Diversity and Polling response thread: county level USA map of ethnic diversity. Notice the swing states MI PA and WI have pockets of diversity. South and SW mostly diverse

Diversity lowers neighborhood social capital and social trust -even within the same ethnicity-

As the US has become more diverse - with desegregated neighborhoods - trust has collapsed

This lack of trust shows up in the polls. Working class whites aren’t picking up or revealing intentions

This response absence combined with tech changes makes polls unreliable at best and wildly wrong at worst. This trend looks like it will only increase along with diversity

Important: pollsters can’t fix this by weighting for education. Intentions can vary widely. They would have to weight for county level diversity and social trust levels

Most pollsters aren’t doing this even if they could! This explains the “shy voter” and “silent majority”. Only a few like @RobertCahaly @trafalgar_group @Peoples_Pundit @BigDataPoll and Democracy Institute try to uncover that I know of

Look at the Midwest polling spread. Huge margins. Only a couple like @trafalgar_group show a close race. ABC/Wapo WI margin out of this world!

Any model that relies on polling like @RealClearNews @NateSilver538 @Nate_Cohn @Redistrict @SamWangPhD is an exercise in futility and bound to be error prone

And no national polls didn’t fix 2016 error. Many way off in 2018. @trafalgar_group blew GA and TX. @QuinnipiacPoll and @EmersonPolling blew FL

If polling is bad for the foreseeable future what can predict an election? JP Morgan quants and other have an answer: Registration trends seems to be predictive of voter enthusiasm and outcomes. These guys have money on the line so they’ve more incentive to be correct

Registration trends clearly show a tight race #2020Election #Election2020 #ElectionTwitter #ElectionPrediction

Here’s a kicker: polling hispanics is even more difficult. Many don’t speak English well. Polling Spanish speakers poses challenges. But if this Telemundo poll is any indication we could be in for a big surprise this November

In Sum polls are bad. Models are worse. Trumps odds are likely far better and more even than reported

@CottoGottfried @realikamusume