Celebrating 10 years! 2007-2017

Red wave? Blue wave? Cut the noise, says poll of the 60 closest races. All bets are off.

https://www.cnbc.com/amp/2018/10/15 /mixed-signals-for-democr billcarson10/16/18
Here you go. https://fivethirtyeight.com/feature s/electio nycatt10/16/18
Polls are only accurate when done properly and quickly. You qdllc10/21/18
Prediction markets, despite having their own flaws (they can aarongray10/26/18
Polling is now about shaping the election rather than measur massivemissive10/26/18
billcarson (Oct 16, 2018 - 8:28 am)

https://www.cnbc.com/amp/2018/10/15/mixed-signals-for-democrat-blue-wave-in-november-cnbc-survey-says.html


This is all oddly familiar to 2016. Why is polling either (a) so terribly done, or (b) so terribly used.

The polls that focus on regions with actively competitive races show both major parties in the mid-40s. This excludes deep South red meat Trump Voters and the motivated urban voters in Dem districts. Neither of these types matter. Its about winning districts.

I can almost hear the “abolish districts, let the unitarily elected leader appoint!”

Reply
nycatt (Oct 16, 2018 - 5:16 pm)

Here you go.

https://fivethirtyeight.com/features/election-update-an-upset-that-nobody-sees-coming-could-determine-control-of-the-house/

Reply
qdllc (Oct 21, 2018 - 7:15 am)

Polls are only accurate when done properly and quickly. You can skew outcome just by how you word questions (often done to show support or opposition to an agenda contrary to what people really believe).

People can also change opinion quickly. Polls that take days to conduct and analyze are behind the curve on the current mood of the people. In spite of the technology existing to poll thousands in a matter of hours, most big polling services still use 20+ year old methodology...producing errors.

Reply
aarongray (Oct 26, 2018 - 5:55 pm)

Prediction markets, despite having their own flaws (they can become echo chambers, there's meaningful debate as to whether market prices can be reliably read as percentages, amongst others), are a good way to counterbalance the inherent shortcomings of polls. As an aside, it's also much harder to manipulate prices than most some pundits would have you believe.

Prediction markets are accurate more often than not, and much of their derision in the popular press stems from poor journalist understanding regarding both the meaning probability and confusion about the empirical data. Just because they're often misunderstood doesn't mean that people who can accurately access their data shouldn't be leveraging it.

They offer full coverage of outcomes (all districts), and they update and reflect new information in real time, unlike polls. They also incorporate information that can't be shown in polls - historical trends, as one example. Plus, they're obviously incentivized in a manner that polls are not.

Prediction markets, especially when you used in conjunction with aggregate forecasts, are a much better reflection of the likely outcomes than methods like nationwide 800 person polls. I guess the hot takes you can get limited polls better suits our current media model, though.

Reply
massivemissive (Oct 26, 2018 - 7:10 pm)

Polling is now about shaping the election rather than measuring voter preference.

Reply
Post a message in this thread