In the latest issue of American Sociological Review--the flagship journal of U.S. sociology--Andres Villareal finds that even after a number of factors are taken into account (i.e., gender, age, indigeneity, educational level, region, and rurality) darker skinned Mexicans--in Mexico--are poorer than their paler counterparts. The poverty gap is roughly as large as the American black-white difference. Even though the main ethnic division in Mexico is between indigenous people and mestizos, and skin tone variation among mestizos is not socially emphasized, Villareal concludes that the results are strong evidence for intense discrimination based on color.
Keep in mind that this research has appeared in the best quantitative sociology journal in the United States; perhaps in the world. The author is interested in discrimination, but treats skin tone variation as his measure of discrimination. Why not measure discrimination directly? At least ask respondents concerning perceived instances of mistreatment. Is that too difficult to ask about? Why doesn't he save himself a lot of work, skip the data collection and analysis, and simply write a speculative paper asserting discrimination?
The relationship between skin tone and poverty could be explained in many different ways. If I claimed that the link is explained by genes associated both with economic success and skin color, my assertion would be just as plausible as his. But it goes without saying that he wouldn't admit (and doesn't mention in the paper) that a genetic explanation is possible since it is morally out of bounds. This trick of not measuring what you are studying and ruling out other explanations as unacceptable has been used in a thousand similar studies. Instead of relying on data, a whole discipline is based on moralistic faith. If we're going to defund NPR, let's do the same to sociology.