PA: Insider Advantage (B-) Trump +3

[Statistical]

34. So some red flags that might help people spot other dubious polls

Last edited Mon Oct 26, 2020, 04:02 PM - Edit history (1)

https://overland.amgreatness.com/app/uploads/2020/10/PA-POll-Oct-25th-.pdf

First the sample size is only 400 voters. For a race like PA and at this point in the race that is way too low. It means the MoE is 4.9% meaning that the +/- margin could be off by as much as 9.8% meaning not +3 Trump but actually +6.7 Biden.

If you want to run a lot of garbage polls and then just publish the results you like a large MoE is very useful. So you run a poll it shows Biden up 2 and you don't report it, you run it again and it shows Biden up 4 and you don't report it. You run it again and it shows Trump up 3 and you report that.

Most highly rated polling companies are using larger samples with smaller MoE (+/- 2%) in close states. If the race is close a wide MoE makes the results basically useless because most likely the results are going to be within the MoE (statistical tie) like this one is. The only way to avoid that it to use a larger sample with a smaller margin but if you goal is to published outliers that actually works against you.

Second is the tiny number of independents in the demographic. Independents are breaking for Biden so underrepresenting them is a way to push the results. They only makeup 8% of the poll vs the 20% normally. The poll internals show them breaking for Biden 3 to 1.