Race, Phenotype, and Economic Disparities: Evidence from Los Angeles, California

This study builds on the “Color of Wealth: Los Angeles” report by studying the relationship between skin tone, physical attractiveness, and socio-economic outcomes both within and across racial groups in the city of Los Angeles. We use novel face-to-face survey data where interviewers use standardized scales to rate respondents’ physical attractiveness and skin tone in addition to collecting detailed information on financial and health outcomes. We go further than similar studies that estimate racial gaps in socio-economic outcomes by studying racial-ethnic groups (i.e. U.S. blacks, African blacks, Mexicans, Koreans, and Cambodians) as opposed to only racial groups (i.e. blacks, Hispanics, and Asians). Our findings show that across African American, Vietnamese, Korean and Cambodian participants, lighter skin tone correlates with more favorable economic and social outcomes. The opposite pattern is found within the Mexican community in which darker-skinned Mexicans appear to have higher earnings than their lighter-skinned counterparts. This appears to contradict what has been observed about preferences for lighter skin in both the U.S. and Latin America. However, it could be explained based on the immigration patterns of dark-skinned Mexican immigrants who first settled in the Los Angeles area.

Executive Summary

  • The respondents surveyed in the face-to-face interviews are younger than those surveyed by phone: the median age is on average ten years lower for each of the ethnic groups covered in both surveys. As a result, reported income, net worth, marriage rates, foreign-born rates, among other variables, are lower in the face-to-face survey. However, the relative ranks of each of the racial groups—in terms of key financial outcomes—were similar.
  • We compare the face-to-face survey subsamples of those that consented to have the interviewer take a picture of their faces with those that did not. We use the Heckman two-stage selection model and find statistically significant selection bias. We show how to control for self-selection in OLS regression models to obtain unbiased results.
  • Descriptive analyses on skin complexion rating show the face-to-face sample is skewed towards lighter skin tones, skin tones with rates of 3 and 4 account for 43 percent of the observations. We find that skin tone and wealth correlation to be positive only for Other Hispanics and Korean, and it is insignificant for all other groups. The correlation between skin tone and earnings is negative and only significant when accounting for the entire sample. Within group correlation between earnings and skin tone is not statistically significant.
  • OLS results show that the significance for skin tone disappears when controlling for racial-ethnic groups. However, it shows significance when interviewer fix effects are added in the regression – which means unobservables at the interviewer (a proxy employer) level influences the relationship between earnings and skin tone.
  • Regression results indicate that darker skin colors are associated with lower earnings for whites and African Americans, while that was not necessarily the case for Hispanics or Asians.
  • Descriptive analyses on appearance or attractiveness rating reveal that 93 percent of the respondents received a score of “About Average” or above – showing a skewed distribution. Also, skin tones with 3 or below ratings account for approximately 60 percent of the observations that received “Attractive” or “Very Attractive” scores – showing a statistically significant correlation between attractiveness and skin tone. The correlation is higher for Whites (-0.41), Other Hispanics (-0.23), Koreans (-0.20), Mexicans (0.15), and US Blacks (-0.11). We find no significant correlation between attractiveness and skin tone for Africans and Cambodians.
  • We analyze skin color, attractiveness, and earnings together and find that when we control for both attractiveness and skin tone jointly, attractiveness is not statistically significant. In contrast, the coefficient for skin tone rating is significant. However, once we add the race variables as controls along with other demographic variables, skin tone loses its significance. We find that the interaction of skin tone and attractiveness is significant even when controlling for race and different demographics. However, its significance goes away if interviewer fixed effects are added. Thus, supporting the hypothesis that beauty is in the eyes of the beholder.
  • Our wealth gap decomposition results show substantial differences in the average wealth (net worth) gap across races. For example, the difference between the average net worth of whites and US Blacks is 0.80 standard deviations, with 44 percent of this gap explained by group differences in age, education, and gender; and 66 percent unexplained, showing evidence of discrimination. On the other hand, for Africans and Mexicans, most of the gap can be explained by differences in the included covariates. For Africans, the gap is 0.67 standard deviations with 99.9 percent explained by group differences in observables. For Mexicans, the gap is 0.78 standard deviations with 96.8 percent of the gap explained by differences in observables.
  • The wealth gap decomposition results based on skin tone differences within groups show no evidence of a skin-tone-driven wealth gap within racial-ethinic groups. For example, for U.S. Blacks, we find a wealth gap of 0.02 standard deviations, with lighter complexioned individuals having higher net worth than darker complexioned individuals. However, the wealth gap is not statistically significant. We find similar results for Koreans and Cambodians. Although we observe an opposite pattern with dark-complexioned individuals earning more than their light-complexioned counterparts for the two Asian groups, the observed wealth gaps of 0.15 and 0.11 standard deviations are statistically insignificant.
  • The earnings gap decomposition results show substantial racial differences in average earnings for most racial/ethnic groups. The gap in average 2014 earnings between whites and U.S. blacks is $23,631, with only 31 percent of the gap explained by group differences in age, education, and gender. For Mexicans, the gap is $25,209 (45 percent of the gap explained). For Other Hispanics, the gap is $25,129 (46 percent of the gap explained). For Koreans, the gap is $19,237 (4 percent of the gap explained). For Cambodians, the gap is $24,434 (41 percent of the gap explained). For African Blacks, the income gap is $9,611, but it is not statistically significant. When comparing the unexplained part of the income gap, we see that Koreans face the most considerable income discrimination of the racial-ethnic groups, followed by US blacks, Cambodians, and Hispanics. Evidence shows no bias against African Blacks that affect their income.
  • The results for the earnings gap decomposition by skin tone within racial-ethnic groups show an earnings gap of $11,280 for US blacks, with lighter complexioned US blacks earning more than darker complexioned individuals within the same group. We find that 88 percent of the skin-shade earnings gap is unexplained and statistically significant, providing evidence of discrimination or colorism affecting US Blacks. Interestingly, we find that for Mexicans, Koreans, and Cambodians there is a negative differential (-8,540, -$11,060, and -$7,960), which means that darker-skinned members of these racial-ethnic groups earn relatively more than their lighter-skinned compatriots. However, only the skin-shade earnings gap for Mexicans is statistically significant and driven mainly by unexplained factors.
  • Our findings for the racial-ethnic group decomposition on self-reported health show significant differences in self-reported health only for Koreans and Cambodians. On average, Koreans report a health score of 0.50 points higher than whites, primarily due to unexplained factors. In comparison, Cambodians report a score 1.01 points higher than whites, as a result of both explained (43 percent) and unexplained (57 percent) drivers.
  • The results for the health gap decomposition by skin tone within racial-ethnic groups show significant health differences among Mexicans only. On average, light-complexioned Mexicans report a health score 0.59 standard deviations higher than dark-complexioned Mexicans. This effect is driven mainly by unexplained factors.

Leave a Reply

Your email address will not be published. Required fields are marked *