signal strength threshold
Crabtree et al. 275

Figure 2. East German counties with and without WGTV based on Longley-Rice model and Dresden cutoff level

used to separate counties with WGTV from counties Table I shows balance statistics, comparing coun-
without WGTV. ties with WGTV to counties without WGTV. The
Our identifying assumption is that access to WGTV first set of covariates captures broad socio-economic
was idiosyncratic conditional on covariates. In other differences between counties in terms of log(popula-
words, we assume that whether a county had WGTV tion size), population density, sector shares, the
or not was uncorrelated with unobserved county charac- shares of skilled and unskilled labor, and the propor-
teristics related to the risk of protest, conditional on tion of the population that holds a college degree, is
observed characteristics. Kern (2011) already noted that female, and of working age. A district capital dummy
counties with and without WGTV differed in terms of denotes the capitals of the 15 East German districts
a number of covariates, several of which are plausibly (Bezirke), which constituted the urban centers of East
related to the risk of protest. Germany and might therefore have been more prone
276 journal of PEACE RESEARCH 52(3)

Figure 3. Number of protests per county, East Germany, 4 September 1989–18 March 1990

to see antiregime protests. Other covariates capture regard as a rough proxy for the overall desirability
relative deprivation in dimensions that were particu- of living in a county.6
larly salient in East Germany (Kuhrt, Buck & Holz- While the most important aspect of political opportu-
weissig, 1996): housing quality (housing space per nity structure, the change in Soviet policy with regards to
capita and the proportion of apartments with bath- its Eastern European satellites, was constant across East
rooms, interior toilets, and modern heating), the
provision of public goods (residents per medical doc- 6
While residential mobility was very low in East Germany, we still
tor and residents per dentist), and air pollution
see meaningful over-time changes in population size across counties,
(nitrogen oxides, sulfur dioxide, and particulate mat- with 64 counties experiencing a change of more than +5% and 11
ter emissions). We also measure the percentage pop- counties experiencing a change of more than +10% over the 1980–
ulation change over the preceding decade, which we 88 period.
Crabtree et al. 277

Table I. Balance
Covariate WGTV Non-WGTV Standardized Variance t-test KS-test
mean mean difference ratio p-value p-value
Log(population size) 11.008 10.989 0.030 0.704 0.896 0.670
Population density 350.236 474.475 –0.149 1.128 0.478 0.517
% industry 0.357 0.314 0.302 1.146 0.154 0.127
% agriculture 0.174 0.191 –0.112 0.557 0.636 0.652
% crafts and construction 0.094 0.097 –0.118 1.044 0.578 0.567
% services and transportation 0.357 0.379 –0.274 0.770 0.228 0.278
% skilled 0.625 0.613 0.412 0.824 0.071 0.307
% unskilled 0.130 0.135 –0.248 1.346 0.224 0.273
% college 0.201 0.206 –0.167 0.717 0.465 0.599
% female 0.519 0.515 0.523 0.636 0.031 0.003
% working age 0.642 0.641 0.054 0.510 0.812 0.069
Housing space (m2) 27.342 26.152 0.568 0.415 0.028 0.000
% bathroom 0.819 0.795 0.323 0.574 0.178 0.029
% interior toilet 0.739 0.730 0.057 0.831 0.799 0.824
% modern heating 0.427 0.422 0.031 0.487 0.898 0.075
Residents per medical doctor 596.342 616.240 –0.100 1.053 0.637 0.082
Residents per dentist 1,549.489 1,573.600 –0.076 0.459 0.754 0.540
Nitrogen oxides (tons/km2) 6.235 7.630 –0.089 0.925 0.683 0.569
Sulfur dioxide (tons/km2) 81.554 86.248 –0.024 1.327 0.905 0.327
Respirable dust (tons/km2) 33.095 34.122 –0.016 2.227 0.929 0.542
District capital 0.068 0.080 –0.046 0.828 0.834 0.835
% change in population size 1980–88 –0.017 –0.013 –0.073 0.500 0.762 0.680
June 1953 protests 0.189 0.299 –0.304 0.487 0.215 0.753
Exit visas (per 1,000 residents) 3.269 3.274 –0.002 0.628 0.993 0.076
The table shows covariate balance for 217 East German counties with and without WGTV. The first two columns show means, the third
column shows differences in means standardized by the square root of the average variance, the fourth column shows variances in the WGTV
group divided by variances in the non-WGTV group, the fifth column shows p-values from two-sample t-tests, and the last column shows
p-values from bootstrapped Kolmogorov-Smirnoff tests of equality of distributions.

Germany, we also include two covariates that capture The first two columns of Table I show covariate
county-level variation in social capital and social embedd- means for counties with and without WGTV, the third
edness. Informed by work on the long-term persistence column shows differences between covariate means stan-
of patterns of cooperation and social capital (Putnam, dardized by the square root of the average variance, the
1993; Guiso, Sapienza & Zingales, 2008), we control for fourth column shows p-values from two-sample t-tests
historical antiregime collective action during the June and the last column shows p-values from bootstrapped
1953 East German uprising. We regard this covariate Kolmogorov-Smirnoff tests of equality of distributions.
as an (admittedly imperfect) proxy for social capital in For several of these covariates we see larger differences
1989, although it might also capture time-invariant grie- between the two groups of counties than one would
vances not captured by our socio-economic and relative expect by chance alone. Clearly, our natural experiment,
deprivation measures. The June 1953 protest data are while providing us with variation in access to WGTV,
taken from Kowalczuk (2003) and represent the fraction only imperfectly approximates a randomized experi-
of municipalities in a county that experienced protests, ment. Our analyses will adjust for these systematic
strikes, or attacks on representatives of the East German differences.
regime during the uprising. Finally, we also control for Protest event data are taken from Schwabe (1999),
emigration levels just before the East German revolution which provides a detailed country-wide compilation of
(January–August 1989) using data from Grundmann 2,734 county-level protest events between 4 September
(1997) to address the argument advanced by Pfaff & Kim 1989 and 18 March 1990, the date of the first free elec-
(2003) that there was a curvilinear relationship between tion in East Germany. This compilation is based on
exit (emigration) and voice (protest) (see also Hirschman, records of the East German Ministry of the Interior,
1993). which assembled daily crisis reports submitted by local
278 journal of PEACE RESEARCH 52(3)

police forces, the records of the Ministry of State Secu- the best of our knowledge, ours is the first analysis
rity, and numerous published secondary sources. Our that simultaneously accounts for spatial dependence,
outcome data thus avoid common problems with the event dependence, and heterogeneity within a
coding of protest events from news reports (Mueller, repeated events framework.
1997; Baum & Zhukov, 2015); we are confident that
systematic measurement error is negligible here. Results
Figure 3 shows a heat map of protest counts. On aver-
age, counties with WGTV experienced about 12:9 pro- Table II presents estimates from five Cox model specifi-
test events, while counties without WGTV experienced cations. The model in column 1 contains a dummy vari-
about 10:6 protest events. The raw data thus seem to able for access to WGTV but no covariates or frailty
provide some face validity for the claim that WGTV terms. Column 2 adds the covariates listed in Table I,
increased the likelihood of protest events. Of course, and column 3 also adds frailty terms to account for
given the systematic differences between counties with unobserved heterogeneity across counties. Coefficient
and without WGTV documented in Table I, this differ- estimates for covariates and frailty terms are not shown.
ence hardly represents a credible estimate of the causal Exponentiating the coefficient estimate for WGTV gives
effect of WGTV. us the estimate of WGTV’s multiplicative effect on the
Given the spatial and temporal dimensions of our probability of a protest event occurring conditional on
data, a simple non-parametric comparison of counties covariates and frailties.8 Standard errors are shown in
with and without WGTV is not feasible. Any robust brackets. Figure 4 graphically presents the estimates of
analysis of the impact of WGTV on the occurrence of WGTV’s multiplicative effect based on the models in
protest during the East German revolution will need to Table II with 95% confidence intervals. In none of the
account for both time- and space-dependency. More- three models is the impact of WGTV statistically signif-
over, it will also have to account for systematic differ- icant at the 0:05 level, and only in the model without
ences between counties with and without WGTV. We covariates or frailties is it significant at the 0:10 level.
therefore rely on a spatial survival model explaining the This null result could be due to our focus on the East
risk and timing of protest events as a function of access German revolution in its entirety. If WGTV did have an
to WGTV, conditional on covariates. A critical feature effect on protest activities, we would expect its impact to
of the East German revolution was that East German decrease over time as East German media became more
counties were at risk of experiencing multiple protests independent from the East German regime and better
over the course of the revolution, and, in fact, typically able to provide unbiased coverage of the mounting
did. As a consequence, a repeated events modeling regime crisis (Kern, 2011; Grdĕsić, 2014). Such a pat-
approach that accounts for the clustered nature of the tern of decreasing WGTV effects would provide strong
data is needed. Following the advice given by Box- evidence for the impact of WGTV, perhaps even more
Steffensmeier & De Boef (2006), Box-Steffensmeier, convincing than if we had found a positive effect of
De Boef & Joyce (2007), and Box-Steffensmeier, Linn WGTV throughout the entire East German revolution.
& Smidt (2014), we employ a conditional frailty gap Models 4 and 5 in Table II look at over-time hetero-
time Cox model that allows us to simultaneously address geneity in the effect of WGTV. The model in column 4
event dependence and heterogeneity.7 includes an interaction term between WGTV and a
Moreover, in some models we will also investigate dummy variable for the time period before the fall of the
the extent to which our results are robust to spatial Berlin Wall. The model in column 5 allows the effect of
dependency (Darmofal, 2009). We do so through the WGTV to vary even more finely, by including interac-
inclusion of a conditional spatial lag predictor vari- tion terms between WGTV and dummy variables for
able. Specifically, we include a temporally lagged spa- three time periods: 4 September to 18 October
tial lag variable that measures the proportion of (Through 18 October), 19 October to 31 October (Late
county i’s neighbors that experienced a protest event October), and 1–9 November (Early November). (The
during the preceding three days, during the preceding omitted category is the period after the fall of the wall.)
seven days, or during the preceding two weeks. To These time periods roughly correspond to periods of
increased media freedom in East Germany (Kern,

7
Please see the online appendix for a more detailed justification for 8
For example, in column 3, exp(0.130) ≈ 1.139, for an approximate
our choice of estimator. 13.9% increase in the probability of a protest event occurring.
Crabtree et al. 279

Table II. Effect of WGTV on probability of protest event from Cox models
(1) (2) (3) (4) (5)
WGTV 0:113y ½0:067 0:115 ½0:075 0:130 ½0:173 0:222 ½0:102 0:224 ½0:100
Wall 1:728 ½0:152
WGTV  Wall 0:099 ½0:151
Through 18 October 1:719 ½0:283
Late October 1:815 ½0:218
Early November 1:688 ½0:197
WGTV  Through 18 October 0:131 ½0:305
WGTV  Late October 0:058 ½0:222
WGTV  Early November 0:128 ½0:202
Covariates no yes yes yes yes
Frailties no no yes yes yes
Number of events 2,598 2,598 2,598 2,598 2,598
Number of observations 40,365 40,365 40,365 40,365 40,365
Log/I-likelihood 9,077 8,843 8,741 8,583 8,586
The table shows coefficient estimates from five Cox models, with standard errors in brackets. Wall corresponds to the period before the fall of
the Berlin Wall. yp < 0.1, *p < 0.05, **p < 0.01.

Robustness
2.0

In this section of the article we subject our results to a


series of robustness checks. We start by investigating the
sensitivity of our results to spatial diffusion processes. In
1.5

areas of East Germany that had access to WGTV, both


WGTV and social networks could have served as coordi-
WGTV effect

nation devices for collective action by providing East


1.0

Germans with political information not available from


the state-controlled East German media. Survey data col-
lected by Opp and his colleagues (Opp & Gern, 1993;
Before Fall of Berlin Wall
After Fall of Berlin Wall

After Fall of Berlin Wall

Opp, Voss & Gern, 1993) indeed show that social net-
Through 18 October
0.5

Early November
Late October

works played a role in mobilizing the Leipzig protests. It


thus stands to reason that social networks could have
been an even more important source of political informa-
0.0

(1) (2) (3) (4) (5)


tion in parts of East Germany that did not have access to
WGTV. If social networks served as a substitute for
WGTV when WGTV was not available, the estimated
Figure 4. WGTV impact estimates from Models 1–5 in Table
II and 95% confidence intervals effect of WGTV would be biased downwards. We at
least partially control for this possibility by including
temporally lagged spatial lag variables that measure the
2011: 1185). Note that neither model shows heteroge- extent to which protest events triggered subsequent pro-
neity in the effect of WGTV.9 In both models, only the tests in neighboring counties. This strategy does not
impact estimates for the period after the fall of the wall directly capture the existence of social networks, a task
are statistically significant, which is the opposite of what for which we would need individual-level network data.
we would expect. Overall, the results presented so far do However, it does allow us to control for one important
not support the hypothesis that WGTV facilitated anti- effect of social networks, the spatial diffusion of collec-
regime protest in East Germany, in particular during the tive action (Gould, 1991).
early stages of the revolution. Table III shows impact estimates from five Cox mod-
els. Model 1 adds a temporally lagged spatial lag variable
9
The interaction term in Model 4 is not statistically significant. For to the baseline model in column 3 of Table II. This spa-
the model in column 5, we use a LR test to test for joint significance tial lag variable measures the proportion of a county’s
of the three interaction terms: w23 ¼ 0:539; p ¼ 0:910. neighboring counties that experienced a protest event
280 journal of PEACE RESEARCH 52(3)

Table III. Effect of WGTV on probability of protest event from Cox models: Accounting for spatial dependence and alternative
sources of political information
(1) (2) (3) (4) (5)
WGTV 0.105 [0.170] 0.091 [0.171] 0.103 [0.174] 0.097 [0.168] 0.103 [0.182]
Spatial lag three days yes no no yes no
Spatial lag seven days no yes no yes no
Spatial lag 14 days no no yes yes no
Poland border dummy no no no no yes
Trains dummy no no no no yes
Covariates yes yes yes yes yes
Frailties yes yes yes yes yes
Number of events 2,598 2,598 2,598 2,598 2,598
Number of observations 40,365 40,365 40,365 40,365 40,365
I-likelihood –8,693 –8,683 –8,724 –8,669 –8,740
The table shows coefficient estimates from five Cox models, with standard errors in brackets. yp < 0.1, *p < 0.05, **p < 0.01.

Table IV. Effect of WGTV on probability of protest event from Cox models: Subsets of counties or events
(1) (2) (3) (4) (5) (6)
WGTV 0.103 [0.175] 0.062 [0.254] 0.195 [0.237] 0.065 [0.186] 0.204y [0.122] –0.582* [0.347]
Omit East Berlin yes no no no no no
Omit southeast no yes no no no no
Omit northeast no no yes no no no
Omit neighbors no no no yes yes no
Best subset no no no no yes no
First events only no no no no no yes
Covariates yes yes yes yes yes yes
Frailties yes yes yes yes no yes
Number of events 2,521 2,458 2,476 2,340 585 207
Number of observations 40,170 38,220 37,830 36,465 9,165 11,700
I-likelihood –8,599 –8,163 –8,179 –7,592 –1,070 –842
The table shows coefficient estimates from six Cox models, with standard errors in brackets. yp < 0.1, *p < 0.05, **p < 0.01.

on the preceding three days. Model 2 adds such a spatial have increased their awareness of the weakness of com-
lag variable for the preceding seven days, and Model 3 munist regimes more generally. We thus include dummy
does the same for the preceding 14 days. Model 4 simul- variables for counties through which the trains passed as
taneously includes all three spatial lag variables. Across all well as counties bordering Poland. Neither of these pre-
four models, the estimated impact of WGTV remains dictors has a substantively or statistically significant
small and statistically insignificant. impact on protest events (results not shown); more
Model 5 controls for two alternative sources of polit- importantly, the effect of WGTV remains small and sta-
ical information.10 First, the riots that took place when tistically insignificant.
the trains carrying embassy fugitives passed through East Table IV shows that our results do not change when
Germany could have had a particularly strong impact on we focus on specific subsamples of counties or when we
East Germans’ awareness of the weakness of the commu- estimate the effect of WGTV on time to first event.
nist regime. Second, in Poland the communist party was Model 1 in Table IV omits East Berlin from the sample.
ousted in early 1989. East Germans living close to the This does not affect the WGTV estimate or its statistical
Polish border might have been better informed about significance. Model 2 drops the non-WGTV counties in
these events than the average East German, which might the southeast from the sample and identifies the effect
of WGTV only using the non-WGTV counties in the
northeast. Model 3 does the reverse, dropping the non-
10
We thank an anonymous reviewer for suggesting these tests. WGTV counties in the northeast from the sample but
Crabtree et al. 281

keeping the non-WGTV counties in the southeast. Even for the hypothesis that WGTV facilitated antiregime
though the counties in the agrarian northeast are quite dif- protests, in particular during the early stages of the East
ferent from the counties in the industrialized southeast, German revolution. This finding is robust to variation in
both control groups generate small and statistically insig- the time periods we looked at, spatial diffusion, the mea-
nificant WGTV effect estimates. Following Kern (2011), surement of WGTV availability, and the specific samples
Model 4 omits all WGTV counties that border at least of counties or events used in the estimations.
one non-WGTV county (22 counties in total). The effect
estimate remains small and statistically insignificant.
Model 5 uses a reduced sample chosen to maximize Discussion
the comparability of WGTV and non-WGTV counties Our finding that WGTV had no discernible impact on
given the imbalances we saw earlier in Table I. As in protests during the East German revolution raises the
Model 4, we first dropped all WGTV counties bordering question of how protest was mobilized and spread so
non-WGTV counties. We then estimated propensity swiftly across East Germany. One possibility that we
scores using a probit model with all covariates entered already alluded to earlier is social networks. Research has
linearly and dropped all counties for which there was shown that friendship and family ties (McAdam, 1986;
no overlap in propensity score distributions (see Figure Gould, 1991) are important for protest mobilization.
A4 in the online appendix). This left us with 25 non- For the East German revolution, Opp, Voss & Gern
WGTV counties  and 34 WGTV counties. We then (1993) and Opp & Gern (1993) have illuminated the
34 role social networks played in both increasing awareness
evaluated all  52:5 million samples that can
25 of the Monday demonstrations and motivating potential
be generated by choosing (without replacement) 25 participants to join. Moreover, formal work on high-risk
WGTV counties from the pool of 34 available WGTV collective action suggests that merely having access to
counties in terms of the maximum absolute standardized politically relevant information, such as was provided
difference in means across covariates. The sample with by WGTV, is not sufficient to convince people to partic-
the best balance between WGTV and non-WGTV ipate. When the personal risks associated with participa-
counties was used to estimate Model 5.11 In this sample tion in collective action are high, as they undoubtedly
of highly comparable WGTV and non-WGTV counties, were during the early stages of the East German revolu-
the estimated WGTV effect is positive and statistically tion, resolving coordination dilemmas requires multiple
significant at the 10% level.12 interpersonal contacts who reinforce the normative
Finally, Model 6 restricts the sample to the first event importance of taking action (Centola & Macy, 2007).
in each county, which addresses the concern that WGTV WGTV did supply the political information that the lit-
might only have facilitated the first protest event in each erature on ‘revolutionary bandwagons’ has identified as a
county. The WGTV estimate is statistically significant crucial ingredient in revolutionary collective action. But
at the 10% level but negative, implying that availability perhaps the impersonal supply of information by mass
of WGTV increased the time to first protest event. media cannot rival friendship and family ties that not
Overall, based on the results presented in Tables II– only transmit information about successful protests in
IV, we conclude that there is no robust empirical support other places but also persuade individuals to risk their
lives in support of a common cause.
In many respects, WGTV in East Germany was a
11
Balance is exceptional in this sample and much better than in Kern’s ‘most likely’ case for the impact of foreign media on anti-
(2011) matched sample. The maximum absolute standardized regime collective action in authoritarian regimes. For
difference in means across all covariates is just 0.131 standard one, there were no significant language or cultural bar-
deviations; the average absolute standardized difference in means
across all covariates is 0.059 standard deviations. Finally, we also
riers between East and West Germans. Moreover,
estimated Cox models using the 100 best balanced samples among WGTV, which was extremely popular in East Germany
the ≈ 52.5 million possible samples. Effect estimates for these 100 (Kern & Hainmueller, 2009), devoted a great deal of
samples are shown in Figure A5 in the online appendix. WGTV attention to East German politics even before the out-
effect estimates are generally small and statistically insignificant. break of the East German revolution. It is therefore not
12
Note, however, that for almost all other samples among the 100
surprising that WGTV has been heralded as a prime
best subsamples the estimated impact of WGTV is statistically
insignificant (see Figure A5 in the online appendix). We therefore example of the influence of foreign mass media on tran-
do not regard this result as providing strong support for a positive sitions to democracy. Yet given our inability to find
impact of WGTV. empirical support for the claim that WGTV played an
282 journal of PEACE RESEARCH 52(3)

important role in mobilizing antiregime protests in this References


case, we are somewhat wary of broad claims that commu- Adena, Maja; Ruben Enikolopov, Maria Petrova, Veronica
nication technology can facilitate collective action in other Santarosa & Ekaterina Zhuravskaya (2013) Radio and
cases. At the same time, recent work by Adena et al. the rise of the Nazis in prewar Germany. SSRN work-
(2013) on the effects of Nazi radio propaganda and Sha- ing paper (http://ssrn.com/abstract¼2242446).
piro & Weidmann’s (2015) research on the impact of cel- Andrews, Kenneth T & Michael Biggs (2006) The dynamics
lular networks in Iraq convincingly demonstrate that of protest diffusion: Movement organization, social net-
horizontal and vertical communication technologies can works, and news media in the 1960 sit-ins. American Socio-
sometimes affect collective action (in perhaps unexpected logical Review 71(5): 752–777.
Baum, Matthew A & Yuri M Zhukov (2015) Filtering revolu-
ways). While calls for further research have become a
tion: Reporting bias in international newspaper coverage of the
cliche, we do believe that the impact of communication
Libyan civil war. Journal of Peace Research 52(3): 352–368.
technology on collective action is one area in which such Box-Steffensmeier, Janet M & Suzanna De Boef (2006)
calls are justified and where further research is needed. Repeated events survival models: The conditional frailty
Even though arguments linking communication technol- model. Statistics in Medicine 25(20): 3518–3533.
ogy to collective action are plausible enough, the question Box-Steffensmeier, Janet M; Suzanna De Boef & Kyle A Joyce
probably cannot be settled by relying on cross-national (2007) Event dependence and heterogeneity in duration
evidence alone. Detailed case studies with credible causal models: The conditional frailty model. Political Analysis
identification strategies are needed to pin down the effects, 15(3): 237–256.
or lack of effects, of communication technologies on col- Box-Steffensmeier, Janet M; Suzanna Linn & Corwin D Smidt
lective action. (2014) Analyzing the robustness of semi-parametric dura-
tion models for the study of repeated events. Political Analysis
22(2): 183–204.
Replication data Centola, Damon & Michael Macy (2007) Complex conta-
gions and the weakness of long ties. American Journal of
All statistical analyses were done using R 3:0:3. The Sociology 113(3): 702–734.
online appendix as well as the dataset, codebook, and Chebib, Nadine Kassem & Rabia Minatullah Sohail (2011)
do-files for the empirical analyses in this article can be The reasons social media contributed to the 2011 Egyptian
found at http://www.prio.no/jpr/datasets. Revolution. International Journal of Business Research and
Management 2(3): 139–162.
Dale, Gareth (2005) Popular Protest in East Germany,
Acknowledgements 1945–1989. London: Routledge.
Darmofal, David (2009) Bayesian spatial survival models for
We are grateful to Siegfried Grundmann and Steven
political event processes. American Journal of Political Sci-
Pfaff for sharing data. We also thank William Lynn
ence 53(1): 241–257.
Shirley and Kevin Remington of the Department of DellaVigna, Stefano; Ruben Enikolopov, Vera Mironova, Maria
Geography at the University of South Carolina and Petrova & Ekaterina Zhuravskaya (2014) Cross-border
Yosef Bodovski of the Population Research Institute media and nationalism: Evidence from Serbian radio in
at Penn State University for providing invaluable GIS Croatia. American Economic Journal: Applied Economics
assistance. Giovanni Zambotti of the Center for Geo- 6(3): 103–132.
graphic Analysis at Harvard University graciously Eltantawy, Nahed & Julie B Wiest (2011) Social media in the
allowed us to use the Communication System Plan- Egyptian revolution: Reconsidering resource mobilization the-
ning Tools ArcGIS extension. Julie Kubrick of the ory. International Journal of Communication 5(1): 1207–1224.
Institute for Telecommunication Sciences helped Enikolopov, Ruben; Maria Petrova & Ekaterina Zhuravskaya
direct our efforts at an early stage. Ben Olken and (2011) Media and political persuasion: Evidence from Rus-
sia. American Economic Review 101(7): 3253–3285.
Ruben Enikolopov generously shared advice and soft-
Goldstein, Joshua (2007) The role of digital networked tech-
ware. Janet Box-Steffensmeier, Kyle Joyce, and
nologies in the Ukrainian Orange Revolution. SSRN work-
Suzanna Linn provided extremely helpful advice with ing paper (http://ssrn.com/abstract¼1077686).
regards to our spatial survival model. Allan Dafoe, Gould, Roger V (1991) Multiple networks and mobilization
Jason Lyall, and Nils B Weidmann offered extremely in the Paris commune, 1871. American Sociological Review
useful comments on earlier drafts of this article. Any 56(6): 716–729.
remaining errors are our own. All authors contributed Granovetter, Mark (1978) Threshold models of collective
equally and are listed alphabetically. behavior. American Journal of Sociology 83(6): 1420–1443.
Crabtree et al. 283

Grdĕsić, Marko (2014) Television and protest in East Ger- Kuran, Timur (1991) Now out of never: The element of sur-
many’s revolution, 1989–1990: A mixed-methods analysis. prise in the East European revolution of 1989. World Pol-
Communist and Post-Communist Studies 47(1): 93–103. itics 44(1): 7–48.
Grix, Jonathan (2000) The Role of the Masses in the Collapse of Lichbach, Mark I (1995) The Rebel’s Dilemma. Ann Arbor,
the GDR. New York: St Martin’s. MI: University of Michigan Press.
Grundmann, Siegfried (1997) Territorialplanung in der Lohmann, Susanne (1993) A signaling model of informative
DDR: Indikatoren zur Analyse regionaler Disparitäten and manipulative political action. American Political Science
[Regional planning in East Germany: Indicators for the Review 87(2): 319–333.
analysis of regional differentials]. In: Annette Becker (ed.) Lohmann, Susanne (1994) The dynamics of informational
Regionale Strukturen im Wandel [Transformation of Regional cascades: The Monday demonstrations in Leipzig, East
Structures]. Opladen: Leske þ Budrich, 105–146. Germany, 1989–91. World Politics 47(1): 42–101.
Guiso, Luigi; Paola Sapienza & Luigi Zingales (2008) Social Lynch, Marc (2011) After Egypt: The limits and promise of
capital as good culture. Journal of the European Economic online challenges to the authoritarian Arab state. Perspec-
Association 6(2–3): 295–320. tives on Politics 9(2): 301–310.
Gurr, Ted Robert (1970) Why Men Rebel. Princeton, NJ: Lynch, Marc (2012) The Arab Uprising: The Unfinished Revo-
Princeton University Press. lutions of the New Middle East. New York: Public Affairs.
Havel, Vaclav (1985) The Power of the Powerless: Citizens Against McAdam, Doug (1986) Recruitment to high-risk activism: The
the State in Central-Eastern Europe. Armonk, NY: ME Sharpe. case of Freedom Summer. American Journal of Sociology
Hirschman, Albert O (1993) Exit, voice, and the fate of the 91(1): 64–90.
German Democratic Republic: An essay in conceptual his- Mueller, Carol (1997) International press coverage of East
tory. World Politics 45(2): 173–202. German protest events, 1989. American Sociological Review
Howard, Philip N (2010) The Digital Origins of Dictatorship 62(5): 820–832.
and Democracy: Information Technology and Political Islam. Myers, Daniel J (2000) The diffusion of collective violence:
New York: Oxford University Press. Infectiousness, susceptibility, and mass media networks.
Huntington, Samuel P (1991) The Third Wave: Democratiza- American Journal of Sociology 106(1): 173–208.
tion in the Late Twentieth Century. Norman, OK: Univer- Neubert, Ehrhart (1998) Geschichte der Opposition in der DDR
sity of Oklahoma Press. 1949–1989 [History of the East German Opposition Move-
Hussain, Muzammil M & Philip N Howard (2013) What ment, 1949–1989], 2nd edn. Berlin: Ch Links.
best explains successful protest cascades? ICTs and the Olken, Benjamin A (2009) Do TV and radio destroy social
fuzzy causes of the Arab Spring. International Studies capital? Evidence from Indonesian villages. American Eco-
Review 15(1): 48–66. nomic Journal: Applied Economics 1(4): 1–33.
Jarausch, Konrad H (1994) The Rush to German Unity. New Opp, Karl-Dieter & Christiane Gern (1993) Dissident groups,
York: Oxford University Press. personal networks and spontaneous cooperation: The East
Kern, Holger Lutz (2011) Foreign media and protest diffusion German revolution of 1989. American Sociological Review
in authoritarian regimes: The case of the 1989 East German 58(5): 659–680.
revolution. Comparative Political Studies 44(9): 1179–1205. Opp, Karl-Dieter; Peter Voss & Christiane Gern (1993) Die
Kern, Holger Lutz & Jens Hainmueller (2009) Opium for the volkseigene Revolution [The People’s Revolution]. Stuttgart:
masses: How foreign media can stabilize authoritarian Klett-Cotta.
regimes. Political Analysis 17(4): 377–399. Pierskalla, Jan H & Florian M Hollenbach (2013) Technology
Khamis, Sahar & Katherine Vaughn (2011) Cyberactivism in the and collective action: The effect of cell phone coverage on
Egyptian revolution: How civic engagement and citizen jour- political violence in Africa. American Political Science
nalism tilted the balance. Arab Media and Society 14(1): 1–37. Review 107(2): 207–224.
Kopstein, Jeffrey (1997) The Politics of Economic Decline in Pfaff, Steven (2006) Exit-Voice Dynamics and the Collapse of
East Germany, 1945–1989. Chapel Hill, NC: University East Germany. Durham, NC: Duke University Press.
of North Carolina Press. Pfaff, Steven & Hyojoung Kim (2003) Exit-voice dynamics in
Kowalczuk, Ilko-Sascha (2003) 17. Juni 1953: Volksaufstand collective action: An analysis of emigration and protest in
in der DDR [17 June 1953: Popular Uprising in East Ger- the East German revolution. American Journal of Sociology
many]. Bremen: Temmen. 109(2): 401–444.
Kuhrt, Eberhart; Hannsjörg F Buck & Gunter Holzweissig Plog, Ulla (2004) Wer zu spät kommt, den bestraft das Leben
(1996) Die wirtschaftliche und ökologische Situation der [Those who are late will be punished by life itself]. Frank-
DDR in den 80er Jahren [The Economic and Environmental furter Allgemeine Sonntagszeitung 3 October: 16.
Situation in East Germany During the 1980s]. Opladen: Pollack, Detlef (2000) Politischer Protest: politisch alternative
Leske þ Budrich. Gruppen in der DDR [Political Protest: Politically Alterna-
Kuran, Timur (1989) Sparks and prairie fires: A theory of unan- tive Groups in East Germany]. Opladen: Leske þ
ticipated political revolution. Public Choice 61(1): 41–74. Budrich.
284 journal of PEACE RESEARCH 52(3)

Putnam, Robert (1993) Making Democracy Work: Civic Warren, T Camber (2014) Not by the sword alone: Soft
Traditions in Modern Italy. Princeton, NJ: Princeton power, mass media, and the production of state sover-
University Press. eignty. International Organization 68(1): 111–141.
Roscigno, Vincent J & William F Danaher (2001) Media and Warren, T Camber (2015) Explosive connections? Mass
mobilization: The case of radio and southern textile worker media, social media, and the geography of collective vio-
insurgency, 1929 to 1934. American Sociological Review lence in African states. Journal of Peace Research 52(3):
66(1): 21–48. 297–311.
Schelling, Thomas C (1978) Micromotives and Macrobehavior. Whitehead, Laurence (1996) Three international dimensions
New York: WW Norton. of democratization. In: Laurence Whitehead (ed.) The
Schmitter, Philippe C (1996) The influence of the interna- International Dimensions of Democratization. Oxford:
tional context upon the choice of national institutions and Oxford University Press, 3–25.
policies in neo-democracies. In: Laurence Whitehead (ed.) Yanagizawa-Drott, David (2014) Propaganda and conflict:
The International Dimensions of Democratization. New Evidence from the Rwandan genocide. Quarterly Journal
York: Oxford University Press, 26–54. of Economics 129(4): 1947–1994.
Schwabe, Uwe (1999) Chronik [Chronicle]. In: Thomas Ahbe, Zelikow, Philip & Condoleezza Rice (1995) Germany Unified
Michael Hofmann & Volker Stiehler (eds) Wir bleiben hier: and Europe Transformed: A Study in Statecraft. Cambridge,
Erinnerungen an den Herbst ’89 [We Will Stay Here: Recollec- MA: Harvard University Press.
tions of Fall ’89]. Leipzig: Kiepenheuer, 1–78.
Shapiro, Jacob N & Nils N Weidmann (2015) Is the phone CHARLES CRABTREE, b. 1983, MPPA (Northwestern
mightier than the sword? Cell phones and insurgent violence University, 2012); PhD student in Political Science (2013– ),
in Iraq. International Organization 69(2): forthcoming. Pennsylvania State University; research interests: authoritarian
Shirky, Clay (2011) The political power of social media: Tech- regimes, mass media, and social movements.
nology, the public sphere, and political change. Foreign
Affairs 90(1): 28–41. DAVID DARMOFAL, b. 1968, PhD in Political Science
Stiglitz, Joseph E (1994) Whither Socialism? Cambridge, MA: (University of Illinois at Urbana-Champaign, 2003);
MIT Press. Associate Professor, University of South Carolina;
Süss, Walter (1996) Die Stimmungslage der Bevölkerung im research interests: political behavior and political
Spiegel von MfS-Berichten [Public opinion as shown in Stasi methodology.
reports]. In: Eberhart Kuhrt, Hannsjörg F Buck & Gunter
Holzweissig (eds) Die SED-Herrschaft und ihr Zusammenbruch HOLGER L KERN, b. 1976, PhD in Government
[SED Rule and Its Collapse]. Opladen: Leske þ Budrich, 3–55. (Cornell University, 2008); Assistant Professor, Florida
Tarrow, Sidney (1998) Power in Movement. Cambridge: Cam- State University; research interests: domestic politics of
bridge University Press. authoritarian regimes, democratization, mass media, and
Tilly, Charles; Louise Tilly & Richard Tilly (1975) The Rebel- causal inference in observational and experimental
lious Century: 1830–1930. Cambridge, MA: Harvard studies.
University Press.

You might also like