Political communication on social media in Latin America: unequal use of Twitter by members of parliament

This article focuses on the under-researched topic of the use of Twitter by members of parliament (MPs) in Latin America. There have not yet been any thorough comparative studies on this topic, and the majority of publications on the region focus on single case studies. Previous studies have primarily concentrated on presidents, or in the case of MPs, on specific case studies. A total of 2,353,138 tweets were accessed via the Twitter API, and 3,215 MPs were examined. Regression models and correlations were used to answer research questions, and the main variables examined concerned individual characteristics of MPs (gender and age) and socioeconomic indicators of the country (number of people on Twitter , internet access, Human Development Index - HDI). As a result, this paper offers a report on how MPs in Latin America are currently utilizing Twitter . While this social network is used by more than 90% of MPs in some countries (Argentina, Colombia, Ecuador, El Salvador, Peru, and Uruguay), there are countries where fewer than half of MPs use it (Bolivia, Honduras, and Nicaragua). The results show that female MPs are more likely than male MPs to use Twitter . In addition, Twitter is being adopted more by younger MPs. Other results show that country characteristics such as internet penetration, Twitter population, and the HDI are significant predictors regarding the adoption and use of Twitter by MPs. These results are consistent with assumptions based on cost-benefit calculus. Thus, it does not make as much sense for politicians to adopt Twitter in countries where there are fewer people on Twitter and low internet penetration. In particular, if politicians want to be elected or inform citizens about their activities, they have an opportunity to reach voters through Twitter . However, this is only true if Twitter is used in their countries.


Introduction
In recent years, there has been an increasing amount of research looking at social media and politics globally (Casero-Ripollés, 2018; Jungherr, 2016; Matassi; Boczkowski, 2020). Among social media publications, articles focusing on the United States and Europe have consistently dominated, and more recently, research on Latin America has increased. (Matassi; Boczkowski, 2020). This article looks at parliamentarians and their adoption of Twitter in Latin America as a currently unexplored topic. Most publications on Latin America are single case studies, and to date, there is no comprehensive comparative study of the adoption of social media by members of parliament (MPs) in Latin America.
Political actors are increasingly using Twitter for various purposes, including disseminating political proposals, engaging with the public, broadcasting their messages, and shaping public debate (Casero-Ripollés; Alonso-Muñoz; Marcos-García, 2022; López-Meri et al., 2017). In the 2016 US presidential election, Twitter played a significant role in both Hillary Clinton's and Donald Trump's communication strategies (Buccoliero et al., 2020;Enli, 2017). During electoral campaigns, political actors may use Twitter to spread political messages, reference candidate activities, emphasize personal aspects, and mobilize supporters (López-García, 2016). Emerging parties tend to focus on mobilization and generic announcements, while traditional parties are more likely to publish policy proposals (López-García, 2016). However, the use of Twitter for dialogue with citizens remains limited (Alonso-Muñoz et al., 2016).
In the European context, Fazekas et al. (2021) found that most political actors did not engage with the public about EU issues, resulting in less public interaction. Meanwhile, Vergeer, Hermans and Cunha (2013) noted that candidates in the 2009 European Parliament elections primarily used Twitter for electoral campaigning. Members of the European Parliament have been found to use Facebook for national audiences and Twitter for international audiences, with language preferences varying across countries (Haman; Školník; Čopík, 2022). In the Spanish context, Casero-Ripollés, Alonso-Muñoz and Marcos-García (2022) identified ideology, political initiative, and political career as factors influencing the authority and digital influence of political actors on Twitter. López-Meri, Marcos-García and Casero-Ripollés (2017) observed a trend towards hybridization between new and conventional digital media in the platform's usage. Jivkova-Semova, Requeijo-Rey and Padilla-Castillo (2017) found that virality on Twitter did not guarantee electoral victory, but certain communication trends could provide valuable insights. The use of Twitter by Latin American political leaders has also been examined, with Segado-Boj, Díaz-Campo and Lloves-Sobrado (2015) finding no common strategy for using the platform during times of crisis.
Overall, political actors use Twitter for various purposes, including disseminating messages, engaging with the public, and shaping political discourse. However, the extent of their interaction with citizens and the effectiveness of their communication strategies on the platform vary across different contexts and political actors. Latin America has presidential regimes, so the main focus of the research is on the political communication of presidents or during presidential elections and less on members of parliament. However, MPs are very important players in the political system that deserve attention.
The contribution of this paper is that it analyses the unexplored topic of the use of Twitter by Latin American MPs. This study, therefore, provides up-to-date data on the Twitter activity of MPs in Latin America. Previous research has focused almost exclusively on presidents or, in the case of MPs, single case studies. This paper is divided into four parts. The first section presents the research questions and hypotheses. The second section presents the methodology, and the third section presents the results. The last section is the conclusion of the paper.

Research questions and hypotheses
This paper aims to answer two research questions. These are: 1. What are the differences between the adoption and use of Twitter by parliamentarians in Latin America across countries? 2. What factors influence the adoption and use of Twitter in Latin America?
To answer the first research question, it will first be necessary to examine the current state of Twitter adoption, i.e. to identify the individual legislators who use Twitter. This is not an easy task, as no databases exist yet where the usernames (Twitter handles) of legislators on Twitter are present.
Multiple methods and data are used to answer the second research question. The dataset will consist of variables for each legislator and information on the number of citizens on Twitter, the Internet, and the HDI in each country.
In this research, Twitter was chosen for several reasons. Firstly, at the time of the research, Twitter offered the Twitter API for academic research to researchers, which provides access to almost all data on Twitter. In contrast, Facebook provided an API that was considerably limited, and researchers often did not have access to it or only had temporary access through various tools. Overall, Facebook has been criticized by academics for providing limited access to a restricted amount of posts (Ho, 2020). Among social networks, Twitter is most frequently the subject of research concerning digital communication technologies for legislators (Neihouser; Tremblay-Antoine, 2021), which demonstrates that Twitter is used by legislators as a common communication tool. Another issue associated with Facebook is that Latin American lawmakers often have a Facebook profile rather than Facebook pages, from which it is even more challenging to obtain data, and sometimes they use both. Instagram, on the other hand, suffers from similar problems as Facebook and is predominantly used for publishing images. Additionally, there is limited space in a single article, so for all the reasons mentioned above, only Twitter was analyzed in this research.

Gender
Although some of the earliest research on social media use suggested that women generally use social media more actively (Hargittai, 2007) Sullivan, 2021). Differences between studies about politicians may be related to the adoption of the technology at a given time within a given country and among citizens. Since most studies have not found a difference between women and men, the hypothesis is: H1: There will be no relationship between gender and adoption or Twitter activity among MPs.

Age
The adoption of new technology is often related to age according to the diffusion of innovation theory (Rogers, 1962). So, it should be young politicians as early adopters and active users of social media. Early studies showed that young candidates in elections are more familiar with the internet and make more use of its potential.  Straus et al., 2013;. But there are also studies where age is not a significant predictor (Macková; Štětka, 2016; Rauchfleisch; Metag, 2020). Given the majority of studies confirming the relationship between age and social media use, the hypothesis is: H2: Younger MPs will adopt Twitter more and also use it more actively.

Technological development of the country
From a classical microeconomic approach, demand should influence supply. It is the adoption of new technologies by the population that should increase the demand for political information in the online environment and politicians should respond to this fact. Parliamentarians should use Twitter primarily when they believe that the benefits outweigh the costs, at least according to a simple calculation based on rational choice theory. If Internet penetration is very low, then a politician should have less incentive to use the Internet for his or her purposes.  et al., 2013). However, only rarely (Haman; Školník, 2021) has research directly used data on the percentage of residents using a given social media. This is primarily because this data is not publicly available. H3: The higher the number of citizens on Twitter (as well as Internet penetration and the Human Development Index -HDI), the higher the adoption of Twitter by MPs.

Methodology
First, it was necessary to collect data. Unfortunately, there is no up-to-date list of Latin American MPs and their Twitter accounts. Some parliaments allow an MP to put a link to their official profile on the parliamentary website. However, even in this case, not all MPs make use of this. In any case, parliamentary websites were the primary source of data. If a Twitter account was not found, I proceeded to use the Google search engine that offers the Google Knowledge Graph, which often lists the social media accounts of politicians and other known persons: https://cloud.google.com/enterprise-knowledge-graph/docs/search-api However, not all MPs are searchable in this way. Thus, there was also a direct search on Twitter, using various combinations of names. On parliamentary websites, MPs often have their full names listed, which makes direct Twitter searches problematic as some MPs do not use all their names and are therefore not easily traceable. Hispanic names can include five or more names, and a politician may only use two of them. At the same time, some parliaments use Twitter accounts and have lists of MPs on them. In this case, this source was also used.
There are a total of three independent variables in the paper. The first two independent variables are individual and based on data from parliamentary websites. These are age and gender. In the previous section, reference was made to studies that utilized variables such as gender and age. This study employed variables constructed in a similar manner. Gender was coded as 1 for male and 0 for female. Age was measured in years. Not all parliaments reported information on the date of birth and therefore the age of MPs, so some countries have this variable missing. Also, not all parliaments reported the sex of the MP, but gender could be inferred from the name of the MP or other characteristics. Twitter population data comes from the latest Latinobarómetro survey in 2020 (Latinobarómetro, 2022 The year 2021 and the first three months of 2022 were chosen arbitrarily. However, it is not possible to choose a long period during which there would be no elections in all countries. Elections make the analysis problematic because one would have to examine two or more legislative periods in one country, which could cause problems for cross-country comparisons if other countries have only one legislative period. By selecting this time frame, even countries like Argentina had more than three months' worth of tweets, which should be sufficient for analysis. At the same time, as will be shown below, a variable of an average number of posts per week and a dichotomous variable that takes the value of 1 if an MP sent at least one tweet per week on average were created for each MP. Indeed, it is not enough to include mere adoption, it is also necessary to include how much Twitter usage the MP has. The threshold of one tweet per week was chosen. The models also use the number of tweets per week to check the actual activity among the MPs present on Twitter. A total of 2,353,138 tweets were collected via the Twitter API using the rtweet package (Kearney, 2019) and subsequently analyzed in the R programming language, and a total of 3,215 MPs were examined. Table 1 provides some data on the adoption and activity of parliamentarians. The second and third columns give information on Twitter adoption. As already mentioned, in the second column I mean adoption in the form of posting at least one tweet during the period under study, The third column then requires the activity of at least one tweet per week during the period under study. These data are also displayed on two maps. However, the percentages of MPs on Twitter decrease, often significantly, if we look at the MPs who sent on average at least one post per week. Subsequently, only Ecuador, El Salvador, and Uruguay reach values above 90%. The third and fourth columns give more detailed activity information. It can be seen that the MPs from El Salvador have developed the absolute highest activity, with an average of 64.9 and a median of 50.1 tweets per week, followed by Ecuador with an average of 33.3 and a median of 23.3 tweets per week. Several countries have an average that corresponds to sending at least one tweet per week. However, it is important to note that the average and median include only MPs who are present on Twitter. Therefore, these two columns need to be combined with the information from the second column to see the overall state of political communication on Twitter in a given country.

Use of Twitter by MPs
The sixth and seventh columns show the distribution of retweets in terms of the number of hashtags and retweets. The sixth and seventh columns show additional data on the use of Twitter by MPs. The second column provides information on how many tweets had a hashtag in the period under study. The mean for Latin American countries is 31.5%. However, there is considerable variation between countries. While 65% of tweets from Ecuador used a hashtag, only 13.7% of tweets in Uruguay did. These differences suggest that it is hardly possible to speak of a pattern of Latin American MP communication. This fact is confirmed by the third column, which shows what percentage of the tweets were retweets. The average for Latin American countries is 39.8%, but the figures range from 10.4% in Bolivia to 78.8% in Venezuela.

Adoption of Twitter -gender and age-individual level
In this part of the paper, only data at the first level, which is the level of the parliamentarian, will be used. The following section will then use hierarchical models (multilevel models). This is primarily a robustness testing of the results using other models as well. At the same time, for hierarchical models, the lack of cases at the second level, that is, the state level can be problematic. This paper works with a total of 18 Latin American countries, which may be considered methodologically insufficient for hierarchical models. For example, Kreft and Bokhee (1996), Hox (2010, p. 235), or Snijders and Bosker (1999, p. 154) propose a "30/30 rule" whereby there should be at least 30 cases at each level.
There are four models in the following Table 2. The first two models have the dependent variable as a dichotomous variable of adoption (1) of the MP on Twitter, i.e., posting at least one tweet during the observation period. The third and fourth models have the dependent dichotomous variable of whether the MP sent on average at least one tweet per week (1) during the observation period. Because of the dichotomous variable, logistic regression was chosen. There are two variables in the model, namely gender, with the male taking the value of 1 and this variable being present for all MPs. On the other hand, age is not present for all MPs and therefore has separate models. Dummy variables have also been used for each country but are not presented here to save space. Note: *p<0.05; **p<0.01; ***p<0.001 Table 2 shows that in both the first and third models, where there are the most cases, gender is statically significant and with a negative value, hence implying that women used Twitter more, both in the form of adoption and when including requiring a minimum activity of one tweet per week. In the second model, however, gender loses statistical significance, which is due to the fact that a number of countries where women were more prevalent in Twitter use than men dropped out of the model because there is no information on age in these countries. Models 1 and 3, therefore, reject the first hypothesis that there is no difference between the genders. The age variable is statistically significant in both models 2 and 4 and takes negative values, i.e. adoption and activity (in the form of one tweet per week) increase with lower age of the MPs. Models 2 and 4, therefore, confirm the second hypothesis that younger MPs adopt Twitter more and are more active.  In these models, only MPs with a Twitter account are included. In the second and fourth models, age is then added. All four models show that the gender variable is negatively statistically significant, i.e. women use Twitter more actively. At the same time, no statistical significance is found for the age variable. Thus, while the models in the previous table implied that women are more likely to be on Twitter, these models confirm that women are also more likely to be more active among MPs who are on Twitter. In contrast, the models showed that younger MPs are more likely to adopt Twitter. However, when the analysis includes MPs who have a Twitter account and their activity, it loses statistical significance, so there is no way to confirm the second hypothesis in the activity part. Younger MPs adopt Twitter more, but this is no longer the case for activity in terms of the number of tweets.

Activity on Twitter -gender and age-individual level
Graph 1. Correlation between Internet users -population in countries and the percentage of MPs who have adopted Twitter

Technological development of countries
The following Graphs 1 to 6 show the relationship between the presence of MPs in each country on Twitter and the number of people on the internet, the number of people on Twitter, and the Human Development Index (HDI) for that country. The data on the Twitter population comes from Latinobarómetro's latest survey for 2020 (Latinobarómetro, 2022). Information on the number of individuals using the Internet in a given country is provided by the International Telecommunication Union (2022), while not offering data for Venezuela. The HDI is then offered by the United Nations Development Programme (2022). The relationship between the percentage of MPs who, on average, sent at least one tweet during the period under review is then plotted against these three indicators. These three variables are moderately to strongly correlated. The highest correlation is between the HDI and the percentage of people on the internet (r = 0.905), while the relationship between the percentage of people on the internet and on Twitter (r = 0.717) also shows a very strong correlation. The least correlation, but still significant, is between the HDI and the percentage of people on Twitter (r = 0.534). Strong correlations are not surprising, as they give similar information, namely about the socio-economic and technological progress of a given country.  Table 4 shows the values for each country. It can be seen that there are a few exceptions where the Internet is not very prevalent, yet it is very much used by MPs. The most notable exception is El Salvador, where just over half of the people have access to the internet, but almost all of its MPs use Twitter. Other countries with internet penetration of around 50%, while having a lower HDI, are among the countries with the lowest adoption of Twitter by MPs.

Multilevel models
In Table 5, there are six multilevel models, where the variables gender or age are at the first level. Then at the second level are the variables percentage of citizens on the Internet, HDI, and percentage of people on the Internet. The data source is the same as mentioned above. The dependent variable is whether the MP was on Twitter (1) with at least one tweet sent during the period under study. The six models are because the three second-level indicators are always in the model separately due to the strong correlation but also the fact that they give similar information to some extent and the age variable was not available for some countries. The models confirm what the correlations and models in the previous part of the paper already showed. MPs adopt Twitter more when the country's technological development is greater. Younger MPs also adopt Twitter more. At the same time, all three variables are statistically significant. Thus, Twitter is used more in countries where there is a higher demand, i.e. where citizens have adopted new technologies more, in the form of the Internet or Twitter directly, to find political information. This result is in line with the assumption that in these countries MPs will benefit more from using it as it reaches a larger percentage of the electorate.  Strandberg, 2009;2013;. However, in this case, a difference was found, as there is a gender gap in Latin America, with female parliamentarians using Twitter more. This finding represents an original contribution to the field of study, as it significantly contradicts previous findings from other countries, and is based on a large dataset. One potential reason for female MPs being more active on Twitter could be their desire to bypass traditional media, which has been known to portray them less favorably (Heith, 2003;Kahn, 1996).
By using Twitter, female politicians can present a more open, personal, and interactive image (Carlson; Djupsund; Strandberg, 2013) and communicate directly with voters, mobilizing them and targeting specific groups like young women. This direct communication allows female politicians to promote both themselves and their parties more effectively while simultaneously sidestepping gendered coverage (Lawless, 2012). Another possible explanation for the observed results could be the different communication styles adopted by women and men. Studies have indicated that women tend to use technology in a more sociable manner (Walton;Rice, 2013;Lasorsa, 2012), which may contribute to their more active engagement on Twitter. This communication style can help female politicians forge stronger connections with their audience, enabling them to better address their constituents' concerns and needs.
The second hypothesis is confirmed as younger MPs adopt Twitter more frequently. Existing research suggests that younger representatives are more likely to embrace social media (Gulati;Williams, 2013;Larsson, 2015;2015;Schmelzinger, 2017;Strandberg, 2009Strandberg, , 2013Straus et al., 2013). In this regard, the results align with current research. Younger age cohorts in developed nations are often labeled as "digital natives" due to their lifelong exposure to computers, demonstrating a more intuitive command of online technologies compared to "digital immigrants" from older cohorts Schmelzinger, 2017). Individuals from younger generations typically exhibit greater proficiency in using online platforms for political engagement (Bakker; De-Vreese, 2011). This expertise is also likely to manifest among younger politicians (Larsson, 2015). Younger and freshman politicians might be more motivated to leverage all available communication channels to consolidate political support and cultivate their brand (Peterson, 2012), unlike older politicians who might have already built their brand via traditional media.
At the same time, the hypothesis regarding country characteristics was also confirmed when variables such as the number of people on the Internet, on Twitter, and the HDI were statistically significant in the models. Based on assumptions from rational choice theory and cost-benefit analysis, if politicians aim to be elected or re-elected, they should consider the number of citizens on a given social network. If no one uses the social network, no one will read the politician's Twitter posts, making it illogical for the politician to be present on Twitter in a country where almost no one utilizes the platform. In this case, correlations initially revealed a strong relationship between the number of people on the internet in a given country, the quality of life as a socioeconomic indicator, the number of Twitter users, and the adoption of Twitter by MPs. Subsequently, these variables were statistically significant in regression models, confirming the hypothesis. Of course, the results do not imply that there are no exceptions with a smaller percentage of internet users where MPs simultaneously use Twitter. These findings corroborate a previous study comparing the relationship between Twitter usage by citizens and MPs in Europe (Haman; Školník, 2021).The most significant exception is El Salvador, which has a low standard of living and fewer people on both the internet and Twitter; nevertheless, nearly all its MPs are present on Twitter. This intriguing phenomenon is worth investigating further.. Consequently, it does not solely depend on the number of people on the internet in a given country, but other factors are also involved.

Conclusion
This paper has provided a unique analysis of the use of Twitter by Latin American MPs. More than 3,000 MPs were examined and two million tweets were collected. Research on political communication in Latin America was dominated by research on elections, especially presidential elections, and presidential political communication. The paper, therefore, provides an update on the current state of the MPs' adoption in Latin America. The use of Twitter by MPs in Latin America varies considerably. While in several countries more than 90% of MPs use Twitter (Argentina, Colombia, Ecuador, El Salvador, Peru, and Uruguay). Then there are countries where less than half of MPs use Twitter (Bolivia, Honduras, Nicaragua). At the same time, in several countries there is very low activity, so we cannot speak of much active use, even if a higher percentage of MPs are on Twitter. Differences were also found between the use of hashtags and the proportion of retweets among tweets. While MPs in some countries actively use hashtags, in others they do not use this form of communication at all. A high proportion of retweets means that MPs in a given country are spreading ideas already formulated by someone rather than creating their content.
In terms of factors that may help explain Twitter use, the first hypothesis was rejected when a significant relationship was found in that female MPs were more likely to use Twitter than male MPs. When analyzing MPs on Twitter, it was also found that women are more active. The second hypothesis was confirmed and younger MPs adopt Twitter more. However, when comparing MPs already present on Twitter, the models did not show younger MPs to be more active at a statistically significant level. The third hypothesis was also confirmed when MPs in states with higher numbers of Twitter users adopted Twitter more. Results were similar when using Internet penetration and human development variables. Similar results are consistent with assumptions based on rational choice theory and cost-benefit calculus. Thus, in countries where Twitter or the Internet is less used, it does not make as much sense to use Twitter as in a country where many people are connected to the Internet. Especially if a politician wants to get elected, he or she has a unique opportunity to reach voters through social media. But this is only true if social media are used in the country. One limitation of the article is that it primarily concentrates on the adoption and activity of Latin American parliamentarians on Twitter, without delving deeply into the content of their tweets. This focus may overlook important nuances or trends in the messages being communicated by these politicians. Consequently, the main objective of the article was to map the current adoption of Latin American parliamentarians on Twitter, rather than to examine the precise purposes for which they use Twitter. A more in-depth analysis of the content could have revealed patterns and differences among the parliamentarians, as well as their stance on critical matters in the region. Future research could build upon the findings of this article to explore these aspects more thoroughly, thus providing a more holistic perspective on the role of Twitter in Latin American politics.
Of course, there are several other limitations. For example, a shorter period was observed, but this was primarily to offer the most recent data. However, research looking at longer periods would certainly be useful in the future. At the same time, there is also a problem in identifying MPs on Twitter. As there are no official complete lists, and thus multiple techniques had to be used to collect the data. Therefore, it is not possible to ensure that an MP could have not missed through data collection, especially if they did not use their official name and were not registered on any lists or found through Google search. At the same time, of course, only two individual variables are used in the research, which is primarily due to comparisons across states. Important variables such as an MP's ideology or their position in the government were not examined. While several other variables can be collected on a single country, data on multiple parliaments does not exist in a comprehensive and uniform form. Even in this study, it was not possible to find the age of all MPs. However, the above reasons are relevant in interpreting the paper's results.