Abstract
Recent studies examine politicians’ decisions to use social media, as well as the content of the messages that these political actors disseminate on social media platforms. We contribute to this literature by examining how race competitiveness and a candidate’s position in the race relative to her opponent affect their decisions to issue attacks. Through content analysis of nearly 15,000 Facebook posts for tone (positive or negative), we find that while competitive races encourage both candidates to issue more negative posts, candidates in less competitive races embrace attack messages with more or less frequency depending on whether they trail or lead their opponent. We find that social media negativity is much more likely to be a desperation strategy employed by underdog candidates in less competitive races. We also run separate models examining the factors that drive policy and personal attacks. While underdog candidates are more likely to engage in issue attacks, candidates in competitive races are significantly more likely to use Facebook to make personal attacks.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Sharon Angle (R-NV) is not included in the sample. Though initially reviewed in a pilot study, she subsequently deleted her posts from the timeframe before beginning the full study.
Gainous and Wagner (2014) operationalize race competitiveness as the absolute difference the vote totals of the previous winner and loser in that race. While they analyze both chambers, we use their term “district competitiveness,” which captures state competitiveness as well.
The candidates in this race appear in our data as well as in Gainous and Wagner’s.
Even among the subset of studies that find candidate partisanship to matter, the results are inconsistent. While Republicans have been found to be more critical of their opponents in campaign ads (Kahn and Kenney 1999; Lau and Pomper 2001), Druckman et al. (2010) find that Democrats are more negative on the internet.
While Gainous and Wagner (2014) do find that Republicans post more negatively than Democrats, they are unable to tease out empirically whether this finding stems from an inherent proclivity to go negative or their status as the “out-party” in 2010.
The two additional candidates running as “independents” in our sample are Senator Lisa Murkowski (Alaska) and Charlie Crist (Florida). Murkowski lost the GOP primary to Tea Party favorite Joe Miller but continued in the race as an independent. Crist began the 2010 election as a Republican, but switched to an independent status in May after trailing Marco Rubio in the GOP primary polls. For each of these races, we include the Republican and Democratic primary election winners as well as the independent candidate given their stature in the state/race.
Two independent coders scored a subset of 1483 randomly selected Facebook posts for overall negativity, personal negativity, and issue negativity. We subsequently generate Cohen’s Kappa statistic to ensure interrater reliability for each variable. According to Lombard et al. (2002), the Cohen’s Kappa statistic is a conservative measure of interrater reliability, and is commonly used. It accounts for the possibility of two coders assigning the same score by random chance, and tests against a null hypothesis that the coders’ scores are independent from one another. The results of the Kappa statistic diagnostics suggest the high reliability the dependent variables (negativity—96.22 % agreement, Kappa = 0.8637, Z = 32.18; personal negativity—97.7 % agreement, Kappa = 0.7264, Z = 27.87; issue negativity—95.1 % agreement, Kappa = 0.7619; Z = 29.29). Thus the null hypotheses can be rejected.
When candidates post a link—for example to YouTube—we code the content of the linked material (campaign ads, negative press about an opponent etc.).
Although we did not employ a list of keywords to identify negative posts as Gainous and Wagner (2014) did, our coding scheme would have coded each of their keywords as a negative post as well. Our measure of negativity, therefore, would likely be highly correlated with their Twitter negativity measure.
We have run additional models where the dependent variable is the percentage of posts that are negative in a given week, using OLS regression analysis rather than a negative binomial regression. In these models, the results are substantively the same as those presented here. These additional models are available in an online appendix (available at the following URL: http://jfine.people.clemson.edu).
The Tea Party candidates in the sample are challengers Joe Miller (R-AK), Ken Buck (R-CO), Linda McMahon (R-CT), Christine O’Donnell (R-DE), Marco Rubio (R-FL), Rand Paul (R-KY), Kelly Ayotte (R-NH), Pat Toomey (R-PA), Mike Lee (R-UT), and Russ Johnson (R-WI). Sen. Jim DeMint is the only incumbent classified as a Tea Party candidate. As noted earlier, Sharon Angle (R-NV) is not included in the sample, though she also would be classified as a Tea Party candidate.
While many races have a single incumbent and challenger, open seat races are coded with both candidates set as challengers with values of “1.”.
The CPR is partly driven by polling information in the state as well as the underlying partisanship of the state. Given the regular use of the Cook Political Report by politicians and media members, it is reasonable to expect those in the campaign to use it (or other similar ratings) as one of many barometers about their standing in the race.
The sample contains some posts made during the primary election (for those states with primaries that took place after June 1st). Since differences are expected in posting behavior based on the competitiveness of a candidate’s current race, one must consider that candidates in competitive primaries might behave based on the competition of that race rather than the general election. We therefore account for the competitiveness of the primary as part of the Race Competitiveness variable. While the Cook Political Report does not provide competitiveness ratings for the primary elections, we extend the principle of the CPR ratings to create unique ratings for the primary races that take place within the population of Facebook posts. Primary races whose final vote outcomes were greater than a 15 % margin are given a CPR-equivalent of a ‘3’ (‘safe’ races), between 7 and 15 % the CPR-equivalent of ‘2’ (‘likely’ races), between 2 and 6 % the CPR equivalent of ‘1’ (‘lean’ races), and less than 2 % the CPR-equivalent ‘0’ (‘toss-ups’).
We have run additional models allowing for a non-linear effect of our time (Week) variable, but do not find empirical support for a non-linear relationship.
For the Florida (Crist) and Alaska (Murkowski) independent candidates, we ‘pair’ them with the frontrunner in the race. Therefore, the opponent negativity variable for Crist is the level of negativity from Rubio (R) at t-1, and vice versa. In Alaska, Murkowski (I) is ‘paired’ with Miller (R). The Democratic candidate in each race trailed substantially, and their opponent negativity score is that of the race frontrunner (Rubio and Murkowski, in Florida and Alaska respectively).
We determine the appropriate model by comparing their ability to match observed and predicted values. Because we cluster on candidate in the models, we cannot test for overdispersion using the standard likelihood-ratio test of α = 0. Instead, we compare the observed versus predicted values for the various count models, using Long and Freese’s (2006) countfit diagnostic program. This diagnostic test reveals that both negative binomial regression (NBR) and zero-inflated negative binomial regression (ZINB) outperform both the Poisson (PR) and zero-inflated Poisson (ZIP) models.
In analyses presented in our online appendix (http://dx.doi.org/10.7910/DVN/TGKZ2T), we have replicated our models using a random-effects negative binomial regression with a lagged dependent variable included. In these models, we find a statistically significant and positive effect of proximity to Election Day; candidates post more negative messages as Election Day nears. Our interaction model using this alternate specification parallels the results presented in Table 1, as we find the interaction term to be significant and negative. This suggests that underdog candidates use Facebook for negativity more as Election Day nears. In this alternate specification, we also find evidence to suggest that candidates in toss up races are more negative later in the election cycle.
References
Abramowitz, A. I. (1991). Incumbency, campaign spending, and the decline of competition in U.S. House Elections. Journal of Politics, 53(1), 34–56.
Aldrich, J. H. (1995). Why parties? The origin and transformation of political parties in America. Chicago: University of Chicago Press.
Damore, D. F. (2002). Candidate strategy and the decision to go negative. Political Research Quarterly, 55(3), 669–685.
Downs, A. (1957). An economic theory of democracy. New York: Harper.
Druckman, J. N., Kifer, M. J., & Parkin, M. (2010). Timeless strategy meets new medium: Going negative on congressional campaign web sites, 2002-2006. Political Communication, 28(1), 88–103.
Evans, H. K., Cordova, V., & Sipole, S. (2014). Twitter style: An analysis of how house candidates used twitter in their 2012 campaigns. PS Political Science and Politics, 47(2), 454–462. doi:10.1017/S1049096514000389.
Gainous, J., & Wagner, K. M. (2014). Tweeting to power: The social media revolution in American politics. New York: Oxford University Press.
Geer, J. G. (2006). In defense of negativity: Attack ads in presidential campaigns. Chicago: University of Chicago Press.
Gelman, A., & King, G. (1990). Estimating incumbency advantage without bias. American Journal of Political Science, 34(4), 1142–1164.
Gerber, A. S. (1998). Estimating the effect of campaign spending on senate election outcomes using instrumental variables. American Political Science Review, 92(2), 401–411.
Goldstein, K., & Freedman, P. (2002). Lessons learned: Campaign advertising in the 2000 elections. Political Communication, 19(1), 5–28.
Gordon, A., Shafie, D. M., & Crigler, A. N. (2003). Is negative advertising effective for female candidates? An experiment in voters’ uses of gender stereotypes. The International Journal of Press/Politics, 8(3), 35–53.
Gulati, G. J., & Williams, C. B. (2015). Congressional campaigns’ motivations for social media adoption. In V. A. Farrar-Myers & J. S. Vaughn (Eds.), Controlling the message: New media in American political campaigns (pp. 32–52). New York: New York University Press.
Hale, J. F., Fox, J. C., & Farmer, R. (1996). Negative advertisements in U.S. Senate Campaigns: The influence of campaign context. Social Science Quarterly, 77(2), 329–343.
Kahn, K. F. (1993). Gender differences in campaign messages: The political advertisements of men and women candidates for U.S. senate. Political Research Quarterly, 46(3), 481–502.
Kahn, K. F., & Kenney, P. J. (1999). The spectacle of US Senate campaigns. Princeton, NJ: Princeton University Press.
Klotz, R. (1998). Virtual criticism: Negative advertising on the internet in the 1996 senate races. Political Communication, 15(3), 347–365.
Krehbiel, K., & Wright, J. R. (1983). The incumbency effect in congressional elections: A test of two explanations. American Journal of Political Science, 27(1), 140–157.
Lau, R. R., & Pomper, G. M. (2001). Negative campaigning by US senate candidates. Party Politics, 7(1), 69–87.
Lawless, J. L., & Fox, R. L. (2005). It takes a candidate: Why women don’t run for office. New York: Cambridge University Press.
Lombard, M., Snyder-Duch, J., & Bracken, C. C. (2002). Content analysis in mass communication: Assessment and reporting of intercoder reliability. Human Communication Research, 28(4), 587–604.
Long, J. S., & Freese, J. (2006). Regression models for categorical dependent variables using stata (2nd ed.). College Station, Texas: STATA Press.
Morales, L. (2011). Google and facebook users skew young, affluent, and educated. Retrieved from Washington, DC. http://www.gallup.com/poll/146159/Facebook-Google-Users-Skew-Young-Affluent-Educated.aspx.
Peterson, D. A. M., & Djupe, P. A. (2005). When primary campaigns go negative: The determinants of campaign negativity. Political Research Quarterly, 58(1), 45–54.
Schattsneider, E. E. (1960). The semi-sovereign people. New York: Holt, Rinehart, and Winston.
Sigelman, L., & Buell, E. H. (2003). You take the high road and i’ll take the low road? The interplay of attack strategies and tactics in presidential campaigns. The Journal of Politics, 65(2), 518–531.
Skaperdas, S., & Grofman, B. G. (1995). Modeling negative campaigning. American Political Science Review, 89(1), 49–61.
Smith, A. (2011). 22% of online Americans used social networking on Twitter for politics in 2010 campaign. Pew Internet.
Smith, A., & Rainie, L. (2010). 8% of online americans use twitter. Retrieved from Washington, DC. http://www.pewinternet.org/2010/12/09/8-of-online-americans-use-twitter/.
Stokes, D. E., & Miller, W. E. (1962). Party government and the saliency of congress. Public Opinion Quarterly, 26(4), 531–546.
Straus, J. R., Glassman, M. E., Shogan, C. J., & Smelcer, S. N. (2013). Communicating in 140 characters or less: Congressional adoption of twitter in the 111th Congress. PS Political Science & Politics, 46(1), 60–66.
Stromer-Galley, J. (2014). Presidential campaigning in the internet age. Oxford: Oxford University Press.
Theilmann, J., & Wihite, A. (1998). Campaign tactics and the decision to attack. The Journal of Politics, 60(4), 1050–1062.
Wagner, K. M., Gainous, J., & Holman, M. R. (2016). I am woman, hear me tweet! gender differences in twitter use among congressional candidates. Journal of Women, Politics, and Policy.
Acknowledgments
We wish to thank Kristin Bender for her excellent research assistance. We also would like to thank Laura Olson, Chris Bonneau, Kris Kanthak, Mac Avery, the editor, and the anonymous reviewers at Political Behavior for their helpful comments and suggestions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Our data and the replication code for each our models, as well as our online appendix, can be found online at the following URL: http://dx.doi.org/10.7910/DVN/TGKZ2T.
Appendix
Rights and permissions
About this article
Cite this article
Auter, Z.J., Fine, J.A. Negative Campaigning in the Social Media Age: Attack Advertising on Facebook. Polit Behav 38, 999–1020 (2016). https://doi.org/10.1007/s11109-016-9346-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11109-016-9346-8