Happiness over lifetimes


In a recent study withTony Beatton (QUT), I looked at how happiness changes by age. For the freely downloadable working paper version, see here. 

What got me into this issue is the dominance of the “U-shape” story in the economic literature on happiness. The dominant story is that we get miserable in mid-life as the stresses of life are piled onto us and then get happier again the closer we get to death. It didnt seem right, either from the point of view of raw data nor intuitively, so we set to work. The main findings of our study can be summarised by the graph below, where you can see the way happiness changes over the lifetime in Australia, the UK and Germany for a representative individual starting at a 7 on a 0-10 scale.

This Graph summarises data from over 50,000 individuals in Australia, the UK, and Germany, followed over more than 10 years. It turns out that individuals are happiest early into retirement years (65-75) and at their most miserable close to death (80-90), with relatively little changing in the years before retirement. Our interpretation is that individuals older than 65 no longer have unrealistic expectations of what their life will be like and simply enjoy reasonable health and wealth, leading to a marked surge in happiness. As their health starts to deteriorate after 75, their happiness plunges.

In particular:

  1. For Australia, in a sample of over 10,000 Australians followed for over 10 years (the HILDA dataset), the happiness of a representative person who is a 7 on a 10-point scale hardly experiences any change in his happiness until about age 50 after which happiness climbs to 7.3 at age 65, dropping off increasingly fast as death approaches, being no more than 6.5 when over 90 years old. In the picture it shows up as the blue line which tells you how the happiness of this Australian actually changes as he ages.
  2. For the UK, it is a similar story, but the happiness peak is slightly later (age 70) and the peak itself is less high reaching only 7.2, declining to 6.3 when over 90. Life in old age is clearly relatively better in Australia than the UK, perchance because of the better weather, more generous public pensions, and more space. The BHPS, which tracks about 25,000 individuals since 1991, yields the green curve in the graph.
  3. In Germany, the story is slightly different. There is still a happiness peak at age 65, but it is preceded by a reduction in happiness during early adulthood, meaning that the peak is no higher than 7, with a sharp happiness drop after age 75, leaving individuals over the age of 90 with happiness levels around 5.8. Life simply only gets worse in Germany after the age of 18. Also, we found that there were severe data problems in Germany, with only quite miserable people in middle-age prepared to partake in the sample and respondents becoming markedly more honest (and miserable) as they answered the happiness questions year after year. The 20,000 individuals followed since 1984 reveal the dotted red line.
  4. We found that previous studies severely underestimated the degree to which miserable people in middle-age were over-represented in these datasets: happy people in middle age are busy and don’t have time to participate in lengthy questionnaires, leading previous studies to erroneously think there was a huge degree of unhappiness in middle-age. When you actually follow people over time, no such ‘middle-age blues pattern’ can be found, at least not in Australia or the UK and only to a mild degree in Germany.
6 Responses to "Happiness over lifetimes"
  1. Hi Paul

    I wonder about the old-age tail here. Is there a selection bias? Maybe grumpy-old men (and women) live longer so that the happy ones die young (leaving an apparent increase in unhappiness as we look at an older cohort)? As someone who is both grumpy and intends to live forever, I plan to be stuffing up the tail of the distribution in a few decades.

  2. I was going to make Stephen’s exact point, although I planned on not being grumpy and living for quite a while.

    If your graph is reliable only up to about 65-70 years old, you seem to have confirmed the inverted U-shape once again.

    The median age at death in Australia for men is 76, and women is 82 (in 2002), which is probably a little higher than what is representative of the cohorts in the surveys.

    There is also the whole of childhood missing from your graph. I realise children probably aren’t in the survey, but could it not be that the Germans simply take a little longer to get unhappy? They take until their late teens/early twenties, while UK and Aus might get unhappy earlier in their teens. Something to do with schooling/average length of schooling, age when responsibility kicks in? Or compulsory national service in Germany?

  3. Stephen,

    yes, selection is very important, but both types arise: selectivity on the living and on those doing the survey are important (and both are corrected for in the graph). Not only do the happier live longer but the really unhappy ones close to death stop answering the survey. By following people over time we dealt with both issues. So the question is not just whether you intend to remain alive and grumpy but whether on your deathbed you want to record your misery in an interview too!


    sure, if you take the happiness decline close to death out of the equation you get closer to a U-shape, though even then you will note the lack of a major downswing in mid-life which is what that whole literature is about.
    Yes, there is some evidence that younger kids are happier than 20-year olds in Australia and that this has something to do with the pangs of dealing with change. Though i dont know of literature on German kids, I would expect them to be a lot less happy than Australian kids.

  4. “Not only do the happier live longer but the really unhappy ones close to death stop answering the survey. By following people over time we dealt with both issues.”

    How was that? Did you fill in the missing values these people would have given if they were alive or had responded to the survey, by extrapolating from their previous answers?

  5. Rob,

    for the graph in the post we used individual fixed-effects and time-in-panel. In the paper we do a lot more though, including correctoins for health and the like.

  6. Thanks Paul. Sounds like those techniques would effectively forecast what the missing values would have been.

    You might like to give a title to this post – not having one makes it look strange on the site.

%d bloggers like this: