WI-PPP: Clinton +12, at 50%

[Mr. Morden]

Are they a reputed Pollster

Called every swing state correctly in 2012, with some dead on.

Yes, but they've been criticized for potentially cooking the books in order to match the polling average:

http://fivethirtyeight.com/features/heres-proof-some-pollsters-are-putting-a-thumb-on-the-scale/
https://uselectionatlas.org/FORUM/index.php?topic=197049.0
https://uselectionatlas.org/FORUM/index.php?topic=179353.0



They should not be able to get that close to the polling averages by chance, given their sample sizes.  Same applies to other robo-pollsters, for that matter.

[Mr. Morden]

Are they a reputed Pollster

Called every swing state correctly in 2012, with some dead on.

Yes, but they've been criticized for potentially cooking the books in order to match the polling average:

http://fivethirtyeight.com/features/heres-proof-some-pollsters-are-putting-a-thumb-on-the-scale/
https://uselectionatlas.org/FORUM/index.php?topic=197049.0
https://uselectionatlas.org/FORUM/index.php?topic=179353.0

They should not be able to get that close to the polling averages by chance, given their sample sizes.  Same applies to other robo-pollsters, for that matter.

PPP has been accused of fabricating data and polls by some people.

Much as I admire Nate, he's wrong this time.

How is he wrong?  Also, which Nate are you talking about?  Both Cohn and Silver have criticized PPP. 

No... they failed to release a correct poll because they assumed it was way off.  They didn't fabricate anything. 

A pollster not releasing a poll because they don't think it looks right is absolutely dishonest.

[Mr. Morden]

If they're showing Clinton at 12% while other pollsters show her more like 7% up, then it seems obvious that they aren't cooking their books anymore.

That's totally possible.  My point is that if they were indeed herding, then we can't assess their current accuracy based on how they performed in past elections.

[Mr. Morden]

Are they a reputed Pollster

Called every swing state correctly in 2012, with some dead on.

Yes, but they've been criticized for potentially cooking the books in order to match the polling average:

http://fivethirtyeight.com/features/heres-proof-some-pollsters-are-putting-a-thumb-on-the-scale/
https://uselectionatlas.org/FORUM/index.php?topic=197049.0
https://uselectionatlas.org/FORUM/index.php?topic=179353.0

They should not be able to get that close to the polling averages by chance, given their sample sizes.  Same applies to other robo-pollsters, for that matter.

PPP has been accused of fabricating data and polls by some people.

Much as I admire Nate, he's wrong this time.

How is he wrong?  Also, which Nate are you talking about?  Both Cohn and Silver have criticized PPP. 

No... they failed to release a correct poll because they assumed it was way off.  They didn't fabricate anything. 

A pollster not releasing a poll because they don't think it looks right is absolutely dishonest.

My point is, I think this is more PPP being very Bayesian versus fudging the numbers.

What do you mean by "being very Bayesian"?  I mean, any kind of pollster methodology that takes the results from other polls into account is inherently dishonest, unless I'm missing something here?

[Mr. Morden]

Are they a reputed Pollster

Called every swing state correctly in 2012, with some dead on.

Yes, but they've been criticized for potentially cooking the books in order to match the polling average:

http://fivethirtyeight.com/features/heres-proof-some-pollsters-are-putting-a-thumb-on-the-scale/
https://uselectionatlas.org/FORUM/index.php?topic=197049.0
https://uselectionatlas.org/FORUM/index.php?topic=179353.0

They should not be able to get that close to the polling averages by chance, given their sample sizes.  Same applies to other robo-pollsters, for that matter.

PPP has been accused of fabricating data and polls by some people.

Much as I admire Nate, he's wrong this time.

How is he wrong?  Also, which Nate are you talking about?  Both Cohn and Silver have criticized PPP. 

No... they failed to release a correct poll because they assumed it was way off.  They didn't fabricate anything. 

A pollster not releasing a poll because they don't think it looks right is absolutely dishonest.

My point is, I think this is more PPP being very Bayesian versus fudging the numbers.

What do you mean by "being very Bayesian"?  I mean, any kind of pollster methodology that takes the results from other polls into account is inherently dishonest, unless I'm missing something here?

Basically, PPP takes into account the data from their previous polling and modifies it with new data.

You mean on a horserace question like "Who are you voting for for president?" they actually take topline data and incorporate that into subsequent polls?  I haven't seen any documentation indicating that that's what they're doing.  But if they are, isn't that pretty problematic?  What we as poll consumers are interested in is what new information a poll is providing.  But if their published poll result is actually some kind of weighted average that includes results from previous polls, then we're being fooled as to what the new information is.

If they think a weighted average of polls gives a more accurate indication of where the race stands, then fine.  But they should still publish the individual polls results as separate things, and then show the weighted average alongside it.

[Mr. Morden]

Are they a reputed Pollster

Called every swing state correctly in 2012, with some dead on.

Yes, but they've been criticized for potentially cooking the books in order to match the polling average:

http://fivethirtyeight.com/features/heres-proof-some-pollsters-are-putting-a-thumb-on-the-scale/
https://uselectionatlas.org/FORUM/index.php?topic=197049.0
https://uselectionatlas.org/FORUM/index.php?topic=179353.0

They should not be able to get that close to the polling averages by chance, given their sample sizes.  Same applies to other robo-pollsters, for that matter.

PPP has been accused of fabricating data and polls by some people.

Much as I admire Nate, he's wrong this time.

How is he wrong?  Also, which Nate are you talking about?  Both Cohn and Silver have criticized PPP. 

No... they failed to release a correct poll because they assumed it was way off.  They didn't fabricate anything. 

A pollster not releasing a poll because they don't think it looks right is absolutely dishonest.

My point is, I think this is more PPP being very Bayesian versus fudging the numbers.

What do you mean by "being very Bayesian"?  I mean, any kind of pollster methodology that takes the results from other polls into account is inherently dishonest, unless I'm missing something here?

Basically, PPP takes into account the data from their previous polling and modifies it with new data.

You mean on a horserace question like "Who are you voting for for president?" they actually take topline data and incorporate that into subsequent polls?  I haven't seen any documentation indicating that that's what they're doing.  But if they are, isn't that pretty problematic?  What we as poll consumers are interested in is what new information a poll is providing.  But if their published poll result is actually some kind of weighted average that includes results from previous polls, then we're being fooled as to what the new information is.

If they think a weighted average of polls gives a more accurate indication of where the race stands, then fine.  But they should still publish the individual polls results as separate things, and then show the weighted average alongside it.

I don't think it's necessary a good idea, but it does explain why their results are similar to herding.

What I described above is herding.  It's just herding using their own polls, rather than polls from other pollsters, no?