This weekend, at a dinner party, I brought up one of my favorite topics, Twitter, knowing it can elicit a wide variety of reactions; from exuberance to disdain. While some at the party claimed to not understand the value of Twitter, since its 2006 launch over 100 million people have joined. With so many people participating and so much information being shared, marketers and business analysts alike are eager to try to dissect the data and understand trends and opportunities.

Studies looking into who uses the site and why are pervasive; however, a new study released last week takes a different angle, examining how happy Twitter users are as the day goes on. The study is entitled Pulse of the Nation - U.S. Mood Throughout the Day, as inferred by Twitter, and was conducted by the College of Computer and Information Science at Northeastern University in conjunction with Harvard University.

Results

It is not going to surprise anyone familiar with common stereotypes about New Yorkers (they're rude, they have superiority complexes etc.) that according to Pulse of the Nation, New York is one of the unhappiest places in the entire country (scoring a pathetic 1.5/8). Just to add some salt to that wound, California appears to be one of the happiest (5.2/8). What about the regions where the majority of our Government officials live? Would you assume people living in Maryland or DC would be happier than the rest of the population? The MD/DC/VA regions ranks as middle of the road in comparison to other regions with a score of 3.13, but surprisingly those in West Virginia are even more unhappy then New Yorkers with a paltry score of just 1.4. In fact, the study registers West Virgina tweets at a happiness score of 1, the least happy on the scale, for the vast majority of the day until the hours of 6:00 pm - 11:00 pm where they average a 2.5 (California's score never falls below a 3). Most regions register a slight uptick in happiness during the early morning hours, and in a move I'm sure designed solely to annoy those on the East Coast, Californians are happiest between 5:00 am - 7:00am PST (what is in their coffee?). I would interpret that early morning happiness as a bit of optimism before the day actually hits people in full force. Personally I like to tweet a nice uplifting quote from someone famous to start my day, and my followers, on a pleasant note.

Research & Data

While many of us are inclined to take these results at face value (because they do seem to mirror our societal stereotypes regarding the types of people that live in different areas of the U.S. ), in looking at these studies it is always important to see how the researchers came to their conclusions. Unlike other studies that use only a small sampling of tweets, the researchers here used over 300 million tweets over a three year period. The point at which I began to raise an eyebrow however, was regarding how they applied the ANEW word list (Bradley, M.M., & Lang, P.J. Affective norms for English Words: Stimuli, instruction manual and affective ratings, The Center for Research in Psychophysiology, University of Florida). 

The ANEW list was created as a part of a 1999 study with the goal of developing "a set of verbal materials that have been rated in terms of pleasure, arousal and dominance". This study had participants rate each word in terms of those three categories on a bubble sheet (see image below).

The first red flag for me are the subjects who participated: they were Introductory Psychology students who "participated as a part of a course requirement." Forced participation in this sense may or may not have affected the results, either way it is something to keep in mind. The words are rated on a scale of 1-9, 1 being the least happy, 9 being the most (for the purpose of this article I will only be focusing on the "Pleasure" (or Valance, as they refer to it in the ANEW study) ranking as that's what was taken into consideration for this recent Twitter study). Also, the "Pulse of the Nation" study only shows eight different rankings on their charts, instead of the nine in this study. I point that out not because I think it's particularly relevant, but to avoid confusion.

Lacking Context, Intonation and Modifiers

The word rankings seem to intuitively make sense, "Christmas" and "Friend" score high (7.8 and 7.7 respectively) whereas "grief" and "hatred" score very low (1.7 and 1.9 respectively). However, these words are being scored outside of any kind of context. Imagine two similar tweets; 1) I need a vacation vs. 2) I am on vacation. These two tweets would receive the same mood score, yet are actually two very different mood states. While this is only one example, we could come up with hundreds or thousands more. Furthermore, this method cannot take into consideration intonation and intent. Sarcasm, even in writing, can drastically change the mood.  For example, a friend of mine tweeted this morning "Dear Monday, I LOVE waking up to you. Oy. Wish I was still asleep!". "Love" is rated at 8.6 and "sleep" gets a 7.2, this study infers that this Twitter user is very happy at the time of posting this tweet, I would beg to differ.Finally, one of the most commonly used methods we have adapted since so much of our conversation has become digital is the emoticon. A sentence that might be taken out of context such as "OMG I hate you" is different when you add a simple emoticon at the end, "OMG I hate you ;)". Unfortunately, when this ANEW study came out in 1999 I highly doubt they thought of adjusting the value of words depending on a modifier such as an emoticon.

Bottom Line

The study is interesting, it makes good dinner conversation, but it is certainly not a definitive look at the mood across the U.S. There is far too much in our written speech patterns that cannot be dissolved down and valued in this way at this level, there are too many nuances. It does open the door to more studies and more analysis on Twitter data. When you have such an enormous group of people conversing openly over a medium such as this, there is bound to be fascinating analysis that can be pulled out.  While this study may miss the mark on a definitive answer to the relationship between mood and geography of Twitter users, it does encourage further analysis.