Twitter made me do it! Twitter's tonal platform incentive and its effect on online campaigning

ABSTRACT

Does Twitter trigger negative tones in politicians' digital communication? On social media direct feedback mechanisms such as retweets or likes signal to politicians which message and tone are popular. Current research suggests that negative language increases the number of retweets a single tweet receives, indicating preferences for negativity in the audience on Twitter. However, it remains unclear whether politicians adapt to the logic of Twitter or simply follow the rules determined by the broader political context, namely the state of their electoral race. We use sentiment analysis to measure the tone used by 342 candidates in 97,909 tweets in their Twitter campaign in the 2018 midterm elections for the US House of Representatives and map the ideological composition of each politician's Twitter network. We show that the feedback candidates receive creates an incentive to use negativity. The size and direction of the tonal incentive is connected to the ideological composition of the candidate's follower network. Unexpectedly, the platform-specific incentive does not affect the tone used by candidates in their Twitter campaigns. Instead we find that the tone is mainly related to characteristics of the electoral race. We show that our findings are not dependent on our sentiment measurement by validating our results using hand coding and machine learning.

KEYWORDS:

• Campaigning
• filter bubble
• sentiment
• social media
• platform effects

Acknowledgements

We thank Stefanie Bailer, Pablo Barbera, Britt Bolin, Thomas Breuninger, Marc Debus, Chung-hong Chan and Sebastian Stier as well as numerous members and audiences in panels at ECPR General Conference Wroclaw, DVPW Dreiländertagung Zürich and IC2S2 Amsterdam for their helpful comments. We thank Sarah Lehman, Luca Joppien, and Çağla Ezgi Yıldız for their assistance and support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 A more detailed discussion of the method and our robustness checks including hand-coding and supervised machine learning can be found in the results section and the appendix.

2 We find the same effects for likes as another possible form of reacting on a tweet. Results can be seen in the appendix Table A4 in the third model.

3 Other measurements of model fit like AIC and BIC also suggest that the model with varying slopes performs better.

4 Descriptive statistics of the coefficient ${\text{β}}_{1j}$ can be seen in Table A1.

5 Given the small effects of all coefficients, we ran simulations to see how well our model (3(M2) ${\text{β}}_{1j}={\gamma }_0+{\gamma }_1\cdot Media{n}_j+{\gamma }_2\cdot \log \left(Varianc{e}_j\right)+{\gamma }_3\cdot \log \left(FollowersCoun{t}_j\right)+{\gamma }_4\cdot Part{y}_j+{\eta }_j$(M2) ) can predict the negativity incentive of individual candidates. We find that our model can predict the true distribution of the negative incentive in our sample, however, it tends to predict higher coefficients than the real values. For more details see Figure A1 in the appendix.

6 Using a subsample analysis of only challengers supports the claim that challengers with a higher win probability can be associated with a more positive average tone in Twitter campaigning. For further details see Table A6 in the appendix.

7 Regression tables using both categorical measurements can be found in the appendix.

Additional information

Funding

This work was supported by the University of Mannheim’s Graduate School of Economic and Social Sciences (GESS).

Notes on contributors

Samuel David Mueller

Samuel David Mueller (M. Sc. Political Economy at the University of Konstanz) is a PhD Candidate at the Graduate School of Economic and Social Sciences (Mannheim) and a Research Assistant at the Chair for Political Economy at the Department of Political Science, University of Mannheim. His research centers on strategic information provision and the effect of behavioral information processing on decision making as well as social media usage in electoral campaigning and social preferences more generally [email: [email protected]].

Marius Saeltzer

Marius Saeltzer (M. A. Political Science at the University of Hamburg) is a PhD Candidate at the Graduate School of Economic and Social Sciences (Mannheim) and a Research assistant at the Mannheim Centre for European Social Research MZES and the Chair of Comparative Government at the Department of Political Science, University of Mannheim. His research focuses on social media use of political elites and parties [email: [email protected]].