The Volokh Conspiracy - How much skin color discrimination is there?

This paper by economist Joni Hersch has been getting a lot of play in the blogosphere (e.g. here and here). for apparently proving that, among immigrants to the United States, those with the lightest skin color earn 8 to 15 percent higher pay than those with the darkest possible skin, even after controlling for many other variables such as English proficiency, education, and years since arrival in the US. The skin color effects found in the study go beyond black vs. white differences, since they extend also to other nonwhite immigrant groups such as Indians and East Asians.

Hersch's study is well-designed and has some impressive data. But I nonetheless have reservations about it. In particular, it fails to control for two key variables that are highly correlated with skin color among immigrants: cultural similarity of the immigrant's country of origin with the United States and quality of education. These shortcomings do NOT prove that there is no skin color discrimination against nonwhite immigrants. But they do suggest that its effects may not be as large as the study claims.

I. Cultural Similarity.

Immigrants from majority white nations (primarily Canada, Australia, New Zealand, and various European countries) generally come from societies that are much more similar culturally to the United States than those from which most non-white immigrants originate. Cultural similarity makes it easier for immigrants to assimilate (which is surely a strong predictor of economic opportunity), and also to function effectively within US businesses. This point applies even to different immigrant groups of the same race. For example, immigrants from Western Europe or Canada have higher incomes, on average, than my fellow Russian immigrants, despite the fact that both groups are mostly white (even more so in the case of Russians than Western Europeans). This is in part because those countries' cultures and business practices are closer to those of the US than Russia's, so Western Europeans and Canadians can assimilate faster and more completely.

The author of the study could have partially controlled for this variable by, for example, checking to see if the results hold up if immigrants from Europe, Canada, and Australia/NZ are excluded. Note that controlling for English proficiency (as the author does) does not fully address this issue, because schools in many countries culturally very dissimilar to the US teach English as the primary foreign language covered in school curricula.

II. Quality of Education.

The study controls for quantity of education (years of schooling), but not for quality. Unfortunately, however, the majority-white nations that send immigrants to the US have, on average, much better education systems than most of the majority-nonwhite ones. An immigrant who got 12 years of schooling in Germany is probably much better educated than one who got 12 years in a third world nation. Moreover, even within particular countries, it is often the case that whites get better quality schooling than non-whites, in part because of historical discrimination against non-whites . This is certainly true, for example, in South Africa and many Latin American countries. Even employers completely indifferent to skin color could still rationally choose to take these quality differences into account. The author does note that part of the effect she observes may be due to discrimination in immigrants' home countries, but does not consider the ways in which this affects quality of education.

This effect is much harder to control for than cultural similarity. But it might be possible to take a stab at it by, for example, controlling for how well students from the immigrant's country of origin do on international comparisons of basic educational skills.

As noted above, these points do not prove that there is a complete absence of skin color discrimination against either nonwhite immigrants or other nonwhites. They do, however, suggest that the study's findings may be overstated.

	[Ilya Somin, October 23, 2006 at 1:37am] Trackbacks How much skin color discrimination is there? This paper by economist Joni Hersch has been getting a lot of play in the blogosphere (e.g. here and here). for apparently proving that, among immigrants to the United States, those with the lightest skin color earn 8 to 15 percent higher pay than those with the darkest possible skin, even after controlling for many other variables such as English proficiency, education, and years since arrival in the US. The skin color effects found in the study go beyond black vs. white differences, since they extend also to other nonwhite immigrant groups such as Indians and East Asians. Hersch's study is well-designed and has some impressive data. But I nonetheless have reservations about it. In particular, it fails to control for two key variables that are highly correlated with skin color among immigrants: cultural similarity of the immigrant's country of origin with the United States and quality of education. These shortcomings do NOT prove that there is no skin color discrimination against nonwhite immigrants. But they do suggest that its effects may not be as large as the study claims. I. Cultural Similarity. Immigrants from majority white nations (primarily Canada, Australia, New Zealand, and various European countries) generally come from societies that are much more similar culturally to the United States than those from which most non-white immigrants originate. Cultural similarity makes it easier for immigrants to assimilate (which is surely a strong predictor of economic opportunity), and also to function effectively within US businesses. This point applies even to different immigrant groups of the same race. For example, immigrants from Western Europe or Canada have higher incomes, on average, than my fellow Russian immigrants, despite the fact that both groups are mostly white (even more so in the case of Russians than Western Europeans). This is in part because those countries' cultures and business practices are closer to those of the US than Russia's, so Western Europeans and Canadians can assimilate faster and more completely. The author of the study could have partially controlled for this variable by, for example, checking to see if the results hold up if immigrants from Europe, Canada, and Australia/NZ are excluded. Note that controlling for English proficiency (as the author does) does not fully address this issue, because schools in many countries culturally very dissimilar to the US teach English as the primary foreign language covered in school curricula. II. Quality of Education. The study controls for quantity of education (years of schooling), but not for quality. Unfortunately, however, the majority-white nations that send immigrants to the US have, on average, much better education systems than most of the majority-nonwhite ones. An immigrant who got 12 years of schooling in Germany is probably much better educated than one who got 12 years in a third world nation. Moreover, even within particular countries, it is often the case that whites get better quality schooling than non-whites, in part because of historical discrimination against non-whites . This is certainly true, for example, in South Africa and many Latin American countries. Even employers completely indifferent to skin color could still rationally choose to take these quality differences into account. The author does note that part of the effect she observes may be due to discrimination in immigrants' home countries, but does not consider the ways in which this affects quality of education. This effect is much harder to control for than cultural similarity. But it might be possible to take a stab at it by, for example, controlling for how well students from the immigrant's country of origin do on international comparisons of basic educational skills. As noted above, these points do not prove that there is a complete absence of skin color discrimination against either nonwhite immigrants or other nonwhites. They do, however, suggest that the study's findings may be overstated.