05 November 2012

Obama Loss due to Bradley effect?

Ok, here it comes. Tomorrow. THE Election. The 2012 Presidential Race. Obama verses Romney. Some say good verses evil, some say black African American verses white American American.

The Polls are all over the place, but overall the objective respectable ones show a very tight race. If you want to know how accurate a poll may be, try an experiment. Put aside questions such as the demographics, methodology, question bias, push polls, small sample polls with heavy bias in survey selection (such as one I saw with Dem +10) just for the moment and consider the Bradley Effect.

From Wikipedia: 'The Bradley effect theory posits that the inaccurate polls were skewed by the phenomenon of social desirability bias. Specifically, some white voters give inaccurate polling responses for fear that, by stating their true preference, they will open themselves to criticism of racial motivation. Members of the public may feel under pressure to provide an answer that is deemed to be more publicly acceptable, or 'politically correct'. The reluctance to give accurate polling answers has sometimes extended to post-election exit polls as well. The race of the pollster conducting the interview may factor in to voters' answers.'

Will we see a similar result tomorrow? Are the polls that far off? I analyzed several prominent polls and adjusted them for several factors to reach a relatively equal sample of D and R and independents and also properly weighted Male/Female. These adjustments were based on an average of actual voting patterns over the last few election cycles, voter registration and average turnout. This analysis shows Romney up by 2 points nationally, and up 1/2 point overall in the swing states. It should provide a few percentages points of victory and an electoral college count of 282 - here is the result published 9/23/2012: Romney Wins

The Bradley Effect is most likely at play here for many pollsters.

Expect some surprises Wednesday. And some very red faces.

No comments: