Using Twitter data and state-of-the-art topic modeling techniques, we introduce a method to study agenda setting and public communications in state legislatures.
Payson, Julia, Andreu Casas, Jonathan Nagler, Richard Bonneau, and Joshua A. Tucker. "Using Social Media Data to Reveal Patterns of Policy Engagement in State Legislatures." State Politics & Policy Quarterly, (2022). https://doi.org/10.1017/spq.2022.1
Oct 18, 2022
Area of Study
State governments are tasked with making important policy decisions in the United States. How do state legislators use their public communications—particularly social media—to engage with policy debates? Due to previous data limitations, we lack systematic information about whether and how state legislators publicly discuss policy and how this behavior varies across contexts. Using Twitter data and state-of-the-art topic modeling techniques, we introduce a method to study state legislator policy priorities and apply the method to 15 US states in 2018. We show that we are able to accurately capture the policy issues discussed by state legislators with substantially more accuracy than existing methods. We then present initial findings that validate the method and speak to debates in the literature. The paper concludes by discussing promising avenues for future state politics research using this new approach.
Despite the profound impact that state governments have on people’s day-to-day lives, limited research has been done to understand how lawmakers publicly communicate policies at the state level. Previously, this kind of research was made challenging by data limitations. For example, scholars lacked a systemic and widely applicable way to measure state-level public communications such as press releases. In recent years, a growing body of literature has been developed to shed light on the way members of Congress discuss policy on social media platforms, mainly Twitter, but this has rarely been applied to state lawmakers. Our study begins to fill this gap and advances our understanding of how state policy communications vary across localized political contexts.
Using a combination of data from the Google Civic API and Twitter REST API, our study looked at all tweets made in 2018 by legislators from a diverse sample of 15 states across the U.S. To identify the policy topics discussed in these tweets (a total of 576,461 messages) we introduced a word-embedding model and convolutional neural net trained on a set of 21 policy topics from the Comparative Agendas Project codebook, a comprehensive and widely used framework for classifying political agendas. With the data classified by our topic modeling method, we performed a variety of exploratory analyses to validate our method and uncover patterns in the way state lawmakers in different political contexts discuss their policies on Twitter.
Before examining the policy content of the tweets in our sample, we began by getting a baseline understanding of the presence of state lawmakers on Twitter. Overall, we found that 75 percent of the policymakers in our 15 state sample had a Twitter account and tweeted on average once a day. Our topic modeling approach classified the policy content of tweets in our sample with 80 percent accuracy and found that 70 percent of lawmakers discussed policy-relevant issues in their tweets.
We also looked into differences in Twitter engagement and usage by lawmakers in states with different political contexts. For example, we find that while Twitter adoption is more common among state lawmakers in more professionalized legislatures, the rate of discussing policy areas is roughly equal across more and less professionalized states (conditional on being on the platform). We also discover that, contrary to existing theory, state legislators from competitive electoral districts are more likely to engage in policy debates than those from less competitive districts. Finally, we provide new evidence about which issues individual legislators discuss and whether state legislators focus on different topics than members of Congress. We show that legislators sitting on policy-relevant committees are generally more likely to tweet about the policy topic associated with that committee. Consistent with federalism literature on policy domains, we also find that state lawmakers are more likely than national legislators to discuss the areas traditionally associated with state politics — such as education and crime — and less likely to weigh in on national issues such as foreign trade.
Our method and findings provide a new way to systematically understand state-level political communication and have important implications for researchers studying agenda setting, policy-making, and communication at the state level.