This study relies on data obtained from the 2012 Americas-Barometer public opinion survey, carried out in 25 countries1
in the region by the Latin American Public Opinion Project, LAPOP [36
]. According to LAPOP, “each survey is implemented based on a national probability design. In some cases, oversamples are collected to allow precise analysis of opinion within sub-national regions. Survey participants are voting-age adults interviewed face-to-face in their households, except in Canada and the United States where the interviews are web-based” [36
]. Data was collected between January and May 2012, and details about sample size and sampling errors can be found in the Appendix
Since we are studying support for protests in Latin American countries, Canada and the United States were excluded from the analysis. The final dataset contained 37,102 cases.
5.1. Dependent Variables
Support for protests
is the dependent variable in this study. Five questions were used to assess support for different protest tactics, all asking “How much do you approve of…” on a 10-point Likert-type scale where 1 = Strongly disapprove and 10 = Strongly approve. The questions were:
“…people participating in legal demonstrations” (M = 6.90; SD = 2.91)
“…people participating in an organization or group to try to solve community problems” (M = 7.83; SD = 2.51)
“…people participating in the blocking of roads to protest” (M = 3.89; SD = 2.95)
“…people invading private property or land in order to protest” (M = 2.53; SD = 2.26)
“…people participating in a group working to violently overthrow an elected government” (M = 2.47; SD = 2.22)
Then, a factor analysis was performed to verify how those items relate to each other, and two factors emerged from the data: support for moderate
and support for radical
protests. Table 1
shows the rotated component matrix for the factor analysis.
Support for moderate protests included two items asking respondents how much they approved of “people participating in legal demonstrations” and “people participating in an organization or group to try to solve community problems” (Eigenvalue = 1.622, 2 items; α = 0.71; range = 1 to 10; M = 7.4; SD = 2.4).
Similarly, support for radical protests was measured by three items asking respondents how much they approved of “people participating in the blocking of roads to protest,” “people invading private property or land in order to protest,” and “people participating in a group working to violently overthrow an elected government” (Eigenvalue = 1.622, 3 items; α = 0.69; range = 1 to 10; M = 2.9; SD = 1.9).
The two variables are similar to the concept of “level of deviance” in protest literature [20
]. A group’s “level of deviance” is determined by its protest goals and tactics. While these two dimensions are correlated, Boyle and colleagues argue for their conceptual separation when analyzing a group’s level of “radicalism” [37
]. The authors found that it is the protest tactics, rather than its goals, that determine the valence of its news coverage. As such, this study focuses on support associated with specific protest tactics, which can be moderate
. Moderate protest tactics include peaceful and non-disruptive demonstrations, while radical tactics encompass violence and civil disobedience [18
]. The models in this paper use both the indexes and their separate components as dependent variables in the analysis. This decision was made for two reasons. First, we used the indexes to provide a more parsimonious model for mapping and clustering purposes. Then, in the regression models, we opted to use the five dependent variables separately to provide more nuance to the analysis.
5.2. Independent Variables
To measure people’s online networking use for political information, respondents were asked whether they read or shared any political information on online networking sites, such as Twitter, Facebook or Orkut,2
in the last 12 months. Of all respondents, about 11% used online networking sites for political information in 2012.
Political Satisfaction. We created two variables to account for political satisfaction—satisfaction with government and satisfaction with services. For satisfaction with government, we added two items asking respondents how they rated the job performance of “the president of your country” and “the members/senators and representatives of Congress/Parliament of your country.” Answers were measured on a 5-point Likert-type scale where 1 = Very bad and 5 = Very good (2 items; α = 0.65; range = 1 to 5; M = 3.1; SD = 0.79). For satisfaction with services, we created an index by adding three items asking respondents how satisfied they were with “the condition of the streets, roads, and highways;” “the quality of public schools,” and “the quality of public medical and health services.” Answers were measured on a 4-point Likert-type scale where 1 = Very dissatisfied and 4 = Very satisfied (3 items; α = 0.65; range = 1 to 4; M = 2.5; SD = 0.58).
. Trust here was conceived as institutional trust, which emphasizes trust as an individual’s response to the performance of institutions [38
]. Trust in institutions, both explicitly and implicitly political, is linked to political involvement. Furthermore, scholars have used aggregated scales like the one we use here to measure institutional trust in relation to political action [38
] and corruption in a variety of countries [39
], including a study of Mexico that also uses LAPOP data [40
]. In order to measure trust
in institutions, we added 13 items asking respondents to what extent they trusted different institutions in their countries: the justice system, Supreme Electoral Tribunal, Armed Forces,3
National Legislature, national police, Catholic Church, Evangelical/Protestant Church, political parties, President/Prime Minister, Supreme Court, local or municipal government, mass media, and elections. Answers were measured on a 7-point Likert-type scale where 1 = Not at all, and 7 = A lot (13 items; α = 0.91; range = 1 to 7; M
= 3.9; SD
Efficacy. We accounted for both external and internal efficacies. For external efficacy, we used one item asking respondents if they believed those running the country are interested in “what people like the respondent think” on a 7-point Likert-type scale where 1 = strongly disagree, and 7 = strongly agree (range = 1 to 7; M = 3.31; SD = 1.89). Similarly, we measured internal efficacy using an item that asked if the respondent “feels like he/she understands the most important political issues of their country” using the same 7-point Likert-type scale (range = 1 to 7, M = 3.84; SD = 1.80).
Political interest measures people’s general level of curiosity about politics. Respondents answered the question: “Generally speaking, how interested are you in politics?” on a 4-point Likert-type scale where 1 = None and 4= A lot (range= 1 to 4; M = 2.1; SD = 0.95).
Political knowledge. Respondents were asked two questions measuring their knowledge about politics: “Who is the President of the United States?” and “How many years is the President’s term of office in your country?” Answers were measured as 1 = right answer and 0 = wrong answer, added and averaged to create an index of political knowledge (two items; range: 0−1, KR-20 = 0.48; M = 0.89; SD = 0.23).
Strength of partisanship
was initially measured on a 10-point scale where 1 = left and 10 = right. Respondents were asked to think of their own political leanings and place themselves on this scale (range = 1 to 10; M
= 5.52; SD
= 2.63). Then, results were folded to assess the strength of partisanship of the respondent, following the recommendation of Gil de Zúñiga and Valenzuela [41
]. The final scale measures partisanship on a 5-point scale where 1 = strong and 5 = weak (range = 1 to 5; M
= 3.14; SD
Demographics. This study accounts for four key demographic variables: people’s age (M = 37.84; SD = 14.2), gender (male = 49.6%), as well as respondent’s level of formal education, measured as years of school (M = 8.98; SD = 3.78). Income was understood as the monthly income in the respondents’ household, measured in 17 categories based on the currency and distribution of the country (range = from 1 to 16; M = 8.13; SD = 3.87; Median = 8.0).
5.3. Statistical Analyses
RQ1 asks: How do citizens classify depending on how they support moderate and radical protest tactics? To answer this question, a two-step cluster analysis was performed using both support for moderate protests and support for radical protests as the variable criteria to classify cases. Once the clusters were created, a series of goodness-of-fit chi-squares was run to show each country’s highest positive residual by cluster. The classified proportions for each country were compared to Latin American respondents as a whole. This inspection of residuals provides a way of assessing how different each country is from the average response in the region, and how these differences appear for each protest tactic.
To test the hypotheses posed by this study, zero-order Pearson’s correlations were performed to ascertain the ways in which all variables of interest related to each other. Furthermore, two linear regressions were estimated for each of our five dependent variables—support for legal demonstrations, groups organizing to solve problems, blocking streets to protest, invading private property to protest
and overthrowing elected governments.
These analyses allowed us to test the relationship between online networking
and the dependent variables, while controlling for the effects of a set of key influential variables previously identified by the literature, such as political satisfaction, trust, efficacy, political interest, political knowledge, strength of partisanship and demographics. The models also included a block controlling for each country’s fixed effects (see Appendix Table A2
RQ2 asks whether online networking varies between attitudinal groups, and therefore, affects support for protests differently. To answer this question, chi-square tests were calculated by correlating the created clusters with people’s online networking use.
Finally, to give a better impression of how the data are distributed geographically, we mapped the different attitudinal groups and the relationship between online networking and support for protests by country using ArcGIS4
. Residuals are mapped on a gradient of light to dark in four different colors, depending on the attitudinal group. An additional map displays statistically significant relationships between online networking and protest attitudes, differentiating between moderate and radical protest tactics.
Given the high number of cases in the dataset (more than 37,000) all statistical analyses (correlations, hierarchical regressions and chi-squares) were performed with 5000 bootstrapped bias-corrected resamples [42
Our first research question asks how citizens classify depending on how they support protests. Cluster analysis revealed respondents cluster into four groups based on support for moderate and radical protests. The first cluster accounts for 18.2% of respondents. They support radical protests more than the average (range = 3 to 10; M = 4.61; SD = 1.25) but not more moderate forms (range = 1 to 7; M = 5.15; SD = 1.36).
A second cluster (23.3%) supports all types of protests more than the average–radical (range = 3 to 10; M = 5.07; SD = 1.74) and moderate (range = 6.5 to 10; M = 9.14; SD = 0.96). On the opposite end, the third and largest cluster (29.8%) expressed little support for either type of protest, radical (range = 1 to 3; M = 1.66; SD = 0.69) or moderate (range = 1 to 7.5; M = 5.34; SD = 1.75).
Lastly, the fourth cluster (28.7%) mirrors the first—these respondents only support moderate protests (range = 8 to 10; M
= 9.44; SD
= 0.74), but not more radical tactics like blocking the streets or invading property (range = 1 to 3; M
= 1.53; SD
= 0.65). Table 2
shows the proportions of those in each country who fell into each of the four clusters. With the exception of Chile, Paraguay, Trinidad and Tobago, and Guatemala, all other countries’ largest groups fall either into the “moderate protest” cluster, or the “no protest at all” cluster (see Table 2
Goodness-of-fit chi-squares were run for each country to show each country’s highest positive residual by cluster. Proportions for each country by cluster were compared to the Latin American respondents as a whole based on protest attitudes. Inspection of the residuals from these chi-squares reveals that all countries differed significantly from Latin America as a whole, except Brazil. The largest positive residuals from each country are reported in Figure 1
. The examination of residuals suggests that most countries have significantly larger groups than the average of people not supportive of protests at all or supporting moderate protest only. The residuals suggest that a smaller set of countries differ from Latin America as a whole in their increased support for all forms of protests. Additionally, results suggest that residents of Bolivia, Honduras, and Guatemala support radical protests more so than Latin Americans broadly. The geographic locations of the countries are shown in Figure 1
Hypothesis 1 assessed the impact of online networking on support for moderate and radical protests. Results in Table 3
show correlations between education, income, internal efficacy, external efficacy, strength of partisanship, political interest, political knowledge, satisfaction with services, satisfaction with government, trust in government, online networking and attitudes toward moderate and radical protests. Online networking was positively correlated with legal demonstrations (r = 0.079, p
< 0.001), people organizing in groups to solve problems (r = 0.069, p
< 0.001), blocking streets (r = 0.044, p
< 0.001), and seizing private property to protest (r = 0.013, p
< 0.05). The relationship between online networking and protests aiming to overthrow an elected government is not statistically significant. Taken together, these numbers indicate that support for less radical protests is more strongly correlated with online networking use.
Correlations also suggest that variables associated with strategic resource mobilization theory—higher levels of income, education, political knowledge, and interest—have a higher impact on support for moderate protests such as legal demonstrations and group organizing. Variables associated with collective behaviorism—dissatisfaction with government and services, lower age, income and education—are associated with support for radical protests—blocking streets, seizing property and overthrowing government. One exception occurs when it comes to internal efficacy: the belief that they understand the most important political issues of their country is associated with support for all types of protest tactics. Furthermore, online networking emerged as a variable strongly correlated with other strategic resource mobilization theory variables, such as political interest (r = 0.170, p < 0.001), higher income (r = 0.185, p < 0.001), higher education (r = 0.271, p < 0.001) and internal efficacy (r = 0.128, p < 0.001).
In order to address the set of hypotheses, five linear regression models were estimated, one for each protest tactic. After ensuring the absence of collinearity by examining variance inflation factor scores (<1.48), the models were estimated with all variables entered simultaneously. Table 4
summarizes linear regression models of support for each protest activity for all countries in the data. For readability purposes, the fixed effects for country were omitted from Table 4
and can be found separately in the Appendix
As hypothesized, online networking was a significant predictor of a more positive attitude in regard to legal protests (β = 0.038, p < 0.001); therefore, hypothesis 1a was supported. Males (β = −0.027, p < 0.01), older people (β = 0.028, p < 0.01), those with higher income (β = 0.048, p < 0.001) and higher education (β = 0.077, p < 0.001) were found to be significant predictors of higher support for legal protests. Political interest (β = 0.041, p < 0.001), trust (β = 0.073, p < 0.001), weaker partisanship (β = −0.046, p < 0.001), dissatisfaction with government (β = −0.069, p < 0.001), and dissatisfaction with services (β = −0.023, p < 0.05) were also predictors of positive attitudes regarding legal demonstrations. While external efficacy was a predictor of less support for legal protests (β = −0.050, p < 0.001), internal efficacy was a strong and positive predictor of support for legal protests (β = 0.081, p< 0.001). The full model explained 12.6% of the variance observed.
H1b asks about the relationship between online networking and support for “people organizing in groups to solve problems.” Results reveal that online networking was significantly correlated with support for this tactic (β = 0.34, p
< 0.01). As such, the same patterns emerged for support for legal and group organization, further confirming the results of the factor analysis for support for moderate protest tactics.
The second column of Table 4
reveals that those who are male (β= −0.019, p
< 0.05), higher income (β = 0.048, p
< 0.01), higher education (β = 0.038, p
< 0.01), higher political interest (β = 0.030, p
< 0.01), higher levels of trust (β = 0.073, p
< 0.01), less strength of partisanship (β = −0.066, p
< 0.01), less satisfaction with the government (β = −0.052, p
< 0.01) and services (β = −0.049, p
< 0.01), higher internal efficacy (β = 0.062, p
< 0.01), but lower external efficacy (β = −0.069, p
< 0.01) are more likely to support groups organizing to solve problems as a form of collective action. The model explains 15% of the variance observed. Hypotheses 1b was supported.
also shows the linear regression model of support for more radical protest activities. Online networking had a positive and significant relationship with support for blocking streets to protest (β = 0.034, p
< 0.01) (Hypotheses 1c was supported
). Those who are younger (β = −0.095, p
< 0.01), poorer (β = −0.090, p
< 0.01), have less political knowledge (β = −0.043, p
< 0.01), less partisanship (β = −0.020, p
< 0.05), more political interest (β = 0.038, p
< 0.01), trust (β = 0.053, p
< 0.01), are dissatisfied with both government (β = −0.072, p
< 0.01) and services (β = −0.035, p
< 0.01), and have higher internal efficacy (β = 0.056, p
< 0.01) are more likely to support protestors blocking the streets. The model explains 6.3% of the variance.
As for the second type of radical protest tactic (H1d—Invading property), online networking had a significant relationship (β = 0.030, p < 0.01) (Hypotheses 1d was supported).Those who are younger (β = −0.093, p < 0.001), poorer (β = −0.085, p < 0.001), have less education (β = −0.055, p < 0.001), political knowledge (β = −0.048, p < 0.001), more political interest (β = 0.021, p < 0.05), trust (β = 0.076, p < 0.001), are dissatisfied with services (β = −0.028, p < 0.01) and have higher external efficacy (β = 0.066, p < 0.01) are more likely to support demonstrators invading private property to protest.
Finally, the last column of Table 4
depicts the model for support for protests aiming to overthrow an elected government. In this case, online networking was not significantly associated with support for protests aiming to overthrow elected governments (H1e not supported
). Perhaps not surprisingly to scholars in Latin America, the results from the other independent variables are very similar to the models on blocking the streets or invading private property. Those who support this tactic were younger (β = −0.083, p
< 0.001), poorer (β = −0.081, p
< 0.001), have less education (β = −0.064, p
< 0.001), political knowledge (β = −0.060, p
< 0.001), trust (β = 0.032, p
< 0.001), are dissatisfied with services (β = −0.026, p
< 0.01) and government (β = −0.064, p
< 0.001), have higher external efficacy (β = 0.064, p
< 0.01) and internal efficacy (β = 0.027, p
In addition, we ran the regression models by country in order to verify the relationship between online networking and protest attitudes individually, rather than treating Latin America as a homogeneous entity. After splitting the cases, the relationship between online networking and support for moderate protests remained significant or marginally significant for Guatemala (β = 0.13, p
< 0.01), Chile (β = 0.09, p
< 0.05), Panama (β = 0.07, p
< 0.08), Dominican Republic (β = 0.08, p
< 0.06), and Venezuela (β = 0.09, p
< 0.06). For radical protests, online networking led to more supportive attitudes in Trinidad and Tobago (β = 0.10, p
< 0.09), Chile (β = 0.09, p
< 0.05), Belize (β = 0.08, p
< 0.06), El Salvador (β = 0.14, p
< 0.01), Ecuador (β = 0.08, p
< 0.08), and Venezuela (β = 0.09, p
< 0.06). Peru (β = −0.08, p
< 0.06) was the only country in the sample where online networking led to negative attitudes towards protesting. Results of these country-specific models can be seen in Figure 2
To answer RQ2 about how the clusters may vary regarding online networking use, chi-square tests were run to examine the extent to which each cluster engaged in the use of online networking sites. As a base level, an analysis of the whole sample reveals that 11.1% of the respondents reported using online networking. Results indicate that there are substantive differences in the ways online networking is used by clusters that represent the four typologies of support for protests. Those who support radical protests only report low levels of online networking (8.3%), and use online networking sites significantly less than Latin Americans as a whole (χ2
(1) = 50.31, p
< 0.000). Of Latin Americans who do not express support for any type of protests, only 8.8% use online networking, which is significantly less than expected (χ2
(1) = 54.88, p
< 0.000). However, those who support all protests use online networking significantly more than expected, with 14.7% reporting some use (χ2
(1) = 107.47, p
< 0.000). Likewise, those who support only moderate protests use online networking more than expected (χ2
(1) = 40.13, p
< 0.000)—13.1% of those in the fourth cluster report using online networking sites. See Table 5
for full results.
In the decades since the military relinquished control of the government, Latin American countries maintained relatively stable democracies, with several decades of democratically elected governments being challenged by popular protests. However, studies that have attempted to understand how Latin Americans view protests are rare. As such, this paper makes at least two theoretical and methodological contributions. First, we classified citizens into four distinct groups depending on their level of support for moderate or radical protests: radicals, moderates, all protests and no protests. Despite the increase in protests in the region after military regimes, the numbers presented here suggest Latin Americans tend to only support the right to peacefully protest while rejecting protests that threaten the public order in any way. Then, we assessed the impact of online networking on support for both types of protests. We find that online networking predicts support for both moderate and radical protests. However, Latin Americans who support only moderate protests use online networking more than Latin Americans as a whole, while those who support only radical protests use online networking sites significantly less.
Surprisingly, we find that countries with a tradition of street protests, such as Argentina, Haiti and Bolivia, showed less support for any type of protests than average. One can only speculate as to why. Perhaps these findings come as a result of power shifts in those countries, where historically anti-government forces that led protests in the past are now in power. The findings may indicate a fear that any movement that threatens political order could lead to another coup d’état. We urge future studies to qualitatively investigate the political culture in specific countries in order to understand this phenomenon.
When it comes to online networking, those who support moderate demonstrations spend more time online than supporters of more radical forms, as well as those who do not support protests at all. Evidence suggests that the use of online networking sites, such as Twitter and Facebook, is related to support for the right to protest using both moderate and radical tactics, with the exception of protests aiming to remove an elected government from power.
Our country-by-country analysis also revealed that the relationship between online networking and support for protests is not homogeneous across the region. Figure 1
and Figure 2
illustrate significant differences within Latin America and can be helpful for future research focusing on sub-regional levels of analysis, or those wishing to present their research to a public unfamiliar with the region. In countries like Chile and Venezuela, online networking use is positively associated with positive attitudes towards moderate and radical protests. In other countries like Colombia and Brazil, this relationship is not significant at all. While it is not within the scope of this paper to account for such differences, we believe that the uniqueness of Chile and Venezuela comes from a sharp increase in the number of protests after 2010 in both countries. Our findings echo those of Wolfsfeld and colleagues who recommend scholars take political context into consideration when assessing the relationship between social media and protests [35
]. Future research should analyze the impact of online networking on a country-by-country basis while also accounting for the role of different protest tactics and attitudinal subgroups presented in this study.
Our findings support the notion that peaceful, legal demonstrations have been “normalized” in the region: people tend to overwhelmingly support moderate protests, and online networking is related to this support, which speaks to the predictions of Karatzogianni on the normalization of digital activism [27
]. In tandem with resource mobilization theorists, our findings suggest that moderate protest tactics have not only been “normalized” and accepted as a legitimate form of participation, but also that online networking is a form of resource that can be mobilized to facilitate acceptance of protest behavior. Echoing Norris et al. [19
], we do not find evidence that ascending support for protests have negative consequences for democratic stability in Latin America. If demonstrations are understood as a form of legitimate political expression, then their acceptance in the region indicates the health of democracy. This finding is particularly important because it distinguishes protest attitudes in the region from protests explicitly aimed to regime change, such as the Arab Spring movement. More than ever, the issues raised by Tarrow [16
] are relevant here: what does the institutionalization of protests mean for the future of social movements and repertoires of contention? Will an increase in contentious acts obscure other routine forms of political participation, such as electoral campaigning, strikes and petitions?
It is important to note that the variables tested only accounted for 6% to 12% of the variance observed, a strong indicator of the shortcomings of the variables from the literature developed in Europe and the United States to explain what leads people to support protests in the region. It is also significant to highlight that the effect sizes of online networking on protest attitudes are small. This is in tandem with the recent argument made by theorists that the role of social media for protests has been highly overrated [27
]. Rather than suggesting causality, our results reveal a small and significant relationship between using social media and supporting protest behavior, especially when it comes to moderate tactics. Another limitation may come from the measure used by LAPOP to tap into online networking use, which specifically asks about users reading and sharing political information on social media. It is possible that users get incidentally exposed to political information while using platforms for other activities (e.g., keeping up with family and friends). We urge future scholars using primary data analyses to include nuanced social media activities in their measures.
It is also important to note that obstructing traffic and trespassing are not normally identified as “radical” protest tactics in the United States or Europe. Nevertheless, Latin Americans’ views on such activities were more closely correlated with violently overthrowing an elected government than with legal demonstrations. This is especially interesting since the recent protests in the region involved some level of radical tactics (e.g., Black Blocs in Brazil), a behavior strongly disapproved of by the respondents in our sample.
The models presented in this study are exploratory rather than definitive. In light of the current wave of protests in the region, findings point to the need for further research regarding what makes people protest in the Americas, especially when it comes to radical tactics. While our results support the idea that online networking foments support for demonstrations, how this support translates to actual behavior is yet to be analyzed.