This essay will argue that the paper by Sebastián Valenzuela, Chile Teresa Correa and Homero Gil de Zúñiga from 2017 is not using adequate representative data and that this may influences the conclusions drawn by the paper. This will be done by firstly describing the paper, its data, methods and findings. Subsequently, the research design and analysis will be criticised and means of improving the paper will be proposed. Lastly, it will be concluded that the paper may instead provide indicators for the mechanisms behind the link between social media and political participation but that no firm conclusion can be drawn due to the limited data sample. Given the limited scope of this paper, the technical aspects of the statistical analysis will only be covered briefly.
The paper, named Ties, Likes and Tweets: Using Strong and Weak Ties to Explain Differences in Protest Participation across Facebook and Twitter Use, analyses the link between social media and political participation by focusing on Twitter and Facebook. The paper acknowledges that a mild positive link between social media use and citizen engagement has already been found by other papers, notably Boulianne (2015) and Skoric et al., (2016), Therefore, the authors argue, that further research should focus on finding theoretical and empirical evidence that may explain the underlying mechanisms and conditions leading to the positive link between of social media and different forms of political participation (Correa et al., 2017, p. 121-122). Specifically, the paper investigates whether individuals who use Twitter and Facebook may be influenced by different social ties on the medias to engage in political protest activities.
Social media is defined as “Internet-based applications that build on the ideological and technological foundations of the Web 2.0, and that allows the creation and exchange of user-generated content” with a reference to Kaplan & Haenlein 2010, p. 62.
The paper subsequently refers to Granovetter 1983 and defines weak ties as ‘’acquaintances, such as friends of a friend, who serves as a link to more distant clusters of people and provide more novel information and diverse perspectives’’ (Correa et al., 2017, p. 121). Strong ties are defined, with reference to Kenny, 1994, as ‘’an individual’s intimate connections, such as family and close friends, who may share overlapping information and provide emotional support, trust and reinforcement of ideas’’ (Correa et al., 2017, p. 121).
The paper argues that Facebook users are influenced by their strong ties, more so than their weak ties, when deciding to engage in protest behaviour. The opposite is true for Twitter, where users are influenced by their weak ties, more so than their strong ties, when deciding to engage in protest behaviour. Specifically, the authors present the following hypotheses:
H1: The indirect relationship between general uses of Facebook and protest behaviour is stronger for political information obtained from strong ties than weak ties.
H2: The indirect relationship between general use of Twitter and protest behaviour is stronger for political information obtained from weak ties than strong ties
The underlying arguments presented for these hypotheses are that it is easier to receive information from people who users do not know personally, the weak ties, on Twitter, whereas users often have a closer relationship with the people they befriend and receive information from on Facebook (Correa et al., 2017, p. 125)
The data originates from the Youth and Participation study, an annual face-to-face survey conducted by Universidad Diego Portales in Chile since 2009. It includes a representative sample of 1000 individuals between the age of 18 and 29 who are living in Chile’s three largest urban areas. The three areas that are surveyed contains 64% of the adult population and it should be noted that Chile is mostly an urban country, with 87% of the population living in urban areas (Correa et al., 2017, p. 127) The specific survey used in the paper, was made in the winter of 2014 in conjunction with two of the papers’ authors. The data is weighted with a post-stratification weight, which uses auxiliary information to reduce sampling errors, to make it more representative for the Chilean population. However, the authors note that the results are very similar when using unweighted data (Correa et al., 2017, p. 128).
The survey asks participants about several subjects. This includes control variables such as education, exposure to news media and political interest among other things. However, the most highlighted variables are the following:
Have you participated in the following activities in the past 12 months: (a) attended public demonstrations, (b) boycotting, (c) joined causes in social media, (d) signed a petition and (e) attended in-person political forums and debates.
The data was then turned into binary data with 0 for not engaging and 1 for engaging. Subsequently, the authors created an index by counting the number of positive responses to each item. This meant that responses could range from 0 to 5. The median was found to be 0.89 with 54% of respondent reporting no protest activities.
With what frequency have you received information about issues of public interest on your account from: a) People you know personally are close to you; b) people you do not know personally?
A was considered strong ties and b was considered weak ties. The data was subsequently reverted and rescaled, so that it ranged from never (0) to frequently (1). The following medians were found: Strong ties on Facebook: 0.48, strong ties on Twitter: 0.10, weak ties on Facebook: 0.30, weak ties on twitter: 0.07.
Lastly, participants were asked about their usage of Facebook and Twitter separately.
Responses very then divided into: (1) Every day, more than once a day; (2) every day, once a day; (3) at least three times a week; (4) once a week; (5) two or three times a month; (6) once a month or less; and (7) never.
The data was subsequently reverted and rescaled, so that it ranged from never (0) to every day (1). The median was 0.76 for Facebook and 0.12 for Twitter (Correa et al., 2017, p. 125-126).
The data is subsequently analysed in Stata with a Generalised Structural Equation Model, which is a model that can accommodate the different variables including binary, ordinal, multinomial or count variables. The model is expressed with two equations:
M= i_1+aX+fZ+e_M (1)
Y=i_2+c^’ X+bM+gZ+ e_Y (2)
i1 and I2 are intercepts, eM and eY are error terms, Z is a set of control variables and a and c’ are path coefficients estimating the direct effects of X (usage of Twitter or Facebook) on M (Information of public interest from weak or strong ties) on Y (protest engagement). b estimates the effect of M on Y. The indirect effect of X on Y is estimated by the product of a and b. In layman terms, this means that the equations are analysed to find whether there is a correlation between the data gathered. That is, general usage of the two media platforms, information from strong or weak ties and engagement in protest behaviour. If a correlation is found when doing the regression analysis, the authors may attribute it to the underlying theory presented earlier (Alan Bryman, 2012, p. 349 – 351).
The results do find a positive relationship between general usage of Facebook and protesting. However, this positive relationship itself does not provide evidence about the underlying mechanisms leading to the relation. To explain this, the authors point to a result showing a relationship between general Facebook usage and exposure to politically mobilizing information from strong and weak ties. Moreover, there was found to be a relationship between frequency of mobilizing information received and likelihood of engaging in protest activities.
However, the results found for Twitter show no direct relationship between usage of Twitter and engagement in protest activities. The authors ague that this does not prevent any indirect influencing when using Twitter. In fact, the authors found that respondent who frequently used twitter were far more likely to receiving mobilizing information. However, unlike Facebook, no relation was subsequently found between engagements in protests and information from strong ties. Instead, in accordance with hypothesis H2, a relation between information from weak ties and engagement in protest activities were found. As such, the authors conclude that their findings are consistent with hypothesis H1 and H2 and that political information from strong and weak ties on social media mediates the relation between protest behaviour and usage of the two platforms (Correa et al., 2017, p. 124 – 126)
However, there are several areas where the paper could improve. Firstly, when looking at the underlying data, the total dataset contains 1000 respondents but only 17.5% of responders use Twitter (figure 2). The low Twitter use is also evident from the paper itself, where the median Twitter use is 0.12 on a scale from 0 – 1 (Correa et al., 2017, p. 127). Moreover, only 100 of the 1000 responders reported that they had received information about issues of public interest on their Twitter account from a weak tie (figure 2). When looking at responders that use twitter whilst also having received information from weak ties and engaged in protest activities, we are left with only 6% of the sample (figure 1). This means that the conclusion regarding weak ties influence on protest behaviour through twitter is based on a regression analysis of only 60 people.
A representative sample has strong external validity when compared to the target population the sample is meant to represent (Bryman, 2012, p. 197 – 199). It is difficult to see how 60 individuals from a specific age group, found in strictly urban areas, can provide representative data that may be used to draw the conclusions on causality that the authors claim. The study may have benefited from using more surveys as to ensure that the relevant responses were adequate for them to be considered representative. If we are to assume that 1000 individuals are a representative sample of the population (Bryman, 2012, p. 197 – 199), then the authors would need to increase their sample by a factor of 5.7 just to ensure that 1000 participants actually use Twitter (figure 1).
Lastly, the participating group, answer options and the means of conducting the survey may have benefited from being different or at least expanded. The paper claims that illegal protest activities are rare in Chile and use this as an argument for entirely excluding illegal protest activities from the survey. However, the authors never cite any source that can confirm that such activities are a rarity in Chile. This also highlights the issues researchers face when making face-to-face surveys. Participants may be more inclined to provide controversial or political information, such as engagement in protests and illegal activities, if they are able to do so anonymously and online in the safety of their own home (O’Connor and Madge, 2001, 11.2).
It is also problematic that the survey includes examples of political participation, allegedly to prevent responders from guessing what it might mean (Correa et al., 2017, p. 127), as this may influence the response of the responders. Moreover, the data gathered is from a sample of 1,000 individuals between the age of 18 to 29 living in Chile’s three largest urban areas. Whilst Chile is mostly urban with 87% of the population living in urban areas, excluding the majority of the population, through the age barrier and the exclusion of the rural population, reduces the representativeness of the survey. Furthermore, if we accept the premise that most political activities take place in major towns, limiting the survey to the three largest urban areas may lead to an unrepresentative high amount of political activity among participants. In conclusion, the paper may provide indicators regarding the influence on protest ties by social ties on social media. However, the criticisms above, with an emphasis on the small data sample, means that the authors cannot draw any firm conclusions. If the authors wished to use their sample, they may have benefitted from narrowing the scope of their theories to a specific subgroup that aligns with the sample.
