The Big Ten Battleground Poll released today shows Barack Obama and John McCain virtually tied in Iowa, with Obama at 42.7 percent and McCain at 42 percent when leaners are not added, and Obama and McCain both at 44.8 percent with leaners.
I am not buying it.
It’s not just that Selzer and Associates, which has an excellent track record in Iowa, released a poll four days ago showing Obama ahead 52-40. It’s not just that CNN/Time released a poll two weeks ago showing Obama ahead 55-40.
It’s the fact that according to a commenter at MyDD who has dug into the methodology, the Big Ten Battleground polls did not weight the data according to turnout projections.
The reason Ann Selzer was right about the Iowa caucuses was that she weighted the results to reflect the record turnout she expected (about 60 percent of caucus-goers being first-time caucus-goers).
The Big Ten Battleground poll supposedly has a margin of error of plus or minus 4 percent, but not weighting the results to reflect likely turnout of various demographic groups could introduce far more bias than that.
To cite one egregious example, this Virginia poll showing McCain ahead by 9 points was not weighted for race. As a result, it assumes that black voters will comprise less than 10 percent of the electorate in November.
Despite the fact that blacks were approximately 20 percent of voters in Virginia in November 2004, when John Kerry did nothing to target the state.
Despite the fact that the current Democratic nominee is black and is targeting Virginia with 35 field offices.
It’s worthless to even release a Virginia presidential poll with such ridiculous assumptions regarding turnout.
I suspect that the Big Ten Iowa poll has some similar erroneous assumption, perhaps underestimating Democratic or youth turnout.
On the other hand, I hope the McCain/Palin campaign starts spending lots of time and money in Iowa, hoping that we are a swing state. That will be pouring resources down a sinkhole for them.