https://bit.ly/4aSXcMp https://bit.ly/3KpGWIe https://bit.ly/3yXajPm https://bit.ly/3V6mYHn https://bit.ly/3V6mZuV https://bit.ly/4aMfOxB https://bit.ly/3V6n1D3 https://bit.ly/3V6n2a5 As you can see, the deviations from a simple linear prediction are much smaller than the deviations between specific rows and columns. In particular, the rows and columns for whites and black actually have much smaller variance than the others, so attributing racial attitudes to any particular race seems to be misplaced. Most of the apparent pattern you were picking up on appears to be due to thresholding effects in your color scheme. The chart for men responding to women shows similar homogeneity (the response rate here is a lot higher, so you have to halve these deltas to get comparable values to the previous chart): Note that the response rate of white men to middle eastern women, which you picked up on as being high, is actually 0.9% lower than is predicted by a simple linear model, so clearly your visualization was misleading. The one notably off number is that indian males don’t like responding to islander females, for some odd reason. There is one very interesting data point which you neglected to notice though. If you do a scatter plot of response rate given versus response rate received across races for both men and women, you see an inverse correlation in both cases. (Your weighted averages are probably much more appropriate for this graph, because it’s weighted vs. weighted) The reason is probably the obvious one – people who can get a good response rate from mailing out are less inclined to both responding to messages they receive. This is probably true for people who get better response rates for any reason, not just race. The upshot is that although I think your data analysis is somewhat flawed, and in particular you fault white women for something they aren’t doing, I draw the same basic conclusion, which is that white men are considered more desirable than other ethnicities while black women are considered less.