|
MOE for state polls is understated. Anyone who puts more reliance on state polls than national polls is kidding themselves, even for the best pollsters.
Back to the original question. I looked at state polls in relation to national for both '96 and '00. I used the final net deviation from the national popular vote average for each state to evaluate. For instance, let's say Nevada is 4 points more GOP than the nation, which is close to accurate. I tried to find the date where the averages of reliable national polls and reliable state polls came closest to that margin. For instance, let's say Bush led Gore by 3 points nationally. The Nevada average poll differential should be 7 points.
Admittedly, not all states cooperated in this study. Or not enough state polls, period. 1996 was better than 2000, since Bush's national lead was overestimated in 2000.
But there was a definite verification that state polls DO lag, even late in the race. The average was about 4 days, but the mean closer to 3. In smaller states with fewer reported polls, it could be nearly a week, especially in '96.
|