The changes in the media environment brought up by digital media are expected to have profound effects on politics, from citizen’s participation to elites’ strategies. However, the literature has paid little attention to systemic and contextual factors that might constrain or limit the possibilities for collective action in this new communication context. Using a structural approach, I analyze presidential election campaigns’ use of Twitter in a study of 16 Latin American countries between 2012 and 2015 – Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Panama, Paraguay, Uruguay and Venezuela. To demonstrate the quality of the data set, I show that forecasting models with Twitter Volume can preview election results in most of the cases, and even give more accurate forecasting than polling data in Colombia, Costa Rica, Ecuador and Mexico. In order to fill the shortage of theoretical models on social media, I propose a new concept of campaign spheres for understanding the linkage between social media messages and public opinion associated with campaigns. Campaign spheres are sections of the public sphere. Instead of forming public opinion through rational deliberation, a campaign sphere mostly reproduces a candidate’s political slogans and symbols to mobilize opinion leaders. While the public sphere is inclusive, a campaign sphere has clear boundaries dividing those who support the campaign and those who oppose it. Institutions and social context do affect campaign spheres in Latin American presidential elections. Party systems can predict how intensely people participate, while incumbency increases the number of participants, mainly in left-wing campaign spheres. Two-round systems had a negative impact on the participation of right-wing campaign spheres, and also demonstrated a small negative effect on the prestige of all candidates. Yet challengers had more prestige than incumbents – an effect that was stronger in plurality system countries and after the elections. This research gives a perspective on the constrained context for communication change in Latin America, which can be extended and generalized to other new democracies.