In Defense of GDP

My recent Marketplace commentary focused on the recent Sarkozy Commission report, which re-examined the usefulness of the usual economic indicators, like Gross Domestic Product (or GDP).

The report raises many of the usual shortcomings of GDP. And I agree with each of their criticisms. Much of this was summarized 40 years ago, in a famous Bobby Kennedy speech:

But there’s a funny thing about GDP: despite all of these shortcomings, in practice, it still turns out to be a surprisingly useful indicator of the health of nations. Before you rush to disagree, let me be clear: this is an empirical observation, not a theoretical defense of GDP.

Here’s exhibit one, a chart I’ve discussed here before. The chart shows each country’s GDP (shown on a log scale) and it’s average level of subjective satisfaction, measured from surveys where people are asked to rate their well-being on a zero to ten scale. The correlation between the objective measure of GDP and the subjectively assessed experiences of people’s lives is greater than 0.8. That’s an astonishingly high correlation. (Related paper, here.)


In fact, the correlation is so high that it presents an interesting tension for folks like Betsey Stevenson and myself, who are working on these alternative indicators. The fact that subjective measures are correlated with objective measures adds to their credibility. But when the correlation is this high, it becomes less clear that it is so important to supplement measures of Gross Domestic Product with new measures of Gross National Happiness.

It’s not just life satisfaction that is so closely related to GDP. We’ve also analyzed data on the frequency of a range of other subjective measures, and in each case, there’s a robust correlation between log GDP per capita and the proportion of the population reporting each feeling:


Another response to Bobby Kennedy would be to build broader indices of development. The UN’s Human Development Index is one such measure, combining data on life expectancy, adult literacy, educational enrolment and GDP per capita. But again, it turns out that GDP per capita is also highly correlated with this broader index.


My full Marketplace commentary is available here, or you can listen to it here. (Oh, and if you missed Betsey’s commentary on Christmas Eve, you can catch up on that one, here.)

R.B. Glennie

What's forgotten about this talk about the `GDP being a misleading indicator', is that it was developed (originally as `GNP', gross national product) in order to get a more accurate gauge on the economic health of a country.

It is alleged by its critics that `GDP is the be-all and end-all of everything', but really, who treats it this way?

I would certainly ask if those who say that `just bec. GDP goes up by two percent', doesn't mean everyone is better off: if the GDP goes *down* by two percent, is this too irrelevant to people being better off?

the GDP is the best overall method of measuring economic performance, period. All other proposed measures - i.e. `Gross Domestic Happiness' or whatever - rely on too-subjective measures and thus, are useless.

L F File

So I guess measuring the national debt load as a percentage of annual GDP is a fair way of gauging the effectiveness of policies designed to avoid runaway national debt. This leads us to what some might consider the perverse conclusion that since the end of WWII - and especially over the last 25 years - democratic administrations have been much better at keeping the national budget in line than the republican administrations. So why are the democrats constantly castigated as spendthrifts?

Kevin H

Ah, but correlation isn't causation, as you well know. What about leading indicators? is there a link between GDH and future GDP growth? That relationship would result in a very similar graph (even account for the log relationship) but point to the idea that it would be much more important to maximize GDH than GDP in the long run.

Of course, GDH might equally be well linked to past economic growth, which would bolster your claims. It'd be interesting to try and get a Cross-correlogram of the two data sets, assuming you had good enough temporal resolution.


Wow. "Excitement/interest in something" is negatively correlated with log GDP. Apparently I made a good decision not going into a mainstream career...


"The UN's Human Development Index is one such measure, combining data on life expectancy, adult literacy, educational enrolment and GDP per capita. But again, it turns out that GDP per capita is also highly correlated with this broader index."

GDP better be highly correlated. It's factored into the calculation.

Mark Brucker

Interesting that the correlation stated between GDP and life satisfaction is very high. But the line has a very flat slope. An increase of about 70-fold (7000%) in GDP leads to an increase of a little less than 100% in satisfaction (as measured by that index).


The Hong Kong data point in the first graph gave me a chuckle.


Joseph Wright

I would expect a high degree of correlation across the full spectrum of countries. ie first world country populations are more satisfied then third world. country populations. Simple levels of life expectancy, safety and living conditions ought to drive a strong correlation.

However, GDP is primarily used for domestic policy making to measure increases in "wealth" for domestic citizens. For that reason I think nations should be initially separated by economic status before looking for correlation that could be seen to be meaningful.

Jirka Lahvicka

Considering that Human Development Index actually contains GDP per capita, its correlation with GDP per capita is hardly surprising. How high would the correlation be if you constructed this index WITHOUT GDP per capita?


Mr. Wolfers,

I always learned GDP and well-being are fairly correlated for lower income countries. Beyond a certain point, extra income doesn't mean extra well-being.

Do you find any evidence of that?

Johan Urh

Exactly Kevin, correlation is not causation. What I am thinking is that the increase in happiness is not caused by the actual increase in GDP. It is a lagged response to observations made about the improving state of the economy (growth in GDP). Example: investor confidence is improved by a perceived improvement of the state of the economy. Improved investor confidence is usually corrolated to increased share prices but what is the cause and what the effect is unclear... a more complex version of the: what was first, the chicken or the egg?


@6 No, slope and correlation is not the same.
@7 Thank you for that insight full comment.


What GDP does not consider though are the environmental and social costs of whatever is consumed. The medium and long term costs to health on account of toxic elements in certain kinds of processed food or in cosmetics are not subtracted from the revenues of these industries ; Or for that matter, what about the intangible cost of the deprived childhood of children employed as labor in sweatshops or the more tangible cost of unfair wages paid to workers ? These may be difficult costs to measure or quantify but does that mean that we can ignore them ?


That first plot may be a good fit overall, but it sure looks like there are two distinct things going on there: a vaguely linear and well-correlated rise for up to about 24000 or so (I see NZL as the cutoff for this observation) and then a hard-to-interpret blob of mostly first-world countries above that. I'd certainly like to see what kind of correlation you can get out of just those countries with adjusted GDP >= that of New Zealand (or some other reasonable cutoff in that vicinity).

Mark S

Comparing the two graphs is rather tricky. Sure, they both demonstrate the correlation nicely, but once you've appreciated the correlation, it's the outliers that are interesting.

On the first, higher is happier, further right is richer - so being above the curve means a country's citizens have more happiness per dollar than average.

On the second, lower is more developed, further left is richer - so being above the curve means a country has achieved _less_ development per dollar than average - exactly the opposite of what you'd normally expect.

I was at first surprised to see Cuba so low down on the scale in the second graph, until I puzzled out the counter-intuitive scale...

Ben Ticehurst

Mark (6) identifies one of the key problems.

The log scale hides the fact that a truly massive increase in economic activity is required to bring about a small increase in happiness. So raising GDP is a blunt instrument for achieving this purpose. (That is assuming that high GDP causes happiness, as opposed to happy people buying more stuff. Also assuming that we know what happiness means and how to measure it well).

The true reason we should look for alternate indicators of progress is not to find ways of increasing our moment-to-moment pleasure, but to find ways of flourishing that support and nurture (rather than undermine) the living world around us that is the foundation of all that we enjoy.


"A widening gap between data and reality is distorting the government's picture of the country's economic health, overstating growth and productivity in ways that could affect the political debate on issues like trade, wages and job creation."

"The fundamental shortcoming is in the way imports are accounted for. A carburetor bought for $50 in China as a component of an American-made car, for example, more often than not shows up in the statistics as if it were the American-made version valued at, say, $100. The failure to distinguish adequately between what is made in America and what is made abroad falsely inflates the gross domestic product, which sums up all value added within the country. "

"from the point of view of pure capitalism, that's actually a useful way to do things because if you put lead in people's paint, then you make profit doing that. And then you have an opportunity to make a profit taking that leaded paint out of their houses later. "

Economists Must Learn to Subtract