Weight and economic development: current net nutrition in the late 19th- and early 20th-century United States

ABSTRACT

When traditional measures for material and economic welfare are scarce or unreliable, height and the body mass index (BMI) are now widely accepted measures that represent cumulative and current net nutrition in development studies. However, as the ratio of weight to height, BMI does not fully isolate the effects of current net nutrition. After controlling for height as a measure for current net nutrition, this study uses the weight of a sample of international men in US prisons. Throughout the late 19th- and early-20th centuries, individuals with darker complexions had greater weights than individuals with fairer complexions. Mexican and Asian populations in the US had lower weights and reached shorter statures. Black and white weights stagnated throughout the late 19th- and early-20th centuries. Agricultural workers’ had greater weights than workers in other occupations.

Acknowledgmetnts

I appreciate comments from John Komlos, Lee Carson, and Paul Hodges. Shahil Sharma, Chinuedu Akah, Meekam Okeke, Ryan Keifer, Tiffany Grant, Bryce Harper, Greg Davis, Kellye Manning, and Brandon Hayes provided research assistance.

Notes

¹ $BMI=\frac{w\left(kg\right)}{h{\left(mt\right)}^2}\Rightarrow lnBMI=lnw-2lnh.\Rightarrow dlnBMI=dlnw-2dlnh.$

$E_{BMI,w}=\frac{\mathrm{\%}\mathrm{\Delta }BMI}{\mathrm{\%}\mathrm{\Delta }w}=1E_{BMI,h}\frac{\mathrm{\%}\mathrm{\Delta }BMI}{\mathrm{\%}\mathrm{\Delta }h}=-2$.

² Body mass is also related to health, and mortality risk is high for individuals with BMIs less than 19, remains low for BMIs between 19and 27. And increases for values over 27 (Waaler 1984; Koch 2011; Fogel 1994, pp. 375–377). Costa (1993) applies Waaler’s results to a historic population and shows that this modern relationship between BMI and mortality risk holds historically, and Jee et al. (2006, pp. 780, 784–785) show this relationship is similar across ethnic groups.

³ The total prison sample includes Arizona, California, Colorado, Idaho, Illinois, Kentucky, Maryland, Mississippi, Missouri, Montana, Nebraska, New Mexico, Ohio, Oregon, Pennsylvania, Philadelphia, Tennessee, Texas, Utah, and Washington.

⁴ The occupation classification scheme is consistent with Ferrie (1997); The following nativity classification scheme is consistent with Carlino and Sill (2001): New England = CT, ME, MA, NH, RI, and VT; Middle Atlantic = DE, DC, MD, NJ, NY, and PA; Great Lakes = IL, IN, MI, OH, and WI; Plains = IA, KS, MN, MO, NE, ND, and SD; South East = AL, AR, FL, GA, KY, LA, MS, NC, SC, TN, VA, and WV; South West = AZ, NM, OK, and TX; Far West = CA, CO, ID, MT, NV, OR, UT, WA, and WY.

⁵ There is some concern regarding how prison unskilled workers compared to the general population. As expected, the percent of unskilled workers in the prison sample is greater than the percent in the general population, indicating the prison population represents conditions among the working class.

⁶ Margo and Steckel (1992, 518) and Steckel and Haurin, (1994, 122) propose that agricultural workers had greater weight and height because of comparative advantage from size associated with occupational choice. Alternatively, agricultural workers may have head greater weight associated with rural agricultural net nutrition because the relative price of rural nutrition was low.

⁷ There is some concern over the value of decomposing dependent variable differences into returns to characteristics and average characteristics because coefficient estimates vary with respect to the choice of the omitted category (Oaxaca 1973; Oaxaca and Ransom 1999). There is little concern about explaining the dependent variable gap $\left({\bar{X}}_b-{\bar{X}}_w\right){\theta }_w$. However, because the intercept is sensitive to the omitted category, identification of black and white weight decompositions are considered first, followed by foreign-native weight decompositions. $\left({\theta }_{0b}-{\theta }_{0w}\right)+\left({\theta }_{1b}-{\theta }_{1w}\right){\bar{X}}_b$ is less clear, and there is some degree of arbitrariness that is unavoidable (Yun 2008; Fortin, Lemieux, and Firpo, 2011, pp. 40 and 45).