Americans ‘support’ the idea of tuition-free college: an exploration of sentiment and political identity signals otherwise

This blog details my recently released sentiment analysis research regarding the online reactions to America’s College Promise (ACP). Although this research captured commentary in 2015, the analysis remains critically important as once again national tuition-free college is a platform for many 2020 Democratic hopefuls and reactions to Sen. Warren’s plan mimics what my team captured in this study.

In 2015, when I was taking a set of methods courses at the iSchool at the University of Illinois at Urbana-Champaign, President Obama announced America’s College Promise. ACP was envisioned to be a Tennessee Promise-like tuition-free policy for the nation. The public debate surrounding the policy was abnormally robust, given usual debates on higher education policy. For several days, social media was active with many providing an opinion on the viability and nature of ACP.

With my newfound data-scraping skills and understanding of sentiment analysis*, I decided to capture the social media commentary from four media sources — Fox’s Hannity, CNN Money, NBC News, and The White House. These national sources were chosen in part due to national audiences of millions of followers, the volume of comments available, and the belief the sources would provide a wide range of ideological commentary. In total, these sources housed over 66,000 comments.

Often in sentiment analysis, researchers would apply an already developed corpus to the commentary and develop unsupervised or supervised analyses. However, a corpus did not yet adequately exist for this topic and we wanted to take our analysis a step further by combining multiple sentiments for greater nuance. We also decided to capture attributes of the commentators, which Facebook makes nearly impossible to automatically capture unless you’re Cambridge Analytical or have a boatload of money. To make the process manageable, we decided to code the top 1,000 comments per source as determined by “likes.”

My team coded each comment based on three sentiments: (1) being for or against the policy, (2) engaging in civil or uncivil communication, and (3) staying on-topic. We then employed a Bag of Words model and used these codes to capture the tokens (words and phrases) that are most likely to appear in each label. We also generated combined sentiment models, such as being Against/Off-Topic/Uncivil. After coding the commentary, we then manually visited commentators’ Facebook profiles to capture gender, race, age, and likely political leaning. The paper has clear examples of our full methodology including samples of our coding. As a side note, I learned where many of these people lived, where they worked, places they frequently visited, and more — I would suggest more of you find your privacy settings and activate those.

We found most comments were situated against (62%) ACP and that commentators against the policy employed free-ride rhetoric to cement their opposition. Free-ride rhetoric has long been used to oppose social-support policies, especially policies that may advantage lower and middle-earners and persons of color. On the other hand, affirmative commentators seemed to lack a unified message of support. Instead, the tokens suggest general affirmation for the policy and President Obama but little substance in attempting to combat the free-ride narrative used by the opposition. Shocking nobody, we found that commentators who leaned conservative, were men, were White, and commentated on the Fox News site were all more likely to be against the policy. To be noted, we do not know if these groups were against ACP because of the social-support nature of the policy or because President Obama was the politician to offer it — we just saw the existing correlation with being opposed to the policy and these social categorizations. Today, a cursory glance of social media commentary on Sen. Warren’s national tuition-free policy suggests the same patterns are replaying several years later.

Evident to me, those who oppose tuition-free policy have a common narrative, a rallying cry and target rhetoric to adhere to. Yet, those who support tuition-free college appear to be more disjointed preferring to discuss what other countries have and shying away from the historical changes of higher education in the U.S. or the potential economic returns. I encounter this issue often with my Kalamazoo Promise research in that those who support tuition-free policy think it is a “good idea” or could vaguely help with college, but are unable to conceptualize the total value of the Promise in terms of benefits to the K-12 system, to college access, retention and persistence, or post-college economic benefits for the individual and surrounding areas.

The inability to conceptualize these discussion points may partly exist because limited research exists on Promise policies, see W.E. Upjohn for some and here for Tennessee Promise. And, as I see it, for decades conservatives have used free-ride rhetoric to pry higher education away from the social good and liberals have never developed an effective, unified counternarrative. As I intend to reapply our models to recent commentary on Sen. Warren’s plan, I will be curious whether supporters have become more centralized around concrete themes of “investment” and “opportunity” or if general praise for the politician promoting the policy remains more dominant.

Next, we found that half of the comments were civil. Therefore, suggesting the existence of pockets of commentators genuinely interested in having a conversation surrounding the pros and cons of a national tuition-free policy. When I discuss this study, many are shocked that only half of the comments are civil. Such a response suggests to me that we have grown too accustomed to uncivil discourse. The expectation now is that liberals will call conservatives “uneducated” or “ignorant” and conservatives will call liberals “idiots” and “communist.” I would expect that with increased political polarization that overall engagement in uncivil commentary may now consist of a higher percentage than what we captured.

Since starting this project, I have become more interested in uncivil discourse as I feel the breakdown in the social fabric that binds us is being driven by the normalization of uncivil commentary on social media. Normalizing uncivil responses have pushed us further away from each other and made us vulnerable to social discord generated by non-American entities. Personally speaking, this research has forced me to reexamine my online interactions so that I am not contributing to the uncivil communication we have come to expect. Additionally, this introspection helps me facilitate stronger online conversations on the value and drawbacks of tuition-free policy and my research. I no longer choose to engage in online dracarys.


I leave you with this - if you’d like to chat about tuition-free college policy hit me up on Twitter at @Dcollier74.

Daniel Collier is a recent alumnus of the University of Illinois at Urbana/Champaign and current Senior Post-Doctoral Research Associate for the Center for Research on Instructional Change in Postsecondary Education (CRICPE) at Western Michigan University. Email him at

#SentimentAnalysis #BagofWords #TuitionFreeCollege #AmericasCollegePromise #Warren2020 #Civility #SocialMediaAnalysis #FoxNews #CNN #NBC #WhiteHouse #Dracarys

*My man Shub Mishra was the Sentiment Analysis All-Star here though.