Dick Morris actually has NO IDEA how public polling is done [View all]

[cthulu2016]

This is odd because you would think he would... and perhaps he only pretends to have no idea how polling is done since his real job is to predict a Romney landslide on FOX every night and understanding polling might interfere with that.

But either way, he and all his fellow "un-skewers" all operate under the belief that the average poll has a predetermined demographic model of the 2012 electorate, and is weighted to match that model.

The "un-skew" theory is that pollsters have a model of the 2012 electorate that assumes a lot of black people and young people and Democrats will vote.

This is not how public polling generally works. Pollsters have a model of what constitutes a likely voter and they call people on the phone. Those people on the phone answer questions and are classified as a likely voter or not.

Then they are asked what Party they identify with.

The reason there are "too many" Democrats in these polls is that a lot of people who pass the likely voter screen happen to be identifying themselves as Democrats. "Democrat" is not a fixed demographic like white or female or over 65. It is just what people say, and people planning to vote for Democrats are likelier to say (that day) that they are Democrats. Democrats planning to vote for Republicans are probably likelier to say they are independents.

These party ID numbers change constantly.

Almost all public polling is NOT weighted to a turn-out model. The likely turn-out is deduced from the responses.

So pollsters think there are too many Democrats in the likely electorate because when they phone people at random a lot of Democrats pick up the phone and say they are certain to vote.

That's polling.

A pollster who weighted Party ID in the sample based on a theory of how many Democrats will vote would be predicting, not polling.

Some pollster might do that (and will tend to be wrong a lot) but doing so would be abnormal for a public poll.

(Internal campaign poll samples, on the other hand, can be subjected to several different turn-out models as an analytic tool—something to learn from. That is not the same as imposing a particular turn-out prediction that you assume to be right. Campaigns are all about scenarios, like, "With 60% hispanic turnout in precinct 3 we are in good shape. At 50% we're in trouble.&quot

The public polls, of the sort Morris "un-skews" and says people like Nate Silver "do not understand" are not weighted to a theory of the electorate.