One of the big questions in the debate over economic growth is whether it enhances human happiness. One of the sources of information commonly used in this debate is derived from answers to questions of the general form ‘On a scale of 1 to 10, how happy are you?’. This and similar questions have been asked in many countries and over a period of some decades. The ‘stylised facts’ (this is economists’ jargon for the generally accepted summary that characterised the data) are that at any given time and place, people with high incomes are, on average, happier than those with low incomes, but that there’s not much difference between poor countries and rich ones and no significant trend in happiness levels over time.
This evidence has been used to support two kinds of claims
(i) Growth in economic output has been offset by losses in other, equally important sources of happiness such as clean environments or social cohesion
(ii) People’s happiness is primarily determined by their relative position rather than by absolute levels of consumption
I don’t want to debate the merits of these claims right now, but to point out that data of the kind I’ve described is of no use in assessing them. Starting with the second claim, some sort of relative assessment is forced on respondents by the form of the question. The only sensible way to answer the question is to assign 10 to the happiest people you know or can imagine, and compare yourselves to them (the point at which you would consider yourself better off dead arguably provides a natural zero).
So, if I go from a place where everyone is gloomy and depressed to one where everyone is happy (and adjust so that I no longer think about the gloomy people I used to know) my stated score is likely to decline. This will be true even if I’m actually happier myself – apart from the fact that I’m altruistically happy that other people are happy, I just find that gloomy people get me down. That is, the form of the question makes it look as though people are concerned about their relative position, even if they’re not.
So, if the measure provided is inherently local and relative, it can’t be used for comparisons over time or between groups of people with different reference points.
Jason Soon recently cited a piece by Richard Layard in which he tried to bolster the status of claim (ii) by observing that the results of happiness surveys were consistent with ‘objective’ measures obtained by brain scans. This argument doesn’t work, at least in the absence of time-series measures of brain scans. All it shows is that self-reported assessments of happiness give a local and relative answer consistent with the local and relative answer given by brain scans.