Life satisfaction, GDP and sense

Normally, I’m a sunny-natured optimist but every so often I get grumpy. A recent post on Vox made me grumpy – it’s about life satisfaction and GDP. At first I thought I’d ignore it. But this morning I was dipping into A Century in Books, the little volume from 2005 celebrating the centenary of Princeton University Press, and it fell open at the page on Clive Granger’s 1964 book, Spectral Analysis of Economic Time Series. My PhD used a lot of time series econometrics – Mark Watson, now at Princeton, was one of my supervisors. I did my PhD so long ago that it’s now Once Upon A Time, and it is simply depressing that a generation after time series econometrics matured so many economists fail to think about the statistical properties of different kinds of time series data –  like GDP and reported life satisfaction.

(Actually – this is a rant for another time – it’s depressing that so many economists aren’t interested in data and statistics at all, but just expect to be able to download data files and run them through packages that churn out impressive-looking test statistics, without ever pausing to think about how the statistics are constructed or what the regressions might really mean. See Deirdre McCloskey eg [amazon_link id=”1843761742″ target=”_blank” ]Measurement and Meaning in Economics[/amazon_link].)

One of the first things you learn about in time series econometrics is the importance of understanding whether your data have the property of stationarity or not. (It’s on page 3 of my antique textbook, Granger and Newbold’s [amazon_link id=”0199587159″ target=”_blank” ]Forecasting Economic Time Series [/amazon_link] and no doubt equally early in [amazon_link id=”0521634806″ target=”_blank” ]Hendry and Clements[/amazon_link].) In other words, does the series drift over time far away from where it started? If so, it is non-stationary. GDP is like this; it is an analytic construct with no theoretical upper limit. Life satisfaction, however, is a stationary time series – in the World Values Survey it is measured on a scale of 1 to 10 – so it can never go above 10 over time.

So you don’t need to do any fancy econometrics at all to know that the correlation between GDP and life satisfaction over time is zero. You just need to plot the two separately on a chart over time. One is an almost flat line, one goes up a lot.

Or just engage the brain a bit. Over the course of many decades, average height in most developed countries has increased, thanks to better nutrition, healthier mothers, public health measures etc, all the fruit of economic prosperity. There is certainly a link between GDP and height – but you would not expect average height to have increased in proportion with GDP or we’d all be many metres tall; our average height in the UK would have roughly trebled since 1955. Life satisfaction is similarly an organic kind of characteristic and there is no reason at all to expect it to increase proportionately with GDP. That does no mean economic prosperity has no bearing on happiness.

Think of it another way. There has been next to no growth – indeed, falling GDP in some cases – since 2008. Has this really not diminished life satisfaction?

The Vox column gives the appearance of addressing some recent work challenging the idea of no links between GDP and life satisfaction – this by Stevenson and Wolfers is the best known but there are several papers – but it misrepresents them. The Vox authors write: “This last interpretation [ie the no-link interpretation] has been questioned by Deaton (2008) and Stevenson and Wolfers (2008), who claim that there is a positive relation between GDP and life satisfaction in developed countries.” In fact, this is exactly what they do not claim; they agree there is none. However, they find strong evidence for a positive relation between life satisfaction and GDP growth. GDP growth is a stationary time series (ranging between say -10 and +10 percentage points), so this positive correlation can be meaningful.

The psychology of adaptation might well help explain why the level of GDP has no relation to life satisfaction, which could be reflected in the statistical properties. There are equally strong psychological reasons for expecting the change in GDP to be positively associated with life satisfaction.

As Hobbes put it in [amazon_link id=”0199537283″ target=”_blank” ]Leviathan[/amazon_link]: “There is no such Finis Ultimus, no summum bonum as is spoken of in the books of the old Moral Philosophers. Nor can a Man any more live, whose desires are at an end, than he whose senses and Imaginations are at a stand. Felicity is a continual progress of the desire from one object to another.”

GDP is often thought of as just more of the same stuff, and of course how could having one more car or handbag or house make you happier once you already have a certain number? But this is to misunderstand fundamentally what GDP growth indicates (see my forthcoming [amazon_link id=”0691156794″ target=”_blank” ]GDP: A Brief but Affectionate History[/amazon_link]) – which is in fact variety and innovation, new services and goods, from new medicines to graphene, or the internet, that speak to the fundamental human curiosity identified by the Enlightenment philosophers.

6 thoughts on “Life satisfaction, GDP and sense

  1. A good article, and I say that as someone who is not a real fan of GDP growth as a path to happiness. I think what happens too often is that general, national, GDP growth is proposed as a “good” before rather simpler things, like a good breakfast and a run in the park.

    Put differently, GDP growth is “life hacking” the hard and inefficient way.

    • GDP is certainly a different concept from well-being or social welfare. I think the question is really about what are the aims of policymakers, and what information and levers do they have. I come down on the side of GDP growth being a pretty good proxy – but not the only thing policy needs to pay attention to.

  2. Diane:
    Your argument is that our measure of life satisfaction is arbitrarily scaled, and bounded above, whereas GDP (or log GDP) is not; therefore, we should not expect correlation between the two.

    This point is a reasonable criticism of the entire literature on the relationship between GDP and life satisfaction, and you are right to raise it. Most scholars working in the field recognize it, but do not consider that it is binding in most specific applications. The reason is that few people declare themselves at the top of the happiness scale, even in the “happiest” countries.

    If the upper bound on the happiness measure were binding, you would not see the curved income/happiness relationship that Rusticchini and I report in our paper. You would see a horizontal flat line.

    On another issue, I disagree. Sacks, Stevenson, and Wolfers do not just address the happiness/income growth relationship. As you have recognized subsequently (in a tweet), they also analyse the influence of income levels on life satisfaction.

    Thanks for your comment.
    Eugenio Proto

  3. The whole debate is peculiar. The desire for GDP growth and the desire for “life satisfaction” have no obvious or direct relationship. Surely, we don’t need deep spiritual knowledge to recognise that increasing prosperity is at best only one part of a “better life”. It’s hardly profound to suggest that a decline in infant mortality or an increase in literacy improve “life satisfaction” more than a more energy efficient refrigerator. But I am puzzled by the unwillingness of people even to notice that their material prosperity is actually increasing. See my .

Comments are closed.