Selection and historical height data: Evidence from the 1892 Boas sample of the Cherokee Nation

Abstract

Abstract:Economists have increasingly turned to height data to gain insight into a population's standard of living. Because height measures are used when other data is unavailable, testing their reliability can be difficult, and concerns over sample selection have lead to several vigorous debates within the heights literature. In this paper, I use a unique contemporaneous census to gauge the extent of selection into a contested sample of American Indian heights. I have linked people from the 1892 Boas sample of the Cherokee Nation to the 1890 Cherokee Census. An initial analysis finds evidence of negative selection into the Boas sample. A detailed examination of those measured reveals a more complex story. Two distinct groups are present within the data. The first group consists of 64 members of the Cherokee elite. Their households owned more land, had invested more in improvements to their land, and had higher literacy rates. The remainder of the Boas sample is poor relative to both the elite and the rest of the Cherokee Nation. Part, but not all, of this difference is due to their residential location. Forty percent of the Boas sample lived in poorest district of Cherokee Nation. These differences in wealth between the two groups were mirrored by a fairly dramatic difference in average heights. The average height of all men in elite group was 173.9 cm while the non-elite were several centimeters shorter at just 171.2 cm.

Age misreporting: An empirical investigation using the New Zealand contingents in the Second Boer War

“…Miller (2016) is a smaller study that uses census matching to investigate the pattern of selection in a contested sample of Cherokee Nation heights. The author matches the 1892 Boas sample (with 239 observations) to the contemporaneous 1890 Cherokee Census (of approximately 27,000 citizens), by name and age (within 2 years).…”

The Slaughter of the Bison and Reversal of Fortunes on the Great Plains

“…Recent work comparing the Cherokee in Boas' sample to the Cherokee census suggests that the Boas sample is representative on average, but that it may over-represent the upper and lower classes (Miller, 2016). The Cherkoee are likely to be di↵erent from the average Native American nation for a number of reasons, so generalizing from Miller (2016) to the entire Boas sample may be inappropriate.…”

“…Recent work comparing the Cherokee in Boas' sample to the Cherokee census suggests that the Boas sample is representative on average, but that it may over-represent the upper and lower classes (Miller, 2016). The Cherkoee are likely to be di↵erent from the average Native American nation for a number of reasons, so generalizing from Miller (2016) to the entire Boas sample may be inappropriate. Further, other scholars have questioned the representativeness of Boas' sample specifically (Komlos and Carlson, 2014), and of height data more generally (Bodenhorn, Guinnane, and Mroz, 2017).…”

The Slaughter of the Bison and Reversal of Fortunes on the Great Plains

“…Table 1 presents summary statistics for Boas' sample in panel A and Dippel (2014)'s sample in panel B, as well as some of the additional controls we discussed above. 13 We display summary statistics by classifying a nation as bison-reliant if 60 percent of its ancestral territory overlapped 12 While there have been questions regarding the representative nature of Boas' sample (Komlos and Carlson, 2014), and of height data more generally (Guinnane, Bodenhorn, and Mroz, 2014), recent work comparing the Cherokee in Boas' sample to the Cherokee census suggests that Boas sample is representative on average, though it may over-represent the upper and lower classes (Miller, 2016). What is important for our empirical strategy is that, conditional on our set of covariates, over-or under-representation does not vary between age groups or between bison-reliant and non-reliant nations.…”