Sunday, July 07, 2013

Stevenson and Wolfers' Flawed Happiness Research

Russ Roberts had a podcast a couple weeks ago where he interviewed Betsy Stevenson (below) and Justin Wolfers (right), primarily about their research on the Easterlin Paradox, and it highlighted what's wrong with so many academic debates.  

To review, in 1974 USC professor Richard Easterlin found that within a given country people with higher incomes were more likely to report being happy. However, between developed countries, the average reported level of happiness did not vary much with national income per person. Similarly, although income per person rose steadily in the United States between 1946 and 1970, average reported happiness showed no long-term trend and declined between 1960 and 1970. Theoretically, utility is generally assumed to be increasing at a decreasing rate (eg, log(x)). So, if you have twice as much GDP/capita, you should be happier, but in practice it doesn't seem to work this way.

I agree with Easterlin, and the relative-status utility function is the key to my book, The Missing Risk Premium.  Utility as Stevenson-Wolfers see it is a necessary and sufficient condition for an omnipresent risk premium, that is, it exists in a symmetric if--then relation (if one exists then the other does).  Yet, the risk premium seems to show up in only three places, and is usually missing (thus, my book title), if not going the wrong way.  Furthermore, evolution favors a relative utility function as opposed to the standard absolute utility function, and the evidence for this is found in ethology, anthropology, and neurology.  Economists from Adam Smith, Karl Marx, Thorstein Veblen, and even Keynes focused on status, the societal relative position, as a motivating force in individual lives (this was before mathematical utility functions in the 1950s made the profession ignore relative position). So, this isn't just a crazy idea championed by a wacky Easterlin guy, or just wacky me.

Stevenson and Wolfers are married coauthors, and they have published at least three papers on the topic, all refuting the Easterlin finding. Wolfers states 'most economists have our view, that there is no Easterlin paradox and there probably never was.' I'm sure he is correct, that most economists share his views, but only because they always have: if economist used a relative utility function many (most) seminal models would become ambiguous, and the whole field loses much of its foundation. Interviewer Russ Roberts is the Merv Griffin of economics interviewers--agreeable to a fault--and so never presses them on what specifically causes the Easterlin crowd to see things so differently. As Stevenson is now part of the prestigious yet irrelevant Counsel of Economic Advisors and Wolfers has more affiliations than your average CFA (University of Michigan, Brookings, CAMA, CEPR, CESifo, IZA and NBER), these two represent best practices in economics. It would be useful to see what the 'best of the best' do when applying their laser-logic.

First, there's the paper that made this May's American Economics Review, Subjective Well‐Being and Income: Is There Any Evidence of Satiation? Here they document two things. First, that cross-sectionally, higher GDP/capita generates higher happiness. Strangely, they find that the income-happiness effect is at least twice as strong among richer countries, which no one thinks is true (the effect should decrease as wealth increases), and further using one set of data the effect of income on happiness is negative. The authors note, however, that this is merely because of one country, the Phillipines. Strange that 2 of the 5 observations here were significant in the direction no one argues is true.  If this was the effect of one single country, could such an explanation be responsible for the positive effect for the rich countries?  The data look disputable (one chart shows Denmark and Norway above Italy and Spain, which would be unusual).  That said, by itself it does support their assertion.

 Their second set of findings concern cross-sectional data within a country. Easterlin did not dispute this, however. Given positional goods like mates and lakefront property, relative wealth should matter. Thus, S-W spend a lot of time refuting findings that looks somewhat relevant to the Easterlin Paradox, and definitely supportive of their view, but if you are a smart economist you should understand this is irrelevant to anything but a caricature of the Easterlin idea. I don't think they are fools or consciously disingenuous, just really good at playing the game: they have convinced themselves that their academic confabulations are objective science as opposed to tendentious rhetoric.  This doesn't move the debate forward, but it does help their status in their tribe, which is what most economic research is really about and why you don't have to follow most of it.

 So, what about the original Easterlin note, that among developed countries, where people are more worried about obesity than malnutrition, as GDP/capita rises we aren't getting happier? Well, Sacks, Stevenson, and Wolfers (2013) adress this point directly, and show this chart, where happiness is on the y-axis (vertical), and log GDP normalized for country and 'waves'.  Now, 'waves' is the name for the particular set of years a specific survey tended to use identical phrasing and protocols, which usually last a handful of years, thus each such set was assigned a fixed-effect.  Here is the resulting data, and their 'effect' in the line just in case it isn't obvious to you.

When an economist tells you a symmetric ovoid contains a highly significant trend via the power of statistics, don't believe them: real effects pass the ocular test of statistical significance (ie, it should look like a pattern).  Here's another view of the data we are interested in--change in log(GDP)/capita over time within a country--versus change in happiness, using a variety of surveys:

Again, for each its happiness on the y-axis, income on the x-axis.  S-S-W add little lines trying to show a pattern that they are sure is there.

Now, two can play this game, as from 2010 Easterlin and co-authors have data with similar blobs, but they draw downward-sloping lines over them.

I think it's best to say, no relation, and to stop drawing lines on blobs.

In any case, the biggest problem with the Sacks, Stevenson and Wolfers analysis is that they estimate a short-term relationship between life satisfaction and GDP, rather than the long-term relationship. Surely over an economic cycle, say between 2007 and 2009, or 1999 and 2002, income is correlated with general anxiety in the predictable way. Only over decades does the null effect of income on happiness arise, and this is basically taken out via 'wave-fixed-effects', which are basically time-dummies for 5-year groupings.

While I think people who aren't fighting for basic necessities are focused primarily on status and the things it can buy, I don't think this implies we should be indifferent to growth.  That would be the naturalistic fallacy, that 'is' implies 'ought.' We should aspire higher than envy, which paradoxically seems to elevate greed, but really just forces us to be grateful for things like the internet, strawberries in winter, and five-blade razors that we take for granted once everyone has them. I note many writers I otherwise admire, usually libertarian leaning, are quite averse to the Easterlin conclusion, thinking it will lead us to adopt a luddite policies because growth would not matter in such a world (see Ron Bailey here, or Tim Worstall there).

The key is that while I admit that my relatively impoverished grandfather was probably as happy as I am, I'm also very glad I live now: growth is good in spite of my envious homunculus.  Further, as productivity growth is the natural consequence of free minds and markets, flattening growth means not merely focusing on 'more important things' but rather squelching freedom, and liberty is more important than equality because it's feasible while allowing a great deal of the latter.  In contrast, true equality is only possible via force because people are not equal in effort or ability.  I mean, how would one prevent Larry Ellison or LeBron James from being richer than everyone else? The only way would be to destroy new companies or merit-based systems, why the worst rise to the top in hierarchies based on non-pnl signals, with examples from smarmy politicians, clueless executives in large regulated corporations, and of course genocidal socialists.


Anonymous said...

I never understood these surveys. If you ask me, "how happy are you?" I'm going to answer the question relative to how happy I think people around me are-they are my counterfactual. If you ask me, "how happy are you compared to people in Ethiopia?" I'm going to say pretty darn happy.

It's not that my happiness is a function of the people around me, but that my *reported* happiness is a function of the people around me. This is a subtle distinction and has more to do with the psychology of survey-taking than happiness. I think this would be a first-order effect that would be difficult to control for, but maybe they do.

Anonymous said...

I lked the residualised x-axis in fig 1.
How can you argue with residualised log GDP? huh?

Quartz said...

At first I thought those trended blobs were perfect for the Ignobel prize, but of course they aren't: at ignobel there is respectable research.

I also see no contradiction in having both relative and absolute drivers... why do so many need b&w so desperately?

But please make no more fun of 5-bladed razors, there are people who really believe in that and deserve respect just like normal beings :-)

Also, beware not to assume merit-based systems just because of a hierarchy (even letting aside the Peter principle)! Gaming the system works as good as improving it, isn't it so? We have no reliable measure to distinguish which one is prevalent in any country/activity, even though my gut feelings can give you a roughly precise estimate for a few countries.

Glad to see you didn't stop posting after all!

Happy Birthday said...

i want to second anon's claim that happiness is understood in a context-dependent way. the meaning of words depends entirely on how they're used, and how and when others describe themselves as happy will inform my usage. there are situations where i inflate my reported happiness because it makes others view me more favorably, and there are other circumstances where i understate my contentment because i know it will irritate my listener. there are situations where my usage is closer to 'pure' description, but i'm sure my intentional uses have subtlely perverted how i understand the word in 'neutral' contexts.

Jim Glass said...

"Russ Roberts is the Merv Griffin of economics interviewers--agreeable to a fault--and so never presses them..."

Oh, that seems a little ungenerous. I'd say Roberts is the Charlie Rose of economics interviewers, always agreeable and never pressing so he can keep getting interesting guests with a wide range of views, including those he doesn't agree with. It's not like his couch is always loaded up with Zsa Zsa and the obese wine salesman version of Orson.

There's far far too much confirmation bias on the econ blogs, and IMHO Roberts deserves credit for bucking this pattern and bringing alternate views (even Stiglitz's) to a libertarian- leaning audience -- even at the steep price of being agreeable.

As to "happiness", frankly I don't see what it has to do with econ at all and why all the fuss about it.

Economic advancement is about welfare, not happiness. Let's stipulate I'm not any "happier" today than were people who lived back in the Middle Ages (or back in the cave era) who just enjoyed a good day and pleasant social encounter. Does that mean I'd be "just as happy" to swap lives to live an existence back then with half the life expectancy, enduring one disease and physical affliction after the other, in a thatched hut, without an internet connection or even a TV set?

I think not, because welfare has a very real value beyond happiness. And that is what economic advancement is about.

Happiness ain't everything, as the decent philosophers have always known.

Mercury said...

[“In contrast, true equality is only possible via force because people are not equal in effort or ability. I mean, how would one prevent Larry Ellison or LeBron James from being richer than everyone else? The only way would be to destroy new companies or merit-based systems.“]

I think that's called "Hope and Change".

cig said...

I suspect a merit-based system would work well to make Larry Ellison poorer (relatively at least). After all his claim to fame is more than he's good at maximising his advantage given random circumstances and a set of rules, bending them as much as he can get away with, than having much externally useful merit. He would have been most certainly a very successful apparatchik in the Soviet Union had he been born there.

On force, all systems require the use of it anyway. Every social arrangement is going to create winners and losers (absolute or relative) and you won't find one where 100% of the losers just accept they deserve their lot.

If liberty matters, should successful systems be those that minimise the proportion of their population they need to keep in prison to function?

Eric Falkenstein said...

and while economists talk about welfare and utility these would be too abstract for a survey, but it turns out not to matter too much: you can substitute 'happiness' with 'life satisfaction' and the debate is pretty much the same. The key is that it clearly matters cross-sectionally in any group, and it's not clear it has risen over time, even though objectively life is 10 times easier than in the 18th century.

I don't think you judge a society by the size of its prison population. The US has its problems, but it's clearly in the top 30 nations today even though its prison population is relatively large (and many failed states don't have large prisons, Haiti's is relatively low for the Caribbean).

What bothers me about Russ is that he had at least 20 big names talk about the 2008 financial crisis. Priorities matter, so a debate about which factors were most important is important, and unfortunately he squandered that opportunity. I bet all but 2 guests who talked about the 2008 crisis thought Russ agreed with them. I understand it's a principle of 'how to win friends and influence people', but that's the difference between science and being a talking head.

S said...

I always hoped you would make it on EconTalk. Robin Hanson's are by far and away the best.

Mercury said...

["I don't think they are fools or consciously disingenuous, just really good at playing the game: they have convinced themselves that their academic confabulations are objective science as opposed to tendentious rhetoric."]

Asking people abstract questions about their feelings is pretty much the antithesis of objective science all by itself.

Anonymous said...

One interesting counterexample to the envy hypothesis is immigration. Why would someone immigrate from China or Mexico to the US where he is going to suffer a drop in relative status but an increase in absolute status?

Eric Falkenstein said...

that's an interesting point. I would say 1) that's one big reason why most people don't immigrate regardless of how screwed up their home country is and 2) that's why immigrants tend to aggregate into ethnic enclaves.

But this point really highlights that social comparisons happen on many levels. Some people compare themselves within an ethnicity, while the 'leaders' of those groups might compare themselves with the leaders of other groups. I mean, it doesn't bother me when a random person gets a big promotion, but when someone who is doing something I do gets a big reward, I feel diminished, because I am, relatively.

cig said...

I don't think the immigration point invalidates the envy hypothesis much: if you remain connected to your departing country, migration allows you to get up in the social rank there (especially if you're stuck with little way up due to local rigidities) because even a modest position in prosperous country will give you a nice position at home in a poor country, which you can show off coming back on holidays, in retirement, or to restart a career after a while with host country skills or savings.

this is pretty common behaviour among migrants, and critically it's likely to be how they think at the time they make the emigration decision. if after some years they lose touch with the source country and start ranking themselves within the host country scale, it's back to scare one, but too late.

cig said...

Re prison population, it may be a poor measure in a general sense, but I meant in the narrow sense of liberty. Being locked in a cell is the epitome of being unfree, and it is a strong evidence of the existence of laws which by construction are about restricting freedoms.

If liberty is what someone values most, is Haiti not a better place than the US? Libertarians often complain about too much regulation, a failed state usually fails to regulate much (by definition!), so is it not liberty-maximising? You might not have running water, sewage or nice highways (but then a free man can just get their own off grid water system and a helicopter, which they can fly without pesky government licensing!) and they won't be bogged down in regulation and unlikely to have to pay taxes (failed states tend to fail to collect taxes, and if they try, petty cash paid to the collecting officer should be enough to buy oneself out of it), there'll be no government body listening to their electronic communications (for there is no public infrastructure to tap), etc.

Mercury said...

When considering the envy hypothesis in the context of immigration look at it from the other side of the fence too.

The US has made it pretty easy for low skilled, third world peasants to immigrate here and scratch out a half-decent living but relatively hard for skilled, Euro-origin people of greater means no become naturalized citizens and operate at their desired socio-economic level. I think that the middle-America perception has generally been that the first group poses much less of a near-term threat to most American’s status/quality of life (or actually improves it) while the second group seems to be more obviously a bunch of direct competitors for jobs, houses and status position.

Also, even if envy/relative status considerations rule human decision making most of the time it’s not really out of character for people to actively choose a different status game to play. The history of America is filled with stories of people who dropped out of the class system or religious system game back home with the express purpose of playing a different class system or religious system game in the New World.

Unanimous said...

Good article as usual apart from the Larry Ellison/Lebron James line.

No one argues that everyone should earn exactly the same money, or thinks that everyone earning exactly the same money is what equality means. Most people who argue for more equality think that we don't have a merit based system now, and that we should have a more merit based system. They argue for equality of opportunity, or equality of treatment, or somethinig along those lines. I've never heard anyone argue for equality of total yearly income regardless of behaviour.

How does a coneptually silly non-merit based system show that the current non-merit based system is correct? It just shows that either you don't understand what people who argue for more equality are talking about, or your thinking is sloppy.

R.Jones said...

Unanimous, I've heard plenty of people talking like that. They never cease about talking about closing the gap.
It is presumed that if equality of opportunity is achieved that equality of outcome will ensue. Still a gap in outcomes? Must be something wrong. Progressives push this point incessantly.

Unanimous said...

R.Jones - name one. Or better, find a link to a site in which someone seriously and explicitly argues that everyone should recieve the same yearly income regardless of their effort and ability. I'd be interested to see such an argument.

By "closing the gap" people are usually talking about reducing the size of the gap, or eliminating a gap in treatment or unfair advantage.

I think you are incorrrectly presuming what other people's presumptions are. The size of a gap in outcomes, if it is out of proportion to a plausible difference in ability or effort, can be one of the indicators of unfair treatment. That doesn't mean that the gap should be eliminated, only looked at further, or maybe reduced. Do you think that any sized gap must be fair just because it exists?

A lot of "Progressives" seem to want to go back to the 1950s and 60s where many of the gaps that exist today didn't exist. "Conservative" might be a better word for those people - it depends which gaps you are talking about.

James Oswald said...

Equality is just as unachieveable through force as it is through trade. People differ in their ability to use violence perhaps even more than their ability to serve one another through trade. Look at any system in which violence is the primary way resources are distributed and you sure as heck won't find an equal one.