Replies

I would love to see your summaries as you have them plus any adjustments for polls that over/under sample based on gender, race and political affiliation. That would be very interesting so we can see the level of adjustments these pollsters make to generate their desired outcome.

I've played with adding a bias adjustment in the past, but have turned it off for now. I was simply adding a +1 or +2 to the totals. I once even made it a random uncertainty, since it wasn't backed up with any evidence.

I don't have the resources to look at in that detail, and I can't really compete with sources like fivethirtyeight.com. Nate Silver looks at polling "house effects," as well as historical quality, turnout predictions, and an extra special sauce.

I am taking the polls as they are, and assuming that the polling companies have already taken into account various turnout assumptions. I hope that by averaging different polls together, that the bias can be smoothed out somewhat.

The best I can do is show the changes over time to the polls as they are released, and leave it the reader to assume the rest. If state polls lag too much behind national polls, or they don't align with people's on-the-ground experiences, then let that show up as comments on websites and calls to radio shows.

-PJ