The hidden (and not so hidden) messages your congressional representative is communicating.

Spread the love

There are plenty of good ideas about how to solve complex social problems. One of the major levers is policy. And the people responsible for brokering between local needs and large-scale, national implementation are the U.S. Congressional Representatives.

As the primary connector between the local and the national, communication is essential. These days, Twitter is one of the primary mechanisms that representatives use to get messages out.

I personally live in the 9th District of PA, which is represented by Republican Dan Meuser, just starting his second term. Less than a mile down the road, the boundary switches to the 6th District, represented by Democrat Chrissy Houlahan, also starting term number two. In a few short weeks, PubTrawlr headquarters will be moving from the 9th District to downtown Reading, within the boundaries of the 6th. Because it’s really important for us and other civically-minded people to be connected to the messaging, we developed a little app to analyze the messages coming out of Twitter accounts.

For this post, we’re going to be comparing Rep. Houlahan and Rep. Meuser’s Twitter feeds, as of last week. Both started their official Twitter accounts within two weeks of each other, and Rep. Houlahan tweets almost twice as much.

We have to acknowledge that many of the tweets are likely not written by the representatives themselves, but by their communication team.

Word and Word Phrases

This first set of graphs use the bag-of-words approach to see what word and word phrases come up most frequently in their tweets. I removed what are known as “stop words” first. These are words like, “the” and “of” that add little informational value. You can click on these and the following graphs to expand them.

The network plots show a bit more nuance about the types of things that each representative communicates. For both, we see shoutouts to constituents and process notes (“bipartisan bill”). Rep. Houlahan talks more about specific issues than Rep. Meuser. We can see tweet patterns in her tweets around health care, background checks, COVID relief, and climate change.

Rep. Meuser tweets more about process than specific issues, though he notes border security and COVID. In the right-hand network plot below, the cluster of numbers in the upper left corner corresponds to tweets where he mentions the TV channels where his media appearances can be found (see also the mentions of Newsmax and Volpe Report, and tomorrow morning.)


Words don’t just communicate ideas. They can communicate emotion. Sentiment Analysis looks at the emotional content behind the words. I used the bing lexicon to look at what positive or negative words the representatives used. I averaged the net emotionality of tweets in a given month (whether it was positive or negative), then plotted this over time. Each dot in the graph represented the net emotion for the month.

Both representatives have pretty positive messaging, or at least more positive than negative. Rep. Houlahan has a wide spectrum within the positive, with Rep. Meuser being a bit more subdued.

The NRC lexicon tries to be a bit more nuanced by assigning more specific emotions to words, like anger, joy, and trust. The below graphs show the proportion of tweets with a given emotion over time. Like above, these are clustered by month.

I expected more distinction between the representatives, but both are pretty similar. Rep. Meuser uses words with fear and trust more, whereas Rep. Houlahan has more of an even distribution.


I then used a clustering algorithm to see what topics the representative tweeted about. The words on the side of each plot are the “most distinguishing keywords.” Rep. Houlahan’s top three topics appear to be about her service as a veteran, educational issues, and violence. Rep. Meuser’s top topics are about his media appearances (recall the network plot about), border security, and veteran’s affairs.

Most Popular Tweets

It’s interesting for posterity, but not necessarily for representativeness, to identify the “most favorited” tweet. I’ve embedded these below.

What to do with this information

Twitter is an important method of public discourse for our elected officials. It’s one of the main methods that they use to communicate policy position and their perspectives on issues that concern their constituents. By looking at these feeds (and the feeds of other officials), we can get a sense of what they think is important for others to know about. And, if this misaligns with what their constituents think is important, it can be a way to start holding them accountable.

Taking it a step further

Could we maybe use emotion as a predictor of engagement? For example, are posts that convey fear more likely to get the attention of readers? In future posts, we’ll dig more deeply into this. This tool will be going live on our sister site, Readiness Learning Systems by the end of next month, so you’ll be able to try it yourself!

Oh, and how about PubTrawlr?
Who has time to stay on top of the most recent and current edge scientific research?
Well, you do, with PubTrawlr.

Head over to our homepage to search for topics and trends relevant to the work you do in the organizations and community you live in.
Hit the button below to head to our homepage.

Leave a Reply

Your email address will not be published. Required fields are marked *