Here is Roper in 2020
https://ropercenter.cornell.edu/how-groups-voted-2020
They cite CBS, and CBS does not even cite their source lol.
2012
https://ropercenter.cornell.edu/how-groups-voted-2012
-Final exit poll results from interviews of randomly selected voters as they exited voting places across the country on Tuesday, November 6, 2012.
Florida, North Carolina, Ohio and Wisconsin polls included telephone interviews with early voters. The poll was conducted by Edison Media Research for the National Election Pool, the Washington Post and other media organizations. Typical characteristics have a margin of sampling error of plus or minus four percentage points. “Don’t know” and “other” responses not included.
Hell this one is only 4 states, LOL. And you talk about being consistent? Yikes.... <_<
2008 -
https://ropercenter.cornell.edu/how-groups-voted-2008
Notes: Survey by Edison Media Research/Mitofsky International for the National Election Pool (ABC News, Associated Press, CBS News, CNN, Fox News, NBC News). Sample of 18,018 voters consisted of 15,640 voters as they left the voting booths on Election Day November 4, 2008 and a telephone absentee/early voters survey of 2,378 respondents conducted October 24-November 2, 2008. “Don’t know” and “other” responses not included.
Again, Edison Research, which only uses English and Spanish, and NPR already commented on being less accurate.
edit: this is the problem with these "fact checkers" like on Forbes. They draw from so many pots and are not consistent.