Monday 27 December 2010

Policy implications of happiness

If you spend time in the dodgier parts of the interwebs, you'll see arguments that we should replace GDP statistics with gross national happiness statistics and that government should be targeting happiness rather than GDP growth.

We economists are utilitarians; we tend to like GDP growth because it correlates with increased happiness. Sure, you'll see odd things like results showing low incremental happiness gains with increased income beyond a certain threshold, but think about how the happiness numbers are collected: by self-reported happiness on a scale from zero to ten. You start running into right truncation issues pretty quickly. So the guy who answered "seven" when asked how happy he was a decade ago while a student without much money may well answer "seven" when asked again today because he can still imagine being even happier - who can't? Alternatively, imagine if we stopped measuring folks' income directly and started asking, as one Twitterer [update: Tim Harford, thanks for the reminder, LemmusLemmus] suggested a while back, how rich people felt on a scale from zero to ten: "Oh, I just got paid today, so I'll say a 7."

But suppose that we take the happiness directive seriously. What policy implications flow from this kind of work:
Here we explore the extent to which baseline happiness is influenced by genetic variation. Using data from Add Health, we employ a twin study design to show that genetic variation explains about 33% of the variation in happiness, and that the influence of genes varies by gender (women 26%, men 39%) and tends to rise with age. We also present evidence that variation in a specific gene predicts happiness. Individuals with a transcriptionally more efficient version of the serotonin transporter gene (SLC6A4) are significantly more likely to report higher levels of life satisfaction; having one or two alleles of the more efficient type raises the average likelihood of being very satisfied with one's life by 8.5% and 17.3%, respectively. Finally, using data from an independent source (the Framingham Heart Study) we show that a linked single nucleotide polymorphism (rs2020933) in the SLC6A4 gene also predicts life satisfaction. These results are the first to identify a specific gene that may be associated with baseline levels of happiness.
The policy conclusion seems obvious to me: provide a baby subsidy to folks with the transcriptionally efficient alleles. Make the magnitudes big enough to matter. Within a few generations, we have an expectationally happier population. Look again at the magnitude of the differences reported: they're big relative to the effects of other shocks, and they're baseline rather than transitory. If what we care about is national happiness, it's hard to think of anything that would be more effective. Now, I wouldn't advocate the policy. But neither would I advocate the other policies that get pushed by the happy people.

The excellent James Fowler is one of the authors of the paper and sensibly avoids policy conclusions. He instead makes even more interesting suggestions about using genes as instruments in disentangling endogeneity problems.


  1. Leaving policy implications aside, it may also be that any such genetic effects are wholly or partly effects on reporting style rather than actual subjective well-being, as suggested by Judith Rich Harris more than ten years ago. (This is not acknowledged by the authors in the intro, discussion or conclusion of the paper, but I have not read the rest of it.)

    Somewhat related, see this post about self-reported psychological well-being as associated with self-reported vs. objective measures of health: "The large apparent influence of stress on incident angina was probably seen because the people who reported high stress also reported other forms of discomfort in their lives, including chest pain. This was obviously not due to there being any actual stress-related coronary disease, otherwise it would have been revealed in incident ischaemia and cardiovascular disease mortality."

    P.S.: "one Twitterer".

  2. When I was studying econometrics at university I was told that trying to to explain a stationary process (like the output on a scale of 1 to 10) with non-stationary processes that aren't cointegrated (like GDP) then you've made a fundamental specification error.

    I find it troubling that so much happiness research is making such a fundamental error.