Reply #16: This is What I Had in Mind [View All]

[On the Road]

"How does one ascribe different "weights" to the two sample sizes?"

You ascribe different weights for various demographic variables between the first and second versions of the exit polls. If the weightings are done assuming 50% of voters are women and the actual turnout shows 60% women, the exit polls will change even if N is constant. This is how I understood the pollsters' explanation.

In and of itself, this is perfectly reasonable. The question is whether this is all that was done, and where the data came from that was used to reweight.

There actually IS another variable in this election, namely large and uneven increases in turnout. If there was monkey business going on, turnout was used to mask and explain away trends that would have raised eyebrows otherwise.