Convergence (and divergence) in the biological standard of living in the USA, 1820–1900

Abstract

Standard economic indicators suggest that the USA experienced long-run economic growth throughout the nineteenth century. However, biological indicators, including human stature, offer a different picture, rising early in the century, falling (on average) mid-century, and rising again at the end of the century. This pattern varied across geographical regions. Using a unique data set, consisting of mean adult stature by state, we test for convergence in stature among states in the nineteenth century. We find that during the period of declining mean stature (1820–1870), heights actually diverged. Later in the century (1870–1890) we find a type of “negative” convergence indicating that stature among states tended to converge to a new, lower steady state. Only towards the end of the century (1880–1900) do we find classic convergence behavior. We argue that the diversity of economic experiences across regions, including urbanization, industrialization, and transportation improvements, explain this pattern of divergence and then convergence.

Appendix

The state-level estimates of mean adult stature were derived by inverting a technique developed by Craig and Weiss (1998) and Haines et al. (2003) to estimate the height of individuals. The base data consist of a sample of Union Army recruits from data originally collected by Fogel, Engerman et al. (ICPSR). They include recruits born in the nineteenth century for whom information was available on, among other things, place of birth and adult height. Following Craig and Weiss (1998) and Haines et al. (2003), observations on individual (white) adult male heights were estimated based on the underlying economic relationship between adult stature, net nutrition, and the economic environment in which the recruit matured. The basic equation, as estimated by Craig and Weiss is:

$$ \begin{aligned}{} HEIGHT & = \alpha + {\mathop \beta \nolimits_1 }MOVER + {\mathop \Sigma \limits_{j = 1862}^{1865} }{\mathop \beta \nolimits_{2j} }{\mathop {YEAR}\nolimits_J } + {\mathop \beta \nolimits_3 }NUTRITION \\ & + {\mathop \beta \nolimits_4 }WEALTH + {\mathop \beta \nolimits_5 }TRANSPORT + \varepsilon \\ \end{aligned} $$

HEIGHT is the height in inches of the ith Union Army recruit. MOVER is a dummy variable, which takes the value 1 if the recruit enlisted in a county other than the one in which he was born, 0 otherwise. YEAR _j is 1 if the recruit enlisted in the jth year, 0 otherwise. NUTRITION is the marketable surplus of protein production in the county in which the recruit spent infancy. WEALTH is the sum of agricultural and industrial wealth per capita in the recruit’s county. TRANSPORT is 1 if the county was on a navigable waterway, 0 otherwise. Haines et al. adjusted this basic framework by adding a labor force variable, FARMER, which equals 1 if the individual was a farmer, 0 otherwise; HINDEX, an index for the concentration of agricultural production in the county in which the recruit was born; URBAN, the proportion of the county’s population residing in an urban area; and CDR, the county’s crude death rate.

Since we are estimating mean height at the state level, we have dropped the YEAR and TRANSPORT variable. We also dropped HINDEX and CDR, because we did not have comparable state-level data for the entire nineteenth century. That leaves us with the MOVER, NUTRITION, FARMER, WEALTH, and URBAN variables. We calculated each of these variables at the state level from various primary and secondary sources. We then transformed each variable by subtracting the national mean from it. Thus we have:

$$ \begin{aligned}{} HEIGHT = \mu + \alpha DMOVER + \beta DNUTRITION + \gamma DFARMER + \delta DWEALTH + \phi DURBAN \\ \end{aligned} $$

Where HEIGHT is the mean adult (white) male height in the ith state for birth cohort born in year t; μ is the mean US height in year t; DMOVER is the difference between the proportion of the resident population not born in the ith state and proportion of the US population not born in the USA in year t; DNUTRITION is the difference between the marketable surplus of protein produced in the ith state and US production in year t. DFARMER is the difference between the agricultural share of the labor force in the ith state and the US share in year t. DWEALTH is the difference between the sum of agricultural and industrial wealth per capita in the ith state and US wealth in year t. DURBAN is the difference between the proportion of the population residing in an urban area in the ith state and US proportion in year t. The coefficients α, β, γ, δ, and ϕ are taken from column 1 of Table 7 in Haines et al. (2003, p. 407).

The estimated heights, by birth cohort, are reported in Table 8. Note that only 26 states had complete time series dating back to 1820, and thus they are the only ones used in the regression analysis above. Of course the estimates assume that the relationship estimated by Craig and Weiss and Haines et al. was stable across the century. While this is clearly a strong assumption, note that at the bottom of the table, we compare a linear combination of estimated heights (weighted by population) to the US average. Until the end of the century, differences are quite small by almost any reasonable standard. However, at the end of the century the relationship begins to breakdown. The most problematic variable was URBAN. The relationship between urbanization and height is highly non-linear, and for the five most urban states, we had to decrease the weight on the URBAN variable at the end of the century. Although there are stature data for certain populations, for some states, for some years (see for example Komlos 1987; Steckel 1995; and Sunder 2003, 2004), the estimates reported in Table 8 are only for native-born white males born in year t. For consistency we have used the estimated figures even when a sub-sample might have been available.

Table 8 Mean height in the USA, white adult males, by state, by birth cohort, 1800–1900