How do we know that they used the likely voter model to estimate the 2016 turn out in this poll? That's pure speculation in my view and probably false. Why couldn't they have asked all respondents this question? After all, they don't estimate the likely voter turnout until after the poll has been executed. When they ask a question related to 2016, there's no reason they'll spend heaps of resources to use a 2014 electorate to calculate such an outcome instead of just using the raw numbers.
"PPP surveyed 1,266 likely voters from September 11th to 14th."
And since the focal point of the poll is the senate race, the likely voters are of the senate race.
Exactly. The likely voters are of the senate race. It doesn't say whether they used likely voters or registered voters for their 2016 calculation. I would be very surprised if they calculated a likely voter outcome this far in advance.
Every single question in the poll is answered by the 2014 likely voters.
PPP really should know better than to test 2016 matchups among 2014 likely voters.
How do you know that if it's not stated anywhere?
I don't consider this sentence to answer the question: "PPP surveyed 1,266 likely voters from September 11th to 14th."
It just tells about the poll in general (the main theme is after all the 2014 election), but don't say anything about specific questions within the larger poll. Do you think they only ask the baseball/American football questions to likely voters? I think not. I'm pretty sure it's only 2014 related questions which are screened using likely voters. This does of course also include policy questions that might influence their 2014 decisions, for instance their stance on minimum wage. Yet the 2016 universe is a whole 'nother ball game, so it doesn't apply really.