29 January 2013

It's Time to Bury the Easterlin Paradox

Over the past several years as I have discussed The Climate Fix and the "iron law" of climate policy, someone often brings up the Easterlin Paradox as a counter argument. There are several reasons why it fails to contradict the iron law, the most important one being -- it is just wrong.

... Richard Easterlin of the University of South[ern] California, has been studying the concept of national happiness since the 1970s, when he formulated his "Easterlin Paradox".

"Simply stated, the happiness-income paradox is this: at a point in time both among and within countries, happiness and income are positively correlated," he said. "But, over time, happiness does not increase when a country's income increases."
The Easterlin paradox suggest that in terms of human happiness -- a squishy concept to be sure -- there is a limit to economic growth beyond which there really is just no point in attaining more wealth. Further, a decoupling between income and happiness at some threshold would imply that GDP would not be a good measure of welfare, we would need some other metric.

A recent paper (PDF) by Daniel Sacks, Betsey Stevenson and Justin Wolfers argues that the Easterlin paradox is also wrong. They explain why this matters:
This conclusion has important implications for policy and for science. If raising income does not raise well-being, then policy should focus on goals other than economic growth. And given the central role of relative income, researchers have spent a great deal of time and energy understanding why relative concerns are so important.
They ask "But is Easterlin correct?" and they answer as follows, with a resounding "No":
The accumulation of data over recent decades shows that Easterlin’s Paradox was based on empirical claims which are simply false. In fact rich countries enjoy substantially higher subjective well-being than poor countries, and as countries get richer, their citizens experience ever more well-being. What’s more, the quantitative relationship between income and well-being is about the same, whether we look across people, across countries, or at a single country as it grows richer. This fact turns Easterlin’s argument on its head: if the difference in well-being between rich and poor countries is about the same as the difference in well-being between rich and poor people, then it must be that absolute income is the dominant factor determining well-being.
Their paper is non-technical and worth reading in full, also Derek Thompson at The Atlantic has a great summary.

Sacks at al. argue that the Easterlin paradox was accepted mistakenly, based on a misapplication of statistical reasoning of the sort that is quite common in academic studies. They explain:
When scholars began studying comparative well-being in the 1970s, data was only available for a handful of countries. Consequently Easterlin (1974) failed to find a statistically significant relationship between wellbeing and GDP—although in fact the estimated relationship was positive. This failure to obtain statistically significant findings reflected the limited power of a test based on a small sample of countries, rather than a finding of a precisely estimated nil relationship. Indeed, Easterlin’s original data also fail to reject the null that the cross-country relationship equals the cross-person relationship (Stevenson and Wolfers, 2008). In other words he could reject neither the presence of the Easterlin Paradox nor the complete absence of any such paradox.
I'd add to that explanation the fact that the Easterlin paradox fits in very well with a Malthusian, limits to growth world view. No doubt many have wanted it to be true, regardless of the data.

The Sacks et al. paper complements an analysis that I discussed last summer on the relationship of GDP to proposed alternatives to GDP., by Delhey and Kroll. They found that GDP does a surprisingly good job of reflecting outcomes in more complex, non-GDP metrics. I concluded:
Calls to replace GDP are common these days. Any such metric should meet the basic empirical test of doing better than GDP in its relationship with outcomes that people care about. Most proposed metrics fail this basic test.

Ultimately, GDP will all but certainly remain core to efforts to measure well being. That said, there are important dimensions that it misses. Bringing those dimensions into view will not, in my view, be accomplished by inventing a better single metric, but by realizing that no one metric captures all that matters and recognizing that understanding well being is complex, multi-dimensional and involves trade-offs that people do not agree about. Multiple metrics will thus aid in both clarity and focus, and help to avoid the risk of experts seeking to impose their values on the broader community via stealth accounting.
The Easterlin paradox, like many urban myths, won't be going away soon. However, research provides a compelling case that the paradox is empirically not supported. Money certainly isn't everything, but it does help to explain an awful lot.


  1. “There are some ideas so absurd that only an intellectual could believe them.”

    George Orwell.

  2. Unfortunately, the "compelling case that the paradox is empirically not supported" is a bit out of date!

    "When languages first developed, human utilization of planetary resources were so nearly
    negligible that barter, and then, monetary tokens, quickly became convenient substitutes, as
    ‘economic’ motivators, for rewarding reciprocal cooperative behaviors that were intended to
    improve perceived quality of life. Two hundred years go, Malthus proposed a model that
    predicted that, since populations can increase exponentially, resulting over-utilization of nonrenewable planetary resources must ultimately limit economic growth. His ‘mistake’, and that of
    his more recent disciples, was in prematurely predicting when this might happen. Unfortunately,
    it seems that his prediction is now finally beginning to be realized. As productivity per man-hour relentlessly increases as a result of automation, Information Technology and computers replacing humans as ‘workers’, this problem is exacerbated. There are fewer people who can afford to consume what might be produced. The growing numbers of unemployed reduce demand for
    non-essentials. As a result, the secondary, economic, token motivators – money, capital-growth
    and gross domestic product – are losing their convenience. As a part of a ‘cure,’ they now may
    need to be replaced (as the motivators of economic activity) by the original primary stimulus for cooperation – the desire to increase quality of life."



    Note: Quality of Life is not equal to "happiness".

    It includes survival, safety, convenience and comfort – but not the latest craze of the moment.

  3. From Steve F. via email:

    "Thanks for another interesting post. While it is clear that money is important (the slope of happiness vs money is similar over a wide range of countries), the thing that I find most interesting is the large difference in happiness between countries at the same income level. This is consistent with what I have seen in my (many) travels around the world over the last 20 years: Brazilians, even poor ones, are in general happier than Japanese, even rich ones. (There are many similar comparisons which can be drawn.) I have concluded that the limitations imposed by society on control of ones' own life (severe restrictions on person choices for actions and activities) offers the best explanation for happiness that is independent (orthogonal to) money... for lack of a better description, "personal liberty" may be an appropriate description. My guess is that a 3-D graphic with X and Y being income and personal liberty, and Z being happiness, would be very instructive. Of course, an objective measure of personal liberty may be hard to develop."

  4. I read a book a few years back that postulated in part that happiness mapped very well with relative prosperity level in your peer group, much better than total prosperity. There were quite a few quoted studies to back this up.

    This immediately made sense to me, because how else to explain the tendency for people to keep turning to socialism, despite its inevitable destruction of total prosperity? It explains that paradox quite nicely.

  5. The best refutation of Easterlin was written in verse by Ogden Nash. it's called 'The Terrible People', it's short, and well worth reading.

  6. This is an interesting paper. However, after reading it over carefully, I'm not sure it really buries Easterlin's Paradox, at least not as stated above by Easterlin. He emphasizes that income and happiness ARE correlated (exactly what this paper shows), but improvements in happiness over time are not.

    Figure 4 shows the data related to this claim. Note that over time the majority of countries show little improvement or even decline in satisfaction (most data points are below zero on the y-axis) despite improving GDP (above zero on the x-axis). The only exception to this is the 1996 - 2005 graph, which is the shortest temporal sample. There's really little to no evidence that satisfaction generally increases over time.

    Earlier in the paper the cross-country and within-country comparisons look impressive and seem to support the equating of life satisfaction with economic well being. But there seems to be a problem with the methodology. Specifically, the well-being measure is based on this ladder analogy, where the top rung represents the "best possible life for you." It's not asking about happiness, per se.

    I can't speak for anyone else, but the ladder imagery is strongly associated with economics in my mind, such as "climbing the corporate ladder." Personally, I don't know that I would rate myself very highly on the ladder scale because I know I could really make a lot more money (hence I haven't achieved the best possible life), but I WOULD rate myself very high on the life satisfaction scale (I'm generally happy and don't really care whether I make more). In other words, if I am not particularly odd, the ladder analogy may be uniquely capturing people's view of their economic standing and not their general life satisfaction. In this case, high correlation would not be surprising. It would also not be particularly relevant to Easterlin's claims.

    Evidence that the ladder analogy is skewing the results is seen in Figure 3, where life satisfaction is measured differently. Here, 2/3 of the countries show a positive, significant correlation between satisfaction and GDP, but the average coefficent is less than half that found with the ladder method (0.16 vs 0.34) and only one of the countries has a coefficient as large as the average ladder coefficient.

    In any case, one might expect both absolute income and relative income to matter and this paper doesn't seem contradict that expectation.

  7. I'm with Mark B. - and Orwell.

    That the breadth of human experience can possibly be reduced to a linear expression is positively absurd.

    The qualities that define human happiness change radically - by entire ontological levels - as cultures advance. The requirements for happiness as defined by a member of hunter-gatherer society can scarcely be compared to that of a post-post modern person in the fully developed and post-industrialized West - even if 'happiness' is essentially the same state of consciousness.

    For a Kung bushman in the Kalihari, who lives in an essentially childlike world his whole life, and for whom happiness is defined more or less by whether he and his family can find the game or not is a VERY different affair compared to a Sorbonne educated executive in Paris finding happiness in his or her pursuits - or even a Pashtun tribesman for that matter.

    And BTW, this isn't a Daniel Quill call to return to humanities animist roots as remedy for our post-post-modern angst, that would require the extinction of almost everyone to realize - I'm against that.

    Maslo's pyramid is not equally steep at all levels of societal evolution, it gets steeper and steeper at every successive level.

    With every new advance into complexity, new failure modes emerge, new sources of unhappiness as far as this discussion is concerned.

    At a certain level, happiness is an electro-chemical state that is common to us all clear back to our proto-homonid roots, on another level it is completely unique to each culture and society. In so far as in the modern world it is an increasingly more complex and expensive task to keep oneself alive - much less happy - and in so far as the variables required to control the factors that affect our lives become further removed from individual control, it makes good sense that in Maslovian terms happiness should correlate pretty closely with income.

    In my opinion.


  8. The metric we should measure is peace of mind, not happiness. There is no perpetual state of bliss in a finite existence. There is no universal conception of bliss.

    A person can have peace of mind when poor or rich. Their state of mind correlates with, but is not uniquely dependent on, their material wealth.

  9. The core of misunderstanding I think is in the definition of happiness - the "squishy concept" part. If we don't understand it well, it's hard to decide what are its causes. And to give a definition would be too long winded here.
    What seems to me quite obvious, and here I agree with Roger, is that of the things that are measurable that contribute to happiness, money (GDP), is the best metric.
    But this is precisely where GDP also finds its limitations, since we can think/know of so many cases of poor people who are happy and rich people who are sad (Steve F. made the same point above). Antoine Saint Exupery categorized this intuition well when he said the following in the context of friendship: ": It is only with the heart that one can see rightly; what is essential is invisible to the eye."
    Yet in a culture with strong positivistic tendencies that tends to overvalue what is observable the legitimate importance of measurable things, GDP for example, will tend to become absolutized to the point it is equated with happiness itself. It's so extreme a relation that a euphemism needs to be created, "well-being" or whatever that means, to describe what GDP leads to in human beings. This "well being" can contribute to happiness, but I don't think, nor anyone really, that it equates to happiness.
    It seems Easterlin was trying to react to the rationalistic strain in modern culture that only values what is measurable and ignores the other more humanitarian dimensions of knowledge and being, but he tried to do so using the wrong methodology - a measurable one. The "essential things" can't be quantified well, since they are "invisible to the eye". For the more essential things we need dialogue and friendship, literature, philosophy, religion, history etc. to make the point.
    Point is: Yes, probably GDP is the beast way to measure happiness, but that only speaks of how poorly happiness can be measured. Happiness belongs much more to the realm of most important things, which are not measurable.

  10. A person can have peace of mind when poor or rich. Their state of mind correlates with, but is not uniquely dependent on, their material wealth.

    We can be poor and happy. It's just a lot easier to be happy when you are rich.

  11. Mark:

    Unfortunately, there is a strict limit for the number of people who will enjoy a "beachfront property in Hawaii." Any attempts to ignore that reality imposes inviolable limits on accessibility and availability of resources, both natural and human, will sponsor corruption and its attendant consequences for individual and general Welfare.

    So, we compromise, and establish a market system designed to "fairly" distribute finite resources based on individual productivity, demand, and need. All the while cognizant that preserving individual dignity, a pervasive value of human life, and controlling corruption are boundary conditions for any schemes we invent to modify the behavior of an emergent, organic system.

  12. From a friend of mine:

    "Very interesting. The Easterlin phenomenon is a fairly well-established empirical finding over 40 years of analysis. So it quite surprising that these authors think it is unambiguously false. A couple of remarks:

    1. First a totally pedantic point. The authors say that their data shows "no evidence of satiation at high levels of income". But considering that they are measuring "well-being" on a 10-point scale, obviously there has to be "satiation" at some point ---otherwise you would go past the 10 point scale. It's just that right now we aren't close to the satiation point.

    2. In their conclusion, the authors claim that the previous findings by Easterlin and others are essentially due to insignificant sample size. In short, they seem accuse Easterlin of confusing "failure to reject the null hypothesis" with "proving the null hypothesis". This would be an astonishingly elementary statistical mistake if it was true. But this is very noisy data, and the sample sizes are pretty small. I wonder what the "R" values on these regressions are. It is a bit alarming that, in the conclusion, the authors say, "And, sensitive to the difference between a precise zero and a large but statistically insignificant number, we focus on quantitative comparisons, rather than statistical significance." What is that supposed to mean? "Our correlations are statistically insignificant, but we don't care"?

    3. The authors claim that their data shows that relative-income effects are statistically insignificant. But this is based on the assumption that people only make within-country comparisons. In 1970 (when Easterlin did the original work) this might have been true. But it isn't true now. These days, even poor people in third world countries have access to plenty of Western media. TVs are cheap. (Thank you, LG!) And even if you are too poor to own a TV, you will spend plenty of time at the local bar or coffee shop, where the TV is on 24 hours a day, playing reruns of American sit-coms. So these days, even poor sweatshop workers in Manilla and dirt-farmers in Nigeria know exactly How the Other Half Lives. Who says they aren't comparing themselves to wealthy Westerners when they answer these well-being surveys?

    4. For these reasons, the within-country time-series analysis is the most important, because this is the only way you can isolate a relation between income and well-being which is not potentially contaminated by people comparing themselves to other people around the world. This is the weakest part of the whole study. When they turn to timeseries analysis, they find very weak correlations. Indeed, the discussion of the U.S. on page 10 is particularly interesting ---here there is exactly no correlation between GDP growth since 1970 and reported levels of wellbeing. To explain this finding, the authors say, " something about increasing inequality." I think this reference to "inequality" is rather peculiar in a study which claims that it's big finding is that "there are no relative income effects in wellbeing"."

  13. The remainder of his comment:

    "5. Continuing on the same line, have they controlled for income inequality in their analysis? Poor countries tend to have higher levels of income inequality. Rich countries tend to be more equal. (The U.S. is an outlier, as usual.) So if income inequality per se is detrimental to well-being, then we would expect to find a positive correlation between per-capita GDP and well-being, but mediated through the income inequality effect. (Indeed, the book The Spirit Level is entirely about this effect. It finds a very strong inverse correlation between inequality and well-being).

    To compensate for this effect, you need to introduce a third variable specifically representing income inequality (e.g. Gini coefficient), and then control for this variable when computing the correlation between per capita GDP and reported level of well-being. I doubt that this would completely wipe out the correlation the authors found, but it would certainly reduce it.

    The problem is that, with a dataset this small and this noisy, controlling for a third variable may completely wipe out all statistical power, so that the remaining results are statistically insignificant. Why do the authors not comment on this issue?

    In short, I think it is quite plausible that (contra Easterlin) there is a strong correlation between "absolute income" and well-being. Indeed, it would be quite peculiar if there wasn't ---it would mean that the whole field of Development Economics is a big waste of time. In fact, it would completely undermine the entire Enlightenment notion of "progress". So I wouldn't be surprised if the data show this. But the authors have not convinced me of their other claim ---that there is "no" correlation between relative income and well-being. Based on what I know of human psychology, this claim is prima facie absurd. And their data doesn't convince me."

  14. I've always felt that to attain the highest accolades in his field a statistician should be required to prove his worth by supporting himself on the proceeds of his sports bets in Las Vegas.

    That said.

    As a statistically insignificant individual who has lived a long life on a financial rolled coaster my observations have been:

    To the extent that fiscal excess affected my mood, the direction I was headed in if was far more import than the height of the plateau I was departing from.

    Once the lowest levels of Maslow's slippery edifice have been scaled, there is a satisfaction (dare I say happiness?) that accrues from accruing slightly more than others within your immediate peer group. The old saw about the perfect friend being almost as tall, almost as bright and only a few pounds heavier than yourself extends into the financial sphere as well.

    Mark's observations re. the scarcity of Hawaiian beach front property are undone when ownership of a cottage by Lake Simcoe or a Condo overlooking Hyde Park provide the same sense of "well-being".

    Because we change our financial status far more rapidly than we change our peer groupings a rising GDP, assuming a reasonable GINI index is in place, should see a greater number of people moving in a fiscally positive direction & therefor answering positively to the questions posed while a falling GDP, even though the absolute number may be far higher than in the first instance, should elicit more negative responses.

    The periods I recall as being the happiest in my own life certainly aren't congruent with those periods in which I'd acquired greatest disposable income. I think that most, reflecting on Easterlin's Paradox have similar experiences and, by extrapolating are able to intuitively sense that he was in fact correct.


  15. Those interested in this topic should also look at http://unsdsn.org/happiness/, which collects and adds data on cross country and time series correlates of happiness.