Is RealClearPolitics trying to make things looks tighter in Pennsylvania??

[Roland99]

Link to tweet

@RealClearNews making things look tighter than they really are.
Why is the Reuters poll of Likely Voters ending 10/11 excluded? The MoE? Well, others just below ended at the same date as the last 2 included had 2 polls w/MoE LOWER than those included

Biden +5.9 in last 10 polls

[Fiendish Thingy]

1. 538 PA avg: Biden +6.5; 538 PA odds: 87/12 Biden, almost the same odds as national. Nt

[Indykatie]

2. It's Common Knowledge that RCP Picks and Chooses Which Polls to Include in its Average

I heard Nate Silver comment on this a couple of weeks ago. I actually like that because it understates Biden's lead. I feel the same about Rassmussen which has a strong Repub lean. I figure Biden is doing at least 3 points better than whatever their result is.

[Roland99]

3. yeah...that's one reason I've been tending toward them lately

even they are having trouble hiding Biden's lead

[Thekaspervote]

7. The poll aggregator for the Economist has nothing good to say about RCP

[Drunken Irishman]

9. RCP, warts and all, continues to prove fairly accurate...

Compare their final average in 2016 to 538's final projection:

In Pennsylvania, RCP had Clinton winning 46.8-44.7 (+2.1), while 538 had Clinton winning 48.9-45.2 (+3.7). Hillary lost 47.5 to 48.2. RCP underestimated Clinton's total by .7 points, while 538 overestimated Clinton's total by 1.4. RCP underestimated Trump by 3.5. 538 underestimated Trump by 3. So, RCP actually was .5 worse for Trump but more in line with Hillary's total than 538 in Pennsylvania.

In Wisconsin, RCP had Clinton winning 46.8-40.3 (+6.5), while 538 had Clinton winning 49.6-44.3 (+5.3). Hillary lost 46.5-47.2. So, the margins were closer than RCP's, which favored a bigger Clinton win. However, RCP overestimated Clinton's support by just .3 points, while 538 overestimated it by 3.1 points.

I hate to say it, but I'll default to RCP most the time because I think they've proven pretty accurate.

Case in point: the final RCP 2016 Florida average had Trump winning the state 48.6-47.4. He won it 47-46.6. 538 had Hillary a 55% favorite, and winning it 48.1-47.5. Close, of course, but 538 overestimated Hillary's support by 1.5, while RCP overestimated it by just .08.

RCP's average is pretty dang accurate, IMO, despite the crap it gets.

[Drunken Irishman]

10. RCP was closer to the actual results in 2016 than Nate Silver's model.

Compare their final average in 2016 to 538's final projection:

In Pennsylvania, RCP had Clinton winning 46.8-44.7 (+2.1), while 538 had Clinton winning 48.9-45.2 (+3.7). Hillary lost 47.5 to 48.2. RCP underestimated Clinton's total by .7 points, while 538 overestimated Clinton's total by 1.4. RCP underestimated Trump by 3.5. 538 underestimated Trump by 3. So, RCP actually was .5 worse for Trump but more in line with Hillary's total than 538 in Pennsylvania.

In Wisconsin, RCP had Clinton winning 46.8-40.3 (+6.5), while 538 had Clinton winning 49.6-44.3 (+5.3). Hillary lost 46.5-47.2. So, the margins were closer than RCP's, which favored a bigger Clinton win. However, RCP overestimated Clinton's support by just .3 points, while 538 overestimated it by 3.1 points.

I hate to say it, but I'll default to RCP most the time because I think they've proven pretty accurate.

Case in point: the final RCP 2016 Florida average had Trump winning the state 48.6-47.4. He won it 47-46.6. 538 had Hillary a 55% favorite, and winning it 48.1-47.5. Close, of course, but 538 overestimated Hillary's support by 1.5, while RCP overestimated it by just .08.

[helpisontheway]

4. A poll was just released showing Biden leading by 3 😬nt

[Drunken Irishman]

5. It's being excluded because a new Reuters/Ipsos poll has replaced it.

From what I can remember, RCP never allows one pollster to double-dip.

Reuters/Ipsos just released a poll today that had Biden +4, which was a narrowing from their last poll where he led by +7.

It would not have been excluded if it was a completely different polling outfit. But because it's the same, they dropped the last one. It's why RCP also doesn't include a week-plus worth of tracking poll results in their national averages.

If you look at their national results, you'll see they only count the IBD/TIPP once - despite the fact it's a daily poll.

[Roland99]

6. Ah, yes. I missed that. Still...the cutouff at 6 polls? With others ending the same dates ?

[Drunken Irishman]

8. I think six is their cut off if there aren't enough recent polls.

Florida has more polls in their average than Pennsylvania right now.

But Nevada only has five - but the oldest poll was from September.