Happy to be corrected. Where did I go wrong?
“At first blush, yes. But these are just registered voter polls. And in both polls, the CBS pollster assumed a 99% voter turnout rate!”
They interviewed a total of 1100 Regestered Voters of which 1091 were considered to be Likely Voters. Likely Voters are simply respondents who are Registered Voters and have reported that they are likely to vote or have a past history of voting on a regular basis.
CBS is not assuming a 99% turnout rate, but simply that almost all of 1088 of 1100 Respondents reported that they were likely to vote.
From looking at the cross-tabs, it appears that they, in most cases, Terminated interviews with “Maybe Will Vote,” “Probably Will Not Vote,” “Definitely Will Not Vote” and the DK’s.
The did retain the a few, of the respondents that were not in the Definitely Will Vote cohort.
I would have tossed those, but they likely belonged to cohorts that were difficult to reach or complete interviews with.
From the best numbers I have seen, ‘around 75-80% of the Respondents who tell you they definitely will vote actually do vote.
If you want to spend a whole lot more money on a survey and better identify the Likely Voters, you ask about their voting history.
Did you vote in 2012?
If so, who did you vote for?
Did you vote in 2008?
Who did you vote for?
The best questions to ask to determine the likelihood that a person will vote is to ask them about their voting in the Primaries and The Mid Term Elections.
At that point, you spending tons of cash on a poll and while the poll will be more accurate, that accuracy would be well within the Margin of Error.
Trump and Clinton are probably fielding polls with Voting History questions.