Sunday, September 30, 2012

Overtaxed by Polls

Just finished watching Meet the Press, with David Gregory.  He and his guests talked about little else than polls.  Polls that say Obama is winning.  Depressing stuff, except that the good polls, Rasmussen and Gallup, say it's a tie. 
   Thing about polls, is that you gotta weight them.  If you do telephone polling (and all of 'em do) you quickly find out that people who answer their home phones in daytime are mostly retired elderly.  Every one else is at work.  And everyone knows that the retired elderly are conservative and vote Republican.  So they weight the results by what they think the population truly is.  If the pollster thinks the population is 41% Democratic to 34% Republican, he throws out  excess Republican polls until he gets  down to 41% Democrat and 34% Republican.  And so on for what ever other categories ( age, income, whatever) that the pollster wants to correct for. 
   The better pollsters, the ones with a reputation for accuracy that they want to protect,  are pretty good at weighting.  Rasmussen brags that his polls came out within 1% of the actual election results in 2010.  On the other hand, plenty of polls commissioned by politicians and newspapers come out the way the politician or newspaper wants them to.  Pollsters who merely want to get paid, will produce the results their customer wants.  The customer is always right.
    When I see a headline "Some and so is ahead in the polls by 1%" I know the newsie is flimflamming me.  The polls ain't that good.  When it's within 1%, it's a dead heat, no matter what David Gregory calls it.  It's gotta be more like 5 %, and it's gotta stay there for more than a day before I'm gonna believe it's in the bag for anyone.    
  So I look at the Rasmussen and Gallup websites from time to time, and ignore the newsie's poll chatter.

No comments: