[extropy-chat] The Long Now: Bets and Predictions Over Time

[Mike Lorrey]

--- Hal Finney <hal at finney.org> wrote:
> 
> Electoral-vote.com is an interesting site but the data is extremely
> "noisy" and seesaws up and down a great deal.  You can't draw a very
> strong conclusion from where the polls are on any given day.

I don't. I look at its trending and final vote tally, which there is a
link to.

> 
> There are also increasing methodological problems with telephone
> polling.
> For example, pollsters can't or don't call cell phones.  I think it
> might be against the law.  

Cellphone number directories are not public information specifically
because the owner pays for every minute, generally speaking, even
incoming calls. Land lines don't charge the recipient for the cost of
the call, so land lines are legal to telemarket.

> That's a potentially enormous source of bias.

The bias is that more young people use cellphones. Problem is that
young people don't vote as frequently either. A good pollster will have
studies done to determine how many people are in such situations and
adjust projections accordingly.

> Plus with all the telemarketing in recent years it's getting harder
> to reach people at home.  I read recently that land line use has been
> dropping precipitously.  My guess is that we are heading towards
> another "Dewey Defeats Truman" polling debacle in the next few
> election cycles.

I think they'll just figure out a way to do it online, or they will
tell cellphone providers they can't bill call recipients.
 
The increasing numbers of people on unlimited calling plans is
indicative of the trend.

> 
> Probably the results at IEM,
> http://128.255.244.60/graphs/graph_Pres04_WTA.cfm ,
> and TradeSports,
>
http://tradesports.com/jsp/intrade/trading/t_amd.jsp?selConID=11738&z=22/09/2004-18:43:18#
> ,
> are more meaningful, since people are betting real money there.

Betting 'real money' isn't useful information if the number of people
involved are too small a statistical sample, which is the same
predicament that pollsters are in: they need x number of random
contacts to develop an accurate picture of the whole population. 200
people can accurately tell you the opinions of a 20,000 person
population. The smaller sample size you use, the less accurate your
projection (hence plus or minus 4% error rates). The less random your
sample, the less accurate your projection, hence internet polls.

If only 200 people are involved in your futures market, they are still
only going to tell you the accurate opinions of 20,000 people who are
accurate represented by the cross section of those 200.

What would be really useful is to sell virtual votes for Bush and Kerry
on eBay, supported by an ad campaign to promote the auctions.

=====
Mike Lorrey
Chairman, Free Town Land Development
"Necessity is the plea for every infringement of human freedom.
It is the argument of tyrants; it is the creed of slaves."
                                         -William Pitt (1759-1806) 
Blog: http://www.xanga.com/home.aspx?user=Sadomikeyism


		
_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com