sentiment analysis made so simple, you'll think we're mental.

Modern President

The present administration has been widely commended for its modern approach to media.

Since the 2008 campaign, President Obama's team has mounted a comprehensive online strategy unmatched by predecessors and GOP opposition. This includes widespread use of social media, revamping of the White House website, deep engagement of the blogosphere, and even an AMA appearance on Reddit in August 2012.

Note: The “Barack Obama” page was created in 2007 and is officially managed by Organizing for Action. This non-profit was responsible for 2008 and 2012 Obama campaigns. The page states that the White House maintains a separate page, but that one sees an order of magnitude less activity and will presumably live on past the current administration.

A tale of two posts

The following analysis uses Sentimental to look more closely at the response to a sample of President Obama's social media activity.

Analyses below reference 2 sample posts, but the commentary and conclusions are widely applicable.

Timeline Photos

We must bring balance to the Force.
3/1/2013 10:41 PMBarack Obama

Timeline Photos

"Now, if Congress allows this meat-cleaver approach to take place, it will jeopardize our military readiness; it will eviscerate job-creating investments in education and energy and medical research." —President Obama on automatic budget cuts scheduled to hit next Friday if Congress doesn't act:
February 20, 2013 03:53 PMBarack Obama

Tremendous engagement

The Obama team generates tremendous engagement through Facebook activity. The page averages just under two posts per day and has received over 35m likes. That puts Obama's in the top 100 of all Facebook pages in terms of total likes, bested mostly by pop culture icons like Justin Bieber with >50m.

Obama posts routinely generate more than 5k comments and 50k likes. Samples below show the vitality of comments in the 1st hour after the post.

Created with Raphaël 2.1.0vitality1.0 hours# commentscumulative49011160
Created with Raphaël 2.1.0vitality1.0 hours# commentscumulative3007850

Facebook as a swing state

One may expect the sentiment of commentary to vary randomly around some average value. Instead, you often see large swings in the cumulative sentiment due to herd behavior on Facebook.

In the case of Obama and other politicians, many posts lead to positive comments early on and then a long tail of negativity. Samples below exemplify this behavior.

Facebook's news feed feature may play a role in exacerbating these swings. Under this explanation, positive comments are expected early on, given the likelihood of Obama fans to follow his page directly. You might call this a 1st order response.

Negative comments would come later, as friends of commenters are referred to the original post by their news feed. These friends of commenters are less likely to be Obama fans, and perhaps most likely to actively contribute when they disagree. You might call this a 2nd and higher order response.

Created with Raphaël 2.1.0sentiment1.0 hoursfeelingcumulative17-3148-145
Created with Raphaël 2.1.0sentiment1.0 hoursfeelingcumulative19-2560-136

Buzz words

One major learning is that comments reflect much more context than simply the specific post to which they are associated. Commenters use Facebook as a forum for discussion as opposed to a place where they can respond directly to posts by the page owner.

An implication is that careful curation of posts can only do so much to control the conversation that ensues.

Samples below show that in the case of Obama's page, sentiment is driven by words covering a wide range of political hot-buttons no matter the content of his original post.



We expect polarization of likes and comments because users are more inclined to contribute time and energy when they strongly agree or disagree with a post. This is consistent with what we observe for posts with very few comments, while posts with many comments (i.e. >1000) typically show a more normal distribution.

Samples below have many comments and show nearly normal distributions.

Created with Raphaël 2.1.0likabilitymost negativemost positive# likes
Created with Raphaël 2.1.0likabilitymost negativemost positive# likes
Created with Raphaël 2.1.0moodmost negativemost positive# comments
Created with Raphaël 2.1.0moodmost negativemost positive# comments

Election results

Posts made in close temporal proximity tend to converge in their summary statistics. This makes sense if you assume (1) the same population of commenters and friends of commenters and (2) that conversation is reflection of external factors as well as the post itself.

Samples below show posts only days apart with similar breakdowns of neutral, positive, and negative comments and identical sentimentality (average sentiment per post). This reinforces the conclusion that posts on this page serve as a forum for discussion but do not necessarily direct the content therein.

Interestingly, these posts and others on Obama's page closely mirror the most recent election results. 2012 general election was split 51% to Barack Obama and 47% to Mitt Romney. Voter turnout was 59% of those eligible. If we take non-voting as a neutral vote, the split becomes 41% neutral, 30% positive for Obama, and 29% negative for Obama.

Created with Raphaël 2.1.0summary-0.1sentimentalitynegative31%neutral38%positive31%
Created with Raphaël 2.1.0summary-0.1sentimentalitynegative33%neutral33%positive34%