So I maybe agree with your intuitive conclusion that "the polls started out relatively unbiased, but have been getting more biased towards Democrats over time".
But because performance, and thus poll bias, is going to be hugely correlated across the nation, the sample size is even tinier than it appears. So e.g…
So I maybe agree with your intuitive conclusion that "the polls started out relatively unbiased, but have been getting more biased towards Democrats over time".
But because performance, and thus poll bias, is going to be hugely correlated across the nation, the sample size is even tinier than it appears. So e.g., a bias/unexpected advantage across 50 races could all be explained by Trump driving the vote up at the same time.
2.
> it also raises the possibility that the likely Republicans who cooperated with the poll were much more likely to support Biden than the likely Republicans who did not respond.
This could be checked by doing polling on Republican primaries, and seeing if the share of votes in the Republican primary is similar to the share predicted by polls.
3.
> It’s important to note that we haven’t seen anything resembling a fair fight between pollsters and betting markets. That’s because polling operates with the support of major media...
Also, prediction markets have access to Nate Silver, but Nate Silver doesn't use his access to prediction markets. Back in the 1900s, when polls first appeared, they did handily beat the prediction markets of the day
4.
> The fact that Salem/CSPI and PredictIt tend to be extremely similar despite them having very different incentive structures implies that they are converging on a true market price in each race or forecast.
Great post, some comments:
1.
So I maybe agree with your intuitive conclusion that "the polls started out relatively unbiased, but have been getting more biased towards Democrats over time".
But because performance, and thus poll bias, is going to be hugely correlated across the nation, the sample size is even tinier than it appears. So e.g., a bias/unexpected advantage across 50 races could all be explained by Trump driving the vote up at the same time.
2.
> it also raises the possibility that the likely Republicans who cooperated with the poll were much more likely to support Biden than the likely Republicans who did not respond.
This could be checked by doing polling on Republican primaries, and seeing if the share of votes in the Republican primary is similar to the share predicted by polls.
3.
> It’s important to note that we haven’t seen anything resembling a fair fight between pollsters and betting markets. That’s because polling operates with the support of major media...
Also, prediction markets have access to Nate Silver, but Nate Silver doesn't use his access to prediction markets. Back in the 1900s, when polls first appeared, they did handily beat the prediction markets of the day
4.
> The fact that Salem/CSPI and PredictIt tend to be extremely similar despite them having very different incentive structures implies that they are converging on a true market price in each race or forecast.
Or that Salem is copying PredictIt