I think it is 90% selection bias of various types. I doubt any other possible contributing factors, would be nearly as significant a viewer selection bias
The most obvious is “beauty selection bias”. Go to an impoverished urban center, and you’ll see many “ugly” people of mixed parentage, many more roughly ‘average-looking’, quite a few with the transient beauty of youth (lower SES correlates with high birth rate). However, we tend to disregard unattractive people (and depending on private nature, often average-looking people as well). Of the exceptionally attractive people we see (leaving aside personal tastes), those who are unusual for a clearly defined reason (like mixed race) stand out, and stick in our minds with a label under them.
There’s are also social selection of parents. (The “attractive people” you are admiring were probably born in the 1960s-80s, when mixed race mating was more stigmatized than it is today especially in the middle and upper classes; presumably a greater degree of mutual attraction was required on average to bring such couples together – pretty parents tend to produce pretty kids: they are likely to share the generally preferred proportions and features. That’s one reason I cited the lowere SES urban centers, where the barriers began to break earlier (due to propinquity) and where, therefore, mixed couples of average appearance (or less) are more common. This happens almost anywhere that the barriers are lowered: Hawaii, the Carribean, etc. have many attractive people, but I’ve never seen any evidence that they are more attractive on avergage
While there is probably some merit to the idea of “averaged” faces, it doesn’t apply to individuals. An individual is the mix of his/her parents, not their “races” and therefore not much more likely to be pretty or ugly [on average] than the child of any othe parents of the same degree of attractiveness as their own.
When it comes to research on facial averages, I much prefer the research that focuses on mathematical models like “eigenfaces” or socio-, psycho-, anthropological papers that are written with a good understanding of the findings of studies based on those models (I often check the bibiographies before analyzing a paper - it’s good to know what major work they considered or ignored). Too many papers on this subject fixate on methods or some specific conclusion. Some skew is unavoidable, but mathematical modeling studies seem to eliminate many possible complicating factors
For example, marked signs of age or obesity are often deliberately eliminated from the sample pool of photo evaluations studies. These are not ‘complications’, but precisely the kind of “reality” that justifies photo studies over CGI models. Any study that eliminates them is skewed at the outset. I know many attractive people with very distinct signs of aging or excess weight. Any robust theory or principle of attractiveness must account for them as easily as the young or slim. The older and heavier populations of the US, combined, are easily the majority, yet many papers treat youth and slenderness as the norm. One can only speculate why.
Of course, these are just my offhand thoughts and personal opinions, even less reliable than my usual posts