Back To Schedule
Wednesday, July 13 • 10:46 - 12:15
Detecting Well-Established Trends about Political Affiliation and Affect in Facebook Microblogs

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Location: PSH (Professor Stuart Hall Building) - 314, 
Goldsmiths, University of London, Building 2
Campus Map 

  • Roxanne B. Raine, University of Hawaii at Manoa, United States
  • Scott P. Robertson, University of Hawaii at Manoa, United States


It is well-established that conservatives report higher life satisfaction than liberals (Napier & Jost, 2008; Alesina, Di Tella, & MacCulloch, 2004; Taylor, Funk, & Craighill, 2006), even when controlling for potential confounds such as household income, age, education, and numerous other factors. Although the finding that conservatives tend to seem happier than liberals is not new, our research contributes to the existing body of literature in two ways. First, the detection of this trend in microblogs is novel, as previous findings were based on surveys about life-satisfaction. And second, by analyzing the content of the microblogs, we gain insight into the reality behind this well-known trend. Further plans for research are also discussed. 


We investigate whether these life-satisfaction differences are detectable between Facebook status updates of liberal versus conservative Americans.


Our source of Facebook updates was www.myPersonality.org, which provides over 4,000,000 individuals’ Facebook profile information (Kosinski et al., 2015). The myPersonality project is affiliated with 250 researchers and over 32 publications (e.g., Youyou, Kosinski & Stillwell, 2015; Lamiotte & Kosinski, 2014).

The dataset of 16,906 users’ self-proclaimed political affiliations in the myPersonality database is comprised of 144 categories because Facebook does not have any constraints on what can be entered in this field. For some of the affiliations (e.g., “democrat”), a user’s status as liberal is clear. However, some are less objectively categorized, so we had 13 people interpret the affiliations. We narrowed the users into 3 types of voters: conservatives (3,622), liberals (5,333), and either (2,496). Probable non-voters were excluded. Rationale for this will be explained in the presentation.

Example affiliations in each category:

Liberal - Obama baby, Democrat, Liberal

Conservative - Conservative, Nobama, Republican

Either -  Depends, I don’t know, Middle of the road

Neither - Who cares?, Bullshit, Anarchy

We then compared groups using Linguistic Analyisis and Word Count (LIWC) emotionality data (Tausczik & Pennebaker, 2010). LIWC measures a corpus’ emotionality based on word frequency. For example, positive emotion words such as love, nice, and sweet increase a body of text’s positive emotion score, whereas hurt, ugly, and nasty increase negative emotion scores. We compared these two LIWC scores for our users in the three voting groups: liberal, conservative, and either.


An ANOVA comparing positive and negative emotion words in status updates of conservatives, liberals, and swing voters showed significant differences for positive emotions F(2, 11,448) = 25.93, p < 0.001 and for negative emotions F(2, 11,448) = 25.93, p < 0.001. Fisher LSD post-hoc analyses showed that liberals and conservatives were significantly different in both ANOVAs. The “either” group did not have any significant differences.

Future Work: 

There are a number of possible explanations for why liberals are less happy than conservatives. Although these possibilities have been discussed in previous literature, no conclusions have yet been made. One possibility is that one tends to associate more with members of the same political party. Over time, the peer group’s affect could converge, causing this emotional heterogeneity between groups. Another alternative is that the worldview that leads one to become liberal or conservative is at the root of the language differences we have found. If that is the case, the content of the status updates could provide more insight, which will also be discussed in the presentation.  

Although numerous other studies have found that liberals tend to be less happy than conservatives, to our knowledge, this is the first study to show that the everyday Facebook language of liberals versus conservatives reflects the previously found affect differences between groups. We plan to continue this research with empirical studies to determine whether we can reverse the effect. If the effect is reversible, that would provide evidence that the peer groups are driving the language. We will also continue our text analysis by evaluating the content of the updates, which will lead to a deeper understanding of the types of positive and negative statements being made by each group. These findings could influence politics, commerce, and social media design.


Alesina, A., Di Tella, R., & MacCulloch, R. (2004). Inequality and happiness: are Europeans and Americans different? Journal of Public Economics, 88(9-10), 2009–2042. http://doi.org/10.1016/j.jpubeco.2003.07.006

Kosinski, M., Matz, S., Gosling, S., Popov, V. & Stillwell, D. (2015). Facebook as a social science research tool: Opportunities, challenges, ethical considerations and practical guidelines. American Psychologist, 70(6), 543-556.

Lambiotte, R. & Kosinski, M. (2014). Tracking the digital footprints of personality. Proceedings of the Institute of Electrical and Electronics Engineers (IEEE), 102(12), 1934-1939.

Napier, J. L., & Jost, J. T. (2008). Why Are Conservatives Happier Than Liberals? Psychological Science, 19(6), 565–572. http://doi.org/10.1111/j.1467-9280.2008.02124.x

Tausczik & Pennebaker, J.W. (2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology 29(1), 24-54.

Taylor, P., Funk, C., & Craighill, P. (2006). Are we happy yet? Pew Research Center social trends report. (M. A. Motes, Ed.) PloS one (8). http://doi.org/10.1371/journal.pone.0083143

Youyou, W., Kosinski, M., & Stillwell, D. (2015). Computer-based personality judgments are more accurate than those made by humans. Proceedings Of The National Academy Of Sciences (PNAS), 112(4), 1036-1040. http://www.pnas.org/content/112/4/1036.full

Wednesday July 13, 2016 10:46 - 12:15 UTC
PSH (Professor Stuart Hall Building) - 314 Goldsmiths University, Building 2