Arnold Kling  

Surveys Vs. Revealed Preference, II

Comment of the Week, 2003-04-0... Internet Bubble and Growth...

In a previous post, I mentioned Richard Layard's critique of economics, based on survey research. Now, I have written an extended response to Layard. An excerpt:

[Layard] is saying that you cannot trust people's behavior as an indicator of their preferences, because their tastes may be socially constructed.

However, if tastes are socially constructed, then it seems to me that social context is at least as likely to affect survey research as it is to affect behavior. For example, Layard reports that marriage increases happiness, based on survey could easily be that the difference in reported happiness reflects respondents' views of how they are expected to feel about being single or married.

In a shorter comment, Zimran Ahmed writes

I had the rare privilege of meeting Nobel Prize winning Chicago economist Ronald Coase today...

I asked him if the questionnaire based data sets behavioral empirical economists come up with were any good, and he said they were worthless because they don't test what people do when it matters, and that's when folks start to think seriously about costs.

For Discussion. Consider the popularity of the movie "Titanic." Why are so many people willing to pay to see something that makes them sad? Does this complicate the relationship between money and happiness?

Comments and Sharing

COMMENTS (3 to date)
scott writes:

If your objective is to predict what people will say, then by all means use a survey. Since economists normally want to predict behavior, then to begin by observing behavior makes a lot of sense.

As to your question, people go to movies to be entertained. Exercising the emotions, as with the movie Titanic, seems to provide that entertainment. I don’t go to horror flicks, but lots of people like them. It’s all just entertainment.

And that makes people happy.

Jake writes:

My favorite course during college was in the study of Communications, titled Persuasion. If one looks into the remarkably well-researched and documented analysis of human behavior, they will find that Richard Layard's critique is on point. People will underreport socially "disowned" behaviors and overreport socially "preferred" behaviors.
I've always thought that the acceptance in society of misrepresenting oneself is so interresting. Watching politicians and advocates on news programs provides a clear case-in-point that demonstrates that we believe that it is okay to say anything as long as there is some justification for it -- not necessarily one that is true or correct. This applies to surveys.
There are so many ways to distort surveys that it makes them meaningless in my eyes. Are the types of people who take surveys biased to certain views? Does the survey process create biases in answers? Etc... The answer to these questions is yes, the real question is to what degree.
So, if you take a process that will provide flawed data and then extrapolate it to cover entire populations, you derive false conclusions. Granted, its the best we can do, but what some people aren't understanding is that its not good enough. I think chaos theory holds pretty well here. Small changes in the underlying fact patterns have large consequences for the conclusions. Just because its the best we can do doesn't mean its giving us answers that are good enough.
So what is the communication theorists answer to this problem? Observe the behavior, not the self-representation.

Mark I Lueck writes:

There is so much to say on this topic so I will focus on one excerpt from Kling's response to Layard. The excerpt is:

For example, we rarely see affluent people move into poor neighborhoods in order to enjoy higher relative incomes. We often see immigrants come to the United States knowing that they will be relatively poor. Thus, their behavior appears to suggest that people value absolute income rather than relative income.

The issue I have with this argument is that it assumes that the primary determinant in creating a set of people to whom I would compare myself is physical proximity. I believe the more important determinant would be psychic proxmity. If, for example, I were to move into a neighborhood in which everyone earned half what I did I imagine I would continue to measure myself against my high school rival, as I do now. In addition, I imagine that the people who immigrate to the United States continue to compare themselves to their friends and relatives in their native country.
Fundamentally, the theory of revealed preference assumes an "autonomous self" nature of human beings. It assumes that a person's behavior is "revealing" or uncovering some ontological truth about the person's nature. However, it seems to me that all we can really say when someone behaves a certain way under certain circumstances is, "That person behaved a certain way under certain circumstances."
There may be patterns to people's behavior and these may or may not be indicative of future behavior. I'm not saying that we throw behavior out altogether but that it be taken as one component in the discernment of what we call rational behavior; a component that may be tempered with survey research and a whole host of other data.

Comments for this entry have been closed
Return to top