American Social Media Response to January 2018 Government Shutdown

Washington DC – On Monday, January 23, 2018, the U.S. Congress reached an agreement to reopen the government for a few weeks while they fight over spending priorities. So, this will be the last release (for now) of our tracker of the American mood via social media. Again, this is not a survey, and we cannot extrapolate how all Americans feel about the issue. For survey findings, check out our recent poll release with Reuters (forthcoming). 

One full day into the shutdown, and Americans are increasingly frustrated with both parties. The number of Americans on social media blaming both Republicans and Democrats has climbed to 78% on Sunday from 60% Saturday. Fewer than a quarter blame one party more than another. Among these, Republicans and Democrats receive roughly equal amounts of blame at 13% and 9%. 

Social media discussion of the shutdown continues across the entire United States. The coasts continue to see more anti-Republican messaging but anti-Democratic messaging has surged in several Republican strongholds such as Texas, the Deep South, and Florida. 

Anti-Republican messaging (hashtags) dominates Sunday with significantly more #trumpshutdown and associated compared to #schumershutdown. However, activity is relatively balanced in the Southern United States. 

Anti-Republican messaging (i.e. #TrumpShutdown) continued to outpace anti-Democratic messaging (#ShumerShutdown) on day two of the shutdown, but the gap narrowed throughout the day. Towards the end of Sunday, anti-Democratic messaging saw a notable uptick putting it slightly ahead of anti-Republican messaging. 

METHODOLOGY

• Step 1: A random sample of ~420,000 tweets, Facebook and Reddit statements about the government shutdown was collected from 0000 EST/2400 EST on 1/21/2018. The sample included both Democrat-leaning (e.g. #Trumpshutdown), Republican-leaning (e.g. #Schumershutdown) and overtly neutral (e.g. #govtshutdown) hashtags and thematic statements. The sample represented an estimated 1/10 of the entire conversation taking place during this time. 
• Step 2: The average compound sentiment score (CSS) of comments mentioning different parties was calculated. The compound sentiment score is a continuous variable going from negative to positive, and takes into consideration frequency of appearance of actors, organizations or topics next to, or associated with negative, positive and neutral comments. The lower the number, the more negative the sentiment. The CSS was calculated utilizing VADER lexicon in NLTK toolkit. (https://github.com/cjhutto/vaderSentiment) which has been specifically designed to extract measures of sentiment from social media. The CSS provides a useful characterization of whom did the American social media public blame for government shutdown.    
  • Reference: Hutto, C.J. & Gilbert, E.E. (2014). VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Text. Eighth International Conference on Weblogs and Social Media (ICWSM-14). Ann Arbor, MI, June 2014.

• Step 3: The entire CSS-enriched sample was geo-coded and aggregated at the state level in order to provide a nation-wide view of the blame-related discussion. The characteristics included: 1) blame Republicans 2) blame Democrats 3) blame both 4) anti-party/anti-government messaging.  

• Step 4: Velocity of the anti-Republican (e.g. #Trumpshutdown) vs. anti-Democratic (e.g. #Schumershutdown) messaging was calculated for entire time frame of the sample. Velocity is defined as speed of messaging per hour. Velocity provides a useful characteristic as to which side's narrative is winning and also helps to identify as to when a particular digital narrative is going on offensive.