Nonviolence in a Time of Interconnectedness
A study on globalization's effect on the outcome of nonviolent campaigns
Mikaela Hansson
Peace and Conflict Studies C Bachelor Thesis
Supervisor: Annekatrin Deglow
Department of Peace and Conflict Research Uppsala University
Autumn 2019
Introduction ... 4
Previous Literature ... 6
The outcomes of nonviolent campaigns ... 7
Theoretical framework ... 10
Globalizations effect on society ... 10
Globalization and the success of nonviolent campaigns ... 11
Research design ... 14
The dependent variable ... 15
The independent variable ... 16
Control variables ... 16
Validity & Reliability ... 18
Method ... 19
Results & Analysis ... 20
Descriptive analysis ... 20
Results & Analysis ... 21
Robustness check ... 27
Discussion ... 28
Summary & Conclusion ... 30
Bibliography ... 30
Appendix ... 35
KOF Globalization Index 2012 ... 35
Robustness check ... 36
Figure 1: Visualization of Causal Mechanism ... 14
Figure 2: Predicted Probability of Success for a Nonviolent Campaign ... 24
Figure 3: Predicted Probability of Success for a Nonviolent Campaign after the Cold War ... 26
Table 1: Descriptive statistics ... 20
Table 2: Logistic Regression ... 21
Table 3: Predicted probability of Success ... 23
Table 4: The predicted Probability of Success after the Cold War ... 25
Table 5: Descriptive Statistics robustness check ... 36
Table 6: Logistic regression robustness check ... 36
Introduction
Demonstrations, protests, and strikes are all actions to challenge the power of the regime. In this writing moment, nonviolent campaigns are ongoing in Hong Kong, Chile, and Spain and some are even arguing that the occurrence of the second wave of the Arab spring is about to begin due to new protests in Sudan, Algeria, Lebanon and Iraq (Fahmi, 2019). Nonviolent and violent actions have been used throughout history to alter political progress, but it is only recently that the scope of the two phenomena has been compared to each other in the question of efficiency and what causes its onset (Gleditsch & Rivera, 2015; Karakaya, 2016; Stephan & Chenoweth, 2008; Sutton, Butcher, & Svensson, 2014). Major nonviolent campaigns have shown to be more successful than violent campaigns, as the cost to participate in a nonviolent campaign is lower than in a violent campaign, making widespread, crosscutting and decentralized mobilization possible (Stephan & Chenoweth, 2008).
But the world develops, and our societies change. The world has become more connected and the flow of goods and information has increased as we have become more globalized.
A positive correlation between globalization and the onset of nonviolent campaigns has been established (Karakaya, 2016), but this pattern has not been investigated for the outcomes of nonviolent campaigns. That is the puzzle this thesis will try to solve with the research question:
does globalization affect the probability to succeed in a nonviolent campaign? A nonviolent campaign will be considered successful when 100% of its stated goals are fulfilled and the nonviolent campaign was the main driving factor behind the success of reaching those goals.
In previous research on nonviolent campaigns, it has been established that there are three key aspects to a successful nonviolent campaign, which are mobilization, leverage, and resilience. All these aspects need to be in place for a nonviolent campaign to be successful (Schock, 2013;
Stephan & Chenoweth, 2008). With the use of these three key aspects, the patterns on how nonviolent campaigns are successful are starting to be explored, but there is still little research done on how different factors can affect a campaign’s likelihood of fulfilling these three key aspects. The effect on a nonviolent campaign’s outcome by geographical location (Butcher, 2017), the emergence of radical flanks(Tompkins, 2015), and regime type (Celestino & Gleditsch, 2013) are the ones that have been explored. For all these examples, these surrounding conditions have affected the possibility of the nonviolent campaigns to create and fulfill these key aspects, which have been directly connected to the outcome of the nonviolent campaign. Only two
studies have studied the effect globalization has on the onset of nonviolent campaigns, whereas both studies found a positive relationship between the two phenomena. To my knowledge, no study has explored the possible relationship between globalization and nonviolent campaign outcomes. By filling that gap, another puzzle piece will be added to the understanding of the dynamics of nonviolent campaigns.
As there are several effects of globalization on society that can be directly linked to the creation and fulfillment of the three key aspects, I argue that nonviolent campaigns are more likely to succeed in a more globalized country due to the assets’ globalization bring society. Globalization increases the information structure in the country and provides the campaign with more ways to reach possible new supporters both domestic and international. It also provides the campaign with information on other campaigns' choice of tactics and methods and how it affected the campaign concerning the outcome. The liberal norms that come with globalization will favor nonviolent action over violent action, providing an environment for the campaign were
nonviolent tactics are presented as the most favorable option compared to violent ones, making the usage of nonviolent actions more likely. Liberal norms also bring tolerance, creating a possibility for a nonviolent campaign to mobilize a bigger following over the boarders of different groups in society. Tolerance can also create leverage, as with a more tolerant view towards society, actions performed by the regime to shatter the nonviolent campaign can be overcome as the campaign is not easily disrupted by a division of groups within the nonviolent campaign itself.
Analyzing all the major nonviolent campaigns between 1970 to 2006 through a multivariate logistic regression, a positive relationship between globalization and the probability of a successful outcome for nonviolent campaigns was found. The results do come with a caveat relating to the generalizability of the result. The analysis is based on a relatively small sample size which gives a large standard deviation, leading to a wide confidence interval of what the predicted probability of success for a nonviolent campaign would be for different levels of globalization.
The predicted probability of success for a nonviolent campaign increased from 42% to 95% from the minimal to the maximal value of globalization in the sample, but the range was for the
minimal value 70 percentage points and 36 percentage points for the maximal value. A
preformed robustness check, with a change in the operationalization of the dependent variable, produced contradicting results to the main analysis but should be interpreted with caution due to
statistical weaknesses in the analysis. But the robustness check has still left interesting question marks for future research.
In the pursuit of exploring the relationship between globalization and the outcome of nonviolent campaigns, this thesis will be structured as follows. First, a summary of the previous literature defining the field of nonviolent campaigns. Secondly, a theoretical framework will be presented connecting the concept of globalization to the success of nonviolent campaigns. Then the research design will be discussed together with the operationalization of the variables included in this study. Followed by the analysis of the results connecting back to the theoretical framework, and lastly, a summary and conclusion of the thesis results and contribution to the field of nonviolent campaigns.
Previous Literature
Nonviolent campaigns have been defined in several different ways, but the main components from the different definitions can be combined as: unarmed civilians organizing to challenge regime authority with the use of nonviolent action through protest and persuasion,
noncooperation, and intervention, without the use of threats or physical harm towards the opponent (Chenoweth & Cunningham, 2013; Schock, 2013; Stephan & Chenoweth, 2008;
Svensson & Lindgren, 2010). This definition is derived from Gene Sharp’s foundational work on nonviolent campaigns in his trilogy The Politics of Nonviolent Action (1973) but has also been used in the research on revolutions, social movements, and civil resistance (Schock, 2013). All these fields have contributed to the research on nonviolent campaigns but due to different assumptions which their definitions are constructed upon, the fields have enhanced different aspects. The studies on revolutions and social movements generally build on the idea that nonviolent
campaigns are a part of a chain reaction, starting with contentious politics followed by nonviolent resistance to end in violent resistance if the perceived goal has not been fulfilled. Civil resistance studies distinguish themselves here with the notion that nonviolent and violent actions are opposites, as depending on which action used, the advantage will lie in the hands of the civilians or the regime. With nonviolent actions, the comparative advantage lies in the hands of the civilians (ibid).
The outcomes of nonviolent campaigns
A nonviolent campaign has been defined as successful when it fulfills 100% of its stated goals and is the driving factor behind the success of reaching these goals. Even though the field of nonviolent resistance builds on different assumptions, the different branches have agreed on three key aspects a nonviolent campaign needs to fulfill and create to be successful; mobilization, resilience, and leverage (Schock, 2013). Mobilization is the process to acquire resources.
Resilience is the ability to bounce back from repression by the regime, but also the ability to use different kinds of actions when needed. Leverage indicates the movement’s capacity to alter the power balance between themselves and the regime by affecting the necessary power supply for the opponent (Gleditsch & Rivera, 2015; Schock, 2013; Stephan & Chenoweth, 2008).
Mobilization is often highlighted as the most important factor, but without the possibility to be resilient when faced with resistance, leverage is hard to gain over one's opponent, leading to lower possibilities of success for the overall campaign. This conclusion has been strengthened by Stephan & Chenoweth (2008), who were the first to test these key factors and compare the success rate of nonviolent campaigns to violent campaigns in a large-N study. In their study, they conclude that 53% of nonviolent campaigns successfully reach their goals compared to 23% of violent campaigns. Stephan and Chenoweth (2008) conclude that the strengths of nonviolent campaigns lie in their ability to mobilize in a crosscutting, widespread and decentralized manner while being resilient against repression, but also at the same time create leverage in the form of loyalty shifts from the regime to the campaign. A main reason why nonviolent campaigns manage to fulfill these aspects to such a degree that they succeed more than twice the amount the times violent campaigns succeed is argued to be the lower cost to join a nonviolent campaign as you do not risk your life in the same extent by joining a nonviolent campaign (Stephan & Chenoweth, 2008). The incorporation of a violent flank into a nonviolent campaign has shown to lower the participation level for the campaign (Chenoweth & Schock, 2015) further supporting the argument that violent actions increase the cost for participation.
For mass mobilization to be possible, especially cross-cutting, widespread and decentralized mobilization, the media has been pointed out as an important factor (Stephan & Chenoweth, 2008; Sutton et al., 2014). Sutton, Butcher & Svensson (2014) point out the process of political jiu-jitsu, which is when a regime faces backlash after using violent repression towards the campaign. They highlight especially impartial media as important in this process to make mobilization possible again after violent repression by the regime to silence a protest. Impartial depictions of the violent repression event give the campaign the possibility to gain leverage over
the regime, as using violence against the population is not seen as legitimate by human rights legislation. This shift of legitimacy, from the regime to the nonviolent campaign, is argued to be vital to the correlation found between backlash against the regime and increased mobilization for a nonviolent campaign (Stephan & Chenoweth, 2008). A specific group that the nonviolent campaigns usually do not reach is the regime’s personnel. But in these situations of backfiring when legitimacy is shifted, it is possible to reach them through medias depiction of the
repression. The possibility to reach and mobilize this group is important to create leverage for the campaign (ibid).
To create leverage for a nonviolent campaign, loyalty shifts within the regime has shown to be crucial. Such as the mobilization of the military and the bureaucracy from the regime to the campaign. With loyalty shifts, the legitimacy of the regime will be undermined and transferred to the campaign as the regime's source of power - their control of their people and system - has been obstructed (Celestino & Gleditsch, 2013; Stephan & Chenoweth, 2008: 42). International involvement can also affect the leverage of a nonviolent campaign over their opponent. In the situation of violent repression by the regime towards the nonviolent campaign the risk of sanctions from international actors increases due to the regime's legitimacy being questioned for using violence against their people. The sanctions in themselves can undermine the power position of the regime by restricting their availability of resources and obstructing trade to gain financial power. Power loss by the targeted regime of the campaign is power won for the nonviolent campaign (Stephan & Chenoweth, 2008: 43).
The last key aspect for a positive outcome for a nonviolent campaign is resilience, the ability to strategically choose the best fitting action for the time being, but also the process of being able to bounce back when faced with hardship. In the process of creating resilience for a nonviolent campaign, education is argued to be an invaluable resource. With education, information on historical campaigns, their tactics used, and the following outcome of these decisions can be analyzed and utilized for the benefit of the ongoing nonviolent campaign. Resilience can be created through education as inspiration on strategic tactics can be drawn from empirical
examples (Stephan & Chenoweth, 2008: 43). The other end of resilience focuses on the ability to bounce back after hardship, such as violent repression from the regime. To be able to be resilient in the situation of repression, the nonviolent campaign needs to be able to uphold their unity as a movement against their opponent but also needs to be able to publicize their perspectives to undermine the legitimacy for the regime based on their repressive action. Svensson and Lindgren
(2010) have studied how different goals of a campaign affect its identity, which in turn affects its outcome. In a nonviolent campaign, the legitimacy of the state is questioned, and different goals question this legitimacy in different ways. When the goal is territorial change, the horizontal legitimacy of the state is questioned, and thereby also its identity. In comparison, the goal of a regime change challenges the vertical legitimacy of the state, which is challenging the current way of handling the state apparatus. In a society divided along ethnic lines, success will be harder to achieve for a nonviolent campaign when facing goals connected to identity (horizontal
legitimacy), as the already existing division of the population’s identity is likely to be strengthened through social-psychological constraining mechanisms which can be triggered by the regime.
Mechanisms that create fear, lack of trust and societal polarization will divide the movement and thereby also hinder the possibility for mass mobilization. Without the power in numbers, the process of withdrawing consent from to regime to create leverage for the campaign will not be as powerful as with mass mobilization (ibid: 110-115).
In several case studies comparing successful and failed nonviolent campaigns, the presence of these three key aspects could be connected directly to the success or failure of the nonviolent campaign. In the cases where the process to mobilize and create leverage and/
or resilience failed throughout the campaign, the goals were never reached. In nonviolent campaigns where these processes to obtain and create these three key aspects succeeded, the campaigns also succeed with their stated goals and were also the main driving factor behind this success (Stephan & Chenoweth, 2008). However, there are additional factors that have been found to affect the campaign’s ability to obtain these three key aspects. The geographical location of the nonviolent campaign is one, where the proximity to the capital of the country affects the probability to successfully create a regime shift. If a large nonviolent campaign occurred in the capital, the likeliness to succeed with a regime transition increased noticeably. The capital is argued to be a symbolic place associated with power regarding both economics and demography.
A nonviolent campaign is argued to be “…effective at exploiting the representational value of this space” (Butcher, 2017: 1467) during the process of undermining the power and legitimacy of the regime, which also can be connected to the three key factors.
The existence of violence, while the campaign is active, has also shown to affect its success.
Violence mostly in the form of a radical violent flank to the campaign, which can be both officially connected to the nonviolent campaign or freestanding, has been studied. The presence of simultaneous violence has produced contrasting effects on campaign success, with some
positive effects short-term, but long-term results being mostly negative. This is due to the negative effects violence has on popular participation, hampering the ability for a nonviolent campaign to obtain mass mobilization (Bayer, Bethke, & Lambach, 2016; Tompkins, 2015).
Theoretical framework
Another possible factor that could have an impact on the success of a nonviolent campaign and the possibilities to achieve mass mobilization, create leverage and resilience is globalization. A positive correlation between globalization and the onset of nonviolent campaigns has been confirmed (Flaten & de Soysa, 2012; Karakaya, 2016), and this thesis will focus on analyzing the effect globalization has on the possibility for a nonviolent campaign to succeed, thereby
contributing to research on the factors affecting nonviolent campaign success.
Globalizations effect on society
Globalization is a thick concept with many facets and includes several different spheres of society: the political, the social, and the economic (Potrafke, 2015: 510). In these three spheres, globalization affects the flow of goods and money, as well as the flow of information, norms, people, and culture (Karakaya, 2016: 316). Thereby, globalization has been defined as the interconnectedness around the globe between various international actors and states in these societal spheres, but also the flow through these connections and spheres. This definition follows the patterns of previous definitions, emphasizing the flow and the connectedness as especially defining, and thereby breaking the assumption that globalization only evolves around capital flow and trade (Flaten & de Soysa, 2012:630; Gygli, Haelg, Potrafke, & Sturm, 2019:544; Potrafke, 2015:510).
As globalization ties together several areas in society, the effects of globalization on society are extensive. In this thesis, the focus of globalization’s effects on society will be divided into the three spheres used for its definition; the political sphere, the economic sphere, and the social sphere. In the political sphere, the biggest effect globalization has had is the increase of interconnectedness between regimes and international actors. As interconnectedness increases between international actors through trade agreements and membership in international
collaborations, the sovereignty over national politics is weakened as more decisions are made on an international level, overriding the national decisions (Karakaya, 2016: 319).
In the economic sphere, globalization brings economic development based on liberalism. With a developed economy, the country will modernize, and based on modernization theory, this modernization will bring urbanization, industrialization, and development of the communication structure and education provision. The emergence of a middle class together with the
advancement of liberal norms will also occur. The concept of liberal norms is mostly focused around the social liberal norms, for example, human rights, in this thesis. This development of society makes the country as a whole richer, increased GDP, but economic development also produces losers as economic development can lead to increased segregation. To announce the discontent produced by modernization, the usage of contentious politics increases as it is more legitimate to challenge the government in a globalized country. For example, protests are seen as legitimate in countries were liberal norms are present, compared to an authoritarian state where protests are not seen as legitimate, as the freedoms of the people mostly are not the central concern for the authoritarian regime. Another reason for increased usage of contentious politics in a globalized country is a decrease in individual attachment towards the nation-state due to increased international involvement in national affairs. This detachment produces a loss in legitimacy for the regime, leaving room for the will of the people to be expressed through contentious politics (Karakaya, 2016: 321-322).
In the social sphere, there are several bigger changes in the process of globalization. Firstly, increased connectedness brings increased flow of goods, ideas, and people. As people travel and settle down in new countries, new perspectives intermingle with new lifestyle norms. Most prominent are the liberal norms that come with globalization. Two prominent parts of liberal norms will be highlighted, which is the denouncement of violence, favoring nonviolent actions, and the need for tolerance towards other people, their culture, and differing opinions. To be able to collaborate and trust other international actors, these norms need to permeate the societies (Karakaya, 2016: 321-322), especially human rights as they are legislated by the United Nations (UN).
Globalization and the success of nonviolent campaigns
The success of a nonviolent campaign is connected to its ability to fulfill and create the three key aspects of success; mobilization, leverage, and resilience. Below I argue that in a globalized country these aspects are more likely to be fulfilled which creates a bigger possibility for the nonviolent campaign to succeed as a whole.
Globalization brings four different tools to enhance the possibility of mobilization. The first tool is the media. With globalization, based on modernization theory, the communication structure develops (Karakaya, 2016: 321-322), and the media will reach more members of society. This means there is a bigger possibility for more people to hear about the campaign through, for example, news updates, leading to more people becoming possible supporters of the campaign. It needs to be highlighted that it is the presence of impartial media that can favor the nonviolent campaign (Stephan & Chenoweth, 2008; Sutton et al., 2014). The media can be manipulated by the regime, thereby the importance of the clarification that the media should be impartial to see this effect. Secondly, with a more developed communication structure, mobilization can also be easier as geographical location or difficulties traveling no longer disrupts the communication of the campaign’s existence and its goals. Increased awareness, while not guaranteeing support, nonetheless increases the probability of recruitment. Thirdly, without geographical restrictions, recruiting supporters across national borders is now possible and thereby the pool of potential supporters is now bigger. The fourth and last tool is the liberal norms that come with
globalization favoring nonviolent actions over violence (Karakaya, 2016: 321-322). Previous work demonstrates a lower cost to join a nonviolent campaign compared to a violent one, providing nonviolent campaigns with an easier task recruiting than violent campaigns (Stephan &
Chenoweth, 2008). This chain of logic should be even stronger if the norms of society explicitly denounce violence and favor nonviolent action when the legitimacy of the regime needs to be questioned. To summarize, the increased possibility of information flow, provided by the technological development globalization has brought, together with the interconnectedness internationally and the liberal norms for nonviolent action, the ability to establish decentralized, crosscutting, and widespread mass mobilization have been made easier as old restrictors have been overcome.
As for leverage, globalization will affect the possibility to create loyalty shifts from the regime to the campaign through international involvement and disruption of the power supply to the regime. Here, international involvement mostly refers to sanctions issued against the target regime of the nonviolent campaign. When a country is more globalized, it will be more connected with the international community through trade deals and joint political projects.
When a regime challenges the international liberal norms, the legislated human rights, in particular, it is more likely that another international actor will denounce the regime and announce sanctions towards it, due to that the regime having broken the norms, and possibly international laws, which the collaboration is built upon (Stephan & Chenoweth, 2008: 43). The
sanctions are used to punish the government for its behavior, leading to a shortage of necessary resources for the regime to keep their power position, but also to a questioning of the regime's legitimacy. If the trust in the regime decreases to such a degree that its employees and allies question the regime's legitimacy, then the possibility for loyalty shifts increases. When loyalty shifts occur, the legitimacy and power of the regime are weakened even further (Celestino &
Gleditsch, 2013; Stephan & Chenoweth, 2008: 42-43). With globalization, the process of creating leverage and decreasing the legitimacy of the regime is now divided between the domestic sphere and the international sphere, attacking the targeted regime from two fronts. As nonviolent campaigns build on questioning the legitimacy of the regime, having an international actor also questioning the regime's legitimacy should increase the likelihood of a successful outcome for the nonviolent campaign as more people are questioning the regime's legitimacy.
The last key aspect is resilience. Through modernization theory, education is expected to increase in a country developing economically, and globalization leads to economic development
(Karakaya, 2016: 321-322). To create resilience for a nonviolent movement, as stated before, education is crucial as it provides the possibility to analyze and utilize the actions and strategies used by previous nonviolent campaigns (Stephan & Chenoweth, 2008: 43). Together with the increased information flow, it should be easier for a nonviolent campaign in a globalized country to make strategic decisions on actions that will favor them and thereby create more leverage for the campaign compared to one without the access to education and better information flow. The last argument for why a nonviolent campaign in a globalized country should be more able to create resilience is connected to the presence of liberal norms in a globalized setting (Karakaya, 2016: 321-321). Liberalism is the ruling ism within globalization and liberalism urges for tolerance. As presented earlier, divisions in society can lower the probability of success for a nonviolent campaign as the campaign cannot withhold their unity when faced with hardship (Svensson & Lindgren, 2010). The tolerance for other people should make it easier for a nonviolent campaign to keep united when faced with difficulties as the differences between people should not mean as much in a society with liberal norms. This argument also connects back to mobilization, as more tolerance for other people should also make it easier to recruit more supporters as people will not be excluded based on, for example, age, religious beliefs, or sexual orientation.
Figure 1: Visualization of Causal Mechanism
With these arguments in mind, for why globalization should increase the possibility to increase mobilization, leverage, and resilience and thereby also the probability of a nonviolent campaign to succeed, I argue that the causal mechanism for this relationship is divided into three different paths which all are important and need to be fulfilled for the perceived outcome to be true. The three paths go from globalization through mobilization, leverage, and resilience which then together affects the probability of a successful outcome for a nonviolent campaign (see Figure 1).
Together with the causal mechanism, the main hypothesis for this thesis will be:
Hypothesis: The more globalized a country is, the more likely is it for a nonviolent campaign to succeed.
Research design
The research goal of this study is to determine what effect globalization has on the probability of nonviolent campaign success. The cases used are major non-state resistance campaigns with maximalist goals, such as self-determination, expelling foreign occupation and regime change, but also a few cases of major non-state campaigns with objectives concerning bigger social changes, like the anti-apartheid campaigns. The data for these campaigns comes from the Nonviolent and Violent Campaigns Outcomes Data Set (NAVCO v.1.1). In the NAVCO 1.1 data set, all
campaigns are coded either as nonviolent or violent, but as it is rare that a campaign is solely one or the other the coding is based on the predominantly used actions throughout the campaign (Chenoweth, 2011a). The cases used in the analysis are the ones coded as nonviolent. The NAVCO 1.1 data set is complemented with data from the Konjunkturforschungsstelle (KOF) index for globalization for the independent variable (Dreher, 2006). The time range for the cases will be from 1970 to 2006 as the NAVCO 1.1 data set includes data from 1900 to 2006 while the KOF index includes data from 1970 to 2019. The created data set for this study includes 65 nonviolent campaigns, were of which 48 of these cases have data on both the dependent and the
Increased mobilization Higher probability of a Globalization Increased leverage successful outcome for
Increased resilience a nonviolent campaign
independent variable. The sample size for this study is small, which has consequences for the statistical power of the analysis, but as the sample of cases in the data set is the whole population of existing cases of major nonviolent campaigns with existing data between 1970 and 2006, the decision has been made that the sample is still large enough to provide evidence on the
relationship between globalization and the success of nonviolent campaigns. The consequences of the small sample size will be discussed in the Methods section below.
The unit of analysis for this study is the peak year of the campaign, where it reached its largest following. The campaign is defined as “… a series of observable, continuous tactics in pursuit of a political objective” (Stephan & Chenoweth, 2008: 16), which is the definition used by the NAVCO 1.1 data set. This distinguishes campaigns from single events, as the actions need to be continuous to be defined as a campaign. For those campaigns where the membership numbers are missing, the year the campaign ended, succeeded or failed, is coded as the peak year, as the majority of the campaigns had their peak in the end year of the campaign (Chenoweth, 2011b:
10-11).
The dependent variable
The dependent variable is the success of a nonviolent campaign. A campaign is defined as successful when it has reached 100% of its stated goals and was the driving factor behind the fulfillment of these goals. Data on the outcomes of the campaigns will be received from the NAVCO 1.1 data set (Chenoweth, 2011a). To be able to confirm if the goal has been fulfilled, the change the goal demands needs to have happened within two years after the campaign ends.
If the two requirements from the definition were fulfilled within two years of the end of the campaign, the campaign will be classified as successful. If no goals where fulfilled, then the campaign will be classified as failed. For those campaigns where the goals were not met but significant concessions were made by the regime as an answer to the demands of the campaign, a third classification is used: limited success. Even though limited success is a “fuzzy category”
(Chenoweth, 2011b: 2), it can be valuable to distinguish these cases from successes and failures to create the nuance of what an outcome is. The variable was originally coded as a categorical variable but due to the methodological reasons concerning reliability and method applicability, discussed below, the dependent variable is recoded as dichotomous for this thesis. Success is coded as 1 and limited success together with failure is coded as 0. The nuance of the outcome for the main analysis, with the classification of limited success, is lost with a binary coding of the
outcome, but this nuance is used for a robustness check of the result from the main analysis (discussed in the method section).
The independent variable
The independent variable for this study is globalization. Globalization has been defined as the interconnectedness around the globe between various international actors and states in the societal spheres of politics, economics, and society, but also the flow of goods, ideas, people, and norms through these connections and spheres. As the definition indicates, globalization is a thick concept and can thereby be hard to fully grasp in a quantitative study as nuance might be lost in the collection and aggregation of data. The data on globalization was collected from the KOF Globalization Index from 2012 (Dreher, 2006). The indices used for this study to measure the concept of globalization, also used by the KOF index, is by dividing globalization into three spheres, the social, the political, and the economic and measure the flow and interconnectedness of each. In the social sphere, the focus lies on personal connectedness, information flow, and cultural proximity. In the political sphere, it is the political globalization that is measured. And for the economic sphere, it is the global perspective of actual flows and restrictions of financial means that are measured. These independent values are then weighted against each other to be combined and then aggregated into the globalization score (see appendix for full details and weighting). In the data set created, the globalization score has been aggregated to create an average across all active campaign years. Before the average was created, the globalization scores were lagged 1 year to control the time order for the dependent and independent variable,
securing the globalization impact to have occurred before the nonviolent campaign end.
Control variables
Three control variables are included in this study as previous research has established the effect of these specific variables on the outcome of nonviolent campaigns (Stephan & Chenoweth, 2008; Svensson & Lindgren, 2010; Tompkins, 2015). All control variables are from the NAVCO 1.1 data set as these variables have been tested several times in different studies and were shown to be possible confounding variables explaining variation in the outcome of nonviolent
campaigns.
The first control variable is violence occurring simultaneously in the same country as the active campaign. This violence can be connected to the campaign as a radical flank, or not be connected to the campaign as violence for another cause. The presence of a radical flank to a nonviolent
campaign has shown to produce a negative relationship with the probability of success for a nonviolent campaign (Chenoweth & Schock, 2015; Tompkins, 2015), and the presence of ongoing violence within the country, whether it is minor armed conflict, inter- or intrastate war, has shown in previous research to have both positive and negative effects for the regime and the nonviolent campaign (Chenoweth & Schock, 2015; Svensson & Lindgren, 2010). The presence of simultaneous violence is coded as a dummy variable with 1 as present and 0 as not present.
The second control variable is the campaign target regime type. Previous research has shown that whether the regime is autocratic or democratic, and whether it is stable or not will have an influence on the campaign. The specific influence of regime type depends on the goal. It has been suggested that democracies should be more tolerant towards the display of dissent, creating bigger possibilities for a campaign to show their dissent publicly (Stephan & Chenoweth, 2008:
19). On the other hand, it is also argued that the instability of non-democracies creates good possibilities for nonviolent campaigns to root. These variations in regime type will affect the possibilities and struggles present for a campaign (Svensson & Lindgren, 2010). Democracies also have an easier time establishing themselves on the global market than autocracies, leading to autocracies having less access to credit than democracies (Beaulieu, Cox, & Saiegh, 2012). The regime type can then affect the effects globalization has on societies. This variation of regime type is measured with the Polity IV scale, scoring countries from -10 to 10, on the year before the end of the campaign.
The last control variable controls for the presence of the Cold War. During the Cold War, there were systemic differences to the global system compared to after the Cold War, as the global system was divided with two leading hegemonies, the USSR and the USA as leading parties.
These power conditions have been shown to lower the probability of nonviolent campaign success during the war (Stephan & Chenoweth, 2008: 24). The Cold War also affects the presence of globalization, as the Cold War was a war of ideologies between liberalism and communism, and as globalization is part of liberalist ideology, the Cold War may have had an impact on the globalization of the communist countries during the Cold War. The Cold War variable is coded into a dummy variable indicating if the campaign was active during the Cold War (between 1949 and 1991) or not (after 1991).
Validity & Reliability
For a quantitative study, external validity is usually set as high due to its possibility to generalize to a bigger population through its random sampling, but this comes at the cost of the internal validity of the results produced, and this study is no exception. As the concepts used in this thesis, especially the thick concept globalization, are defined and operationalized to be able to be quantified, they are simplifications of reality. This is a normal trade-off for quantitative research, but an important caveat.
The dependent variable, the success of nonviolent campaigns, has high validity from theoretical definition to empirical measurement as the success of the campaign is correlated with the fulfillment of the stated goals of the campaign in both theory and measurement. The definition of a successful nonviolent campaign is built on previous research in the theoretical aspect, as well as for the empirical measurement, increasing the validity of the concepts as the definition and operationalization are cumulatively made. The reliability of the dependent variable is also high as the coding for a successful and failed campaign has clear criteria, but as mentioned by the creators of the NAVCO 1.1 data set, the reliability for the coding of limited success is less reliable (Chenoweth, 2011b). The loss in reliability for the classification limited success comes from the less strict criteria of "significant concession", which produces subjective judgments on what significant concessions are. To solve this problem, and strengthen the reliability of the dependent variable, it is recoded to a dichotomous variable. To not lose the nuances of the three classifications of success, the classification of limited success will be used for a robustness check.
For the independent variable, globalization, the reliability is high as all measurements used are clear and easily interpreted. The validity of the concept is generally high, due to the
operationalization of the concept that keeps the connection of interconnectedness and flows as the core attributes of globalization from definition to measurement. However, for a thick concept especially, simplifications will be necessary for it to be possible to quantitively measure the concept. For the concept of globalization, the operationalization of the social sphere, in particular, cultural proximity, can be scrutinized. The operationalization is very western-centric as it is mainly measured through the number of McDonald’s restaurants and IKEA stores present in a country. Even though globalization started in the western part of the world and is a western concept, the concept of globalization is mainly defined as the flow and interconnectedness of different actors around the globe. With only these factors measured it is questionable how well the measurement captures the core attributes of globalization's effect on the flow of culture. To
improve the validity of the measurement of cultural proximity, data on the export of, for
example, films and music could be added as an indication of the flow of popular culture between countries. Due to time restrictions, these additions could not be made here, but the inclusion of more nuanced data could improve the results of future research on globalization.
Lastly, concerning mostly reliability for all variables, are the data sources. The NAVCO 1.1 data comprises different encyclopedic entries, previous research, but also the knowledge and scrutiny of experts for the specific subjects and areas. The biggest problem for the NAVCO 1.1 data set concerning sources is the underreporting of a failed nonviolent campaign, creating a data set with a skewed population of cases towards successful outcomes. This problem exists for violent campaigns as well. To balance source bias, data collection was selective: only mature campaigns, mature here meaning major campaigns, with a following of at least 1000 people are included in the data set (Chenoweth, 2011b). Furthermore, outcome coding is made as strict as possible for what a successful nonviolent campaign is to balance the biased reporting, thereby the choice to code limited success as failure.
Method
The method used to test the hypothesis is a large-N statistical analysis using a logistic regression model. As the dependent variable is coded as a dichotomous variable, whether the campaign succeeds or fails, and the independent variable, a globalization score for each country where the campaign is active, is a continuous variable, a linear relationship cannot be distinguished between the two variables. An increase by one in the value of globalization does not produce an increase of one in the value of the outcome, thereby is the relationship between the variables nonlinear.
With the use of logit transformation and a logistic regression model, the problem of nonlinearity can be solved and an interpretable result can be produced (Pampel, 2000).
The dependent variable was recoded from categorical to dichotomous for this study due to reliability and methodological reasons, as stated above. By creating a dichotomous variable where only success would be counted as success and limited success and failure would be counted a failure, a truer to reality measurement can be made where bias and reliability are accounted for.
But to not lose the nuance of the outcomes, for which data exists, the classification of limited success will be used for a robustness check for the results of the analysis. For the main analysis, limited success is classified as a failure but in the robustness check, the limited success will be coded as success to see if this alters the outcome of the main analysis.
A last, but very important, methodological issue for this study is the sample size of the analysis:
48 cases for both the dependent and independent variable. For a quantitative study, this is a small sample size and has statistical consequences for how generalizable the results will be. As the sample size decreases, the standard deviation of the variables increases, leading to a bigger range of possible values that can be generalized from the results. This affects the certainty of
generalization as the effect of the relationship can take on many different values (Kellstedt &
Whitten, 2018: 157-158; King et al., 2001). A small sample can also affect the statistical power of the analysis, as when the number of cases is lower only larger effects can be found. Smaller effects are harder to detect due to low variation in the sample.
Results & Analysis
Descriptive analysis
Table 1: Descriptive statistics
Statistic N Mean St. Dev. Min Median Max
Success 65 0.538 0.502 0 1 1
Globalization 48 42.123 13.430 20.144 45.383 69.252
Cold War 65 0.523 0.503 0 1 1
Polity IV 56 -3.571 5.257 -9.000 -6.000 9.000
Violence simultaneously 65 0.431 0.499 0 0 1
The descriptive statistics include all variables used in the regression analysis, including sample size and the range and composition of values for each variable (see Table 1). Two variables do not contain data for all cases: globalization and Polity IV. Polity IV has a skewed composition, with most cases on the negative side of the scale, meaning more autocratic than democratic. For the other variables, the composition of the different cases is somewhat even in their distribution, only globalization is a little skewed towards lower scored cases.
Results & Analysis
In the logistic regression analysis, three different models are used (see Table 2). M1: bivariate regression with the dependent and the independent variable to establish if any correlation can be found between the two different variables. M2: a multivariate regression with the three control variables included. M3: a model without Polity IV to see how the correlation between the
dependent and independent variable changes when the control variable which reduces the sample size to only 40 cases are not used.
Table 2: Logistic Regression
Dependent variable:
Successful outcome of nonviolent campaign
M1 M2 M3
(1) (2) (3)
Globalization 0.062** 0.065** 0.058**
(0.025) (0.032) (0.026)
Cold War 0.695 -0.013
(0.915) (0.662)
Polity IV 0.114
(0.086)
Violence simultaneously -0.879 -0.632
(0.762) (0.648)
Constant -2.243** -1.902 -1.761
(1.068) (1.517) (1.265)
Observations 48 40 48
Log Likelihood -29.063 -22.075 -28.578
Akaike Inf. Crit.
Pseudo R2
62.126 0.6093382
54.150 0.7699903
65.156 0.6204891
Note:
*p<0.1; **p<0.05; ***p<0.01 Standard error in parenthesis Pseudo R2 is the Veall-Zimmermann Pseudo R2
M1 finds a positive correlation, which is statistically significant at the 95% confidence level, between globalization and a successful outcome for a nonviolent campaign. Likewise, M2 finds a positive correlation, statically significant at the 95% confidence level, between globalization and a successful outcome for a nonviolent campaign while controlling for possible confounding variables. However, M2 has one caveat. The decreased sample size from 48 to 40. This a loss of 17 % of the cases, which is a considerable decrease in cases. The loss of cases will affect the confidence of the generalizability of the results. But on the positive side, none of the control variables are statistically significant, meaning that none of the control variables can explain the variation in the dependent variable with such certainty that the result can be generalized. Even though the controls are not statistically significant, the effect they do have has been controlled for, making sure that what has been measured as the relationship between globalization and a successful outcome for a nonviolent campaign is explained by the effects of globalization and not the control variables. As the relationship between globalization and a successful outcome of the nonviolent campaign is statistically significant and positive, we can say that globalization affects the probability of success for a nonviolent campaign, with the caveat that all possible
confounding variables have not been tested. The ones included though are the ones that theory and previous research have pointed out as the most important ones that could explain the variation of success for a nonviolent campaign.
M3 was included to explore what effect the decreased sample size could have on the relationship between the independent and dependent variables. No large difference in the outcome of
statistical significance or direction of the effect can be seen comparing M2 and M3. The difference between the models can be found in the goodness-of-fit measures, the Akaike score and the pseudo R2 score. The Akaike score has increased and the pseudo R2 has decreased, both indicating that M3 explains the relationship between the variables not as good as M2, thereby showing that increasing sample size while not controlling for regime type is not favorable for exploring the relationship between globalization and nonviolent campaigns.
Model goodness-of-fit is measured using pseudo R2, producing a value between 0 and 1 measuring how much the independent and control variables can explain the variation in the dependent variable, with a higher value indicating a better fit. M2 is the best fitting model at a value of about 0.77 explaining 77% of the variation.
As all models are logistic regressions, no more than direction and statistical significance can be interpreted from the models presented above as the coefficients are presented in log odds. To further interpret the coefficients, statistical measures need to be taken to transform the
coefficients. As the relationship between globalization and a successful outcome of a nonviolent campaign was statistically significant in the models above, the predicted probability of this relationship was calculated. To calculate the predicted probability for a successful outcome for a nonviolent campaign, three scenarios were created where the level of globalization changes. In the three different scenarios, the value for globalization was set to its minimum, mean and maximum value to explore how the predicted probability for success changes as globalization changes. The Polity IV score was set to its mean of -3.571, the control variable for the Cold War to 1 (active during the Cold War), and the control variable for violence occurring simultaneously to 0 (no violence simultaneously present). The control variables have been set to their mean and mode value to be able to control for their effect on the relationship between globalization and a successful outcome of a nonviolent campaign when the predicted probability is calculated. The mean and the mode value for the control variables where chosen as they represent the average scenario of the values for the control variables in this sample. The predicted probability for success for a nonviolent campaign depending on the level of globalization present is presented in Table 3.
Table 3: Predicted probability of Success
Level of globalization Min Mean Max
Predicated Probability of
success 42% 75% 95%
The probability of success for a nonviolent campaign increases with the increase of globalization, which supports my hypothesis. The values in Table 3 have been calculated with the results from M2, but the standard deviation has not been accounted for in this calculation of the predicted probability. As the sample size is low, and standard deviation high, the confidence interval is
quite larger for the predicted probability of nonviolent campaign success. To see what the actual value would be when generalized, a graph on the predicted probability including the confidence interval is presented in Figure 2. In Figure 2, the red line represents the predicted probability of success for all the values of globalization included in the analysis, and the dotted blue lines represent the confidence interval (CI) at the 95% confidence level. The lines in the graph follow the predicted relationship purposed by the hypothesis, but as the lines for the 95% CI shows, the range of the predicted probability for nonviolent campaign success is large. From the minimum to the maximum value of globalization, the range changes from including 70 percentage points (11% to 81% probability of success) to 36 percentage points (63% to 99% probability of
success). For the median value of globalization, the CI is 46 percentage points large (48% to 94%
probability of success). Due to the large range for the predicted probability of success for a nonviolent campaign, the certainty of the generalizability is quite low as the predicted value can vary quite a lot within the confidence level of 95%.
Figure 2: Predicted Probability of Success for a Nonviolent Campaign
For the calculation of the predicted probability, the score of the control variable Cold War was set to its mode of 1, but as the Cold War is no longer active and was a unique occurrence in history, it would be interesting to see how the predicted probability would change if the value of the Cold War variable would be set to 0 (not active) to match the reality of today. As just above half of the cases (53%) were active during the Cold War, the calculated predicted probability for the time after the Cold War is calculated from an even sample giving an accurate representation of the predicted probability of success statistically. Thereby the predicted probability of success for a nonviolent campaign after the Cold War was calculated. The predicted probability of success calculated when the control variable of the Cold War is set to 0 (not active during the Cold War) is presented in Table 4.
Table 4: The predicted Probability of Success after the Cold War
Level of globalization Min Mean Max
Predicated Probability of
success 27% 60% 90%
In this calculation, the predicted probability of success has decreased somewhat, the minimum and the mean have decreased by 15 percentage points and the maximum by 5 percentage points.
Even though they have decreased they are still quite high for the mean and maximum value for globalization, both with a margin above the 50% limit. In Figure 3, a graph with the predicted probability, together with the 95% CI, for nonviolent campaign success after the Cold War is presented. As the calculated predicted probability (Table 4) indicates, is the angel of the predicted probability slope less steep in Figure 3 compared to Figure 2. Another difference is the
confidence interval, which has increased in Figure 3 compared to Figure 2. The minimum value has a range of 5% to 72% probability of success (67 percentage points), the mean value ranges from 34% to 87% probability of success (53 percentage points), and the maximum value ranges from 52% to 99% probability of success (47 percentage points). Otherwise, the positive trend in the relationship between globalization and the success of nonviolent campaigns is seen here as well, providing support for my hypothesis.
Another interesting finding with the results of the predicted probability of success calculated after the Cold War is that it contradicts previous research. As stated in the Research Design section, the presence of the Cold War should decrease the probability of success for a nonviolent campaign compared to occurring after the Cold War (Stephan & Chenoweth, 2008: 24). But the
results here imply the direct opposite. However, if the confidence interval of the two results is compared to one another it is clear that the majority of the CI overlap for both results. Thereby, as most of the predicted values for the probability of success are the same for both during and after the Cold War, nothing can be said with certainty of how the Cold War affects the success of a nonviolent campaign. The results of previous research on the Cold War can thereby neither be strengthened or weakened with the results produced by this analysis.
Figure 3: Predicted Probability of Success for a Nonviolent Campaign after the Cold War
If we then would compare the results of the analysis with the main hypothesis for this study, the more globalized a country is, the more likely is it for a nonviolent campaign to succeed, we would find support for the hypothesis as in all models the relationship between globalization and a successful outcome of a nonviolent campaign is positive and statically significant. The calculated predicted probability, for both scenarios of active and not active during the Cold War, also provides support for the hypothesis, as when globalization increases the probability for success increases.
Even though support has been found for the hypothesis, margins of uncertainty exist, especially because of the small sample size, regarding how certain we can be in the generalizability of the result. This is portrayed in the predicted probabilities (Figure 2 and Figure 3) with wide
confidence intervals. In spite of the small sample size, support has been found for the hypothesis, as the relationship between globalization and the success of nonviolent campaigns is positive and statistically significant at the 95% confidence level.
Robustness check
A robustness check is used to test the robustness of the correlation found in the main analysis trough changing some parts of the main analysis. A methodical decision was made in this study to recode the dependent variable from an ordinal variable to a dichotomous variable where limited success was coded as a failure. The reason for this was to create a stricter measurement of what counts as a successful outcome for a nonviolent campaign to balance the underreporting bias of failed nonviolent campaigns and strengthen the reliability. To see if this operationalization of the dependent variable made a significant impact on the regression analysis results, a
robustness check was performed where limited success was coded as a success instead of as a failure. With this change, the distribution of successful and failed campaigns was 38 versus 9 cases with values for all variables. This distribution of outcome is skewed, and together with an already small sample size the regression would be founded upon, the results of the regression could be misleading and underestimated (King et al., 2001:137-139). Even if the problems of a small sample size are present for the main analysis, the sample of cases is more evenly distributed between failed and successful cases (17 versus 23 cases) for the main analysis compared to the robustness check. As the sample is small from the beginning, the distribution of cases is important to be able to have enough variation in the cases to conclude the relationship (ibid).
Nonetheless, the regression with the recoded dependent variable was made (see Appendix).
The robustness check results differ from the regression in the main analysis. First, the correlation between globalization and a successful outcome for a nonviolent campaign has lost its statistical significance. Secondly, the direction of the coefficients for the control variables has changed, in some cases remarkably, but none is statistically significant in this analysis either. The difference in the results comes from the change in the operationalization of the dependent variable, the
inclusion of limited successful campaigns to the successful campaigns. Theoretically, this change in operationalization should produce the same results as the stricter operationalization of what success for a nonviolent campaign is, as they both build on the same theoretical argument. That globalization will make it easier for a nonviolent campaign to obtain the three key aspects which will increase the probability of a successful outcome. There could be different reasons for why the results have changed, firstly, it could be due to methodological weaknesses, such as a small
and skewed sample size which effects are underestimated in the logistic regression (King et al., 2001: 137-139) and/or an operationalization that does not longer balance for the underreporting bias of failed nonviolent campaigns. Alternatively, it could be that the relationship between globalization and a successful outcome for a nonviolent campaign is not fully explained by the theory proposed. It might be that the relationship is weaker than believed or that globalization only has a significant effect on the cases which reached full success and limited success is better explained by other factors. It is though not fully surprising that the robustness check has
produced different results as the sample has changed drastically, with a loss of nearly 50% of the failed cases. The robustness check has left some question marks on what has caused the results to differ, but these unanswered questions are beyond the scope of this thesis to explore and will be left to future research. It is good to have in mind the results of the robustness check when the main results are analyzed, but as there are many question marks around the methodology and results of the robustness check their results will be used with caution in this thesis.
Discussion
In the main analysis, a positive statistically significant correlation was established between globalization and a successful outcome for a nonviolent campaign. This result is in line with the hypothesis for this thesis which predicted that the more globalized a country is, the higher the probability is for nonviolent campaign success. It should be noted that this indication for the hypothesis to be true was drawn from a quite small sample of cases, and as the sample size decreases so does the certainty of the result as the standard deviation of the variables increases the smaller the sample size is. A large standard deviation means that the range of possible values the variable can take when generalized is larger, giving us less certainty in our prediction to, for example, future cases. Even though the problem with the small sample size, the indications found from the result of the logistic regression analysis points towards a positive correlation between globalization and a successful outcome for a nonviolent campaign, strengthening the hypothesis the increased globalization leads to a higher probability of success for a nonviolent campaign.
If we then look at the results from the robustness check, where globalization do not have a statistically significant effect on the probability of success for a nonviolent campaign, compared to the main analysis where globalization does have a statistically significant effect on the
possibility of success for a nonviolent campaign, we can see that this change of operationalization on the dependent variable could mean something for how globalization affects the probability of
success for a nonviolent campaign. But as the confidence in the results for the robustness check is low, due to the skewed and small sample size, this difference might not indicate anything. If the results are correct and not misleading, the result from the robustness check could indicate that the relationship between globalization and the outcome of nonviolent campaigns is weaker then perceived from the main analysis, or if the results are characterized with underestimating bias then it could support the hypothesis. As the confidence of the robustness check is low, the effect of it should be interpreted with caution.
If there would be another theoretical explanation to why this difference between the results of the main analysis and the robustness check has occurred, a possible explanation for why the results varied based on the change of operationalization, could be that the effect of globalization could be one of the differences between major concessions and full success. The campaigns where the effects of globalization are utilized have a higher probability of full success, but in those cases where limited success was reached the full potential of globalization could not be used. To be able to say something more certain, as this possible explanation only is theorizing, a qualitative study, possibly a structured focus comparison, or more statistical analysis focusing on the different factors affecting limited success and success would have to be done. The resources to continue the statistical analysis to explore the differences between limited success and success, and possibly receive some clarity on why the results from the main analysis and the robustness check contradict one another, is not available for this study and thereby is this puzzle left for future research.
The notion that none of the control variables included in the analysis where statistically significant should be addressed. As none of the theorized confounding variables where statistically significant, none of them can explain the probability of success for a nonviolent campaign. Nonetheless, there might be some other variables that could be possible confounding variables that were not included in this study. Another possible change that could create a difference in the result is the operationalization of globalization. As it is a thick concept and is widely disputed of what it includes, the inclusion of other factors in the measurement could alter the results of the analysis.
To conclude the findings in this thesis, the results produced by the different analyses have supported the hypothesis, but the results should be generalized with some caution as the confidence interval is large. The wide confidence interval has come from the small sample size
the analysis is based on, but as the sample includes all the cases of major nonviolent campaigns from 1970 to 2006 the sample size is hard to change. In all, support has been found for the proposed hypothesis study even though it contains important caveats.
Summary & Conclusion
This thesis explores the relationship between globalization and the success of nonviolent campaigns, to see if globalization positively affects the probability of a nonviolent campaign to succeed. The main argument the hypothesis builds on is that globalization will bring changes to society promoting the possibility for a nonviolent campaign to fulfill the three key aspects of success: mobilization, leverage, and resilience. These aspects have been indicated in previous research to be the aspects of securing success for a nonviolent campaign.
The results found in this thesis do strengthen the support for the hypothesis that increased globalization will increase the probability of a nonviolent campaign to succeed. The predicted probability of success increases from 42% to 95% when globalization increases from the minimal to the maximal value. But the results come with a caveat produced by the small sample size, affecting the generalizability of the result negatively. When the sample size decreases the standard deviation increases, leading to less confidence in the exact effect which can be generalized. The confidence interval at the 95% confidence level ranges from 70 percentage points, for the minimal value of globalization, to 36 percentage points, for the maximal value of globalization.
The size of the sample is not something that could have been changed for this study as the sample is the entire population of major nonviolent campaigns occurring between 1970 and 2006, where data of this type of nonviolent campaigns exist. A robustness check of the main analysis where performed, but the results are dubious due to loss in statistical power and
methodological problems and are thereby interpreted with caution. Even though these caveats, a positive relationship between globalization and the success of nonviolent campaigns have found statistically significant support in this thesis.
These results are congruent with previous research as globalization has shown to have a positive relationship with the onset of nonviolent campaigns. But to further understand how globalization affects the success of nonviolent campaigns more research is needed. Both quantitative and qualitative research is needed to answer the questions that have arisen with this thesis. Do all factors and spheres within globalization positively affect the success of nonviolent campaigns? Is there a difference in the factors causing different outcomes such as success, limited success, or
failure for a nonviolent campaign? As well as more evidence for the presence of the proposed causal mechanism of this study is needed to further understand the effect globalization has on the success of nonviolent campaigns. Provided with more knowledge on how globalization affects nonviolent campaign success, policy implications can be made to strengthen nonviolent action over violent action when the legitimacy of the regime needs to be questioned.
Bibliography
Bayer, M., Bethke, F. S., & Lambach, D. (2016). The democratic dividend of nonviolent resistance. Journal of Peace Research, 53(6), 758–771.
https://doi.org/10.1177/0022343316658090
Beaulieu, E., Cox, G. W., & Saiegh, S. (2012). Sovereign debt and regime type: Reconsidering the democratic advantage. International Organization, 66(4), 709–738.
https://doi.org/10.1017/S0020818312000288
Butcher, C. (2017). Geography and the outcomes of civil resistance and civil war. Third World Quarterly, 38(7), 1454–1472. https://doi.org/10.1080/01436597.2016.1268909
Celestino, M. R., & Gleditsch, K. S. (2013). Fresh carnations or all thorn, no rose? Nonviolent campaigns and transitions in autocracies. Journal of Peace Research, 50(3), 385–400.
https://doi.org/10.1177/0022343312469979
Chenoweth, E. (2011a). Nonviolent and Violent Campaigns Outcomes Dataset, v.1.1. Retrieved from https://www.du.edu/korbel/sie/research/chenow_navco_data.html
Chenoweth, E. (2011b). The Nonviolent or Violent Campaigns Outcomes Data Set , v.1.1 Appendix and Codebook. Retrieved from
https://www.du.edu/korbel/sie/research/chenow_navco_data.html
Chenoweth, E., & Cunningham, K. G. (2013). Understanding nonviolent resistance: An introduction. Journal of Peace Research, 50(3), 271–276.
https://doi.org/10.1177/0022343313480381
Chenoweth, E., & Schock, K. (2015). DO CONTEMPORANEOUS ARMED CHALLENGES AFFECT THE OUTCOMES OF MASS NONVIOLENT CAMPAIGNS? *. An
International Quarterly, 2(4), 427–451. https://doi.org/10.17813/1086-671X-20-4-427
Dreher, A. (2006). Does globalization affect growth? Evidence from a new index of globalization.
Applied Economics, 38(10), 1091–1110. https://doi.org/10.1080/00036840500392078
Fahmi, D. G. (2019). Are We Seeing a Second Wave of the Arab Spring? Chatham House.
Retrieved from https://www.chathamhouse.org/expert/comment/are-we-seeing-second- wave-arab-spring
