A group of folks at the University of Vermont measured average happiness on Twitter over a three year period running from September 9, 2008 to August 31, 2011.
While Saturdays (red dots) were typically and not surprisingly high points each week, what is a surprise is that at the end of 2010, excluding holidays, people began getting less happy, a trend which began worsening in 2011.
Here’s a chart (simplified from the original figure 3), showing the end of 2009 through to August 2011 (click to enlarge), with a downward sloping arrow added to the background to show the trend:
To see the full image from mid-2008 through August 2011, go here and click on the left to see “figure 3.” By my analysis, the biggest drop compared to previous years was in late spring/early summer weekdays in 2011. Could a lack of summer jobs for students be a major cause of the drop?
46 billion words from over 63 million users
46 billion words from nearly 4.6 billion expressions over a 33 month span by over 63 million unique users were examined. Here’s a closeup of 2010, where happiness began to trend downwards:
But, what is “Happiness?”
To understand what constituted “happiness,” they conducted a survey to obtain “happiness evaluations” of over 10,000 individual words, which they claim was a tenfold size improvement over similar existing word sets.
As part of the evaluation they obtained 50 independent evaluations per word, asking users to rate how a given word made them feel on a nine point integer scale.
The report is full of fascinating charts, such as these:
- Figure 16. Ambient happiness time series and word shift graphs for tweets containing the keywords ‘Tiger Woods’ and ‘BP’.
- Figure 17. Time series and word shift graphs for tweets containing the keywords ‘Pope’ and ‘Israel’.
- Figure 3. Overall happiness, information, and count time series for all tweets averaged by individual day.
- Figure 5. Average happiness as a function of day of the week for our complete data set.
- Figure 10. Average happiness level according to hour of the day, adjusted for local time.
In Figure 12, shown below cropped for simplicity, the shift in happiness from the happiest hour (5 am to 6 am) relative to the least happy hour (10 pm to 11 pm) is plotted. Go here for the full chart.