about me
art
biz
Chess
corrections
economics
EconoSchool
Finance
friends
fun
game theory
games
geo
mathstat
misc
NatScience
... more
Profil
Logout
Subscribe Weblog

 
From the Pop!Tech 2004 conference over at IT Conversations,

in the Q&A with Joel Garreau and Malcom Gladwell, Frans de Waal has this riff on the value of questionaires in social science (around minute 45 from this mp3):
We should never trust questionaires at all ... There’s another experiment ... Let me tell you of another experiment, this is on sexual preferences of males and females on college campuses. If you ask young men how many sex partners do you have, let’s say, whatever, they say ten. If you ask women, how many do you have, or whatever it is, they say five, it's always half of what the men have, which is very puzzling, because [laughter] ... there must be something going on. And then recently an experiment was done by a group of psychologists where they hooked them up to fake lie detector tests, not a real lie detector machine, but a fake one, with all sorts of tubes and graphs and all of this, and all the sudden the women had as many partners as the men.

And so I usually use it as an example I don’t trust questionaires at all ... you need to pay attention to behavior, behavior is the only thing that tells us what the real preferences are.
AleksJ meinte am 25. Apr, 01:20:
The Source
From New Scientist:


Women who thought their responses might be read said they had had an average of 2.6 sexual partners, compared with 3.4 partners for those who thought their answers were anonymous. But those who thought they would be caught out by the polygraph reported an average of 4.4 partners.

In contrast, men's answers did not vary significantly. Those attached to the lie-detector reported an average of 4.0 partners compared with 3.7 for men who thought their answers would be read.



Fisher and Alexander (Journal of Sex Research, vol 40, p 27) surveyed over 200 unmarried, heterosexual college students aged 18 to 25. The disparity between women and men is due to young males' relative lack of appeal. 
Taggert J. Brooks (guest) meinte am 3. May, 16:18:
Argh
This is getting to be very frustrating. The fact is that economists use surveys/questionnaires all the time. The problem is we hardly know it since we don't collect the data ourselves. The CPI is a survey, expenditures are determined by a survey, and prices are collected by a survey. All subject to the same social desireablity bias noted in the post. IRS data, BEA, census? All surveys. Surveys are not on their face good or bad. Yes they are prone to bias as the previous comment notes, but how would you have us gather that data otherwise? Follow someone around with a camera 24/7 making note of who they sleep with?

The important thing is to try to mitigate any bias, and understand when the respondent is not likely to know thier motivations. Simply deriding the method is not so useful.