Arnold Kling  

Behavioral Science vs. Happiness Research

Honest Measurement... Economic Theory for the Social...

Bryan writes,

If happiness research can't make behavioral predictions, then behavioral research can't make happiness predictions.

Which is as it should be. Books that are based on research designed to predict behavior belong in the Social Science section. Books that tell you how to be happy belong in the New Age/Self-Help section.

If you build a model that predicts the length of marriages, you are doing social science. If you talk about whether people are happy in their marriages, then you are doing self-help.

Next, Bryan asks,

Do we calculate the unemployment rate by observing behavior? No, we call people on the phone and ASK them if they are unemployed. Is that a problem?

Actually, the household employment survey asks behavioral questions. How many hours did you work last week? Did you search for work in the last two weeks? It does not say, "How are you feeling this week: very employed, moderately employed, somewhat unemployed?"

The unemployment rate is calculated by a survey as a matter of convenience. In principle, the information could have been obtained by a third-party observer, except that would be expensive.

I think that a stronger defense of happiness research can be found in Will Wilkinson's comment entry on my last post.

There is no way out of relying on self-reports in psychology, and they have taught us a lot. When people who say they feel very stressed out turn out to have high cortisol levels, that means something. A lot of times you simply can't tell what bits of the brain do unless people tell you what they are feeling at the same time as you are looking at which bits are active. People are not VERY reliable about reporting how they are feeling, but they are reliable ENOUGH to pin down meaningful correlations between brain activity, hormone levels, etc. And this provides a basis for deeper study of the physical processes that do not depend so heavily on self-reports.

The neuroscientists may some day be able to put me on the defensive. I might have to resort to saying that what they can measure is an instantaneous chemical state, and it is a stretch to label such a state as "happiness."

What if it turns out that different people achieve the same chemical state in different ways? Maybe some people prefer being single to being married, no matter what the "happiness research" shows.

And what if it turns out that different people seek different chemical states? That is, I might prefer chemical state X to chemical state Y, even though happiness research has determined that chemical state Y is correlated with self-reported happiness in others.

Comments and Sharing

CATEGORIES: Economic Methods

COMMENTS (10 to date)
R.J. Lehmann writes:

I'm with Arnold all the way here. The difficulties here involve not just the intersubjectivity of potentially widely divergent predicates, but whether "measurement" is any real sense applicable to purely mental entities. Even ordinal systems of measurement need a concept of zero that isn't wholely arbitrary, and it's not clear (to me, at least) whether such a point can be found in introspective consideration of one's "happiness."

Even the results from neuroscience, I'd say, should be taken with a grain of salt until we have a better sense of whether what they are measuring is *cause* (happiness is the feeling that results from chemical state x) or *effect* (chemical state x is the reaction that results from when subject feels happy.) My materialist instincts would lead me to strongly suspect the former is true, but it would seem we're a long way from definitively ruling out the latter.

Frankly, it's been surprising to me that as committed a Szaszian as Bryan would be so obstinate on this point.

harry clarke writes:

Seems to me using a happiness measure (based on neuroscience or other) is no different from using GDP. We can analyse its determinants and decide whether it correlates with things we can experience.

Think you might be a little tough on happiness researchers.

Good measures of agent welfare are valuable.

Bob Knaus writes:

I don't think we need neuroscientists to tell us there are people who prefer chemical state X even though chemical state Y is defined as "happy". All of us can think of people we know who fit the mold "She's only happy when she's complaining" or "He really enjoys argueing with his neighbors". Ms. Complainer and Mr. Grouch both give every evidence of unhappiness, yet it is the state they prefer, so we say they are happy with it.

The interesting questions related to happiness are not those regarding personal happiness in isolation, but how your interactions with others affect their happiness and yours. Think about it. Would you rather be around people who are happy, or people who enjoy making others happy? The former could easily describe an opium den, while the latter is more likely to be a functioning society.

conchis writes:

Arnold seems to be getting desperate: his definition of social science now excludes anything normative; policy advice is "self help". That's fine, I'll grant Arnold the right to use terms in any way he pleases, even if it does seem a little ridiculous. But it should be apparent that, by his definition, "not being social science" isn't much of a criticism any more.

As for the final two "what ifs", they're exactly the sort of questions happiness research should be (and I think is) seeking to answer.

Adam writes:

As for the final two "what ifs", they're exactly the sort of questions happiness research should be (and I think is) seeking to answer.

And how exactly could they do that?

It's one thing to seek to answer something. It's quite another to have a method, or a standard to judge success by.

Zac writes:
And what if it turns out that different people seek different chemical states? That is, I might prefer chemical state X to chemical state Y, even though happiness research has determined that chemical state Y is correlated with self-reported happiness in others.

Please read a book on neuroscience. Perhaps a short book; that should be enough. Then come back and continue blogging.

You need to realize that you sound like a creationist defending the idea that the world is only a few thousand years old. You are digging yourself into a real hole here by spouting ignorance.

I have a theory about the hostility of some people I like and admire and generally agree with (including Arnold) to happiness research. It's that happiness research sometimes seems inevitably to lead to econ-political policies that they don't like.

This is probably unfair and ungallant of me. Yet I'm not sure how else to explain the hostility. We shouldn't look into happiness? Why not? I think, for reasons Arnold and others spell out, we should be careful in how we take the research, though. And we should also keep in mind that happiness is by its nature a slippery subject. Still, why not poke around?

You aren't doing social science if you aren't coming up with a predictive model? But I've read long passages in Adam Smith that I guess would qualify as "self-help" instead of "social science" then, including many of what I found to be the best passages. If Adam Smith violates the distinction, then shouldn't more of us?

I dunno, I find happiness research pretty interesting.

Then comes the the leap to policies ... Richard Layard's book, for instance, is organized in two parts. Part one is a (very snappy and good, IMHO) laying-out of what happiness research is and what it has found. Part two is, wouldn't you know it, the "what we should do politically" part.

I've got two responses to that. One is that you can take note of happiness research without thinking about it in terms of what econ-political policies you want to institute. It's interesting in its own right. Since when is learning a bit about human nature not enriching? And why shouldn't economics benefit from taking in some findings? Perhaps economists will learn a bit about people's motivations and experiences. How could that ill serve economics?

The other is to get local and particular. An example: If a happiness research-type thinks that social-democratic policies are the way to go, why not point out that the US tends to do a lousy job of aping European countries? Specific policies do not universally follow from specific policy findings. The specific place, culture, and time need to be taken into account.

So far as Layard's book, for instance, goes: Why not react in two ways? Hey, part one is pretty good, while part two is kinda naive and in any case makes no sense in an American context. So far as happiness research generally goes, why not say: Hey, pretty interesting! But let's be wary about the policy conclusions some are drawing from it ...

conchis writes:


The research I'm aware of uses self-reports to get at these questions, rather than chemical states. But given that I'm far less sceptical of self-reports than Arnold is, I don't think that's a huge problem. (I do think it's a small one.)

On the first issue, Andrew Clark and others have used latent class analysis to examine heterogeneity in how income affects happiness. There are problems in extending this approach to deal with highly endogenous binary variables like marriage, but I'm not sure they're insurmountable, and in general this seems a promising strategy.

The second issue has been a focus of much of the happiness literature, which suggests numerous disconnects between what people choose, and what would make them happy. See e.g. the work of Dan Gilbert.

Of course, the conclusions you draw from such disconnects are separate matter. Arnold says that people clearly "prefer" whatever they choose, while others would argue instead that they may be mistaken. One test of this would be to see whether people would change their behavior upon learning of their "mistakes". (Though if they don't you might merely conclude that they are poor learners.)

Arnold Kling writes:

My problem with happiness research is that it does not measure anything. There are plenty of lines of research in economics that provide support for policies that I do not like but which are methodologically sound. If somebody came out with a strongly libertarian paper based on happiness research, I would not take it any more seriously than a paper with other policy implications or no policy implications whatsoever.

They aren't measuring anything? Really? I've read a couple of the books and a lot of the happiness-research stuff online that's in plain English. And, though I could certainly see some methodological challenges, and though it's clear that happiness is by its nature pretty slippery, my "this is complete and utter bullshit" warning bells never went off.

But, in your view, the happiness-research crowd is measuring nothing? So it's all a mirage, or tea-leaf-reading, or something?

I can understand a "there are some problems here" reaction, or a "this has no place in economics" argument. But the claim that nothing is being measured strikes me as so extreme as to be peculiar. Which is why I go looking for motives instead ... But maybe you're right!

Comments for this entry have been closed
Return to top