Running Uphill: Political Opportunity in Non-Democracies. Revised, with a new methodological...

PART FOUR

INSTITUTIONS IN COMPARATIVE PERSPECTIVE

sasaki_f10_271-308.indd 271sasaki_f10_271-308.indd 271 7/9/2008 9:57:55 AM7/9/2008 9:57:55 AM

RUNNING UPHILL: POLITICAL OPPORTUNITY IN NON-DEMOCRACIES

Maryjane Osa and Cristina Corduneanu-Huci

Introduction

How is it possible for oppressed groups to mobilize to act on their grievances in a repressive, authoritarian state? The dominant political process theory of social mobilization and related studies of contentious politics primarily focus on protest in democratic societies (Diani 1996; Kitschelt 1986; Klandermans 1997; Meyer 2004; Rucht 1996). In democracies, civil rights are constitutionally protected, the mass media are uncensored, and dissent is tolerated. However, in non-democratic states, the political system is relatively closed, dissenters are persecuted, and the media is controlled. Can the political process framework equally illuminate mobilization in non-democracies?

Empirical studies of non-democratic mobilization over-sample cases of popular contention at the moment of democratic transition. That is, social scientists select their cases from the tail of the distribu-tion—instances where the authoritarian state already has been fatally weakened (Glenn 2001; Kuran 1991; Mueller 1999). Since data on repressive regimes are scarce, social action that takes place despite the risks imposed by a stable authoritarian state has been diffi cult to study. Recently, however, researchers have been making progress, examining a variety of non-democratic regimes in Latin American, Africa, and Asia (Bob 2002; Boudreau 2004; Brockett 2005; Schock 2005). Some have argued that the preconditions of a successful transition to democracy are found in courageous, often abortive, attempts by social actors to challenge non-democratic institutions and repressive practices (Noonan 1995; Pfaff 2006; Sandoval 1998).

An investigation of conditions that facilitate mobilization in stable authoritarian regimes can thus correct for defi ciencies in two separate (sometimes overlapping) scholarly literatures. The Western bias of social movement theory, particularly the political process framework, can be ameliorated through comparative analyses of political opportunity conditions in non-democracies. The democratization literature can be


274 maryjane osa and cristina corduneanu-huci

extended by examining popular contention in repressive regimes where the ruling elites block, rather than succumb to, the democratic forces. Drawing on the comparative empirical studies, scholars are proposing theoretical formulations that are more sensitive to varying state and institutional formats (Hess and Martin 2006; Meyer 2004; Osa and Schock 2007).

Even monolithic tyrannies may lie on cracked foundations. Fissures in the edifi ce of oppressive rule require inspection. The purpose of our research is to identify political opportunities that enable non-elites to resist, and to contend with, authoritarian governments at the peak of their power. We have studied twenty-four political situations in fi fteen stable non-democracies, two thirds of which resulted in signifi cant socio-political mobilizations, and one-third of which did not. Through qualitative comparative analysis, we seek the minimal set of conditions necessary for high risk/high cost activism and popular contention in non-democratic contexts.

The Dimensions of Political Opportunity

Social movement scholars have identifi ed a number of different factors that create openings for political opportunity in an authoritarian setting. Most often mentioned are: elite divisions, changes in state repression, media access, and infl uential allies. There is no consensus regarding which factor or combination is necessary or suffi cient to initiate mobi-lization in non-democratic settings. For example, a divided elite was judged to be an important factor creating political opportunity for chal-lengers in 1989 China, the Philippines 1986, and South Korea 1987 (Calhoun 1994; Liang, Nathan, and Perry 2001; Schock 1999; Yun 1997; Zuo and Benford 1995). However, system-challenging mobiliza-tions also occurred where the authoritarian elite was unifi ed, e.g., the 1980 mobilization in Poland and the 1989 Leipzig demonstrations in GDR (Mueller 1999; Osa 2003).

Complicating the analysis of political opportunity is the “repression/protest paradox” which stems from the contradictory effects of political violence. Under some circumstances, regime violence successfully quells popular protest, but other times it provokes even larger mass collective action. Repression often has its intended effects, decreasing mobilization, as happened with the suppression of the student movement in China, or with the crackdown on Romanian miners 1977 (Calhoun 1994; Deletant 1998; Vasi 2001; Zuo and Benford 1995). On the other hand,


political opportunity in non-democracies 275

there are also cases like Burma, where the bloody suppression of student demonstrations led to an increase in protest mobilization as by-standers became outraged at the military regime (Shock 1999).

Social scientists have attempted to resolve the repression/popular pro-test paradox (Brockett 1995; Mason and Krane 1989; Opp and Roehl 1990). From the rational choice perspective, Mason and Krane (1989) argue that the targeting strategy of a regime discriminates among poten-tial victims of political violence, thus differentially affecting individual calculations regarding the decision to participate in popular protest. They posit that such regimes are strategically repressing either: 1) leaders of the opposition, 2) the rank-and-fi le members of opposition groups, or 3) the mass public, through indiscriminate state violence. Brockett (1995) disagrees with them, demonstrating that Mason and Krane’s thesis is empirically falsifi ed by the experience of El Salvador and Guatemala (Brockett 1995:119–130). In particular, Brockett’s research shows that the third targeting strategy of indiscriminate state violence sustained over time does not increase mass support for oppositional col-lective action. He proposes instead that the popular response to state repression is infl uenced by the dynamics of the protest cycle. Brockett argues that intense repression will accelerate mobilization only during the ascendant phase of protest, i.e., during the opening of political opportunity. Later in the cycle, as exhaustion or cooptation takes a toll on movements, repression is likely to decrease protest participation. By considering repression as a joint factor of political opportunity during the upswing of the cycle, our study examines whether Brockett’s proposi-tion holds true beyond the Central American cases he has examined.

Finally, social networks are cited for their contribution to mobi-lization, but usually they are considered as an intervening variable between political opportunity and social movement formation. Most social movement research treats networks as “mobilizing structures,” or “meso-level groups, organizations and informal groups” that form coordinating loci “through which people mobilize and engage in col-lective action” (McAdam 1999). We have included it in our model of political opportunities because of an anomaly that we found during the collection and coding of the case study materials. Specifi cally, our study of the Romanian coal miners’ strike in 1977 turned up none of the hypothesized political opportunity variables. This dramatic and puzzling case is worth recounting here in some detail.

In July-August 1977, thirty-fi ve thousand Romanian miners organized a work stoppage, took a government minister hostage, and threatened



First Secretary Nicolae Ceau escu with further actions if the government did not accede to their demands for improved working conditions and benefi ts. This protest took place at a time when there was no internal challenge to Ceau escu from within the party or military, there was increasing repression and deep infi ltration and control of society by the Securitate secret police, external governments were supporting Ceau escu because of his independent foreign policy and determination to pay off foreign debts, and the dictator manipulated a far-reaching propaganda and media control operation. With so many obstacles to collective action arrayed against challengers, how could the Romanian miners organize such a broad-based and effective strike? To answer this question, we delved further into our source materials and investigated other miners’ strikes for clues (Ashwin 1999; Crowley 1997; Dennis 1969; Nash 1992). What we found pointed us to social networks.

In the Valea Jiului mining region of Romania, Ceau escu’s eco-nomic policies created a unique and volatile environment. To increase Romania’s hard-currency exports of coal, Romanian authorities pro-vided various incentives to attract thousands of workers to the mining region. Once resettled, they faced many diffi culties if they attempted to return to their former homes. Indeed, Ceau escu’s policy included the abandonment—even the leveling—of many thousands of traditional villages. (This implemented Hrusciov’s 1959 rural systemization plan, eliminating the so-called “non-viable villages” (Deletant 1998:284–289). When miners received work assignments, they were also allocated hous-ing. As a result of the government’s placement of similarly assigned workers in the same housing blocs, the workers’ children also associated together in the schools. In addition, inter-marriage among these miners was a necessity, since further in- or out-migration from the region was seriously constrained. Consequently, the by-product of the regime’s economic policy was a social reality of multiplex networks: overlapping occupational, social, and residential affi liations. In addition, the specifi c organization of work for the miners concentrated them below ground in hazardous conditions that necessitated a high degree of solidarity and mutual concern for survival’s sake. The party operatives and supervisors exercised their authority over the workers from the mining enterprise offi ces. Thus the workplace itself offered a site for workers’ autonomous organization, deep below the earth’s surface (Vasi 2001).

In the summer of 1977, the government cancelled previously gen-erous retirement benefi ts for the miners and raised the retirement



age from fi fty to fi fty-fi ve. On August 3, at a general meeting of the strikers at the large Lupeni mine, a list of demands was formulated to present to the authorities. They included: reduction of the work day from eight hours to six, restoring the retirement age to fi fty, jobs for dependents (wives and daughters), improved medical care, and media coverage of the strike.

The actions of the miners provoked a government crisis. First Sec-retary Ceau escu appointed a commission to negotiate with the miners that included top ranking government ministers, Politbureau members, and local mayors. When the commission arrived at the Lupeni mines, the offi cials were assaulted by workers when they tried to enter the mine manager’s offi ce. The miners particularly distrusted the ranking member of the commission, Ilie Verde‚t, the Politbureau member in charge of the economy and Ceau escu’s close friend. They demanded that he call Ceau escu to appear in person to negotiate with the strikers. The miners then took Verde‚t hostage to put pressure on the dictator.

Ceau escu, fearing that the unrest would spread beyond the mining sector, appeared at Lupeni to negotiate with striking miners. He prom-ised to instate a six hour work day, build new factories to employ min-ers’ dependents, and take no retaliatory actions against strikers. These concessions brought the miners’ protest to an end. Shortly thereafter, the government began an extensive campaign of repression. The min-ing valley was sealed off by the army and secret police, strike leaders mysteriously disappeared, four thousand miners were relocated, and some were sent to forced labor camps (Deletant 1998:233).

Thus, in the Romanian case, absent any other source of political opportunity, the unusual social circumstances of the miners fostered highly multiplex social networks and gave the laborers autonomous space below ground to plan their actions. The networks provided channels for uncensored communications and the social contacts needed to locate material resources. Analysis of the Romanian case leads us to propose the following hypothesis: under some conditions, networks can gener-ate internally the resources to leverage political opportunity. Thus, we have included social networks as an independent variable to explore whether these contribute directly to political opportunity, rather than (as is usual in the literature) operate as an intervening variable within the mobilization process.



Data and Methods

The method of qualitative comparative analysis (QCA 3.0, Drass and Ragin 1986) developed by Charles Ragin (1987) is well-suited for an exploratory analysis of political opportunity conditions in non-demo-cratic settings. First, it allows for detailed, contextual comparisons of more countries than is usual in comparative case studies. Second, QCA allows us to utilize a diverse set of data sources. For example, we use quantitative measures, such as event counts, to estimate governmental repression levels and we draw on the expertise of area scholars to inter-pret the qualitative evidence of domestic political divisions. Third, QCA assumes that social (and political) causality is non-linear, interdependent, and heterogeneous—a much more realistic set of assumptions for our cases than those underlying regression analysis. The use of standard sta-tistical methods in this type of research is problematic for four reasons. First, since the boundaries of the population are unknown, sampling is impossible. Second, for quantitative analyses, the World Handbook data set is unreliable (Diani 1992; Diani and della Porta 1999). Finally, the protest event counts represent discrete values. For statistical analyses, we would have to transform these values into continuous variables as indices measuring magnitude or intensity. Since we are interested in the occurrence, or outcome, of mobilization, measurements of protest intensity or magnitude will not further the analysis. In short, we chose the method of qualitative comparative analysis in order to respect the contextual integrity of the cases, assume causal heterogeneity, and ana-lyze moderately large and diverse sets of empirical instances (Amenta and Poulsen 1994; Ragin 1987, 1995).

Case Selection

Non-democracies obviously encompass a number of types: military dictatorships, monarchies, one-party states, and Leninist (or state socialist) regimes, to name a few. Yet, with the exception of traditional monarchies, they share an important characteristic: these societies lack a mechanism for regular, legitimate transfers of power sanctioned by those subject to the state. As a result, the basis of rule in non-democ-racies always involves a much higher degree of coercion than that in democratic states. And the institutions of coercion vary little among the types of authoritarian states.



Three features of authoritarianism act as barriers to collective action. First, political activity is usually confi ned to membership in a single mobilization party (e.g., the communist party in China or the USSR, the Ba’ath party in Syria). Second, the mass media are usually under some form of state control. Third, individual citizenship rights and due process are not guaranteed. By contrast, the democratic state facilitates social movement organization and mobilization: freedoms of speech, association, and the press are constitutionally guaranteed, and oppositional forms of political activism are considered legitimate (Osa and Schock 2007:139).

For our study, we selected twenty-four cases of occurrence/non-occur-rence of mobilization in fi fteen stable non-democratic regimes. Case selection was not randomized. To obtain a random sample, we would need to enumerate, or defi ne a population, of occurrence/non-occur-rence of political mobilization in non-democratic contexts. This type of list does not exist, nor would it be feasible to compile. Like Goodwin (2000), who recognizes this problem with randomization, we select cases to maximize variation and we refrain from attempting statistical tests on our sample. All cases from our data set occur during the Cold War (1948–1989) but vary according to type of non-democratic regime (e.g., military regimes, state socialist regimes, bureaucratic authoritarian regimes) and region (including Eastern and Southern Europe, Latin America, the Middle East, sub-Saharan Africa, East Asia).

What is a case? To set boundaries around the episodes of mobilization (or their absence) we began by graphing event counts from the World Handbook III (1985) to identify mobilization peaks and troughs. For fi fteen non-democracies, we plotted two variables, political contention (a composite variable that includes strikes, demonstrations, riots, and armed attacks), and net repression (the difference between the number of political sanctions imposed and the relaxation of sanctions). Once we had located the periods of contention and quiescence, we turned to the descriptive area studies literature to learn more about catalytic events (or events of major signifi cance) that would shed light on the mobilization contents. This secondary literature usually contained area experts’ periodizations that further explained the different phases of repression and contention. Finally, where area specialists contradicted the World Handbook data, we coded our cases based on the area ana-lysts’ writings because of well known problems with data coding in the World Handbook (Diani 1992).



Variables

dependent variable. We defi ned “social mobilization” as sustained collective action opposing state policies by participants drawn from non-elite or repressed segments of society. This is a broader defi nition than Shock (1999) uses; he requires a high numerical threshold for participation (more than ten thousand participants). Since we wanted to include semi-clandestine attempts at oppositional mobilization (e.g., human rights organizing in Latin America) and neighborhood-level mobilizations (e.g., poor people’s movements in Honduras and Iran), we opted for a more inclusive defi nition. Although human rights organizing and the occupation of buildings may seem qualitatively different than the traditional indicators of mobilization (e.g., strikes, demonstrations), in authoritarian contexts, they are all manifestations of high risk/high cost political action.

independent variables. The four dimensions of political opportu-nity that were fi rst implied by Tarrow’s (1994) formulation are now regarded by many scholars as crucial factors of mobilization under authoritarianism. These are: state repression, elite divisions, infl uential allies, and media access/informational fl ow (Osa 2003; Shock 1999, 2005; Zuo and Benford 1995). These four variables were used for the initial model. After our analysis of the Romanian miners case, and cognizant of the literature’s emphasis on mobilizing structures in non-democratic contexts, we decided to include social networks in the full model (Deletant 1998; Loveman 1998; Osa 2001; Oxhorn 1991).

Since the QCA method proceeds by comparing and reducing condi-tions to their most parsimonious confi gurations, the correct selection of conditions is paramount for achieving reliable results. Thus, a short discussion on how the independent variables were identifi ed is neces-sary.1 Four factors were coded as dummy variables: elite divisions,

1 For the fi rst independent variable, dynamics of repression, we plotted the times series from the World Handbook, and we took the fi rst difference (n– (n–1)). We took into account both long-term trends and annual value changes. The coding was then corroborated by the area literature’s fi ndings on changes in repression. However, for the fi ve cases after 1982 which were not covered by the World Handbook, we had to use area-specifi c secondary sources to identify changes in repression. The same coding protocol was followed for media access, since the quarterly version of the World Hand-book detaches this variable from the aggregated repression index. For the qualitative variables (infl uential ally, elite divisions, social networks, mobilization), we coded 1 for presence and 0 for absence based on information from case studies by area experts. Each of these variables was coded independently by two coders and coding discrep-ancies were subject to further investigation. We did not impose thresholds in order to assess the degree of presence/absence for given variables.



infl uential allies, media access/information fl ow, and social networks. One condition was coded in terms of its dynamics: 0, if repression was decreasing; 1, if repression was increasing. This accords with extant research suggesting that it is not the repressive apparatus itself, but rather a change in its actions that determines the masses to mobilize (Shock 1999; Tarrow 1998).

Elite divisions (divelit) were considered present if, after a review of the literature, evidence was found of competing factions within the ruling elite. Competitors within the ruling party or junta had to have a realistic chance of supplanting the faction in charge. The presence of a symbolic opposition within the polity or of powerless moderates was not enough to signal a “divided elite.”

Repression is the only variable that was measured in terms of its dynamics (dynrepr). For each case, we calculated the net repression as a difference between the number of political sanctions and the count of relaxation of sanctions calculated from the quarterly data. We used the variable defi nitions specifi ed in the World Handbook III codebook. Variable 18, “imposition of political sanctions,” includes actions taken by the government to neutralize or suppress a perceived domestic threat to the regime. The “relaxation of sanctions,” Variable 20, involves the modifi cation or elimination of these restrictions.

Infl uential allies (inflally) were identifi ed also through a review of the area literature. These allies could include: religious organizations (e.g., Catholic Church in Brazil), foreign governments (US support of anti-Marcos demonstrations), international organizations (European Commission for Human Rights, UNHRC), transnational social move-ments (Comite para la Paz), or foreign political parties (Italian Com-munist Party support of Spanish opposition).

Media access/information fl ows (medacc). All of the regimes included in the study tried to control the fl ow of public information, often by institutionalizing censorship and permitting only state-sponsored news-papers. We assessed the degree of media access, fi rst, by examining the World Handbook data on the imposition/relaxation of censorship. In the cases where the World Handbook indicated that censorship controls had been relaxed or imposed, we went back to the original sources of the Handbook-coded data. The original newspaper reports provided more detailed information on imprisoned journalists or other press restrictions. Finally, the secondary literature provided fuller descriptions of the political context and information fl ows. We coded media access as present, if: there was a sustained relaxation of state censorship, or foreign journalists publicized opposition activities through factual



reporting that was rebroadcast into the country by Radio Free Europe and Voice of America, or there was an active underground press that circulated uncensored materials. Most often, we found a combination of factors increasing information fl ows. For example, if a regime relaxed censorship of print media, authorities usually also put less effort into jamming Radio Free Europe broadcasts. Uncensored information on contention had to be available to the general public for media access to be considered present.

Different types of social networks (socnet) were included in this rough measure. We coded networks present if case studies had shown that mobilization resulted from coordinated activities of specifi c, linked groups. For example, in the South African and the Poland cases, inter-organizational networks were implicated in mobilization (Herbstein 1979; Hirschsohn 1998; Hirson 1979; Karis and Gerhart 1997; Osa 2003). In the Honduran and Chilean cases, area experts pointed to inter-connected, neighborhood groups relying on activists’ interpersonal and community ties to motivate collective action (Brockett 1994; King 1989; Schneider 1991).

Data Set and QCA Procedures

The Political Opportunity Structures in Non-Democracies (POSND) database (Osa and Corduneanu-Huci 2001) was constructed in MS Access and structured as fi ve inter-linked tables. The fi rst table con-tained information on state/political regime characteristics, such as regime structure, and notes concerning the political parties, unions, and economic organizations that were permitted under the regime. The second table coded case characteristics, including the temporal span, number of protest events, and graphs of the quantitative variables (e.g., repression). The third table was the event protocol, detailing individual events and recording bibliographic sources for the event information. The fourth table specifi ed how variables were coded for each of the cases; binary values for the dependent and independent variables were set and coding justifi cations were noted. The fi nal form was the table source, which contained author/date information for sources that documented each variable. Our database includes twenty-four cases, drawn from fi fteen non-democratic countries. Eighteen of the cases resulted in social mobilization (socmob = 1) and six cases resulted in the non-occurrence of mobilization (socmob = 0).

QCA takes the raw data matrix of nominal-scale variables coded in a binary form and reconstructs it as a truth table. (For the data



matrix, see Appendix 1.) The truth table comprises a number of rows (2n where n = the number of independent variables) representing all logical combinations of values on the independent variables. The num-ber of empirical cases corresponding to each row can vary between 0 and n. The program fi rst takes the raw data matrix and transforms it into a truth table in which the non-existent logical combinations are added to the empirical instances. The minimization algorithm starts with the raw data matrix and, by eliminating redundant factors via a step-wise comparison of case pairs, it produces the “prime implicants,” the minimum sets of conditions considered necessary for producing a certain outcome (0 or 1 on the dependent variable). The underlying logic of comparison for this algorithm is derived from John Stuart Mill’s methods of agreement and difference (Mill 1974).

A potential problem with this method is the existence of contradic-tions, i.e., a pair of cases where the same set of causal conditions leads to opposite outcomes. Ragin (1995) suggests a number of options in dealing with this problem. The fi rst option ignores the contradiction, given that the same set of conditions causes both a positive and nega-tive outcome. In practice, the two contradictory cases are removed from the truth table. The second option, more generous analytically, is based on the assumption that the probability of occurrence of any given outcome in the contradiction is 0.5. Only one case from the contradictory pair is included in the truth table. So when the positive outcome is considered, the positive case is included; when the negative outcome is analyzed, the negative case is included. The third alternative is purely qualitative and consists in reexamining the contradictory cases for missed idiosyncrasies. It has the potential for adding new indepen-dent variables to the analysis. Since we had a single contradiction in our data set, we took option one and allowed the computer to ignore the contradiction for the analysis; consequently there were twenty-two cases used in confi guring the truth table.

In sum, by using the Boolean methods of qualitative comparative analysis, we can manage the complexity of our cases while consider-ing the relevant factors in all their logical combinations. We can thus identify the political opportunity confi gurations in authoritarian coun-tries that appear with the occurrence of the dependent variable, social mobilization. By analyzing these conjunctions, or “prime implicants,” we generate empirically grounded hypotheses that can be tested when good quality data become available.



Results

Since the evidence of social networks’ direct contribution to political opportunity was strong for only the Romanian case, we ran the analysis twice, once with the four political opportunity variables, and a second time with social networks added to the model.2 With the inclusion of social networks we achieve a model that is superior to the four political opportunity variables for the following reasons. First, when the social networks variable is excluded, the number of contradictory cases increases to eleven. According to De Meur and Rihoux (2002), a large number of contradictions—46% in the case of this model—indicates that the variable selection is problematic. Second, when we include the social networks variable and the logical cases (for which there are no empiri-cal data) and run the strict minimization procedure, the QCA results are extremely parsimonious, resulting in only two prime implicants. Moreover, the empirical complement of the two prime implicants is identical to the logical complement. This means that the logic predicts what actually happens in our twenty-two cases. In short, when we exclude social networks, the number of contradictory cases jumps to eleven, almost half of the entire dataset, and there is a disjunction between the logical and empirical complements. For these reasons, we present below the results from the expanded model of the four political opportunity variables plus social networks. The results from the four variable model that we rejected are given in Appendix 2.

Outcome One: Social Mobilization

We fi rst consider whether social mobilization (socmob) resulted from any combination of the fi ve independent variables: dynamics of repression (dynrepr), divided elites (divelit), infl uential ally (inflally), media access (medacc), and social networks (socnet). In our dataset, we have eighteen cases with positive outcomes: East Germany 1953, Poland

2 The fi rst model with the four political opportunity variables is specifi ed in QCA notation, where the plus sign represents the logical operator “or,” as follows:

SOCMOB = DYNREPR + DIVELIT + INFLALLY + MEDIAACCThe results of QCA analysis using this model were uninterpretable, non-parsimoni-

ous, and laden with contradictions. According to De Meur and Rihoux (2002), this is an indication that the selection of variables for the model is poor. The results from this analysis are provided in Appendix 2.

The fi nal model adds social networks and is specifi ed as: SOCMOB = DYNREPR + DIVELIT + INFLALLY + MEDIAACC + SOCNETThis is the model that is used in the analysis presented in the body of the text.



1956, Portugal 1958, Portugal 1962, Spain 1962, Greece 1968, Hon-duras 1968, Poland 1968, Chile 1973, Argentina 1977, South Africa 1976, Brazil 1977, Romania 1977, Iran 1980, Chile 1983, Romania 1987, Burma 1988, and China 1989.3 The results of QCA truth table minimization identify four prime implicants, or combination of condi-tions, that are associated with the outcome, social mobilization in these eighteen cases.

This fi rst prime implicant shows that social mobilization occurred when challengers lacked media access if an infl uential ally (or allies) and social networks were both present. This scenario is covered by six cases in our database: Brazil 1977, Chile 1973, Iran 1980, Poland 1968, Portugal 1958 and Portugal 1962. State control of the media was total in most cases (Chile, Brazil and Portugal); already strict censorship was increasing in Poland where foreign journalists had been expelled from the country. Ironically, censorship was mildest in Iran: although the revolutionary government controlled what was published, reports of squatters’ protests did appear in the newspapers.

3 Since East Germany is one of the contradictory cases, it is dropped from the truth table. Consequently, only seventeen cases are included in the QCA analysis of the mobilization outcome.

Table 1.

Prime Implicantssocmobe = 1

N = 17

1st medacc INFLALLY SOCNET1,2 or

2nd dynrepr MEDACC DIVELIT SOCNET or

3rd DYNREPR MEDACC divelit infl ally or

4th One of the following: DYNREPR medacc divelit SOCNET DYNREPR infl ally divelit SOCNET

1 Lower case notation indicates absence of the condition; upper case notation indicates presence of the condition.2 DYNREPR = Dynamics of RepressionINFLALLY = Infl uential AllyMEDACC = Media AccessDIVELIT = Division of ElitesSOCNET = Social Networks



Challengers’ infl uential allies varied somewhat across the cases. In Brazil and Poland, the Catholic Church hierarchy offered support to grass-roots mobilization. International human rights organizations were active in Chile. Iranian “houses of labor,” were local institutions backed by labor unions and left-wing parties; these community-based organizations supported the poor people’s housing movement (Bayat 1997). In Portugal, suppressed political parties aligned with anti-Salazar groups (Fryer and Pinheiro 1961). In general, the external allies were long-standing institutions: local, such as the Iranian houses of labor; national, as in Portuguese historic political parties; or with international connections, such as the Church and human rights organizations in Latin America. Similarly, the networks involved in these mobilizations ran the gamut, from the “passive networks” of neighborhood residents in Iran, to loose coalitions of Brazilian students and workers linked through Catholic base communities, to inter-organizational networks of faith-based groups and human rights organizations in Chile.

In sum, all the six cases share the same overall confi guration of politi-cal opportunity but vary according to how the factors are expressed, especially in the types of allies and networks available to challengers. We expect that this variation would dictate differences in how allies and challengers interact, and it would also be related to the scope, size, duration and type of mobilization. Thus, while the absence of media access, availability of allies and social networks is associated with social mobilization under the fi rst prime implicant, the form that this collective action takes varies. Specifi cally, social mobilization ranges from street demonstrations (Portugal, Iran), student protests (Poland, Portugal), strikes (Brazil), land and building seizures (Iran), hunger strikes and riots (Portugal), and underground human rights organizing to assist families of political prisoners (Chile).

The second prime implicant shows that social mobilization occurred when repression decreased, the ruling elites were divided, challengers had access to media and social networks were activated. This seems to be a rather common scenario found in the social movements literature (Calhoun 1994; Osa 2001). Six cases in our database exhibited this confi guration: Chile 1983, China 1989, Greece 1968, Honduras 1968, Spain 1962, and Poland 1956.

This confi guration starts with a lessening of repression. Decreasing repression often takes the form of a temporary reduction of regime violence due to international attention or a “thaw” in political rela-tions stemming from such processes as de-Stalinization, détente, or



perestroika. Elite divisions may occur within the ruling party (China, Poland, Spain) or the military (Chile, Honduras, Greece). Information fl ows increase because of coverage by the foreign press (China, Poland), Radio Free Europe or BBC broadcasts that are not being jammed (Poland, Spain), loosening restrictions on local press as part of the overall reduction of repression (Chile, Greece, Honduras, Spain), and publication of uncensored materials by émigrés abroad (Greece, Poland). Again, social networks can take a number of forms: local neighborhood networks organized by social movement groups (Chile), student networks (China, Poland, Spain) underground networks of banned organizations (Greece), and coalitions of peasant organizations (Honduras). The col-lective mobilizations associated with this second prime implicant took the following forms: demonstrations and protest marches (Chile, China, Greece, Poland), riots and arson (Poland), strikes (Spain, Chile), hunger strikes (China), student protests (China, Poland, Spain), and property seizures (Chile, Honduras).

The Chinese student mobilization and occupation of Tiananmen Square is a good illustration of this second prime implicant. First, in 1989, repression decreased as a result of the economic reforms of Hu Yaobang and Zhao Ziyang and a relaxation in the ideological fi eld. Second, there was a strong division within the Communist Party elite between reformists and hard-line party elders. Third, large numbers of foreign journalists came to Beijing for the fi rst Asian Development Bank meetings and Gorbachev’s upcoming state visit. There was some softening of state censorship; students also made use of “large character” posters and loudspeakers to broadcast information. In addition, Hong Kong news services, CNN and VOA were broadcasting reports of the students’ hunger strike and other protest events. Finally, the creation of cohesive student networks was facilitated by Beijing’s social geogra-phy that concentrated academic populations in small dormitories and neighborhoods (Calhoun 1994; Liang, Nathan and Link 2001; Zuo and Benford 1995).

The third prime implicant shows that social mobilization occurred when increased repression was imposed by a unifi ed (at least nominally) ruling elite, and challengers lacked infl uential allies but had access to media. Only three cases exhibit these more restrictive conditions for mobilization: Argentina 1977, South Africa 1976, and Burma/Myanmar 1988. The tightening of already restrictive conditions in these cases by a unifi ed elite increased a sense of injustice and focused challengers’ outrage on the authorities. The strong impetus for mobilization in the



confi guration stemmed from the communications media informing the public of extensive injustices and the ability of oppositional groups to use the media for organizing. The expansion of information fl ows took the following forms: the foreign press exposes regime violence in detailed reports (Argentina and South Africa), shortwave broadcasts of opposition radio station publicized injustices and opposition demands (South Africa), the circulation of underground press or independent publications expands rapidly (Burma, South Africa), and news of protest demonstrations appears in the offi cial press when editors are pressured by employees at state publishing houses (Burma). The outcomes of social mobilization in these cases occurred as: demonstrations (Argentina, Burma, South Africa), boycotts (South Africa), and strikes (Burma).

The South African experience illustrates this set of conditions. Mobilization in Soweto in 1976 took place during a period of severe repression. Black leaders such as Nelson Mandela had been impris-oned since 1964 and the activities of the African National Congress were severely limited. The white ruling elite were united in support of apartheid, although leaders had begun to recognize that such a policy harmed the regime’s legitimacy. Protests began in Soweto when the government mandated the study of Afrikaans for public school students. News of the student demonstrations were broadcast on Radio Freedom, the short-wave station operated by the ANC. Besides uncensored radio broadcasts, information fl ows increased though the distribution of leafl ets and underground literature. Foreign journalists also covered the resistance of secondary school students to the study of the Afrikaans language (Herbstein 1979; Karis and Gerhart 1997).

The fourth prime implicant is similar to the third in that an unifi ed elite is combined with increasing repression. However, instead of media access providing a mechanism for mobilization, social networks form the basis for oppositional activity. There are two variants of this prime implicant: version a) shows the absence of media access, version b) is absent the infl uential ally.

Three cases are covered by the 4a variant: Chile 1973, Romania 1977, and Romania 1987. Four cases are covered by the 4b function term: Argentina 1977, South Africa 1976, Romania 1977, and Roma-nia 1987. This confi guration begins in the same fashion as the third prime implicant: increasing repressive actions taken by an undivided elite creates social tension and enhances collective feelings of injustice. Social networks are the means through which opposition to the regime coalesces. Networks take a variety of forms: informal women’s networks



(Argentina), inter-organizational networks of religious and human rights groups (Chile), multiplex labor/residential networks (Romania), and local networks of school-based, labor, and community groups (South Africa). Social mobilizations that occurred under this prime implicant included: organized assistance to victims of the regime by underground groups (Chile), strikes (Romania), protest marches and demonstrations (Argentina, Romania, South Africa), and boycotts (South Africa). An illustration of this confi guration is Romania 1977, the anomalous case discussed at the beginning of this article. Here we found that despite severe and increasing repression by the Ceau escu regime, Romanian coal miners were able to use their multiplex (overlapping) social, occu-pational, and residential networks to mobilize mine workers for an extensive strike. This occurred in the absence of media access, so the distribution of information (and other resources) had to occur within social networks.

Notice that a number of cases are covered by more than one func-tion term. That is, different prime implicants can be associated with a single case. For example, the 1976 South African case is included as evidence for both the third prime implicant and the fourth prime implicant. This means that South African mobilization in 1976 resulted both from increasing repression, unifi ed elite, media access, and no ally and from increasing repression, unifi ed elite, social networks, and no ally. In short, the QCA analysis identifi es two different confi gurations, or two sets of suffi cient conditions, associated with mobilization in the South African case. Thus, if either the networks or media access had been absent, mobilization in Soweto still would have occurred.

Outcome Two: Social Quiescence

We now consider whether non-occurrence of social mobilization resulted from any combination of the fi ve independent variables: dynamics of repression (dynrepr), divided elites (divelit), infl uential ally (inflally), media access (medacc), and social networks (socnet). In our dataset, we have six cases with negative outcomes: Spain 1952, Poland 1960, South Africa 1964, Uruguay 1973, Romania 1978, and Chile 1983lcp.4 The results of QCA truth table minimization identify two prime implicants,

4 Since Chile 1983lcp (low combative poblaciones) is the second contradictory case (paired with East Germany), it is dropped from the truth table. Consequently, only fi ve cases are included in the QCA analysis of the no mobilization outcome.



or combination of conditions, that are associated with the negative outcome, non-social mobilization in these six cases.

The fi rst prime implicant shows that no mobilization occurred when repression was decreasing, elites were unifi ed, there was a lack of media access and social networks, but an infl uential ally was present. This confi guration describes a context where a non-democratic regime whose elite is unifi ed reduces overall repression but maintains strong control of the media. Even if an infl uential ally is available for challengers, absent social networks, this confi guration of political opportunity is insuffi cient to foster mobilization. This conjuncture is represented by one case in our database, Poland 1960. In this case, Poland’s Commu-nist leader, Władysław Gomułka, had consolidated his regime following the upheavals of 1956. By 1960, Gomułka had eliminated his rivals within the Party, lessened the repression of 1958–1959, and although the Catholic Church was available as an ally, opposition networks that had formed in 1956 had been liquidated. Poland was quiet until 1966 (Osa 2001).

The second prime implicant shows a combination of political oppor-tunity variables that are common in repressive, non-democratic regimes. A unifi ed elite increases repression and maintains strict control over media and information flows; challengers lack allies, oppositional networks, and media access. Sixty-seven percent of our cases of non-mobilization are covered by this function term: Romania 1978, South Africa 1964, Spain 1952, and Uruguay 1973. This fi nding shows that if the authoritarian state is able to maintain a unifi ed elite and use the repressive apparatus to control the media, fragment society, and pre-

Table 2.

Prime Implicantssocmobe = 0

N = 5

1st dynrepr medacc INFLALLY divelit socnet1,2 or

2nd DYNREPR medacc infl ally divelit socnet

1 Lower case notation indicates absence of the condition; upper case notation indicates presence of the condition.2 DYNREPR = Dynamics of RepressionINFLALLY = Infl uential AllyMEDACC = Media AccessDIVELIT = Division of ElitesSOCNET = Social Networks



vent challengers from fi nding external support, the regime will not be subject to contentious events. Such a conclusion would have come as no surprise to Hitler, Stalin, Mao, Franco, Pinochet and the numerous dictators who have historically employed this strategy.

Conclusion

QCA minimization identifi ed four prime implicants, or specifi c combi-nations of political opportunity variables resulting in mobilization, and two prime implicants associated with its non-occurrence in authoritarian regimes. Beyond the identifi cation and discussion of the various POS combinations, some more general conclusions can be drawn. First, we are satisfi ed that the fi ve variables—dynamics of repression, divided elite, infl uential ally, media access and social networks—capture the signifi cant dimensions of political opportunity in the non-democratic cases. In all the instances of social mobilization some combination of our variables were found; in all the cases of non-occurrence of social mobilization the variables were missing, with the lone exception of the infl uential ally. That the QCA program found only one contradiction in our data set also suggests there were no specifi cation problems.

Second, what makes mobilization in a non-democratic context a high risk/high cost proposition is the propensity of such a regime to increase repression in order to quell contention. Paradoxically, mobiliza-tion results under increasing repression as well as under more relaxed conditions. We found that decreasing repression plays a role in creating political opportunity in one out of four possible conjunctures (prime implicant 2), although this condition might account for a larger percent-age of mobilization episodes. On the other hand, increasing repression is a factor in two prime implicants (prime implicants 3 and 4 a,b). Our data provide additional cases to support Brockett’s contention that increasing repression at the beginning of a protest cycle will contribute to mobilization. However, since we do not collect data from the end of the cycle, his proposition is not falsifi able with our data set. In sum, although non-democratic regimes rely on repression to curtail political contention, our data show that this is only successful if the authorities are able simultaneously to prevent challengers’ media access and to inhibit the formation of politically-oriented social networks.

Third, social networks clearly create political opportunities as well as respond to them. Although most of the scholarly attention to networks has been on their role as “mobilizing structures” for social movements,



our research suggests that networks can produce opportunities as well as spread mobilization. (See also, Osa and Schock 2007.) Where com-munities contain cohesive, interlocking or multiplex social networks, feel-ings of group identity and solidarity are enhanced. If harsh, repressive measures are imposed upon such a community and a sense of injustice grows, these solidary groups become motivated to protest, despite the odds. The networks generate resources internally (information, material supplies, protection) that are deployed in highly politicized contexts to create opportunities for action.

Finally, our analysis shows no necessary conditions of political opportu-nity in non-democracies. However, the four prime implicants discussed above identify different combinations, that is, four different sets of conditions considered suffi cient to allow mobilization in authoritarian contexts. It is possible to reduce the prime implicants even further. We did a fi nal QCA procedure to fi nd the maximum reduction of political opportu-nity conditions. The program algorithm facilitates this by considering all the empirical cases (including the contradictions) and the subset of the logical cases that do not contradict empirical instances in the data set. This minimization algorithm identifi es the key factors involved in producing the phenomena. For our dataset, these are: media access and social networks.

In conclusion: our study examined the preconditions of democratiza-tion by identifying conditions under which political opportunities arise for challengers living under (seemingly unassailable) authoritarian gov-ernments. We found that the conventional strategies of non-democratic regimes for maintaining quiescence have both a logical and empirical foundation. Yet the greatest danger to stability in non-democracy occurs when society manages to overcome internal divisions and creates links between individuals and groups. Free and uncensored media also have an important role to play in confronting authoritarianism.

The policy implications of this research are clear: if democratic gov-ernments want to encourage challenger movements within authoritarian countries, they should support cultural and educational exchanges that stimulate information fl ows and create networks, subsidize uncensored broadcasts such as Radio Free Europe and BBC, support émigré pub-lications, and provide assistance to non-governmental organizations such as women’s rights groups and human rights associations that operate internationally. Conversely, our research suggests that efforts to engage non-democratic regimes economically may have little effect in stimulating democratic oppositions. Corporate leaders may fi nd doing



business with autocrats easier than negotiating with short-term elected offi cials. And authoritarian governments easily channel profi ts through state-controlled organizations to the military or to regime cronies.

Even if challengers are supported by external allies, political oppor-tunity conditions are not easily produced when authoritarian govern-ments are at the peak of their power. In fact, all the mobilizations we have studied ended with the suppression of contention. Nevertheless, the long-term outlook is hopeful: despite the numerous “failures” of collective action catalogued in our database, the majority of the coun-tries we studied are now functioning democracies.

Methodological Addendum (2007)

Some social scientists remain skeptical of results obtained via QCA. Since the method relies on the logical sorting of conditions to identify causal confi gurations, it is highly sensitive to variable and case selection, as well as to data dichotomization (Amenta and Poulsen 1994; Becker 1998; Bollen et al. 1993; DeMeurs and Rhioux 2002; Goldthorpe 1997; Ragin 1987, 2000). Moreover, QCA has been criticized for not taking into account the temporal dimension of causation (Boswell and Brown 1999). Given these diffi culties, how can we be certain that we have identifi ed the relevant POS dimensions? Could there be some systematic bias operating in case selection of which we are unaware? Does the static nature of analysis miss important causal components? These concerns led us to further exploration of our cases with more conventional statistical methods in order to test the validity of the results presented in the original “Running Uphill.”

Statistical Analysis on POSND Data Set

In this Addendum, we report on the results of a statistical analysis using Maximum Likelihood Estimation (MLE) techniques on mobilization events in our fi fteen countries for the period 1955 through 1982. Spe-cifi cally, we test the marginal probability of the number of anti-regime protest events by using a negative binomial regression model. In order to utilize these techniques, we have had to reorganize and supplement our original data sets. The World Handbook III remains the source for event counts for Argentina, Brazil, Chile, China, East Germany, Greece, Honduras, Iran, Myanmar, Poland, Portugal, Romania, South Africa, Spain, and Uruguay. For the independent variables, we used data



from the State Failure Task Force (Goldstone et al. 2000) and Polity IV (Marshall and Jaggers 20002), in addition to our original data set.

For the QCA analysis, our cases were delimited temporally by identifying mobilization episodes; this required periodization of event sequences. Such a process is subject to error, even when rigorous proto-cols are followed. To check for error in case selection (or periodization), we have now completed a statistical analysis of all occurrences of anti-regime mobilization, and not just the events that make up our cases. In the MLE analysis, we used all the event data for our fi fteen countries (1955–1982), disaggregating our initial dependent variable “social mobilization” into four separate dependent variables: social mobilization (total number of protest events), anti-regime protests, strikes, and riots. All of the reconstituted dependent variables were categorical variables in the form of event counts. Our panel contains 365 country-year observations (15 countries × 28 years—55 missing data points on two of our control variables, Log [GDP/capita] and economic growth).

The fi ve political opportunity variables—divided elites, infl uential ally, media access, repression dynamics and social networks—have been included in this analysis, although we have had to specify them differently in order to give them a quantitative basis. To operation-alize the dimensions of repression and media access, we used the World Handbook data in a time series analysis of repressive measures enacted by the regime generally (dynamics of repression) and against journalists, in particular (media access). Net repression was computed from the WH variables by taking the number of repressive measures imposed by the regime and subtracting the number of relaxation of sanctions. We computed a similar measure for media access, by fi rst reconstructing two yearly variables recording imposition of censorship and censorship relaxation respectively, based on the two WH quarterly series. Second, we performed the same computations for obtaining our media access variable as we had for repression, subtracting the relaxation of censorship incidents from the yearly number of censoring measures undertaken by the government. Thus, a lower value of this variable denotes greater access to a free media.

Divided Elites and Infl uential Allies are equally challenging to quan-tify. The best approximation we found for divided elites uses a measure of elite homogeneity as a proxy for elite unity, which we code as the absence of elite divisions. The State Failure Task Force (Goldstone



et al. 2000) specifi es ten categorical variables (ethnicity, clan member-ship, common ideology, bureaucratic membership, military status, entrepreneurial, mass revolutionary, intelligentsia, land ownership, no shared class characteristics) taking the values 1 (low) and 2 (high) if the ruling elite comprises a homogeneous group on a given dimension of homophily and 0 otherwise. We accept the argument that shared social characteristics increase the probability of elite unity (McPherson, Smith-Lovin, and Cook 2001). Thus, we compute our elite homogene-ity/absence of divided elite measure as a dichotomous variable taking values of 1, if any of the nine dimensions of homophily records a value greater than 0, and 0 otherwise. If the tenth dimension—measuring no shared class characteristics of the ruling elites—is coded as 1 and the remaining nine homophily variables as 0, our elite homogeneity value has been coded as a 0. Since this is an indirect way to get at elite divisions, we are also considering the economic variable “trade openness” as a supplementary proxy measure for “divided elites.” Cur-rent research shows that trade openness, an economic measure of a country’s participation in international trade, is associated empirically with elite divisions in authoritarian regimes, although the mechanism of causation is not identifi ed. Scholars suggest that the correlation between divided elites and protectionism (i.e., low values of trade openness) may stem from the inability of a heterogeneous ruling class to cooperate on upholding and enforcing the property rights and, more generally, the rule of law (Goldstone et al. 2000:10–11; Hegre 2000; Solingen 1998; Osa and Corduneanu-Huci 2005:192–193).

To operationalize “infl uential ally,” we combine three SFTF variables: the number of a country’s memberships in universal, intercontinental and regionally defi ned organizations at a given moment in time. We total the number of all international organization memberships and dichotomize the new variable. If the level of participation in interna-tional organizations exceeds the sample mean (28.8), the variable takes value of 1; otherwise, it is coded as 0. As with “divided elites,” we can only measure the impact of infl uential allies on mobilization indirectly. The argument for considering an authoritarian regime’s participation in international organizations as a proxy for the “infl uential ally” is as follows: a non-democracy’s participation in IOs raises the salience of domestic human rights protections and the operation of the rule of law as the regime’s performance comes under greater international scrutiny. As violations and other abuses of autocrats become more



widely known, the international organizations (or sub-divisions within them) can become potential allies for human rights and other activists.5 For example, Thomas has shown that non-democracies’ participation in an international normative framework articulating “respect for human rights and fundamental freedoms,” creates a “Helsinki effect” which strengthens democratic social movements and undermines the authoritarian state (2001:4). Non-democratic regimes participate in international institutions to enhance their status and legitimacy; however, the resulting international agreements “tend to entrap governments in commitments that exceed their original intentions or contradict their ideological claims” (Thomas 2001:266).

The social networks variable is the most diffi cult one to measure across populations. Given the tendency of mobilization networks to be multiplex, the diffi culty of identifying and coding all possible forms of mobilizational infrastructure, and the paucity of quantitative data, we can only indirectly measure the presence of social networks in anti-regime mobilization. Since we assume that each episode of contention will provide opportunities to create or strengthen network connections, we operationalize this factor as the lagged value of the dependent variable (number of mobilization events). The variable values were determined using the event counts for our fi fteen countries over the twenty-eight year period to create a pooled time series (unbalanced panel model).

The fi ve dimensions of political opportunity in non-democracies are important for anti-regime mobilization, we have argued. But we have not been able to test other independent variables within the structure of the QCA method. To challenge our POS thesis, we have included two other sets of variables—economic conditions and other political conditions—in the maximum likelihood estimation model. The eco-nomic factors that could affect social mobilization are: trade openness, GDP per capita, population size, GDP growth, infl ation and inequal-

5 On the other hand, a selection effect may be at work here: the number of claims against the autocrat and the probability of participating in international organizations might have a common latent cause. In this case, one would expect the two variables to be correlated: for example, the weaker the autocrat, the higher the level of international participation of a regime, and, in the same time, the higher the marginal probability of a mobilization event. If the selection bias were severe, a Heckman procedure would be appropriate to correct for this problem (Heckman 1979). However, when the selection bias is not extreme—which we believe to be the case here—OLS (or MLE) outperforms the Heckman technique (Stolzenberg and Relles 1997).



ity (Osa and Corduneanu-Huci 2005; Schock 1996). Since reliable and systematic data for infl ation and inequality are unavailable for a majority of observations, we test only the fi rst four variables. The trade openness variable is a special case. In addition to capturing elite homogeneity (as described above), trade openness has been found to impact internal political alignments and political stability (Goldstone et al. 2000; Katznelson and Shefter 2002; Milner 1997; Milner and Kubota 2005; Rogowski 1989). Consequently, the trade openness measure does “double duty” in this model. Finally, the two political variables control for the level of democracy, as well as for the periods of prolonged transition between succeeding governments (Marshall and Jaggers 2000).

Social mobilization = exp (β0 + β1* Lag Social mobilization + β2* Repression + β3*Censorship + β4* Divided elites + β5*Infl uential ally + β6* Log GDP/capita + β7* Log population + β8* Growth + β9* Trade openness + β10* Democracy score + + β11* Transition dummy + ε)

The negative binomial regression model tests the infl uences of our fi ve POS/nondemocracy variables, and controls for other (potentially rel-evant) economic and political conditions on four dependent variables.6 These statistical controls are included to assess the internal validity of hypothesized causation so that we can determine whether—net of the other economic and political variables’ effect on our dependent vari-able—the POS variables are still signifi cant.

The dependent variables in this model disaggregate the types of contention represented by our initial dependent variable from the QCA analysis, the presence/absence of social mobilization. We have two strong fi ndings regarding POS from the regression. First, the statistical analysis supports the relevance of our fi ve POS dimensions. An increase in the level of repression and networks, as well as the presence of infl u-ential allies are signifi cantly and robustly associated with an increase in the total number of mobilization events in the fi fteen countries. (See Table 3.) Repression is statistically signifi cant for all four dependent

6 We performed a series of alternative estimations for the same model: fi xed-effects, negative binomial with clustered standard errors, GEE population-averaged model, etc. to control for fi xed unit effect, heteroskedacity and AR(1) autocorrelation. We have also tested an alternative specifi cation with fi rst lags on the potentially endogenous variables: net repression and media access. The results presented here are generated by the random-effects estimation technique that was found to outperform fi xed effects when submitted to a Haussman test.



Table 3. Negative Binomial Regression Model (MLE)

Dependent variables

Independent variables

DV ISocial mobilization (total event count)

DV IIProtests/Anti-government demonstrations (event count)

DV IIIStrikes(event 1count)

DV IVRiots(event count)

Lag Social Mobilization(Proxy for Social Networks)

.002*** (.0005)

.003*** (.0009)

.0005(.002)

.003***(.0009)

Net incidence of Repression(Proxy for Dynamics of Repression)

.02***(.002)

.007*** (.002)

.02*** (.003)

.03*** (.002)

Net incidence of Censorship(Proxy for Media Access)

–.04***, .04 (ZINB)(.02)

.04**

(.02)

–.01

(.03)

–.08***, .02 (ZINB)(.02)

Elite Homogeneity(Proxy for Divided Elites)

.15, –.27* (ZINB) (.14)

.29

(.19)

–.50**

(.22)

.08

(.20)Participation in international organizations(Proxy for Infl uential Ally)

.37*** (.13)

.07 (.17)

.70*** (.27)

.26 (.19)

Log GDP/capita .28

(.24)

1.28***

(.30)

1.23**

(.52)

.85**, .65 (ZINB) (.34)

Log Population .44***

(.14)

.53***

(.18)

.26,

.55**(ZINB) (.29)

.36*

(.20)Growth –2.75***

(.85)–2.32***, –.93(ZINB)(1.18)

–3.34** (1.62)

–2.58** (1.23)

Trade openness(also proxy for Divided Elites)

–.007*, –.001(ZINB)(.004)

.003, .01**

(.005)

–.02***, .008(ZINB) (.008)

–.01**, –.005(ZINB) (.006)

Polity democracy score .06*** (.009)

.04*** (.01)

.046*** (.02)

.07*** (.01)

Transitional period –.21, –.56*(ZINB)(.20)

.49*, .13(ZINB)

(.25)

–.17

(.35)

–.20

(.28)Constant –5.08***

(1.40)–9.66*** (1.80)

–7.39** (2.96)

–7.29*** (2.02)

Number of observations 365 365 365 365Log-likelihood –1259.06 –790.58 –460.90 –680.85

Note: random-effects negative binomial raw coeffi cients reported; standard errors shown in parentheses; *** p<0.01; ** p<0.05; * p<0.1.We are also reporting the coeffi cients and their signifi cance generated by the Poisson process of the Zero-infl ated negative binomial model (ZINB) in those cases in which these results diverged from the standard binomial model.



variables; the networks proxy variable and media access/censorship are signifi cant for all dependent variables except for the number of strikes.7 The results obtained for the divided elites proxies are less clear cut. The elite homogeneity measure is statistically signifi cant at the 0.05 level for strikes (DV III), and signifi cant for total mobilization events (DV I) when estimated with a zero-infl ated negative binomial model (ZINB). When we consider the trade openness measure as a proxy for elite divisions, divided elites becomes signifi cant for total events, strikes and riots. ZINB estimation indicates that divided elites were signifi cant for protests, but not for the remaining three dependent variables. Measure-ment error stemming from the inexact fi t between the concept and the operationalization contributes to this lack of robustness in alternative estimation models. Nevertheless, the overall results for our fi ve POS non-democracy variables provide a strong statistical confi rmation for the selection of these independent variables for the QCA algorithm.

A second fi nding from the MLE regression suggests the importance of disaggregating different forms of contention in order to identify various opportunity structures associated with them. For example, the effects of social networks and media access/censorship are signifi cant for protests and riots, but not strikes (Table 3). It may be that protests and riots need the routine of previous events for network formation, whereas strikes can spontaneously generate the necessary infrastruc-ture for dissent due to preexisting workplace relations. Equally likely, however, is the possibility that the operationalization of the networks variable is insuffi ciently sensitive to pre-existing forms of organization. Further, elite divisions seem to be more important than media access for strike opportunities, while the total social mobilization events (DV I) and strikes (DV III) are the only forms of contention associated with a regime’s increased visibility and participation in the international arena (the infl uential ally proxy). Since our conclusions can apply only to the country cases in our POSND data set (Osa and Corduneanu-Huci 2001), more data are needed for social movement scholars to test these results and generate new hypotheses.

7 Media access and elite homogeneity are unstable across different techniques of estimation, possibly because of the large percentage of 0s among its total values. For example, when estimated with a zero-infl ated negative binomial model, media access changes sign and signifi cance for social mobilization and riots, and elite homogeneity changes sign for social mobilization (see Table 3).



Now let us consider the control variables. The four economic and two political variables were included in the full model in order to estimate the effects of all theoretically relevant factors affecting social mobilization in non-democracies. Political economists argue that per capita income, population size, and GDP growth will affect the mobilizing potential of citizens in a non-democratic regime (Alesina et al. 1996; Hibbs 1973; Huntington 1968; Londregan and Poole 1990; Przeworski et al. 2000; Venieris and Gupta 1983). In the MLE analysis, each of these factors is signifi cant for the three types of mobilizations (DV II, DV III, DV IV). Although the procedure does not allow us to compare the POS variables and the economic variables in terms of relative signifi cance, we can say with certainty that GDP per capita and economic growth have a greater substantive effect on mobilization than the POS variables, giving some support to modernization theories (Huntington 1968; Prze-worski et al. 2000). When the GDP per capita rises, on average, the number of protests, strikes and riots also increase. Further, economic growth and trade openness are inversely correlated with mobilization event occurrence. The faster the economy grows and the more open its trade policies become, the fewer are the number of mobilization events (across all alternative specifi cations of the dependent variables). In short, while the regression shows that the economic factors affect the values on the dependent variable, the POS set is still signifi cant. We conclude that, while economic conditions can increase the salience of anti-regime claims, mobilization in non-democracies also needs a set of structural openings in order to occur.

Finally, we consider the two other political variables. Political sci-entists, especially in the democratic transitions literature, have argued that tipping points are important for social mobilization in a non-democratic context (Kuran 1991; Lohmann 1994). Two variables are used to measure this possibility. First, Polity IV codes regimes in terms of their democratic characteristics. The democracy score is a categori-cal variable (range: 0 to 20) that comprehensively assesses the type of regime. Values greater than ten are democracies, and below or equal to ten are autocratic forms of government. Our selected countries are generally nondemocracies, so they are in the lower range but they can fl uctuate over time. Results from the MLE analysis (Table 3) indicate that the closer a country is to the democratic transition (the threshold value of ten), the higher are the number of mobilization events. The second political variable codes whether the mobilization events occur during an interregnum period (between leaders or regimes) or not. The



effect of the prolonged government transition dummy is less clear-cut, showing an impact only on protest demonstrations, although the coef-fi cient is not particularly robust.

To sum up: Maximum likelihood estimation analysis has confi rmed the signifi cance of the fi ve political opportunity variables necessary for mobilization in non-democracies: divided elites, infl uential ally, media access, repression dynamics and social networks. Testing the nega-tive binomial regression model with four dependent variables shows the utility of disaggregating types of mobilization when considering political contention in authoritarian regimes. The full model shows that economic factors do contribute to mobilization but are not independent of political opportunities. The level of democracy does have an effect on mobilization occurrence but the interregnum variable does not.

Combining Methodological Approaches

Improved understanding of causality in social science has been accom-panied by convincing calls for methodological pluralism (Mahoney 2007). Our interest in the multiple interactions of political conditions for social mobilization initially drew us to the QCA method. The big surprise of our foray into standard statistics was to see the strong complementarity between the QCA and MLE approaches. The regres-sion analysis allows us a stronger basis for asserting the relevance of the fi ve POS dimensions proposed in our 2003 article, “Running Uphill.” The MLE model also shows the importance of disaggregating the outcomes into types of mobilization for a fuller understanding of the causal confi gurations. But QCA more than holds its own, methodologi-cally speaking. First, the QCA method does things that the statistical analysis cannot. Regression has well-known problems with degrees of freedom, multicollinearity and interpretability of results when trying to deal with multiple interaction terms—the area where QCA excels. For example, in order to include all the interaction terms among seven preconditions, a statistical regression model would include no fewer than one hundred variables, need millions of data points and lose degrees of freedom (Aiken and West 1991; De Marchi 2005; Jaccard and Turrisi 2003; Ragin 1987). Second, since MLE analysis is mainly interested in the marginal probability of event occurrence, it cannot account for the clustering of events under the same larger wave of mobilization. By contrast, QCA provides relevant information on the structure of opportunities at the onset of the event sequence. Finally,



App

endi

x 1:

Pol

itica

l Opp

ortu

nity

Dat

aset

Cas

e #

ID c

ode

Cou

ntry

Tem

pora

l spa

n of

the

case

Ind

epen

dent

var

iabl

esD

epen

dent

va

riab

le

Dyn

amic

s of

R

epre

ssio

n

Acc

ess

to

Med

ia

Infl u

entia

l A

llyD

ivid

ed

Elit

esSo

cial

N

etw

orks

Soci

al

Mob

iliza

tion

1.

SP52

Spai

n 19

52 to

195

51

00

00

02.

E

G53

Eas

t Ger

man

y19

53–1

to

1954

–30

1 0

10

1

3.

PL56

Pola

nd19

55 to

195

80

10

11

14.

PO

58Po

rtug

al19

58–1

to

1958

–31

01

11

1

5.

PL60

Pola

nd19

60 to

196

50

01

00

06.

PO

62Po

rtug

al19

61–1

to

1962

–41

01

11

1

7.

SP62

Spai

n19

62–1

969

01

11

11

8.

SA64

Sout

h A

fric

a19

641

00

00

09.

G

R68

Gre

ece

1967

–1 to

19

69–4

01

11

11

10.

HO

68H

ondu

ras

1968

–197

60

11

11

111

. PL

68Po

land

1966

to 1

970

10

11

11

12.

CH

73C

hile

1973

to 1

976

10

10

11

13.

UR

73U

rugu

ay19

73 to

198

11

00

00

014

. A

R77

Arg

entin

a19

76 to

197

91

10

01

1



. SA

76So

uth

Afr

ica

1976

–2 to

19

77–2

11

00

11

16.

BR

77B

razi

l19

77 to

197

80

01

01

117

. R

O77

Rom

ania

1977

–1 to

19

77–4

10

00

11

18.

RO

78R

oman

ia19

78 to

198

01

00

00

019

. IR

80Ir

an19

79–1

to

1981

–10

01

11

1

20.

CH

83C

hile

(com

bativ

e po

blac

ione

s)

1983

to 1

987

01

11

11

21.

CH

83–

lcp

Chi

le (l

ow

com

bativ

e po

blac

ione

s)

1983

–198

70

10

10

0

22.

RO

87R

oman

ia19

86 to

198

71

00

01

123

. B

U88

Bur

ma/

Mya

nmar

1988

–1 to

19

88–3

11

00

01

24.

CN

89C

hina

1989

–20

11

11

1



when different tests of signifi cance are applied to the MLE model and generate contradictory results, QCA can help with the interpretation. For example, the Vuong test of model selection indicated that a zero infl ated negative binomial model be considered as an alternative to the standard negative binomial model. (Beta coeffi cients are displayed in Table 3.) For media access and elite homogeneity, the coeffi cients change sign and signifi cance across the two specifi cations (for DV I and DV IV in the fi rst case, and DV I for the latter). The lack of robustness of the average effect on the dependent variables still preserves useful informa-tion about possible signifi cance, but it complicates the interpretation of the results. On the other hand, the QCA Boolean logic allows for the specifi cation of all the confi gurations (and cases) under which the presence or absence of the condition is necessary and/or suffi cient for mobilization. In short, QCA complements the average effect analysis by further specifying the necessary and suffi cient conditions among those statistically signifi cant facilitators of mobilization in non-democracies.

References

Aiken, Leona S., and Stephen G. West. 1991. Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: Sage Publications.

Alesina, Alberto F., Sule Özler, Nouriel Roubini, and Phillip Swagel. 1996. “Political Instability and Economic Growth.” Journal of Economic Growth 1:189–211.

Appendix 2: Model I—Four Political Opportunity Dimensions

Model: SOCMOB = DYNREPR + MEDACC + INFLALLY + DIVELITOutputs Minimized: 1

Prime Implicantssocmob = 1

N = 17

1st dynrepr INFLALLY DIVELIT 1,2 or

2nd DYNREPR medacc INFLALLY or

3rd DYNREPR MEDACC infl ally divelit

1 Lower case notation indicates absence of the condition; upper case notation indicates presence of the condition.2 DYNREPR = Dynamics of RepressionINFLALLY = Infl uential AllyMEDACC = Media AccessDIVELIT = Division of Elites



Almeida, Paul D. 2003. “Opportunity Organizations and Threat-induced Contention: Protest Waves in Authoritarian Settings.” American Journal of Sociology 109:345–400.

Amenta, Edwin, and Jane Poulsen. 1994. “Where to Begin: A Survey of Five Approaches to Selecting Independent Variables for Qualitative Comparative Analysis.” Sociological Methods and Research 23, 1 (August): 22–55.

Ashwin, Sarah. 1999. Russian Workers: The Anatomy of Patience. Manchester: Manchester University Press.

Bayat, Asef. 1997. Poor People’s Movements in Iran. New York: Columbia University Press.

Becker, Howard S. 1998. Tricks of the Trade: How to Think About Your Research While You’re Doing It. Chicago, IL: University of Chicago Press Chicago, Ill.

Bob, Clifford. 2002. “Political Process Theory and Transnational Social Movements: Dialectics of Protest among Nigeria’s Ogoni Minority.” Social Problems 49:395–415.

Bollen, Kenneth A., Barbara Entwisle, and Arthur S. Alderson. 1993. “Macrocompara-tive Research Methods.” Annual Review of Sociology 19:321–351.

Boswell, Terry, and Cliff Brown. 1999. “The Scope of General Theory: Methods for Linking Deductive and Inductive Comparative History.” Sociological Methods & Research 28:154–185.

Boudreau, Vincent. 2004. Resisting Dictatorship: Repression and Protest in Southeast Asia. New York: Cambridge University Press.

Brockett, Charles. 1994. “The Structure of Political Opportunities and Peasant Mobi-lization in Central America.” In Social Movements in Latin America. The Experience of Peasants, Workers, Women, the Urban Poor, and the Middle Sectors, edited by J. Dominguez. New York and London: Garland Publishing, Inc.

——. 1995. “A Protest-Cycle Resolution of the Repression/Popular-Protest Paradox.” Pp. 117–144 in Repertoires and Cycles of Protest, edited by Mark Traugott. Durham: Duke University Press.

——. 2005. Political Movements and Violence in Central America. New York: Cambridge University Press.

Calhoun, Craig. 1994. Neither Gods nor Emperors: Students and the Struggle for Democracy in China. Berkeley: University of California Press.

Crowley, Steven. 1997. Hot Coal, Cold Steel: Russian and Ukrainian Workers from the End of the Soviet Union to the Post-Communist Transformations. Ann Arbor: University of Michigan Press.

De Marchi, Scott. 2005. Computational and Mathematical Modeling in the Social Sciences. Cambridge, MA: Cambridge University Press.

De Meur, Gisèle, and Benoît Rihoux. 2002. L’Analyse Quali-Quantitative Comparée: Approche, Techniques et Applications en Sciences Humaines. Louvain—La—Neuve, Belgium: Academia Bruylant.

Deletant, Dennis. 1998. Ceauçescu si Securitatea. Constrangere si Disidenta in Romania Anilor 1965–1989. Bucharest: Ed. Humanitas.

della Porta, Donatella, and Mario Diani. 1999. Social Movements: An Introduction. Oxford, UK: Blackwell Publishers.

Dennis, Norman. 1969. Coal Is Our Life: An Analysis of a Yorkshire Mining Community. London and New York: Tavistock Publications.

Diani, Mario. 1992. “Analyzing Social Movement Networks.” In Studying Collective Action, edited by M. Diani, and R. Eyerman. Newbury Park and London: Sage Publications.

——. 1996. “Linking Mobilization Frames and Political Opportunities: Insights from Regional Populism in Italy.” American Sociological Review 63:1053–1069.

Drass, Kriss, and Charles C. Ragin. 1986. QCA: A Microcomputer Package for Qualitative Comparative Analysis of Social Data. Center for Urban Affairs and Policy Research, Northwestern University.

Fryer, Peter, and Patricia McGowan Pinheiro. 1961. Oldest Ally. A Portrait of Salazar’s Portugal. London: Dennis Dobson.



Glenn III, John K. 2001. Framing Democracy: Civil Society and Civic Movements in Eastern Europe. Stanford: Stanford University Press.

Goldthorpe, John H. 1997. “Current Issues in Comparative Macrosociology: A Debate on Methodological Issues.” Comparative Social Research 16:1–26.

Goodwin, Jeff. 2000. “Are Protesters Opportunists? Political Opportunities and the Emergence of Political Contention.” Paper presented at the Annual Meeting of the American Sociological Association, Washington, DC.

Heckman, James. 1979. “Sample Bias as a Specifi cation Error.” Econometrica 47:153–161.

Herbstein, Denis. 1979. White Man, We Want to Talk to You. New York: Africana Pub-lishing Company.

Hess, David, and Brian Martin. 2006. “Repression, Backfi re, and the Theory of Transformative Events.” Mobilization 11:249–267.

Hibbs, Douglas A. 1973. Mass Political Violence: A Cross-national Causal Analysis. New York: Wiley.

Hirschsohn, Philip. 1998. “From Grassroots Democracy to National Mobilization: COSATU as a Model of Social Movement Unionism.” Economic and Industrial Democracy 19:633–666.

Hirson, Baruch. 1979. Year of Fire, Year of Ash. The Soweto Revolt: Roots of a Revolution?. London: Zed Press.

Huntington, Samuel P. 1968. Political Order in Changing Societies. New Haven, CN: Yale University Press.

Jaccard, James, and Robert Turrisi. 2003. Interaction Effects in Multiple Regression. Newbury Park, CA: Sage Publications Inc.

Karis, Thomas G., and Gail M. Gerhart. 1997. Nadir and Resurgence, 1964–1979. Vol. 5 of From Protest to Challenge. A Documentary History of African Politics in South Africa, 1882–1990. Bloomington and Indianapolis: Indiana University Press.

Katznelson, Ira, and Martin Shefter (eds.). 2002. Shaped by War and Trade: International Infl uences on American Political Development. Princeton, NJ: Princeton University Press.

King, Peter J. 1989. “Comparative Analysis of Human Rights Violations under Military Rule in Argentina, Brazil, Chile, and Uruguay.” Pp. 1043–65 in Statistical Abstract of Latin America, vol. 27, edited by J.W. Wilkie, and E. Ochoa. Los Angeles: University of California, UCLA Latin American Center Publications.

Kitschelt, Herbert P. 1986. “Political Opportunity Structures and Political Protest: Anti-nuclear Movements in Four Democracies.” British Journal of Political Science 16:57–85.

Klandermans, Bert. 1997. The Social Psychology of Protest. Oxford: Blackwell Publ. Ltd.

Kuran. 1991. “Now Out of Never: The Element of Surprise in the East European Revolution of 1989.” World Politics 44 (October): 7–48.

Liang, Zhang, comp., Andrew J. Nathan, and Perry Link (eds.). 2001. The Tiananmen Papers: The Chinese Leadership’s Decision to Use Force Against Their Own People —in Their Own Words. New York: Public Affairs.

Lohmann, Susanne. 1994. “The Dynamics of Informational Cascades: The Monday Demonstrations in Leipzig, East Germany, 1989–91.” World Politics 47:42–101.

Londregan, John B., and Keith T. Poole. 1990. “Poverty, the Coup Trap, and the Seizure of Executive Power.” World Politics 42:151–183.

Loveman, Mara. 1988. “High-Risk Collective Action: Defending Human Rights in Chile, Uruguay and Argentina.” American Journal of Sociology 104:477–525.

Mahoney, James. 2007. “Approaches to Causation in Comparative Research,” unpub-lished paper. Northwestern University.

Mason, T. David, and Dale Krane. 1989. “The Political Economy of the Death Squads: Towards a Theory of the Impact of State-Sanctioned Terror.” International Studies Quarterly 33:175–98.



McAdam, Doug. 1999. Political Process and the Development of Black Insurgency, 1930 –1970. 2nd ed. Chicago: University of Chicago Press.

Meyer, David S. 2004. “Protest and Political Opportunities.” Annual Review of Sociology 30:125–145.

Mill, John Stuart. 1974. Philosophy of Scientifi c Method. New York: Hafner Press.Milner, Helen V. 1997. Interests, Institutions, and Information: Domestic Politics and International

Relations. Princeton, NJ: Princeton University Press.——, and Keiko Kubota. 2005. “Why the Move to Free Trade? Democracy and Trade

Policy in the Developing Countries.” International Organization 59:107–143.Mueller, Carol. 1999. “Claim ‘radicalization’? The 1989 protest cycle in the GDR.”

Social Problems 46(4): 528–547. Nash, June. 1992. I Spent My Life in the Mines: The Story of Juan Rojas, Bolivian Tin Miner.

New York: Columbia University Press.Noonan, Rita. 1995. “Women Against the State: Political Opportunities and Collective

Action Frames in Chile’s Transition to Democracy.” Sociological Forum 10:81–111.Opp, Karl Dieter, and Wolfgang Roehl. 1990. “Repression, Micromobilization, and

Political Protest.” Social Forces 69:521–547.Osa, Maryjane. 2001. “Mobilizing Structures and Cycles of Protest: Post-Stalinist

Contention in Poland.” Mobilization 6(October): 40–161.——. 2003. Solidarity and Contention: The Networks of Polish Opposition. Minneapolis: Uni-

versity of Minnesota Press.——. 2005. “Linking Economic and Political Opportunities in Nondemocracies,” in

Maria Kousis, and Charles Tilly, eds., Economic and Political Contention in Comparative Perspective. New York: Paradigm Publishers, 171–201.

——, and Kurt Schock. 2007. “A Long, Hard Slog: Political Opportunities, Social Networks and the Mobilization of Dissent in Non-Democracies.” Research in Social Movements, Confl icts and Change 27:123–153.

——, and Cristina Corduneanu-Huci. 2001. Political Opportunity Structures in Non-Democracies Database [computer fi les]. Columbia, SC.

Oxhorn, Philip. 1991. “The Popular Sector Response to an Authoritarian Regime: Shantytown Organizations since the Military Coup.” Latin American Perspectives 18:66–91.

Pfaff, Steven. 2006. Exit-voice Dynamics and the Collapse of East Germany: The Crises of Leninism and the Revolution of 1989. Durham, NC: Duke University Press.

Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub, and Fernando Limongi. 2000. Democracy and Development: Political Institutions and Well-being in the World, 1950 –1990. Cambridge, MA: Cambridge University Press.

Ragin Charles C. 1987. The Comparative Method. Berkeley: University of California Press.

——. 1995. “Using Qualitative Comparative Analysis to Study Confi gurations.” Pp. 177–189 in Computer-Aided Qualitative Data Analysis, edited by U. Kelle. London: Sage Publications.

——. 2000. Fuzzy-Set Social Science. Chicago: University of Chicago Press.Rogowski, Ronald. 1989. Commerce and Coalitions: How Trade Affects Domestic Political Align-

ments. Princeton, NJ: Princeton University Press.Rucht, Dieter. 1996. “The Impact of National Contexts on Social Movement Structures:

A Cross-Movement and Cross-National Comparison.” Pp. 185–204 in Comparative Perspectives on Social Movements: Political Opportunities, Mobilizing Structures, and Cultural Framings, edited by D. McAdam, J. McCarthy, and M. Zald. Cambridge: Cambridge University Press.

Sandoval, A.M. Salvador. 1998. “Social Movements and Democratization: The Case of Brazil and the Latin Countries.” Pp. 169–201 in From Contention to Democracy, edited by M.G. Giugni, D. McAdam, and C. Tilly. Lanham, MD: Rowman and Littlefi eld, Publishers.



Schneider, Cathy. 1991. “Mobilization at the Grassroots: Shantytowns and Resistance in Authoritarian Chile.” Latin American Perspectives 18:92–112.

Schock, Kurt. 1996. “A Conjunctural Model of Political Confl ict: The Impact of Politi-cal Opportunities on the Relationship between Economic Inequality and Violent Political Confl ict.” Journal of Confl ict Resolution 40:98–133.

——. 1999. “People Power and Political Opportunities: Social Movement Mobilization and Outcomes in the Philippines and Burma.” Social Problems 46:355–375.

——. 2005. Unarmed Insurrections: People Power Movements in Nondemocracies. Minneapolis: University of Minnesota Press.

Stolzenberg, Ross M., and Daniel A. Relles. 1997a. “Assessing and Correcting Sample Selection Bias.” American Sociological Review 62:494–507.

Tarrow, Sidney. 1998. Power in Movement. 2nd ed. Cambridge: Cambridge University Press.

Thomas, Daniel C. 2001. The Helsinki Effect: International Norms, Human Rights, and the Demise of Communism. Princeton, NJ: Princeton University Press.

Vasi, Ion B. 2001. “The Fist of the Working Class: The Social Movements of Jiu Valley Miners in Post-Socialist Romania.” Manuscript, Department of Sociology, Cornell University.

Venieris, Yiannis P., and Dipak K. Gupta. 1983. “Sociopolitical and Economic Dimen-sions of Development: A Cross-Section Model.” Economic Development and Cultural Change 31:727–756.

World Handbook of Political and Social Indicators III, 1948–1982. 1985. Charles L. Taylor, Principal Investigator. 2nd ed. Inter-University Consortium for Political and Social Research, Ann Arbor, MI.

Yun, Seongyi. 1997. “Democratization in South Korea: Social Movements and their Political Opportunity Structures.” Asian Perspective 21:145–171.

Zuo, Jiping, and Robert D. Benford. 1995. “Mobilization Processes and the 1989 Chinese Democracy Movement.” Sociological Quarterly 36:131–156.


Running Uphill: Political Opportunity in Non-Democracies. Revised, with a new methodological...

Documents

Transcript of Running Uphill: Political Opportunity in Non-Democracies. Revised, with a new methodological...