Pew: Clinton +9

[Mallow]

The average lead in the RCP average is Clinton +4.7. This poll is 9. It's off significantly. The relevant polling time this poll was taken is mid-to-late June. They released the results over a week later. That is when Clinton was surging a bit from the wrap up of the nomination.

The polls concluded in the past week have the race in the 3-5 point range. They have shown a consistent narrowing.

This poll was roughly int he 5-8 point Clinton ballpark at the time period that it was taken. It is not representative of where the race is right now. That bounce has faded.

Ignoring for now the fact that FiveThirtyEight's polls-only forecast, which takes into account the relative biases, absolute errors, and recentness of all the national polls, suggests that the current state of the race is a Clinton +5.5 margin, even if we go with your +4.7 number, that still leaves Pew off by 4.3 and Rasmussen off by 6.7. So your statement earlier that they're in the same class of wrong seems shaky at best. If we go with the more realistic 5.5 current state, then Pew is off by 3.5 and Rasmussen is off by 7.5.

As for the "polls conducted in the last week" point, there are three on RCP, and two of them are clear outliers. How do you get a "polling average" from one poll? Indeed, if you average all three of them, you get Clinton +4.8, and the non-outlier is Clinton +5, so how you possibly calculate "the 3-5 range" is beyond me. Unless "3-5" is code in your book for "5". Which doesn't change the idea that the Pew poll way closer to reality than the Rasmussen poll.

[Mallow]

The average lead in the RCP average is Clinton +4.7. This poll is 9. It's off significantly. The relevant polling time this poll was taken is mid-to-late June. They released the results over a week later. That is when Clinton was surging a bit from the wrap up of the nomination.

The polls concluded in the past week have the race in the 3-5 point range. They have shown a consistent narrowing.

This polling was roughly in the 5-8 point Clinton ballpark at the time period that this Pew poll was taken. It is not representative of where the race is right now. That bounce has faded.

Ignoring for now the fact that FiveThirtyEight's polls-only forecast, which takes into account the relative biases, absolute errors, and recentness of all the national polls, suggests that the current state of the race is a Clinton +5.5 margin, even if we go with your +4.7 number, that still leaves Pew off by 4.3 and Rasmussen off by 6.7. So your statement earlier that they're in the same class of wrong seems shaky at best. If we go with the more realistic 5.5 current state, then Pew is off by 3.5 and Rasmussen is off by 7.5.

As for the "polls conducted in the last week" point, there are three on RCP, and two of them are clear outliers. How do you get a "polling average" from one poll? Indeed, if you average all three of them, you get Clinton +4.8, and the non-outlier is Clinton +5, so how you possibly calculate "the 3-5 range" is beyond me. Unless "3-5" is code in your book for "5". Which doesn't change the idea that the Pew poll way closer to reality than the Rasmussen poll.

I never equated Pew to Rasmussen except to put both polls in the same trash heap for the reasons I have stated twice now.

I put this in the same ballpark as Trump +2 with Rasmussen.

Yep, gonna call B.S. on that one.

You must be logged in to read this quote.

You don't remember correctly, and I don't want to make it "3-6", I want to make it "4.5-5.5", which is really the only defensible range at this juncture (and the 4.5 is generous). Tacking on 3, or even 4.0, is wholly partisan wishcasting. As I said earlier, using your own "past week" cutoff and your own choice of RCP, there have been precisely three general matchup polls, Clinton +5, Clinton +11, and Trump +2. Extending it back to polls that ended within the past two weeks, we have the two Rasmussens at Trump +4 and Trump +2, two Clinton +2 polls, three Clinton +4 polls, two Clinton +5 polls, two Clinton +6 polls, a Clinton +9 poll, Clinton +10 poll, a Clinton +11 poll, and a Clinton +12 poll. That gives a mean of Clinton +4.9 and a median of Clinton +5.

These are just the facts.

[Mallow]

I'll also point out that, according to a standard analysis of what makes an "outlier" an outlier (that a data point is greater than 1.5 times the interquartile range outside the first or third quartile), the only outlier over the past two weeks is the Rasmussen Trump +4 result.

Specifically, 1.5*IQR is 6.75, the first quartile is at Clinton +3, and the third is at Clinton +7.5. So, any poll over the past two weeks that gives Trump a bigger margin than +3.75 or Clinton a bigger margin than +14.25 is an outlier. So if we re-did the analysis without the Trump +4 outlier, it only looks more favorable for Clinton.