Snapchat and civic engagement among college students

ABSTRACT

During the 2016 presidential election cycle, Clinton and Trump used Snapchat as one part of their overall voter outreach and engagement efforts. This portion of their campaign strategy was disproportionately targeted toward younger voters, since those between 18 and 25 comprise a vast portion of Snapchat’s user base. Did their efforts, those of political parties, or those of interest groups on Snapchat produce higher levels of civic engagement among college students? We utilize a survey that we conducted from a college campus in the Midwest in October 2016 to answer this question. Using a series of matching analyses, we discover that those students who sent pictures or videos about interest groups, candidates for office, or political parties on Snapchat were more civically and politically active than otherwise similar students who had not participated in these activities.

KEYWORDS:

• Snapchat
• civic engagement
• young voters
• political participation
• social media
• 2016 election

Notes

1. This dataset appropriately tests the theory because Snapchat had the youngest user base out of the nine leading social media platforms in 2014, with 45% of its adult users reported by comScore to be between the ages of 18 and 24 (Bump, 2015). Snapchat’s growth has been dramatic – by year’s end 2015, its user base climbed to 100 million and it grew even further to 150 million in 2016 (Nusca, 2017). This user base remains skewed towards young adults (Duggan, 2015).

2. Further, campaigns must rely on Snapchat staff to discover their number of followers and since links cannot be posted, it is difficult to use the app to recruit volunteers or solicit campaign contributions (Chittal, 2015).

3. The margin of error for this survey is 3.79%.

4. That said, it is possible that differences between the sample and population drive our results. To account for this possibility, we weighted the observations by race and sex and re-ran our analyses. The signs and significance patterns of those analyses are identical to those reported in Tables 1 and 2 with one exception: frequently sending pictures loses significance (p = .118). The complete set of results are available in Appendix D, Tables D2 through D6.

5. M denotes the mean for that variable, while SD signifies the standard deviation for that variable.

6. We performed a confirmatory factor analysis to examine whether the measures of activity on Snapchat are consistent with one another (see Hartman, 2017; Rossell, 2012). It is possible that these measures of activity differ so substantially from one another that it does not make sense to analyze them altogether. The confirmatory factor analysis provides strong evidence that these treatments are consistent with one another with a reasonably fitted model such that they can be analyzed in the way that we do so here. These results are contained in Table D1 in Appendix D of the Online Appendix.

7. It is possible that interest in politics, strong partisanship and being a political science major are so interrelated that they add little to the analyses. To consider this possibility, we removed each of these variables one at a time to investigate whether our results change. The results largely mirror what is reported here. The results for removing one of interest in politics, strong partisanship and political science major are furnished in Appendix C.

8. These were the only questions asked about peer civic experiences. Some respondents responded that they did not know for these questions. When constructing our index, we coded values for these variables as missing. Thus, there is no peer civic engagement score for those who answered “don’t know” for at least one of the questions.

9. Our surveys do not contain a measure that allows us to determine the strength of ideological attachment. That said, considering ideology in the manner that we do allows us to investigate the effect of ideological direction on civic engagement.

10. We coded the dummy variable for conservatives using the same process.

11. Collectively, the comparison category for both binary variables is political moderates.

12. Unfortunately, we lack additional, relevant variables about socioeconomic status such as a student’s (or alternatively, their family’s) income.

13. For example, the binary for rarely having monitored is coded one for those who rarely did so, zero for never having done so, and missing for those without Snapchat accounts, declined to answer the question, and those who responded that they had sometimes, regularly, or very often monitored what candidates for office, political parties, or interest groups post on Snapchat.

14. If civic engagement were related to the decisions of others to engage in one or more activities on Snapchat, then we would observe high correlations between one or more of these activities and offline civic engagement. To test this, we correlated our dependent variable (offline civic engagement) and each of our treatment variables (monitoring, sending a picture, sending a video, or chatting with someone on Snapchat). The correlations that we observed are .34, .37, .42, and .37, respectively. These correlations are sufficiently low such that we can say that the first portion of the SUTVA assumption is fulfilled.

15. If we do not do this, then we cannot satisfy SUTVA, as there would be alternative forms of the treatment (see Imbens & Rubin, 2015, p. 10–13).

16. However, we do not perform a regression following the matching analysis (as Ho et al., 2007 suggest), as Abadie and Spiess (2016) state that this is inappropriate when matching is done with replacement as researchers obtain inaccurate standard error estimates.

17. Table D7 contains the complete comparison of activity levels on Snapchat versus those on Facebook.

18. The data for this variable satisfy the matching assumptions that existed in the previous estimations. In particular, each treatment is binary, there is no interference, and there are no hidden variations of the treatment because of the way in which we coded the treatment variables. In addition, each individual’s assignment to one of the Facebook treatments is unconnected with their level of offline civic engagement, given the values of the remaining covariates. Further, we satisfy the common support assumption because students could have opted out of having engaged in politically-oriented Facebook activities. Finally, we satisfy the SUTVA assumption because offline civic activity does not vary with sharing webpages or pictures on Facebook, as the correlation between these variables is .49. This correlation is sufficiently low such that we can say that we fulfill the SUTVA assumption.

19. We acquired data on the other public comprehensive universities in this state from their websites for the Fall 2016 semester, except for two. For one of the two, we used data from the Fall 2017 semester, while we used data for the Fall 2015 semester for the other.

Additional information

Notes on contributors

Laurie L. Rice

Laurie L. Rice is Professor of Political Science at Southern Illinois University Edwardsville. She is co-author of the book Web 2.0 and the Political Mobilization of College Students and co-editor of American Political Parties Under Pressure: Strategic Adaptations for a Changing Electorate. Her research interests include political communication, the presidency, elections, and civic engagement and her work on these topics appears in journals such as Presidential Studies Quarterly and Social Science Computer Review.

Kenneth W. Moffett

Kenneth W. Moffett is Professor of Political Science at Southern Illinois University Edwardsville. He is co-author of the book Web 2.0 and the Political Mobilization of College Students, and has published academic articles in American Politics Research, Climatic Change, Environment and Behavior,Legislative Studies Quarterly, Party Politics, Social Science Computer Review, Social Science Quarterly, and others. His research interests include American politics and policy.