So let's take these one at a time. On number one, this is the line of reasoning that led to the Great Unskewed Polls debacle of 2012. You can read more about it at the link, but Party ID is fluid and susceptible to question wording. 1/
realclearpolitics.com/articles/2012/…
realclearpolitics.com/articles/2012/…
In other words, you'll get a different answer if you ask "how are you registered" versus "what party do you typically vote for" versus "how do you view yourself." At times I have had a different answer for all three of those questions simultaneously. 2/
This makes assessing Party ID tricky. There have been clever attempts to weight by Party ID (Rasmussen has asked a huge national sample then had used results to weight polling), but in any event, I try to avoid paying too much attention to public poll internals, esp party ID. 3/
On the second point, it matters . . . depending what the question is. Since the president is theoretically president of the entire nation, it can be useful to identify his job approval among the nation. 4/
Now, it can be tricky to predict an election from this, which is where registered voters or likely voter screens come in. But if you want to know POTUS JA among adults -- again, there are legit reasons for this -- it makes sense to ask adults. 5/
On the last point, he has a legit point. Not on the job approval question, which is the first question asked, and is the standard "Do you approve or disapprove of the way Donald Trump is handling his job as president? " 6/
But later on, they ask "Mueller prosecuted Paul Manafort, Trump’s former campaign chairman, who was found guilty of tax and bank fraud by a jury last week. Do you think the case against Manafort was justified or unjustified?" 7/
The inclusion of the fact that a jury found Manafort guilty will probably tend to prime respondents who might have weak opinions, or no opinions, to conclude the investigation was justified. 8/
Then immediately following it with a question about whether Manafort should be pardoned will probably prime leaners/no opinions to oppose the pardon. 9/
And so forth. After a series of questions about Manafort, Sessions, Cohen, it then asks the impeachment question. 10/
Does this ruin things? Absolutely not. It gives us a sense of where the American people stand after being presented with a certain set of facts. Where would be with a different set of facts presented or omitted? Hard to say. 11/
But here's the thing, *all* questions are biased. There's no neutrality here. *Omitting* the facts about Manafort/Mueller/Cohen would probably give a different answer, but not the "right" answer. It would just give public opinion when not directly presented with these facts. 12/
It's like the old Rush song: "If you choose not to decide you still have made a choice." Opting for a different question wording or no wording at all loads things, just in a different direction. 13/
So any way, there's nothing wrong with the poll, but there might be problems with the way people interpret it. As always, the best practice is to look at a wide variety of polls and question wordings. 14/14
