II. Nineteenth Century United States Urbanization

Evaluating late 19^th and early 20^th century urban net nutrition offers insight into economic development, and the relative urban population size within the US reflects urbanization’s effect on material welfare during economic development (Gordan, Reference Gordan2015, p. 30). Urbanization in the United States began during the mid-19^th century along its eastern seaboard, and the US Northeast was the first urban region (Smith, Reference Smith, Whaples and Parker2013, p. 295; Troesken, Reference Troesken2003; Haines, Reference Haines2004). By 1870, 25 percent of US residents lived in urban locations (Haines, Reference Haines, Engerman and Gallman2000, p. 156, Table 4.2). In 1840, New York City was the largest urban area and the first US city to surpass 300,000 people. New Orleans and Charleston were the only two large 19^th century top-10 largest US cities in the South, indicating that urbanization was localized to the North. By modern standards, Chicago, Saint Louis, Philadelphia, and Pittsburgh were large urban centers and are included in this study. Individuals from Philadelphia and Chicago experienced conditions affected by large-scale urbanization from rising relative food prices associated with the separation of food production from food consumption, whereas the number of persons incarcerated from smaller counties that later urbanized shows how biological welfare varied as smaller populations concentrated during early development.

Since its founding, Philadelphia was a key US urban and political center and until 1790, was the largest urban center. Relative to the largest US city—New York City—Chicago’s Cook County, Illinois is the largest urban center in the prison sample (Figure 1). Through 1930, Saint Louis was the fourth largest US city, and Pittsburgh was an early industrial center, with a population similar in size to Saint Louis throughout the period under study. Because of their mid-western locations in the late 19^th century, Chicago and Saint Louis populations were important centers as the US developed economically and demographically. However, larger urban center growth rates converged by 1900 (Figure 1; Panel B).

Figure 1. Large Late 19^th and Early 20^th Century Urban Centers: Chicago, Saint Louis, Philadelphia, Pittsburgh, and New York.

Smaller urban areas are those that have high percentage incarcerations within each state prison and include counties with towns that would grow to be large cities: Marcopa (Phoenix), Arapaho (Denver), Douglas (Omaha), Multanah (Portland), Davidson (Nashville), Hamilton (Chatanooga), and Shelby (Memphis). These small municipal populations enter the sample at various dates; however, each grew to comparable population sizes by 1940 (Figure 2, Panel A). Hamilton and Chattanooga Tennessee were large populations early but were rapidly overtaken as settlers made their way West. While the Oregon Trail and the Northwest’s population were early urban centers in the West, incarceration in Multanah County was comparatively small until 1900; moreover, it overtook other city populations between 1900 and 1940. Portland and Denver were sizeable municipalities, while Maricopa started with a small population but grew considerably during the early 20^th century. Like larger Philadelphia, Chicago, Saint Louis, and Pittsburgh populations, smaller municipal growth rates started high and converged over time to lower growth rates (Figures 1 and 2, Panel B).

Figure 2. Small Late 19^th and Early 20^th Century Growing Urban Centers: Maracopa (Phoenix), Arapaho (Denver), Douglas (Omaha), Multanah (Portland), Davidson (Nashville), Hamilton (Chattanooga), and Shelby (Memphis).

Various health measures are related to urbanization. Average urban statures were adversely effected by pollution, and pollution is related to health and net nutrition (Bailey et al Reference Bailey, Hatton and Inwood2018; Clay et al., Reference Clay, Lewis and Severninni2018; Clay et al Reference Clay, Lewis and Severninni2019). Individuals in high disease areas with high mortality rates had greater claims on nutrition (Pope and Miner, 1988; Pope, 1989). Although the causal link is less clear, the use of urban coal generates higher carbon dioxide (CO₂), sulfur dioxide (SO₂), and nitrogen oxides (NO_x), which are related to increased mortality and morbidity rates, and may have been related to shorter statures and poor net nutrition in urban locations (Haines, Craig, and Weiss, Reference Haines, Craig and Weiss2003). Moreover, use of coal in urban areas may have inhibited calcium absorption through increased atmospheric pollutants, which reduced the amount of incident solar radiation (insolation), and sunlight combined with cholesterol in the epidermis, which is the primary source of vitamin D production and is used in calcium absorption for stature growth (Carson, Reference Carson2008; Carson, Reference Carson2009; Carson, Reference Carson2011; Carson, Reference Carson2020). Still, during the 19^th century, US households continued to migrate and reside in urban areas, indicating that the net-benefits of urban living remained positive, and urban economic opportunity was greater than the net nutrition and health effects that urban residents were required to accept.

III. Urban and Rural Body Mass, Height, and Weight Data

Military and prison records are two common sources for historical weight and height data. While there is abundant military stature data, military records do not contain sufficient numbers of older individuals or persons of African descent (Sokoloff and Vilaflour, 1982; Ellis, Reference Ellis2004; Floud et al. Reference Floud, Fogel, Harris and Hong2011; Meinzer et al., Reference Meinzer, Steckel, Baten, Steckel, Larsen, Roberts and Baten2018, p. 239). Many military records also do not include weight records, further restricting the usefulness of military records when evaluating current net nutrition. Because of military stature requirements (Fogel et al. Reference Fogel, Engerman, Trussell, Floud, Pope and Wimmer1978, p. 85; Sokoloff and Vilaflour, 1982, p. 457, Figure 1)—typically 64 inches—taller individuals disproportionately remain in military samples, which downwardly biases BMIs in military samples because BMIs are inversely related to stature (Carson, Reference Carson2009; Carson, 2012; Komlos and Carson, Reference Komlos and Carson2017). Prison records are an alternative to military records and provide greater insight into biological variation across age, race, and socioeconomic status. However, when used as measures for net nutrition, prison records have their own shortcomings. For example, because crime is frequently committed by individuals in lower socio-economic groups, prison records may represent individuals with lower socioeconomic status who committed crime to survive. Individuals with low income and wealth may have also been incarcerated because they lacked legal counsel at trial. As a result, it is likely that prison records represent net nutrition for individuals in lower socioeconomic status who turned to crime out of privation; however, there is greater biological variation with prison records than other sources (Carson, Reference Carson2009; Carson, 2012; Ellis, Reference Ellis2004; Floud et al. Reference Floud, Fogel, Harris and Hong2011; Sokoloff and Villaflor, Reference Sokoloff and Villaflor1982; Bereczi, et al., 2019, p. 190).

Data used in this study are from an extensive effort to collect 19^th and 20^th century prison records to evaluate health and net nutrition during economic development. Records used are from the Arizona, Colorado, Idaho, Illinois, Kentucky, Missouri, Mississippi, Montana, Nebraska, New Mexico, Oregon, Pennsylvania’s East and West prisons, Philadelphia, Tennessee, Texas, Utah, and Washington. Records were enumerated at the time individuals admitted into a facility, therefore, reflect pre-incarceration conditions and not conditions within prisons. Data are recorded from 1860 through1940, and prison enumerators recorded complexion, age, gender, weight, height, period received, residence, crime, and pre-incarceration occupation. Because prison records had legal implications in case individuals escaped and were later recaptured, prison records were recorded with care. Prison records also helped identify persons within prisons. Because few females were systematically recorded, only males observations are considered in this study.

Race is classified from a complexion variable recorded as white, black, mixed-race, Native-American, Mexican, and Asian. Individuals of African descent were described as black, chocolate, light, medium, and dark black. Individuals of European descent were recorded as white, light, medium, and dark. This white complexion scheme is further supported by individuals claiming European birth in American prisons who were recorded with the same white, light, medium, and dark complexions. There was a higher proportion of blacks in the prison sample than the general population (Steckel, Reference Steckel2000; Haines Reference Haines, Engerman and Gallman2000), which was attributable, in part, to vagrancy laws that incarcerated men without occupations designed to prevent recently freed-slaves from becoming dependent on society (Brands, 2010, p. 156). There were individuals of mixed African and European ancestry who were recorded as various shades of ‘mulatto.’ However, in the results that follow, individuals of mixed African and European ancestry are referred to as ‘mixed-race.’ There were individuals of mixed Native Mexican and European immigrants who were Mexican-Mestizos and are classified as Mexicans. Individuals from China, Japan, and Korea are classified as Asians.

Pre-incarceration occupations were recorded in prison registries, and five occupation categories are used to classify occupations in the late 19^th and early 20^th centuries. Bankers, government workers, physicians, and the clergy are classified as white-collar workers. Blacksmiths, shoemakers, and boilermakers are classified as skilled workers. Farmers include general farmers, ranchers, and stockmen. Laborers, servants, and cooks are classified as unskilled workers. Workers with no recorded or illegible occupations are classified as no specified occupation.

Individuals are partitioned in Table 1 by urban and rural location to further assess demographic and economic conditions by residence. In both historic and modern populations, crime is committed by the young (Hirschi and Gottreddson, 1983; Gottfredson and Hirschi, Reference Gottfredson and Hirschi1990; Carson, Reference Carson2009 EHB; Carson, Reference Carson2018 HM; Baten and Steckel, 2019, p. 317), and teenagers were more likely to reside in rural locations (Table 1). Individuals in their 20s and 30s were more likely to reside in urban areas; however, for older ages, results are mixed between urban and rural areas. White-collar and skilled workers were more likely to reside in urban areas, while farmers and unskilled workers were more likely to reside in rural areas (Fogel, 1974, p. 134). Race and urbanization in late 19^th and early 20^th century cities are well represented in the sample. Whites and mixed-race individuals were more likely to live in urban areas (Fogel, 1974, p. 132), whereas Mexicans, Native Americans, and Asians lived in rural locations. Among the native-born, individuals from the Far West, Plains, and Southwest resided in rural locations, whereas individuals’ native to the Great Lakes, Middle Atlantic, Northeast, and Southeast were more likely to live in urban locations.

Table 1. Urban and Rural Late 19^th and Early 20^th century Characteristics

Source: Arizona State Library, Archives and Public Records, 1700 W. Washington, Phoenix, AZ 85007; Colorado State Archives, 1313 Sherman Street, Room 120, Denver, CO 80203; California State Archives, 1020 O Street, Sacramento, CA 954814; Idaho State Archives, 2205 Old Penitentiary Road, Boise, Idaho 83712; Illinois State Archives, Margaret Cross Norton Building, Capital Complex, Springfield, IL 62756; Kentucky Department for Libraries and Archives, 300 Coffee Tree Road, Frankfort, KY 40602; Maryland State Archives, 350 Rowe Building, Annapolis, MD 21401; Missouri State Archives, 600 West Main Street, Jefferson City, MO 65102; William F. Winter Archives and History Building, 200 North St., Jackson, MS 39201; Montana State Archives, 225 North Roberts, Helena, MT, 59620; Nebraska State Historical Society, 1500 R Street, Lincoln, Nebraska, 68501; New Mexico State Records and Archives, 1205 Camino Carlos Rey, Santa Fe, NM 87507; Ohio Archives Library, 800 E. 17^th Avenue, Columbus, OH43211; Oregon State Archives, 800 Summer Street, Salem, OR 97310; Pennsylvania Historical and Museum Commission, 350 North Street, Harrisburg, PA 17120; Philadelphia City Archives, 3101 Market Street, Philadelphia, PA 19104; Tennessee State Library and Archives, 403 7^th Avenue North, Nashville, TN 37243 and Texas State Library and Archives Commission, 1201 Brazos St., Austin TX 78701; Utah State Archives, 346 South Rio Grande Street, Salt Lake City, UT 84101; Washington State Archives, 1129 Washington Street Southeast, Olympia, WA 98504.

To the extent that BMI, stature, and weight represent biological and material inequality. Stature has been used to illustrate biological and material inequality, stature CVs and Gini Coefficients from urban centers were similar (Moradi and Baten, Reference Moradi and Baten2005). However, as a measure for inequality, stature is genetically determined. Stature also follows a normal distribution and is less sensitive to net nutrition variation than income distribution (Sokoloff and Vilaflour, 1982, p. 456). Rural BMIs and weight were distributed more equally than in urban areas. Rural areas were more abundant in net nutrition, disease rates were lower, and their biological and material inequality was more equal. Rural agricultural diets and close proximity to nutrition created environments where nutrition was accessible, disease environments were less virulent, and did not create as much nutritional stress. Subsequently, rural BMIs were higher, statures taller, weights heavier, and nut nutrition distributed more equally than urban areas.

IV. Body Mass, Height, and Weight by Demographics, Socioeconomic Status, and Urban Residence

Late 19^th and early 20^th century urban and rural net nutrition were related to race, demographics, and socioeconomic status. We now test which of these variables were associated with BMI, height, and weight by urban residence. To start, urban and rural BMIs and weights for the i^th individual are regressed on height, race, demographics, socioeconomic status, and observation period. Urban and rural heights are regressed on race, demographics, socioeconomic status, and birth period.

Body Mass Index

(1) $$\eqalign{ BM{I_i} = {\theta _0} + {\theta _c}Centimeter{s_i} + \sum _{e = 1}^5{\theta _e}Rac{e_i} + \sum _{a = 1}^3{\theta _a}Age_i^a + \sum _{n = 1}^{14}{\theta _n}Nativit{y_i} + \sum _{j = 1}^4{\theta _j}Occupation{s_i} \cr + \sum _{r = 1}^{10}{\theta _r}Decade{\kern 1pt} Receive{d_i} + \sum _{m = 1}^{11}{\theta _m}Urba{n_i} + \sum _{s = 1}^{16}{\theta _s}Residence + {\varepsilon _i} \cr} $$

Height

(2) $$\eqalign{ Centimeter{s_i} = {\theta _0} + \sum _{e = 1}^5{\theta _e}Rac{e_i} + \sum _{a = 1}^3{\theta _a}Age_i^a + \sum _{n = 1}^{14}{\theta _n}Nativit{y_i} + \sum _{j = 1}^4{\theta _j}Occupation{s_i} \cr + \sum _{r = 1}^{10}{\theta _r}Decade{\kern 1pt} Receive{d_i} + \sum _{m = 1}^{11}{\theta _m}Urba{n_i} + \sum _{s = 1}^{16}{\theta _s}Residence + {\varepsilon _i} \cr} $$

Weight

(3) $$\eqalign{ Kilogram{s_i} = {q_0} + {q_c}Centimeter{s_i} + \sum _{e = 1}^5{q_e}Rac{e_i} + \sum _{a = 1}^3{q_a}Age_i^a + \sum _{n = 1}^{14}{q_n}Nativit{y_i} \cr + \sum _{j = 1}^4{q_j}Occupation{s_i} + \sum _{r = 1}^{10}{\theta _r}Decade{\kern 1pt} Receive{d_i} + \sum _{m = 1}^{11}{\theta _m}Urba{n_i} + \sum _{s = 1}^{16}{\theta _s}Residence + {\varepsilon _i} \cr} $$

For BMIs and weight, statures in centimeters are included to test the relationship between current and cumulative net nutrition (Carson Reference Carson2009; Carson, 2012; Carson. 2015; Komlos and Carson, Reference Komlos and Carson2017; Carson, Reference Carson2018). Complexion dummy variables are included to assess how net nutrition varied by race. Annual youth age dummy variables are included to account for how net nutrition varied during early ages, while adult birth decade dummy variables are included for how adult net nutrition varied at older ages. Nativity dummy variables are included for birth in the Northeast, Middle Atlantic, Great Lakes, Plains, Southeast, Southwest, and Far West. International nativity dummy variables are included for Africa, Asia, Australia, Canada, Europe, Great Britain, Latin America, and Mexico. To assess the relationship between net nutrition and socioeconomic status, occupation dummy variables are included for white-collar, skilled, farmer, and unskilled occupations. There are two ways to interpret BMI, height, and weight variation over time. Measured in the current period, BMIs and weight reflect the current net nutrition experienced by diverse cohorts at the time of measurement. Measured since birth, stature reflects how the same cohort’s cumulative net nutrition varied since birth. Birth decade dummy variables are included in height regressions, and observation period dummy variables are included in BMI and weight models (Carson, Reference Carson2019, p. 32). For BMI, height, and weight, urban dummy variables are included to account for how net nutrition varied in larger urban relative to rural areas.

Three paths of inquiry are considered when evaluating relationships between net nutrition, demographics, socioeconomic status, and urbanization. First, the early industrial growth puzzle and antebellum paradox are the propositions that net nutrition decreased during early urbanization and industrialization (Komlos, Reference Komlos1987; Zehetmeyer, 2011; Carson Reference Carson2008, pp. 366-368; Sunder, Reference Sunder2011, p. 168; Sunder, Reference Sunder2013, p. 248), and the pattern is robust across interdisciplinary studies (Berecski et al 2019, p. 187; Meinzer et al., 2019, p. 232; Davidson, et al., Reference Davidson, Rose, Gutman, Haines, Condron, Condran, Steckel and Rose2002, pp. 238-241). BMI, height, and weight averages are presented over time to assess net nutrition throughout the 19^th and early 20^th centuries. Because there is concern over unobserved sample selection bias, time trend weights are augmented in Figure 4 with bubble figures, where circle sizes represent sample proportions (Zimran, Reference Zimran2019; Meinzer, 2019, p. 235, Figure 3). Two general patterns over time are present between BMIs, heights, and weights by urban and rural locations: how they varied between urban and rural locations and how they varied over time. First, urban BMIs were comparable to rural values, which occurred because individuals in rural locations had taller statures and heavier weights, and BMIs are inversely related to height squared (Carson, Reference Carson2009; Zehetmayer, 2013, pp. 161, 167, 176, and 184; Carson, 2012; Komlos and Carson, Reference Komlos and Carson2017). Second, throughout the 19^th and early 20^th centuries, urban BMIs remained approximately constant at a little over 23.1, however, increased mildly in the 1880s and 1910, while rural BMIs had a sustained decrease from 1890 through 1940. The result is that urban and rural BMIs varied with early industrialization, and the difference between the urban and rural BMIs were positive after 1880.

Figure 3. Late 19^th and Early 20^th Century Urban and Rural BMIs over Time.

Source: Stature regression coefficients from Table 6, Models 1 and 2, are weighted by sample size in Table 1.

Note: Circle size represents sample proportion.

Figure 4. Late 19^th and Early 20^th Century Urban and Rural Heights over Time.

Source: Stature regression coefficients from Table 6, Models 3 and 4, are weighted by sample size in Table 1.

Note: Circle size represents sample proportion.

Net nutrition varied by urban-rural status, yet not all urban locations had the same physical environments. Tables 3, 4, and 5 further partition urban status BMI, height, and weight by county. The antebellum paradox is the contradictory result that statures decreased while wages and income increased (Libergott, Reference Libergott1984; Craig et al Reference Craig, Goodwin and Grennes2004; Bogart, Reference Bogart2009; Komlos and Coclanis, Reference Komlos and Coclanis1997, pp. 439-441; Craig, Reference Craig, Komlos and Kelly2016). After weighting for unobservable factors, rural statures were taller than urban statures and both decreased between 1840 and 1870; however, the decrease in urban stature was deeper, and preceded the rural stature decrease (Figure 4; Zehetmayer, 2013, pp. 161, 167, 176, and 184). The greatest stature difference between urban and rural statures was during the 1870s, when households urbanized in the post-Civil War era. For example, Davidson et al. (Reference Davidson, Rose, Gutman, Haines, Condron, Condran, Steckel and Rose2002) illustrates that urban statures were shorter and decreased with the separation of food production from food consumption, and European statures decreased with early industrialization (Carson, Reference Carson2008; Carson and Hodges, Reference Carson and Hodges2014, Meinzer et al., 2019, pp. 232-244).Footnote ² Greater population density increased the relative price of food and worsened disease environments (Voth and Leunig, Reference Voth and Leunig1996, p. 559). Wilson (Reference Wilson and Costa2003) illustrates that high and increasing 19^th century chronic respiratory disease levels were associated with urbanization, industrialization, and pollution, and Bailey et al. (Reference Bailey, Hatton and Inwood2018) indicate part of the effects of deteriorating net nutrition were due to urban atmospheric pollution associated with increased demands on net nutrition from morbidity and disease (Haines, Craig, and Weiss, Reference Haines, Craig and Weiss2003; Zehetmayer, 2013, pp. 161, 167, 176, and 184; Clay et al. Reference Clay, Lewis and Severninni2018; Clay et al. Reference Clay, Lewis and Severninni2019). Moreover, Table 3 illustrates that BMIs were lower in counties that had greater population densities, and individuals in Philadelphia—the most urban location in the sample—had lower BMIs than individuals located elsewhere in the United States (Table 4; Cuff, Reference Cuff2005; Hiermeyer, Reference Hiermeyer2010, p. 128; Correia, Luck, and Verner, Reference Correia, Luck and Verner2020).

Table 2. Biological Inequality by Residence

Source: See Table 1.

Table 3. Late 19^th and Early 20^th Century Urban and Rural BMIs by Characteristics

Source: See Table 1.Note: Bold is significant at .05.

Table 4. Late 19^th and Early 20^th Century Urban and Rural Height in Centimeters by Characteristics

Source: See Table 1.Note: Bold is significant at .05.

Table 5. Late 19^th and Early 20^th Century Urban and Rural Weight in Kilograms by Characteristics

Source: See Table 1.Note: Bold is significant at .05.

Throughout the late 19^th and early 20^th centuries, average rural weight was greater than urban weight, and rural current net nutrition exceeded urban net nutrition (Figure 5). Both urban and rural weight temporarily increased during the late 19^th century and returned to pre-1800 average weights. Nevertheless, the urban-rural average weight difference decreased between 1850 and 1870 but experienced a sustained trend-reversal in 1870. Moreover, it was not simply urbanization, but the size and magnitude of stature by residence differences. Individuals in larger urban centers were mostly made worse-off with urbanization and had a greater willingness to accept diminished urban health in exchange for economic opportunity (Tables, 2, 3, and 4). Subsequently, rural current and cumulative net nutrition exceeded urban net nutrition, and the two varied in different ways during early urbanization and industrialization.

Figure 5. Late 19^th and Early 20^th Century Urban and Rural BMIs over Time.

Source: Stature regression coefficients from Table 6, Models 1 and 2, are weighted by sample size in Table 1.

Note: Circle size represents sample proportion.

Second, considerable research illustrates stature differences by race, which varied between urban and rural locations. Steckel (Reference Steckel1979) was the first to demonstrate taller statures for individuals with fairer complexions. Johnson (Reference Johnson1941, pp. 256-257) and Fogel and Engerman (Reference Fogel and Engerman1974, p. 132) show that individuals of mixed African and European ancestry with fairer complexions were more common in urban areas. Because there were external urban agglomeration effects that may have varied by race, higher concentrations of fairer complexioned Africans in urban locations may have also been associated with better living conditions and net nutrition (Higgs, Reference Higgs1977, pp. 35-37). Bodenhorn (2002, pp. 23, 30, and 43) attributes taller statures for Africans with fairer complexions to 19^th century social preference. However, if taller statures accrued to fairer complexioned blacks because of social preferences, whites should have had greater BMIs and heavier weights than individuals with darker complexions. In fact, the opposite is true, and individuals with darker complexions had higher BMIs and heavier weights than individuals with fairer complexions (Carson, Reference Carson2015a; Carson, Reference Carson2015b). Higgs (Reference Higgs1977, pp. 33-35) suggests that urban black net nutrition may have been better relative to rural black workers if there was greater access to low-priced urban diets and more progressive urban institutions that shielded urban blacks from racial prejudice prominent in rural communities. Urban blacks, should have, therefore, had higher BMIs, taller statures, and heavier weights than blacks in rural locations. Alternatively, black net nutrition may have been worse than whites if urban economic and social conditions put pressure on urban net nutrition that foreclosed lower socioeconomic blacks from opportunity. However, darker complexioned blacks had shorter statures than fairer complexioned whites and mixed-race individuals, yet had higher BMIs and heaver weights (Higgs, Reference Higgs1977, p. 31, 34, and 37). Higher BMIs and heavier weights are associated with shorter statures and poorer net nutrition (Carson, Reference Carson2009; Carson, 2015; Carson, Reference Carson2008). Subsequently, urban blacks had poorer cumulative net nutrition but greater BMIs and heavier weight.

Third, biological markers were related to socioeconomic status, and late 19^th and early 20^th century agricultural workers consistently had greater BMIs, taller statures, and heavier weights than workers in other occupations (Cuff, 1993, p. 177; Craig, Weiss, and Haines, 2003, p. 406; Carson, Reference Carson2017, pp. 26-27; Carson, Reference Carson2009, pp. 154-155; Carson, 2015, pp. 951-955). However, because their physical sizes had greater returns in physically demanding agricultural occupations, greater BMIs, taller statures, and heavier weights reflect both net nutrition and occupation comparative advantage, where taller, larger individuals were in agricultural occupations (Margo and Steckel, Reference Margo, Steckel, Fogel and Engerman1992, p. 518; Steckel and Haurin, Reference Steckel, Haurin and Komlos1994, pp. 120-122). Table 6 partitions rural and urban workers and illustrates that urban farmer BMIs were greater than rural values (Gordan, Reference Gordan2015, p. 55). White-collar and skilled rural weights were lower than workers with no occupations, indicating that rural workers with no occupation were probably in agricultural occupations, faced low net prices for nutrition, and benefited from sparse population densities (Table 6, Church et al., Reference Church, Thomas, Tudor-Locke, Katzmarzyk, Earnest, Rodarte, Martin, Blair and Bouchard2011). Therefore, after controlling for residence, rural agricultural workers had better net nutrition and had taller statures than workers in other occupations.

Table 6. Late 19^th and Early 20^th Century Urban and Rural BMI, Height, and Weight by Characteristics

Source: See Table 1.Note: Bold is significant at .05.

Other patterns are consistent with expectations. Nativity within the US indicates that native Northeastern blacks had shorter statures, and early Northeastern urban residence was associated with lower cumulative net nutrition for both blacks and whites (Zehetmayer, 2013, pp. 161, 167, 176, and 184). However, blacks and whites from the Northeast had the heaviest weight. After controlling for observable characteristics and urban residence, men native to the Middle Atlantic, South, and West had lower BMIs than men in other US locations. Lower Northeast and Middle-Atlantic BMIs were attributable to lower weights and current net nutrition, whereas lower Southern BMIs were attributable to taller statures and greater Southern cumulative net nutrition (Carson, Reference Carson2008; Carson, Reference Carson2009; Hilliard, Reference Hilliard1972). International nativity demonstrates that urban and rural Asians and Latin Americans had lower BMIs, shorter height, and lower weight independent of urban-rural nativity.

V. Decomposing the Urban-Rural BMI, Height, and Weight Difference

Decompositions further illustrate net nutritional differences by urban-rural locations. Oaxaca decompositions are a statistical technique used to partition dependent variable differences into structural and compositional differences. To isolate how 19^th and early 20^th century urban and rural net nutrition varied by characteristics, let $\gamma _{h}$ and $\gamma _{l}$ be BMI, height, and weight dependent variable values. $\theta _{0h}$ and $\theta _{0l}$ are non-identifiable high and low value characteristics in the BMI, height, and weight components intercept. $\theta _{1h}$ and $\theta _{1l}$ are high and low coefficients associated with returns to characteristics. $\overline{X}_{h}$ and $\overline{X}_{l}$ are high and low characteristic matrices. High and low BMI, height, and weight are expressed in vectors.

(4) $$\gamma _{h}=\theta _{0h}+\theta _{1h}\overline{X}_{h}$$

and

(5) $$\gamma _{l}=\theta _{0l}+\theta _{1l}\overline{X}_{l}$$

High and low response variable gaps are differenced and the counter-factual $-\theta _{1h}\overline{X}_{l}+\theta _{1h}\overline{X}_{l}$ is added.

(6) $$\Delta \gamma = {\gamma _h} - {\gamma _l} = {\theta _{0h}} + {\theta _{1h}}{\bar X_h} - {\theta _{1h}}{\bar X_l} + {\theta _{1h}}{\bar X_l} - {\theta _{0l}} - {\theta _{1l}}{\bar X_{1l}}$$

which is rearranged into the decompositions:

(7) $$\gamma _{h}-\gamma _{l}=\left(\theta _{0h}-\theta _{0l}\right)+\left(\theta _{1h}-\theta _{1l}\right)\overline{X}_{l}+\left(\overline{X}_{h}-\overline{X}_{l}\right)\theta _{h}$$

(8) $${\gamma _h} - {\gamma _l} = \left( {{\theta _{0h}} - {\theta _{0l}}} \right) + \left( {{\theta _{1h}} - {\theta _{1l}}} \right){\bar X_h} + \left( {{{\bar X}_h} - {{\bar X}_l}} \right){\theta _l}$$

Equation 7 evaluates dependent variable differences at low average characteristics and high returns to characteristics. Equation 8 evaluates dependent variable differences at high average characteristics and low returns to characteristics. Equations 7 and 8’s first right-hand side element, $\left(\theta _{0h}-\theta _{0l}\right)$ , is the difference in the autonomous differences due to non-identifiable characteristics, such as wealth, disease, and diet. The second right hand side element, $\left(\theta _{1h}-\theta _{1l}\right)\overline{X}_{l}$ , is the structural returns difference due to characteristics. The third right-hand side element, $\left(\overline{X}_{h}-\overline{X}_{l}\right)\theta _{h}$ , is the difference in compositional effects, and a large composition difference indicates that dependent variable differences are due to differences in sample compositions rather than returns to characteristics.

Table 7 presents the urban-rural net nutrition decompositions for late 19^th and early 20^th century males in the US. The proportional intercept indicates that independent of characteristics, urban BMIs were greater than rural BMIs (Figure 6). Among observable characteristics, rural stature and cumulative net nutrition had the greatest BMI return differences. Urban age, complexion, and decade received also had significant structural returns, while average compositional returns were smaller, indicating urban residential characteristics were favorable to net nutrition. Urban BMIs were greater than rural BMIs, and level returns to average characteristics were greater than returns to average characteristics, and urban BMIs were greater because of compositional rather than structural differences. Independent of characteristics, rural statures were taller than urban statures, and besides complexions, rural statures were greater than urban stature returns (Figure 7). Independent of characteristics, returns to rural weight were greater than urban weight; however, the weight returns to height were greater in urban relative to rural areas. Urban weight returns associated with height mostly offset identified sources in weight returns, followed by urban weight returns to occupations, nativity, and age. Rural weight structural returns were greater for observation period and complexions.

Table 7. Late 19^th and Early 20^th Century Urban and Rural BMIs, Height, and Weight Decompositions by Characteristics

Source: See Tables 1 and 6.

Figure 6. BMI Rural vs. Urban, Difference in Decompositions.

Source: See Table 7, Panel A and B.

Figure 7. Rural vs. Urban Stature Difference in Decompositions.

Source: See Table 7, Panel B.

Table 8 presents black-white net nutrition decompositions for late 19^th and early 20^th century individuals in urban and rural locations. Black BMIs were greater than white BMIs, and black level returns to characteristics were greater than returns to average characteristics, indicating that returns to black characteristics were greater than average returns because of structural rather than average return differences. White BMI weight returns were greater than blacks for stature, age, occupations, and observation decade. Subsequently, black BMIs were greater than whites associated with genetics and unobserved characteristics in the intercept, such as diets, disease, and percent protein in muscle tissue; however, whites had greater BMI returns associated with cumulative net nutrition, age, socioeconomic status, and observation period.

Table 8. Late 19^th and Early 20^th Century Black and White BMIs, Height, and Weight Decompositions by Urban-Rural Locations and Characteristics

Source: See Tables 1 and 6.

Blacks and whites have the potential to reach comparable statures when brought to maturity under ideal biological conditions (Tanner, Reference Tanner, Harrison, Weiner, Tanner and Barnicot1977; Carson, Reference Carson2009; Carson Reference Carson2020); however, ideal net nutritional conditions and stature varied between blacks and whites. Whites were taller than blacks associated with non-observable sources in the intercept, which includes genetics and nutrition differences between blacks and whites (Carson, Reference Carson2008; Carson, Reference Carson2009). White returns to stature were greater than blacks associated with birth year, occupations, and urban status. Blacks had greater stature returns associated with nativity and age. Like black BMIs, black weights are greater for each unit of tissue mass because of biological differences, which includes blacks having greater protein in muscle tissue, and protein is heavier than fat (Figure 8;Wagner and Hayward, 2000; Schutte et al., Reference Schutte, Townsent, Hugg, Malina and Blomquist1984; Barondess et al., Reference Barondess, Nelson and Schlaen1997; Aloia et al., Reference Aloia, Vaswani, Ma and Flaster1997). Blacks had greater weight returns associated with genetics, nativity, and urban counties, and black stature returns to characteristics offset white stature returns to average characteristics. Whites had greater weight returns associated with stature, ages, observation year, residence, and occupations, indicating that whites had greater current net nutrition associated with cumulative net nutrition, demographics, and socioeconomic status with genetics and urban counties.

Figure 8. Rural vs. Urban Weight Difference-in-Decompositions.

Source: See Table 7, Panel C.

VI. Conclusion

Nineteenth and early 20^th century urban residence imposed costs on worker health and net nutrition, and urban residents had a greater willingness to accept diminished urban health in exchange for economic opportunity. Stature represents cumulative net nutrition, and nativity and residence in large 19^th century US urban areas were shorter than their rural counterparts. Despite disease, high relative food prices, and pollution, urban economic and social opportunities were greater, and the relative gains to net urban living exceeded the health negative externality associated with urban industrialization and the cost of migrating to rural areas. Urban BMIs were comparable to rural BMIs, urban heights shorter, and urban weights lower than individuals in rural locations. Net nutrition varied by race, and blacks had greater BMIs, shorter statures, and heavier weights than whites. White stature variation over time reflects access to nutrition relative to disease and environmental insults, and it is largely accepted that white-American statures stagnated between the 19^th century’s second and third quarters, a pattern known as the antebellum paradox. However, African-American statures did not exhibit the same stature decrease, and explanations suggest that black statures did not decline because they were not subject to the nutritional disruptions associated with industrialization and urban environmental change. Net nutrition also varied by socioeconomic status, and urban farmers had lower BMIs, shorter statures, and lower weight than rural farmers, indicating that urban agricultural net nutrition by socioeconomic status was worse than rural socioeconomic status (Gordan, Reference Gordan2015, p. 55). Urban nativity was the greatest source of structural returns, followed by age, and there was little compositional difference between urban and rural locations. However, there were greater returns to rural occupations, and rural returns to average characteristics offset the advantage to urban occupations, indicating little causal explanation between urban and rural statures by socioeconomic status. Despite the diminished net nutritional opportunities in urban locations, throughout the late 19^th and early 20^th centuries, individuals and households continued to urbanize indicating there was greater willingness-to-accept poorer urban health and net nutrition in exchange for urban economic opportunity.