Australasian Science: Australia's authority on science since 1938

Mapping Happiness

By Tim Olds

Blogs, tweets, news reports and songs can be used to map happiness levels by city, age group and even the day of the week and the time of the day.

Happiness is good for us. Happier people have stronger immune systems, better mental health, live longer lives, earn more money and are more likely to have stable marriages.

These associations are enduring. One study found that people who were happy in their first year at university had higher salaries 16 years later, even when allowance was made for initial wealth differences.

How do we know if someone is happy? The easiest way is to ask them. The typical question uses a scale looking something like this: “In general, I consider myself not a very happy person (1) … a very happy person (7)”. If we ask lots of people, we can work out patterns of happiness in time and space. We find out, for example, that people who exercise or spend a lot of time with friends and family or are religious are happier. But asking people can be a tiresome and costly business.

Recently an ingenious method has been developed to map happiness. “Textual hedonometrics” analyses very large corpuses of words, such as blogs, tweets, newspaper articles and song lyrics. One method is to use a sizeable sample of words that have been rated as happy or unhappy.

The ANEW corpus consists of 1034 words that have been rated on a 9-point happy/unhappy scale by a large group of respondents. For example, the word “triumphant” rates 8.82, while “suicide” scores 1.25. Texts can be analysed by the average score of ANEW words within them. Consider this phrase: “The quick (6.64) brown fox jumps over the lazy (4.38) dog (7.57)”. The average happiness rating is (6.64 + 4.38 + 7.57)/3 = 6.20.

This method can be applied to historical texts such as newspaper articles or song lyrics. One study analysed the lyrics of 232,754 songs, finding that the overall happiness rating has declined consistently since 1960 by about 0.6 points. The most “up” genre is gospel while the darkest is heavy metal. The happiest artists include Perry Como, Diana Ross, Buddy Holly and the Beach Boys (all above 7); the least happy (I have never heard of any of them, probably a good thing) were Slayer, Misfits and Staind (all below 5). These techniques have also been used to analyse State of the Union addresses – Kennedy is the most upbeat, and Madison the gloomiest.

Large corpuses such as blogs also allow “hedonometers” to plot the distribution of happiness across age groups (happiness peaks in your 50s), geographical regions (around latitude 35° is optimal — the Mediterranean, California and Adelaide), time of the day (happiest at 5–6 am, unhappiest at 10–11 pm), days of the week (peaks on Saturdays, bottoms on Tuesdays) and time of year (peaks at Christmas). We can also track the rise and fall of individual words in tweets, with “love”, “happy” and fun” peaking on Saturdays while “bad”, “homework” and “shit” peak on Tuesdays.

An extension of the corpus method is to use collocation: looking at the other words that co-occur with “happy” words. One Swedish study looked at words co-occurring with the Swedish word for “happiness” in a corpus of Swedish newspaper articles. The words most commonly associated included “Daniel” and “Victoria” (the royal couple who were recently married), “Zlatan” (Ibrahimovic, the Swedish and PSG soccer superstar), “life”, “dad”, “grandma” and “mum”. The words least likely to co-occur with “happiness” were “Windows”, “Samsung”, “kroner” and “police”. Hard to disagree with that.

Early next Saturday morning, as we approach Christmas, I, as a 57-year-old Mac user living in Adelaide, anticipate near-perfect bliss.

Professor Tim Olds leads the Health and Use of Time Group at the Sansom Institute for Health Research, University of South Australia.