>Ailan Chubb wrote:
>>
>> In the Oct. 9, 2000 New Yorker (p. 33), James Surowiecki wrote of an "old
>> B-school stunt, in which a professor presents his students with a jar full
>> of jelly beans and asks them to guess how many there are. Their answers are
>> always wildly inaccurate, but the average of those guesses--the class's
>> collective guess-- is invariably within three per cent of the correct
>> number."
>
> I could be wrong, but this sounds to me like a factoid. Among other
>things, I would expect - based on our numerical system and other things
>- that people would be likely to over- or under-estimate by similar
>*proprtions*, giving a logsymmetric distribution whose arithmetic mean
>would only by the greatest coincidence correspond to anything.
>
Following Richard Thaler, I did a version of this to illustrate the
Winner's Curse. I took a small jar filled with 176 pennies and passed them
around the room. I then auctioned the pennies off via a sealed bid auction.
I told the students the highest bidder would win the difference between the
highest bid and the actual number of pennies. If the high bid was greater
than the number of pennies, they would pay me the difference.
Here's the results:
252
140
97
100
121
310
150
275
275
200
55
120
120
150
250
250
125
180
113
130
215
The max is 310 so I made 310 - 176 = $1.34. (I chipped in a few more
pennies and bought everyone Milky Ways :-)) As advertised, the winner was
cursed. (The Texas Rangers should note this little experiment . . . )
For the purposes of this thread, please take note that the average bid was
175.5. Pretty spooky, huh?
"Fluke! Fluke!" No doubt, but I'll try it again next time I teach the
course . . . I think you could argue the guess of the combined group is
better than any individual and that it may be centered on the true value,
but the 3% rule sounds kinda wacky to me. Any logic or other argument
backing this up?
Bert
***********************
Humberto Barreto
Department of Economics
Wabash College
Crawfordsville, IN 47933
Phone: 765-361-6315
FAX: 765-361-6277
Email: [EMAIL PROTECTED]
WWW: http://www.wabash.edu/depart/economic/barretoh/barretoh.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
