Replies

You have to look at the nonrespondents.

Let’s make it clear, statistics in polls are divided into two groups:

1. Respondents
2. Non-respondents

The non-respondents ALWAYS outnumber respondents in political polls.

If it’s a robocall, most people hang up or don’t answer.

If it’s an internet ad, most viewers will scroll or click past.

The question becomes how many Ds and Rs were in the non-response group?

Example:

Robocall is programmed to get 36 Rs to every 37 Ds.

First question asks or database reveals whether respondent voted last election. A yes answer makes respondent a ‘likely voter’.

Robocall calls 50 Ds and gets 37 to respond that are likely voters.

Robocall calls 200 Rs and gets 36 to respond that are likely voters.

Can you see how this can skew things?

Polls these days are almost always wrong and those lucky enough to be right are that way because of luck. A poll that was right is held up to influence your vote like a lottery winner would be held up to give you financial advice.

There is a nice bit of theoretical statistics that adjusts surveys for non-response but political pollsters are too dumb to know how to code it let alone explain it.

Political pollsters are almost always rejects from graduate programs in statistics. They couldn’t gain acceptance or if somehow they did, they didn’t have the right stuff to stay in.

Statistical Science is hard and compared to other sciences it is enormously difficult to master.

Political pollsters are like nerds with short skirts and fishnet stockings winking at you for a little side action. They are scummy but will find customers in politics, of course. Scum meet scum.