Arnold Kling  

Happiness, I Still Can't Get No

Hanushek on Education... I Wish Larry Summers Had a Blo...

In Happiness: A Revolution in Economics, Bruno S. Frey writes (p. 30)

Human beings are unable and unwilling to make absolute judgments. Rather, they constantly draw comparisons from their environment, from the past, or from their expectations of the future.

Hold that thought.

Let Y = a person's happiness, as reported on a classic survey.
Let X = a person's income.

One could take it as given that Y is an absolute, objective measure of happiness. Frey's point would then be that the effect of X on Y depends on how people look at X in relative terms. The inability of people to make objective judgments applies to their assessment of X. One might say that X is socially constructed, at least to some extent. I evaluate my income by looking at the income of others around me.

But I think that the inability to make absolute judgments applies to Y, the dependent variable. I just cannot concede that Y is an objective piece of data. When someone is asked how satisfied they are with life, the response has to be along the lines of, "Well, I ought to feel ____, at least based on how I compare myself to others, including myself at other points in my life."

I think of Y as a subjective, social construct. It is a measure of what a person thinks that he or she ought to feel.

All of the appeal of happiness research is based on presuming, as Frey does, that Y is absolute and objective. However, in order to make such a presumption you have to think of Y as an exception to the inability of humans to make absolute judgments.

Suppose that unemployed men report low levels of Y. If Y is objective, then the policy implication may that they should be hired to dig holes and fill them in again. If Y is subjective, then a cheaper alternative may be to help the unemployed feel better about themselves, by somehow reducing the social stigma of unemployment.

If Y is objective, then when I read that people with children are less happy, my response should be to not have children. However, if Y is subjective, then what I should do is gauge the opinions of people close to me. If they lead me to believe that I ought to feel happier if I have children, then having children probably will raise my score on the happiness scale.

Note that when happiness research is applied, it is under the assumption that everyone is the same. If the majority of people who are unemployed are unhappy, then I must want a job. If the majority of people without children are happier than people with children, then it must be a mistake for me to have children.

In economics, it is given that people have different tastes. In happiness policy, the assumption is the opposite.

Comments and Sharing

COMMENTS (10 to date)
Unit writes:

I had posted a similar thought on Will Wilkinson's blog:

"Could there be problems with averages? For instance, could there be people that are happy with kids and unhappy without, while other people are happy without and unhappy with kids and then somehow because of the ratio between people with kids and people without, that skews the analysis?"

Also this:

"Also divorce rates are pretty high and it could be that couples that are heading towards divorce with kids have a rougher time than couples in a similar situation without kids. However, it seems hard to control for this because you'd have to compare couple that will never divorce but this is not a "stopping time" (probability jargon)."

It's always easy to punch holes in someone else's theory. I just don't know if these are valid critiques. In retrospect a way to get around the stopping time problem would be to poll people throughout their entire life and then compare responses taking into account the entire life trajectory.

Can't people just take drugs for happiness? Why build policy around it (except I suppose if the policy is demonstrated to be cheaper than the drugs).

BGC writes:


'Happiness' in these studies resembles 'quality' in some management-speak. What is actually being measured is undefined and not-very-interesting - it is the name given to the measurement which has created a spurious excitement.

My experience with happiness scales in psychology (samples of around 50-1200) is that lack they the robustness which I would usually expect from eg. depression rating scales or personality scales - which is why I suppose happiness research is done on such huge numbers.

Indeed, with proper controls for personality, illness etc. almost all the 'happiness' effect disappears.

Happiness research is mostly bogus (with some exceptions such as Andrew Oswald), so it is probably sensible to ignore it - expcet we are not allowed to.

conchis writes:

Arnold, I'm glad to see you're finally engaging with real issues in happiness research, rather than just taking swipes at straw men.

A couple of points (some of which I've made before):

(1) "All of the appeal of happiness research is based on presuming, as Frey does, that Y is absolute and objective."

Not necessarily. What exactly needs to be assumed about the measures depends on precisely what you want to do with them. For example, the results that come out of regressions assuming cardinal interpersonal and intertemporal comparability are pretty much the same as those that come out of regressions assuming only ordinal intrapersonal comparability. This suggests that the big issue is whether (or to what extent) people change their reporting scales over time. There's little direct evidence on this, but indirect evidence suggests that it can't be all that's going on.

(2) "In economics, it is given that people have different tastes. In happiness policy, the assumption is the opposite."

Well, representative agent models do assume identical tastes over aggregates, which is pretty close to what a lot of happiness economics assumes. More generally, there's nothing inherent in happiness economics that says it has to focus on averages. Dealing with heterogeneity is a crucial issue, and one people are beginning to deal with now as the discipline progresses.

(FWIW, Dan Gilbert argues explicitly that, because we're so poor at forecasting our future happiness, averages are actually more informative than individual judgments. Even if you don't buy that, it should be clear that some combination of the averages and individual judgments would probably be an improvement on either alone.)

(3) It's interesting to me that you seem to think of these problems as insurmountable critiques that mean happiness research is (and will forever be) worthless rather than as methodological issues that can potentially be overcome if we work at it. Perhaps I am reading too much into this, but it gives the impression of someone who not only thinks happiness research is worthless, but wants it to be worthless. Is this an unfair assessment?

(4) The potential one sees in (undoubtedly flawed) happiness research, will depend in large part on (a) whether one thinks the questions it is trying to address are interesting; and (b) how flawed one thinks alternative methods of addressing those questions are. I'd be interested to know what portion of your hostility is due to either of these considerations, rather than (or in addition to) the internal methodological critiques you focus on.

conchis writes:

P.S. FWIW, I would probably accept an argument that happiness research is over-hyped. What I don't accept is that it's value is anywhere near zero relative to the other means we have of addressing the same questions.

Will Wilkinson writes:

Adding to Conchis's point... How often to you see economic models that mathematically identify income with utility? Well, all the time! But people have different tastes when it comes to money! At some levels, some people would rather have less than more, which I guess is why they give it away, or do things other than try to get more money. So I guess we can say that economics fails to recognize that people have different tastes? And when economic theory is applied, it assumes we must all want more money all the time. I guess I've heard a few anti-economics types make exactly that argument.

Of course, it really means that they don't understand economic theory; there's nothing about economics theory that really requires identifying utility with money. And there's nothing about happiness research that requires denies individual variation in taste. Indeed, there is a whole literature on how personality types mediates the effects of this or that on happiness. Look into it, you might find it interesting!

But yes, you're right. The policy application of most of social science is based on a series of non-sequiturs.

Gary Rogers writes:

How can you spend so much time on happiness research but dismiss macro economics?

Unit writes:

So the question is: is happiness research "macro" in flavor where we are soon going to hear about the need for the government to stimulate aggregate happiness, or is it "micro"?

Barkley Rosser writes:

If there is one thing that all the mountains of happiness research have pretty strongly confirmed (and as editor of JEBO I see gobs of papers on this subject), it is that the happiness of most people is very much dependent on various relative comparisons. This has led to much recent research focusing on how it is that people decide who is their relevant cohort to compare themselves with in order to be happy or not. That is a very hot topic, with no clearly definite results yet, partly because in fact it is somewhat fungible. The group can change. Indeed, this is part of the problem with the question of people not necessarily getting all that much happier when their incomes rise. They may redefine their comparison group upwards when they get that promotion with that pay raise.

I have said this before here, but Robert Frank likes to quote an old wisecrack: "That man is happy who makes more money than his wife's sister's brother."

fiony writes:

that 's very intersting!

Comments for this entry have been closed
Return to top