Twitter has become an increasingly popular tool for American politicians, from the President to AOC. But does how much a politician tweets impact how they do in elections? According to my study, the answer is “not really,” or perhaps, “not yet.”
I, with the help of my dear friend and colleague Wil Dubree, studied the 2020 Congressional Democratic Primary to look at the correlation between the rate at which a candidate tweets and the percentage of votes they got. A logistical regression, including incumbency and percentage of total campaign finance made up by PAC money as control variables, found that average tweets per day and the percentage of votes won do not have a strong correlation.
| Coefficients | Estimate | Standard Error | Z value |
Pr(>|z|) |
| Intercept | -0.10 | 0.23 | -0.46 | 0.65 |
| Average # of Tweets | -0.00 | 0.028 | -0.04 | 0.97 |
| Incumbent | 1.54 | 0.71 | 2.17 | 0.03 |
| % funding made up by PAC money | -0.01 | 0.01 | -0.91 | 0.36 |
One on one comparisons show a similar lack of strong correlation, but give a better visualization of the relationships that do exist.
Overall, there is a negative correlation between average tweets per day vs percent votes, with most of the candidates at the lower end of the extreme.
When we look just at increments we see much the same relationship.
In this data set, only three incumbents lost, and only one of those three tweeted at an above average rate.
However, when we look exclusively at non-incumbents, the trend becomes positive.
It’s easy to see that, on the whole, non-incumbents tweet more than incumbents do, and they have a positive correlation between winning and the rate at which they tweet.
It gets even more interesting when we look exclusively at races where there was not an incumbent in the running.
The relationship becomes more strongly positive, and nearly statistically significant, with a .111 p-value.
However, some important relationships were statistically significant.
Like the strong positive relationship between the percentage of votes a candidate won and the amount of PAC money they had shown above.
And, the severe negative relationship between the amount of PAC money a candidate had and their average tweets per day, which is almost at the point of mutual exclusivity.
So, what does that tell us about Twitter and election performance? For one thing, we learned a lot about the relationship between money and the use of Twitter. Candidates with a lot of funding (PAC money specifically) are really not using Twitter. Candidates with very little PAC money, or without any at all are using Twitter much much more. This could be because candidates with a lot of funding have ready access to more expensive, but tried and true campaigning methods, like paid television advertising. The flip side of that, is that candidates without much PAC money may be forced to choose cheaper communication options, like the free website that is Twitter dot com.
It’s also worth considering where these candidates’ voting bases are. It would follow that PAC money, which often reflects the confidence of invested organizations, could be serving as a proxy for the type of base a candidate has. When you look at Twitter demographics, it’s easy to see that candidates who rely on an older base may not rely on Twitter, whereas candidates who are trying to reach younger voters may see Twitter as an important tool.
Whatever the specific relationship between PAC money and Twitter use is, it needs to be studied in and of itself, because it could shed some much-needed light on the impact of PAC money in elections.
It’s also really interesting that the closest the relationship between Twitter and election performance comes to being statistically significant is when incumbents are taken out of the equation. Twitter may be a more useful tool for candidates who are trying to build a base, than it is for candidates who already have one. It would be interesting to see what this pattern looks like in studies that examine multiple election cycles. It would also be interesting to see what this relationship looks like in election cycles moving forward from 2020, when so much of our communication occurs digitally. I would not be all that surprised to see shifts in this pattern as older incumbents retire and new candidates take power.
Ultimately, I think the most important contribution this study makes is in the realm of quantity versus quality on Twitter. This study was not enough to demonstrate a relationship between sheer quantity of Tweets and election performance. However, I am not the first to suggest that quality may be a different story. It is time to take a hard, qualitative look at the type of content candidates are tweeting, how that content is performing on Twitter, and how those variables relate to election performance.
Abigail Bainbridge is a senior at Indiana University. She would like to give a special thanks to her advisor Dr. Jason Peifer, and PhD student Harry Yan, both of the Indiana University Media School, for their insightful help with this research.
Leave a Reply