The Great African Height Puzzle

What are Africans so tall? They are not just incredibly tall in an absolute sense. They are much taller than they ought to be given everything we know about the determinants of stature. Height is known to be an excellent proxy for health status and living standards. Adult height is partly a function of in utero environment (ie, the health status of the mother during pregnancy), so that, for instance, birth weight is a good predictor of adult height. Height also responds strongly to nutritional intake and disease burdens during the adolescent growth spurt. Above all, adult height is known to be controlled by childhood health insults — the best predictor of mean adult height for populations is the postneonatal mortality rate (the death rate between 1 month and 1 year). Height is strongly correlated with every known measure of health status and living standards we have — outside sub-Saharan Africa. Height is a good predictor of grip strength, lung capacity, mortality rates, morbidity rates, life expectancy, cognitive test scores, wages, income, socioeconomic status, and subjective measures of well-being. Heights increased in tandem with income, life expectancy, and every other measure of wellbeing, in every single nation that modernized and rode up the hockey stick in the twentieth century.

Even if tall stature in Africans was genetic, or in the old language, a racial character, the relationship within sub-Saharan Africa should be the same as elsewhere. That is to say, within African populations, height should be correlated with other measures of health status. But that isn’t the case either. The correlation between life expectancy and height in the cross section of 86 nations outside the region is extremely large and statistically significant (r=0.78, p<0.001). But within the region it inexplicably vanishes (r=0.14, p=0.425). The African height puzzle also works against the grain of the thermoregulatory logic of morphological adaptation to the macroclimate which predicts than variants in colder climes are bigger than variants of the same species in warmer climes. So appealing to environmental adaption only gets you deeper into the puzzle.

Figure 1 displays the scatter plot for life expectancy and height. The red dots are countries from sub-Saharan Africa. Given the linear relationship between life expectancy and height elsewhere, we estimate that Africans are 7.6cm (std error=0.62) or a full 3 inches taller on average that they ought to be.

African_height_puzzle.png

Health economists have been scratching their heads about this for a very long time. Bozzoli, Deaton, and Quintana-Domeque (2009) develop a model that predicts that at sufficiently high levels of malnutrition, selection dominates scarring. In other words, when the nutrition and disease environment is hostile enough, only the strong, those who would grow up to be tall, survive. When things are bad enough, this selection effect dominates the usual scarring effect associated with health insults in childhood. But, as they note, their model explains only a ‘part of the African height paradox’, while worsening the fit elsewhere. So, even after accounting for selection and scarring, Africans are way too tall.

The problem with genetic explanations goes beyond the missing gradient within sub-Saharan Africa noted above. Height is simply too plastic an anthropometric variable. We have data going back, not a few decades or centuries, but thousands of years, thanks to the allometric relationship between femur length and stature. And as we have seen before, height fluctuates extremely rapidly to be deeply canalized, ie under very tight genetic control. That leaves us two speculative possibilities.

One possibility is that height has been under much stronger sexual selection in Africa than elsewhere. This hypothesis would predict greater sexual dimorphism in body size within Africa than elsewhere. This hypothesis seems implausible on its face since African women are also much taller than they ought to be. I don’t have data on male and female height at hand. But I do have ye olde Howell’s craniometric dataset. We know that skull size roughly scales with body size, so this will have to do for now. A quick calculation shows that skull size is not much more dimorphic in Africa than elsewhere. To the contrary, the ratio of male to female skull sizes in that very restricted database is the lowest in the African demes: 1.038 in Africa, 1.041 in the Americas, 1.047 in Asia, 1.047 in Europe, 1.044 in Oceania, and 1.051 in Polynesia. We can check the dimorphism indices for height. But this doesn’t seem like a promising line of investigation.

Another possibility is that height is under tighter genetic control within Africa than elsewhere. This is not as crazy as it sounds. Africa is the mother of mankind. We know from the literature of ‘gene surfing’ that genetic mutations that would almost certainly get culled by purifying selection in the mothership can thrive on the wavefront of range expansions. This looks like selective sweeps in the expanding frontier, even if the genes surfing the waves are neutral. Perhaps the Out-of-Africa founder populations acquired mutations that made their heights more plastic than populations back home in Africa. This is an easily testable hypothesis. For if this speculative theory is right, then femur lengths in the African paleontological record should not only be long going back to before the Out-of-Africa dispersals, they should also be much more stable than elsewhere.

As things stand, no on earth knows why Africans are so unreasonably tall. This is one of the great open questions of the day.


Postscript. The plot thickens. Turns out that while African adults are dramatically taller than they should be, African children are no taller than they ought to be.

African_children.png

Height-for-Age (HAZ) is given in z-scores in terms of the WHO reference population (which captures the average growth patterns in six countries at the nutritional frontier). HAZ is a very strong predictor of adult heights at both individual and population levels. In the following scatter plot, the line displayed is the least squares line for the cross-section of countries outside Africa. We can see that African countries are clustered above that line — meaning that African children somehow catch up after early childhood so that adults end up too tall even though African children are no taller than they are supposed to be.

HAZvsHeight.png

Table 1 displays my estimates for three models of adult height as a function of HAZ. Model (1) is a simple linear model; model (2) includes a dummy for Africa; model (3) includes an interaction term between the Africa dummy and HAZ. Comparing (2) to (1), we can see that the slope of HAZ is attenuated without the Africa dummy, and that the Africa dummy is very large and bears a positive sign. On the other hand, the interaction term in model (3) is small and insignificant. This means that the intercept for African countries is larger than non-African countries but the slope is no different. That is, African adults are much taller than one would expect given the heights of African children, but the cross-country gradient is the same.

Table 1. Regression results. 
(1) (2) (3)
HAZ 0.502 0.591 0.523
0.086 0.090 0.224
Africa 0.548 0.540
0.198 0.188
Africa*HAZ -0.089
0.253
R^2 0.363 0.457 0.459
Adj. R^2 0.351 0.437 0.429
MSE 0.468 0.406 0.412
Response is adult height. HAZ is z-score for height-for-age in terms of the WHO reference population. We report Newey and West’s robust std errors. Source: DHS, Clio-Infra, author’s computations. 

In a series of papers, Alexander Moradi looked at the African height puzzle. He found the same pattern that I just documented. He also showed that it is highly implausible that selection could dominate scarring sufficiently to account for the African height premium. Moreover, assuming that African Americans largely share the genomic endowment of sub-Saharan Africans, if the premium had a genetic basis, it should show up in the same growth pattern within African Americans. But that’s simply not the case. In dramatic contrast to Africans in Africa, African American growth curves are very close to the WHO rich-country reference group. So it is highly unlikely that the puzzle has a genetic explanation.

AA.png

Moradi (2010) offered two hypotheses to resolve the African height puzzle. As Hirvonen and Moradi summarize in the 2015 paper, “The African Enigma: The Mystery of Tall African Adults Despite Low National Incomes Revisited”:

First, the incidence of diseases is much less severe during puberty compared to early childhood. Age-specific mortality rates underline this. Most of the deaths are concentrated in infancy and childhood, not among adolescents. This can be due to a less harsh age-specific morbidity as well as an acquired immunity at puberty. Better health conditions then translate into considerable catch-up growth, when children go through the adolescent growth spurt. Second, puberty is also the age when adolescents start to contribute more and more to the household income. This may improve their intra-household bargaining power leading to increased food shares within the household or can supplement their food portions from their own income. While plausible, there is no strong evidence for these hypotheses as of yet.

Anomalies are extremely important in scientific progress. They point to flaws in the reigning paradigm. And their resolution often requires a completely new way of thinking. The African height anomaly is of great importance not just to health economists and economic historians but also physical anthropologists and paleoanthropologists. The interpretation of a vast body of evidence compiled over the past hundred years depends on it.


Postpostscript. Sexual dimorphism in height is more pronounced in taller countries. The correlation between mean height and dimorphism is 0.7. That basically rules out the sexual selection hypothesis of African heights.


Postpostpostscript. Thinking some more about it, Moradi’s hypotheses are quite unlikely to account for the African height anomaly. Basically, there is no reason to believe that within-household bargaining power or increased immunity to communicable diseases in adolescence is peculiar to Africa. Indeed, these two observations are just as true of South Asia, where both adults and children are extraordinarily small (although the shortfall is not that great given the pattern of nutrition and disease burdens).

Meanwhile, the plot thickens some more. Harvard’s Yoko Agakuchi and David Canning examined the time trends in infant mortality, nutrition, and height across scores of developing countries in Asia, Africa and Latin America. What they found was astonishing. Infant mortality fell in all countries and heights rose as expected, but only outside Africa! Heights in Africa declined even as infant mortality fell. The reason is that per capita protein consumption did not rise as infant mortality fell, in contrast to what happened elsewhere in the developing world.

Trend.png

This means that infant mortality is an insufficient metric for the health status of populations. What seems to have happened in Africa is that, probably due to interventions by international NGOs, a lot more children survived but the health status of surviving children did not improve. If anything, it deteriorated, likely because nutrition did not improve.

Most developed countries underwent a transition in which infant mortality fell and adult height increased. We argue that this pattern continues to hold true today in much of Latin America and Asia where reductions in infant mortality go hand in hand with improved nutrition, reductions in child morbidity, and increases in adult heights. In Sub-Saharan Africa, however, we find a distinctively different pattern: while infant mortality has been falling, adult heights have been stagnating, or even declining, over the last 50 years. … One possible reason behind the divergent trends in infant mortality and adult height in Sub-Saharan Africa is the source of the infant mortality reductions in the region. In most other regions of the nutrition, as measured by energy (calorie) or protein intake per person, has been increasing. Sub-Saharan Africa in contrast has experienced little in the way of increases in nutrition. Rather than broad based improvements in nutrition and public health measures such as access to clean water and sanitation, mortality reduction in Sub-Saharan Africa appears to have occurred through health interventions measures that directly reduce mortality with limited effect on disease prevalence, morbidity, and the physical development of children.

As for the puzzle of African heights, they are as clueless as everyone else.

Adult Africans are tall compared to people in other countries with similar incomes, despite apparent poor nutrition and high mortality rates when young. We argue that these cross country differences in height are due to unobserved fixed factors, not child health. These country fixed effects could to some extent reflect genetic differences between people in different countries. There is evidence that greater genetic variation exists between humans in Africa than in other regions, because of Africa’s large historical population while humans in other continents are mainly descendents of small migrant groups that have less diversity (). Nonetheless other factors, such as selection effects (Deaton 2007, ), underestimation of nutrition (, ), or the distribution of resources () could play a large role. At present we are not in a position to definitively identify the reasons for the lack of consistent relationship between the level of infant mortality and adult height.

The pattern seems ripe to be misinterpreted as African genetic exceptionalism. But mobilizing genetic exceptionalism does nothing to resolve the enigma. For DNA is mere code. Even if potential height of Africans is greater than that for the populations in other continents, where do they get the nutrients to achieve their adult height? We know from genome wide association studies that potential height is extremely polygenic and greatly influenced by rare genes. This makes sense given the plasticity of the height phenotype. Even if the difference is in DNA, that only moves the explanandum to a different level of analysis. For then the question becomes, what explains the exceptional gene frequencies of height-associated genes in African populations? Given that height and body size in general is costly to sustain, why are they not selected against as elsewhere? As usual, genetic explanations turn out to be non-explanations upon closer examination.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s