SUMMARY:
In the two days after Democratic Sen. Heidi Heitkamp courageously voted against the confirmation of Brett Kavanaugh, she was subject to a surge of condemnation on Twitter.
We examined the accounts retweeting her Republican rival’s most popular anti-Heitkamp Tweet, and found that nearly 75% are very likely to have come from out of district – thus, saying very little about the feelings of her actual constituents.
MAIN TEXT:
For the 48 hours after Sen. Heitkamp’s vote against Supreme Court candidate Brett Kavanaugh, the #NDSen district was among Pollchatter.org‘s most active districts.
More than 10,000 accounts retweeted this post from Heitkamp’s GOP rival, Kevin Cramer, accusing her of siding with “Chuck Schumer, and not the people of North Dakota.”
In a district where daily total Tweet activity is more often in the low four figures, this seemed worthy of a second look. Thus, we examined all the accounts that retweeted Cramer’s post, asking whether they were likely to be actual North Dakota voters or not.
No easy answer
Twitter accounts do include an attribute for “location.” In an ideal (or possibly dystopian) world, it would be a simple thing to look at this attribute and immediately glean a user’s actual location.
In practice, this information is unreliable at best, and outright manipulative at worst. Many people leave this blank, or treat it as a joke (e.g., “Mars”, “Your momma’s bedroom”). For those that do identify a location (e.g., “Bismarck, North Dakota”), there is no way to verify they are telling the truth.
A separate Twitter function allows users to add geolocation coordinates to their posts, with precision up to the level of actual latitude and longitude possible. However, this is optional, rarely used, and can presumably also be manipulated.
Looking at behavior
Thus, rather than trying to locate account-holders on a map, we looked at their past behavior. In doing so, we made the following assumptions:
- Any accounts that post only within #NDSen are fairly likely to be North Dakota residents or voters.
- Accounts that post consistently about other districts in addition to #NDSen are somewhat less likely to be North Dakota residents or voters.
- Accounts that consistently post about other districts more often than about #NDSen are considerably less likely to be North Dakota residents or voters.
The numbers
The Pollchatter.org database contains a total of 10,735 retweets of the original Cramer tweet. That’s about 1,000 less than the 11,769 retweets shown by Twitter, as of this writing. The discrepancy is due to the fact that we search for election-related content using Twitter’s free API, which does not provide access to Twitter’s entire body of content.
Our examination of these 10,735 accounts looks at behavior going back to Aug. 26, or about six weeks. The posts in the database include only those referencing districts regarded as competitive in the 2018 U.S. elections – thus, the figures do not include posts made about other topics.
In this six-week period:
- 1,745 of the accounts posted only about #NDSen.
- 1,035 posted about multiple districts, but posted most often about #NDSen.
- 7,955 posted about multiple districts, and #NDSen was not their top district of focus.
Give our assumptions above, we would thus say that 2,780 accounts (25.9% of the total) are quite likely to represent Heitkamp constituents, while 7,955 (74.1%) are quite likely to come from outside the district.
We additionally looked at what districts are most frequently mentioned within this group of 10,735 accounts. #NDSen comes in at the top, of course, since this defines the group. But the others are also interesting:
| Rank | District | # of appearances |
|---|---|---|
| 1 | NDSen | 10,735 |
| 2 | TXSen | 7,614 |
| 3 | MOSen | 4,221 |
| 4 | MISen | 3,725 |
| 5 | AZSen | 3,534 |
| 6 | WVSen | 3,514 |
| 7 | MTSen' | 3,013 |
| 8 | INSen | 2,866 |
| 9 | nv03 | 933 |
| 10 | TNSen | 799 |
Notable here is that 96% of the likely out-of-district posters have also posted about #TXSen in the last six weeks. As our earlier analyses have shown, #TXSen has been the subject of a very strong, consistent and ongoing steam of posts and retweets by accounts with very high “Trollbot” ratings, as rated by the Bot Sentinel service.
Because we haven’t done a bot analysis on #NDSen, we’ll stop short of saying the anti-Heitkamp surge was driven by bots or social-media manipulators. However, this very significant overlap with pro-Ted Crux #TXSen traffic is quite interesting.