I'd like to know how they actually calculated these county figures.
For example, it's accepted that whites become more conservative/Republican as their community becomes more non-white. And yet the map has Obama ostensibly doing better with whites in the Black Belt counties in Alabama.
They likely have the turnout % wrong and underestimated the % black people who voted in those areas, thus giving Obama a better showing in the Black Belt counties among whites. Blacks turned out at a higher % than whites in 2012.
Read this discussion. At least in Georgia's case, basically, it's a potential combination of over/underestimating black support of Obama, and also not properly determining who comprises the "Other" and "Unknown" categories on the voter rolls. The page says that exit polling (which didn't exist in GA in 2012) and Census data were used, but using only Census data wouldn't produce as accurate of a result in GA as my map, imho (which used SoS turnout figures by race). To give one example, there are close to a dozen counties in Georgia which have intensely inflated black populations because state prisons are located there. I tried to take into account various minutiae that affected the numbers, such as varying voting trends - if ever so slight - among various groups.
The first one is the NYT map, with the counties where Obama scored less than 21% of the white vote in red. The second map is my version with the same over/under 21 scenario as the NYT version. The third map is mine with the MoE I had initially determined (in this case, counties in green are where Obama received 19% of the white vote or more). As a whole, the NYT one seems to fit well between my rather cautious baseline assessment of white voters and the best-case scenario within the MoE.
Around the metro area in particular, it's hard to be as accurate either by Census data or by turnout figures when compared to most other counties/regions, since "Other" and "Unknown" (which is debated in that thread above as to who comprises those categories) comprise a larger chunk of the voting bloc and explicitly-identified Latinos and Asians are a larger voting bloc (with more nuanced voting patterns in GA than blacks or whites). I noticed that's where the biggest differences between my map and the NYT as a whole occurred; in many of the other parts of the state where our maps show different gradients (like the chunk of green int he north on the third map and the lack of a couple NE of Macon), they either just barely made or missed the cut-off.
Because of that, I think as a whole, their project is accurate. There are going to be
huge outliers in some cases/counties on the NYT map, though, thanks to how they compiled the projection.
EDIT - FYI: My spreadsheets that were used to make my 2008/2012 maps ultimately spat out Obama receiving 19% of the white vote in Georgia in 2012, and 23% in 2008 (which matched what 2008 exit polling said).