David R. Henderson  

The Problem with (Some) Experimental Economics

PRINT
Blanchard on the Future of Mac... Exit, Voice, and Borders...

On Facebook today, a friend posed the following question that he had heard in a sermon:

Would you rather have $100K for your own use OR $1M to be given away (anonymously to a good cause with whom you have no direct connection)?

Some people chose the second choice. I needed to know more. So I said:
Not enough information. Do I get to choose the cause and the particular organization?

He answered that I do. But because time passed between each question and answer, I had another thought:
My honest answer is that even though I'm wealthy, I've still got to think about it. A big part of the issue: how effective will this $1 million be.

He didn't answer that. Then I went to do an errand and came back with another thought: where does the $1 million come from? So I wrote:
Another question whose answer I need to know before choosing: who's giving the $1 million. What would he/she/they do with $900K if I take $100K

He didn't answer this one either.

The whole thing got me thinking about an objection that I've had to some experimental economics, similar to one that Steve Landsburg expressed in 2002. Steve pointed out that in the particular experiments he was discussing, the experimenters learned that none of the "subjects" seemed to care about the people donating the money. (Steve assumed, probably correctly in some cases, that taxpayers were paying.) But after I thought about it a bit, I did care. And I didn't care just about taxpayers; I cared about the donor and about what the donor would have used money for.

That got me thinking about the issue more generally. Notice that I had time to ask questions. My impression about most experimental economics is that the subjects don't get to ask questions. Also, I had time to think about things. My thinking above happened over a period of about 3 hours. My impression is that most experimental economics subjects get very little time to think and don't get to consult with other people about their choices. Yet in most real decisions we get time to think and to talk to others. Richard McKenzie made that point in this Econlib article. I think that's a big part of what he discusses in his book, Predictably Rational.


Comments and Sharing





COMMENTS (16 to date)
fundamentalist writes:

Excellent points! I find that most economic experiments have to be so constrained that the conclusions can't be generalized and are often just trivial.

Mike Giberson writes:

Hypothetical questions like that posed to you do have a place in research, but experimental economists tend to trust research where actual money changes hands over surveys.

In research with money changing hands, sometimes it seems to matter where subjects think the money is coming from. For example, experimenters in small communities relatively isolated from large-scale commercial society found subjects believed they would be taking the experimenters' money if they prevailed in the experiment, and it made the situation culturally awkward for the subjects. (See generally Joe Heinrich, et al., Foundations of Human Sociality, 2004.)

AMW writes:

I must confess some bemusement at the McKenzie article you link to. He claims that Kahneman and Tversky list risk aversion as an irrationality. I have never heard any behavioralist or experimentalist make such a claim. Typically it is phenomena like the Allais Paradox or loss aversion that are pointed to as irrationalities.

Khoth writes:

I think Steve Landsburg is missing a possibility in his search for explanations of the experiment - that people assume that the point of the experiment is to see whether they'll give up $X in value to do something that benefits someone else to the value of $3X. That's certainly possible in real life without the need to invoke a taxpayer or anyone else to pay a $2X. It's much easier to do an abstract version with money, and I think people recognise that, and that the experimenter is willingly paying $2X in exchange for finding out what the subject thinks.

David R. Henderson writes:

@Kloth,
TANSTAAFL. Even if it's not a taxpayer, someone is paying $3X. Are we to assume that the subjects are stupid? If so, what does that tell us about behavior in the non-experimental world?

dullgeek writes:

This has also been my objection to the experiments in "Predictably Irrational". What would happen, for example, if you reran the experiment with the same participants? Would they change their answers? What if they chatted with other participants in the experiment?

I often wonder if I could devise an experiment which tested this. For example, if I ran Ariely's experiments multiple times and one time, those leaving the experiment filled out post-experiment questionairres in the same room as those about to go into the experiment. And another time when the two were seperate. Would the first case produce the more rational outcome as knowledge passed from those with experience to those without?

If you ran the experiments with the same people multiple times, would they come to the more rational outcome after thinking about it first? If the 2nd time around if they were in a group with others who had been in it the first time around, would they inform others of what happened the first time?

It has always struck me that Ariely's experiments were games designed to fool the participants. Would time allow those participants to learn something? Should we expect some level of irrational response the first time someone is posed with a new problem? Does rationality develop over time?

I do not get the impression from reading Ariely that this possibility ever occurred to him. Perhaps I missed it, though.

Khoth writes:

@David: Sure, in the experiment, someone is paying $3X. Money can't be created out of nothing. But [i]value[/i] can be. It's just harder to do that in the experiment, so they use money instead of value with the implicit assumption that you're not supposed to worry about where the money came from.

It's like those thought experiments where a psychopath will kill four people if you don't press a button and three people if you do - if you say something like "I wrestle the psychopath to the floor and take his gun", you haven't solved the problem, you've just made the person proposing it sigh and make the description longer and more contrived.

Khoth writes:

Oh, if you're worried about the actual $10 or so of possibly-tax-money in the experiment, your options are:
1. Play along, you cost the experiment $3X, they get the information they were spending $3X to find.
2. Look outside what the bounds of the experiment are and minimise the cost to the taxpayer, they pay $X but don't get the information they were paying to get.

Option 1 looks better to me.

maurile writes:

For these types of questions, I think the donor is God. He does not poof cash into existence; rather, He poofs cows into existence and auctions them off to raise the cash. So the money doesn't come out of somebody else's pocket: it reflects genuinely (and magically) created wealth.

Khoth writes:

Or, for the less theologically inclined, the money you don't take is invested in a business which produces some useful good or service that wouldn't otherwise have existed, and the money is the surplus created.

You could (at least in theory) actually run that experiment, it's just that it would be very hard and also pointless since almost everyone already assumes the value is created rather than taken from someone else.

Vince Skolny writes:

Experimental economics aside, when ministers ask questions like that, rational thought is that last thing they're trying to elicit.

So, too, their sh.. er, parishioners that publish said questions on Facebook.

On your second question, Dan Pallotta has a lot of interesting things to say about charities and their effectiveness.

Philo writes:

"My impression about most experimental economics is that the subjects don't get to ask questions. Also . . . that most experimental economics subjects get very little time to think and don't get to consult with other people about their choices. Yet in most real decisions we get time to think and to talk to others." There is a movement in academic philosophy toward "experimental philosophy" (Stephen Stich of Rutgers is one of the leaders); these same criticisms apply to it.

AMW writes:

"My impression about most experimental economics is that the subjects don't get to ask questions."

Prompting subjects to ask questions prior to the experiment is pretty much par for the course. Even during the experiment subjects can ask questions under most protocols. Experimenters would be shocked by a design that didn't allow for subjects to ask for clarification. In some cases subjects are given quizzes on the rules prior to commencing the experiment. And in some of those cases only the highest scoring subjects are actually allowed to participate. There's variation among experimenters, of course. But flat-out subject confusion is probably not a systematic reason for many of the results that get published.

"Also . . . that most experimental economics subjects get very little time to think and don't get to consult with other people about their choices."

That's a more accurate impression, though time pressure probably isn't that endemic in most cases either. Experiments are typically scheduled to last for 60 - 120 minutes. And in my experience it's not uncommon to schedule subjects for a 60 minute experiment that ends up taking them only 30 - 45 minutes to complete. Also, in more complex experiments subjects will come to a training session on one day and then participate in the actual experiment on another day.

Communication is often very limited among subjects, though. But of course, it is often very limited among the agents of economic models as well.

AMW writes:

The whole thing got me thinking about an objection that I've had to some experimental economics, similar to one that Steve Landsburg expressed in 2002. Steve pointed out that in the particular experiments he was discussing, the experimenters learned that none of the "subjects" seemed to care about the people donating the money.

David, I'm curious. What sorts of experiments do you think fall prey to this aspect of the design, and which sorts don't?

David R. Henderson writes:

@AMW,
You wrote, "What sorts of experiments do you think fall prey to this aspect of the design, and which sorts don't?"
I don't know which ones don't because I don't know enough about the field. The ones that I think do are the kind Landsburg described in the article I linked to.

AMW writes:

David,

It appears to me that Landsburg's critique would apply to pretty much any experiment, because they are (almost) all funded with somebody else's money. In this case, even straightforward market and auction experiments simply become an exercise in distributing other people's money to one's fellow subjects. Yet these types of experiments typically find market mechanisms to be highly "rational" and good at allocating resources efficiently. Yet one never hears critiques of experimental methods regarding those results.

At any rate, I find it highly unlikely that subjects in experiments think of what they're doing as moving money from someone outside the experiment to someone inside it. Landsburg may think of it that way if he were in an experiment, but then Landsburg makes a living by thinking of things in a different way than the average man.

Comments for this entry have been closed
Return to top