A new poll from PPP(D) gives Akin 44%, McCaskill 43%. The pollster says there’s “little difference” from before, as does Josh Marshall. Normally I don’t like to comment on one poll. But this interpretation is likely to be erroneous for two reasons.
First, PPP is making a comparison with its own previous poll in mid-May, showing the same numerical margin (Akin 45, McCaskill 44). It is more appropriate to make a comparison with polls conducted after the August 6th primary, since after a primary a nominee’s support firms up considerably.
Akin was +3% (median, n=3) before August 6th, and +11% (n=1) afterward — an 8-point bounce. So if he is unchanged since May, then he must have lost all this gain. An 8-point loss in one day is an enormous drop. Another comparison that makes the same point is that going from +11% to +1% (PPP) is a 10-point drop. Again, this is very large.
Why did PPP fail to notice this? Because they didn’t have an August 7th poll, and no way would they cite another pollster. This leaves them with a May-vs.-August comparison – and a major analysis error.
Second, even if we accept the premise of comparing with May, PPP has made a statistical error in claiming that a 1-point change indicates a lack of change. When comparing a candidate’s support between two polls, the difference has a standard deviation of about 3.5% and the 95% confidence interval is about +/-7%. All PPP can say from their own data is that since May, Akin’s support changed between -8% and +7% (95% CI). This is not very informative. But it does show how useful poll aggregation can be.
In the short term, this should make both Akin and McCaskill supporters more likely to contribute to their candidates, since the race is now probably on the edge. What happens in the long term is less clear. Is Akin too damaged to continue? Or will he soldier on?