race X gender, per CES (which is more reliable than exit polls), presented as % Biden; % Trump

Asian men: 64/28
Asian women: 66/27
Black men: 81/14
Black women: 89/5
Hispanic men: 53/41
Hispanic women: 64/30
WhiteMen: 42/52
WhiteWomen: 48/46

https://bfschaffner.shinyapps.io/CES2020/ 

One thing that struck me about these data is that virtually every group appears to have a within-group gender gap, except for Asian voters.

All the standard caveats apply:
-Although these data are better than a typical exit, no data set is perfect. We'll need more data before being "mostly sure"
-This is esp. true for subgroup data, esp. for minority groups, etc
-There are issues w/ groups like "Hispanic" & "Asian"

Okay, so I am getting a number of questions about how these data compare to exit polls, especially in the domain of white women. CES as % of white women voting for Trump as 46% & CNN exits have this % as 55%.

Before I say anything about the divergence in these results, let me be clear: I am only talking about data. I am not defending white women. 46% is still too many in a normative sense.

So, I am *only* talking about the data. Not making a value judgement.

CES is more reliable than exits for a number of reasons, not all of which I can go into now. But one reason is sample size: CES has a MUCH bigger sample size.

Other issues are specific to exit polls, especially during a pandemic.

CNN's exit polls were posted on Wednesday, Nov 4th, so they are incomplete. I'm also not sure of what the breakdown was of in-person (at poll place) vs. live-caller, which is a new method, given the pandemic. Depending on the breakdown, the sample could skew towards Trump voters.

Exit polls also have issues even when there's not a pandemic going on. We need to wait for better data sets and more fine-grained analyses before we get a real handle on demographic data. This is the best way to think about exits:

So, again, the CES is a better data set than CNN's at this moment in time. But even the CES is not perfect! We need multiple *good* data sets to get a handle on demographic data. This is especially true for minority sub-groups (e.g. Black &/or Latino voters)

I would weight the CES data over CNN's data. One final point: it is numerically almost impossible that 55% of white women voted for Trump. That would represent a huge gain over 2016, when 47% (not 53%) appear to have voted for him. Very hard for him to lose w/ those #s

And, again, I want to be clear: I am not defending white women. In any case, almost 50% voted for Trump. To me, that is far too many.

I am only talking about the data above, not making a value judgment about white women as a group.