Be sure to participate in the 2022 reader survey if you didn’t already!
Some days ago I came across this study:
- Rashad, I. (2008). Height, health, and income in the US, 1984–2005. Economics & Human Biology, 6(1), 108-126.
Height has been associated with better physical health when outcomes such as diabetes, heart disease, and obesity are considered, yet stature is rarely used in predicting comorbidities or as a proxy for physical health when analyzing outcomes such as income. Since height is a more exogenous measure than variables likely to be affected by lifestyle changes, such as obesity, observing labor market outcomes based on height may be revealing. In addition, gender and racial differences must be taken into account when analyzing the effects of height on physical health and labor market outcomes. This study utilizes the 1984–2005 samples of the Behavioral Risk Factor Surveillance System in estimating trends in height over time by gender and race, and in analyzing the relationship between height and physical health and labor market outcomes in the United States. Trends show that height has not changed substantially at a time when physical health, as indicated by the incidence of obesity, Type II diabetes, and cholesterol, has deteriorated, and earnings disparities across racial gaps persist. Results at mean values for males indicate that being 10 cm taller is associated with a 14–47% increase in obesity, an 8–13% reduction in cholesterol prevalence, and a $1874–2306 income premium. For females, results indicate that being 10 cm taller is associated with an 8–18% reduction in cholesterol, a 14% reduction in diabetes for white females, and an $891–2243 earnings premium.
There is a medium correlation between height and intelligence. I don’t know how much that will explain of this income association, but presumably a good chunk of it. But what got me interested in this study is that they plot the heights by US social race groups over time. Here’s for men:
And for women:
Mysteriously, the Black-White gaps have not only closed but reversed over time. In these both cohorts, USA went from being a legally White supremacist country with special rules for Whites, to being a legally Black supremacist country with special rules for Blacks (admission to university, hiring discrimination, extra funding for programs, tutoring etc.). So if anything, we would expect that Blacks height advantage would be increasing over time if this is caused by social privileges, but we see the opposite. Neither can the differences be explained by changes in the ancestry composition of these groups, as Blacks have remained largely the same in this time span due to endogamous mating. Whites have seen some input of European migrants, but of mixed origins so there isn’t any reason to expect the migrants to be strongly height selected, nor were they coming from tall Northern European populations, as they used to do before the immigration quotas were abolished. In the same time span too, nutrition has overall improved for all races, and probably most significantly so for the Blacks as they were the poorest to begin with and nutrition improvements should have strong diminishing returns on height. So by this reasoning, we would expect Blacks to improve the most, but they didn’t. Stunting it appears more or less absent in the USA, even in NLSY1979. One has to look to poor foreign countries to find significant proportions of children with stunting. So differences in stunting cannot explain the changes we see.
The Hispanic and Other race changes are less interpretable because we know these groups are heterogeneous and have been changing ancestry compositions over time. Hispanics appear to not gain much height, presumably because they are increasingly recent, poor, and thus highly Amerindian migrants. Amerindian ancestry makes you short. Being a foreign born person means that you experienced whatever negative effects of the environment in the home country instead of the superior American environment. I think the Other race used to be mainly Asians, but now a days, it’s increasingly a hodgepodge mixture with a lot of European admixture, hence taller.
So what options remain to explain the change? It’s possible that height selection within the White population has been stronger or opposite of that for Blacks, and this could have changed the polygenic scores of the populations over time. Unfortunately, the GSS didn’t ask about height most years, so we can’t use that. NHANES does have some data. I took a very quick look in NHANES 2013. It didn’t have a question about number of children — bizarre for a health survey — but it did ask about number of live births and whether the woman had ever been pregnant. Combining these, we can get fertility for women. These we can regress upon height and its interaction with race:
The sample sizes are not impressive. 1400 women for the completed fertility group (age >= 43), and 2000 overall. (Really, this should be redone using all the NHANES waves.) But insofar as these data are concerned, in neither case do we see any interaction with Black. The standard errors are 0.13 and 0.18, we can’t be too sure. Overall, there was a strong overall negative selection for height, with height correlating -.30 with fertility in the completed fertility cohort! (I think the contrast with age in the first model is that this overcorrects for cohort differences in fertility, which are also related to height.) Negative height selection for women has been found before, but it is not universal (seen in the UK, not seen in Dutch). Often there is a positive selection for height in men, and negative or neutral in women. The sex difference in selection for height is what resulted in the sex difference in height to begin with (but note that it is hard to evolve a sex difference between most genes for height are sex-neutral, only some are sex-linked, and only the latter can be selected upon to make a larger sex difference). Speaking of taller women, these results confirm that there is something to the stereotype about tall women being less feminine, less interested in children, being more career-minded etc.:
Height was significantly negatively correlated with importance of having children (r = .24, p < .001, n = 678) and self-rated broodiness (r = .22, p < .001, n = 678). Height was significantly positively correlated with importance of having a career (r = .11, p = .005, n = 678) and career competitiveness (r = .12, p = .003, n = 668). Height was significantly negatively correlated with ideal number of children (r = .14, p < .001, n = 672) and positively correlated with ideal own age at time of having first child (r = .11, p = .004, n = 636). The maternal tendencies questions are highly interrelated (e.g., child importance and ideal number of children, r = .52; child importance and broodiness r = .57; all p’s < .001, n = 673). The two career questions are also significantly correlated (importance of having career and career competitiveness, r = .51, p < .001, n = 669). Therefore all the dependent variables were entered into a Principal Components Analysis. Two factors with Eigenvalues greater than 1 were extracted (Factor 1: Eigenvalue = 2.2, accounting for 36.1% of the variance; Factor 2: Eigenvalue = 1.6, accounting for 26.2% of variance). Correlation coefficients less than ±0.4 were not considered to load highly on a given factor (Comrey & Lee, 1992). Variables that loaded highly on factor 1 were importance of having children (r = .81), ideal number of children (r = .71), ideal own age to have first child (r = .56), and self-rated broodiness (r = .81). Variances that loaded highly on factor 2 were importance of having a career (r = .87) and competitiveness in career (r = .87). Factor 1 was interpreted as ‘maternal tendencies’ and Factor 2 was interpreted as ‘career orientation’. The two factors were independent (r = .03, p = .50, n = 624). Factor scores were computed using the regression method. Factor 1, ‘maternal tendencies’ was significantly negatively correlated with height (r = .26, p < .001, n = 623). Factor 2, ‘career orientation’ was significantly positively correlated with height (r = .10, p = .017, n = 623).
It is possible that men’s dispreference for tall women is pushing tall women towards non-family values (“men don’t seem to like me, so I will find something else to do with my life”), but it is also possible height in women just inherently relates to masculinization from hormones. We can’t really say with these data, but smart money is on the latter. For good measure, here’s non-heterosexuality by male height from OKCupid:
The non-linearity is very interesting. By the same reasoning as above with tall women being masculinized, short men should be feminized, and thus more likely to be homosexual (most homosexual men are bottoms). I have no hypothesis for the effect at the very tall end of the spectrum. Ideas welcome!
In contrast to these results for USA, the worldwide changes in height have been nearly universally positive at the country level:
Note that we do see the same relative order of height gains, in that European populations have gained more height than African populations. Are there large between country differences in height selection?