The Covid-19 outbreak and the murder of George Floyd have dramatically exposed the racial inequalities and their potentially explosive implications for US society (Baldwin 2020). In a public speech on 3 June 2020, former US President Obama has linked the current protests to the long history of slavery. A connection between slavery and the “tangle of pathology” affecting the American ghettos, and in particular family structure, had been suggested earlier on by Moynihan (1965), then President Johnson’s Assistant Secretary of Labor.1
In a new paper (Bertocchi and Dimico 2020), we study the long-term empirical determinants of the African American family structure, captured by the presence of a black single female household head, with the aim of testing the Moynihan conjecture. To do so, we dig into US history, searching for a potential explanation of the widespread diffusion of single-mother families among blacks.2 We find that the prevalence of black single female headship is not associated with slavery per se, but rather with the experience of slavery in sugar plantations. This is due to living and working conditions associated with the cultivation and production of sugar which led to the most extreme forms of slavery on such plantations, according to Follett 1997 and Rodrigue 2001.
In the Caribbean and in Latin America – where sugar planting was predominant – the slave population persistently experienced a dramatic natural decrease due to low fertility and high mortality. Sugar planters chose to maximize profits by continually skewing their labor force toward males, which could only be achieved through systematic import of male slaves. This had profound demographic and social consequences including matrifocality, forced celibacy for men, early widowhood for women, and absence of fathers for children. These consequences are believed to have shaped family formation even in the long run (Smith 1982). While in North America, the slave population grew rapidly on average, sugar plantations in the US were an exception, exhibiting the same characteristics as sugar plantations in the Caribbean and Latin America. US Sugar plantations were almost entirely confined to a few counties in southern Louisiana and neighbouring states (Tadman 2000, Coclanis 2010). The conditions of slavery and of forced celibacy in the US, with men being unable to find partners in overwhelmingly male populations, have already been linked to the increased occurrence of the single-mother family model (Kaye 2007). A larger prevalence of extended families among blacks can also be traced back to the slavery-induced fragility of the nuclear family model.
Sugar and the black family from abolition until 1940
We test the hypothesis that the unique demographic and social conditions associated with slavery in sugar plantations may have been conducive to the development of the black family. In order to identify the long-run impact of sugar slavery, we exploit the cross-county exogenous variation in suitability to sugar and other crops that were typical of the southern slave economy, such as cotton, tobacco, and rice.3 Figure 1 shows sugar suitability by US counties (1860 boundaries). Sugar suitable regions are shaded with a darker shade indicating higher sugar suitability.
Figure 1 US counties by sugar suitability
Notes: Sugar suitable regions are shaded. A darker shade indicates higher sugar suitability. Counties are shown with their 1860 boundaries.
To test the influence of sugar suitability on family structure, we first look at the period that immediately follows the abolition of slavery in 1865, and collect Census data on household heads (aged 15-89) from 1880 to 1940.4 Our focal variable, as a proxy for family structure, is a dummy that takes value 1 if the household head is a female without a co-living spouse, and 0 otherwise. We also collect information about households’ characteristics as well as county-level geographical and historical controls.
Figure 2 Single female headship and sugar suitability, 1880-1940
Notes: The dependent variable is single female headship. The dots represent the coefficients on sugar suitability obtained from OLS estimates with matching, for each census year. Official data for the 1890 census have been lost. The values of each coefficient are also reported. Panel A includes all household heads. Panel B only includes blacks. Panel C only includes whites. Controls include cotton, rice, and tobacco suitability, the slave share in 1860, soil nutrients, soil pH, malaria endemicity, temperature, precipitation, elevation, water basins, ruggedness, latitude, longitude, population density, matched group and state fixed effects. Robust standard errors are clustered at the county level. Dotted vertical lines represent 95 percent confidence intervals.
Figure 2 shows results for our preferred specification, based on OLS regressions with matching. The dependent variable is single female headship and the dots represent the coefficients on sugar suitability, for each census year. The figure shows that higher sugar suitability is linked to higher likelihood of single female headship and that the effect is driven by the sample of black households in Panel B, while no effect is detected for white households in Panel C. The magnitude of the effect is sizeable: for blacks in 1880 for example, a one standard deviation increase in sugar suitability raises the probability of single female headship by 16% relative to its mean. The figure also reveals that the effect of sugar persists until 1940 but starts fading in 1920, in connection with the first wave of the Great Migration of blacks from the South to the North. It is noteworthy that slavery, measured by the slave share in 1860, is always included as a control variable, and that running a horse race between sugar and slavery confirms that it is the former, rather than the latter, that predicts the prevalence of single female headship.
We complement the above empirical strategy with a variety of alternative ones, including a quasi-experimental approach based on a geographic regression discontinuity design that exploits the well-defined border of the sugar suitable area.5 In the same setting, we also perform a falsification test based on rice suitability.
To deepen our understanding of the persistent influence of sugar planting across generations and throughout the country, we also construct a linked dataset that matches household heads in 1880 and 1930 on the basis of surnames, race, and state of birth. We are hence able to identify a sample of migrants who have left their ancestors' state by 1930. As these migrants are not exposed to the influence of sugar planting through their current environment anymore, we can identify its portable legacy as embodied in cultural beliefs and norms. Our results show that the portable legacy of sugar slavery is even stronger than what we detected in the cross section. Thus, we identify a non-dissipating influence of sugar through the intergenerational transmission, even after accounting for geographic mobility and the consequent spacial spread of the black family model.
When we move on to contemporary (county-level) Census data,6 we find that the legacy of sugar planting has faded by 1990. It is instead replaced by the contemporary black share. This is consistent with the patterns of migration and intermarriage experienced by former slaves after the second, more intense wave of the Great Migration. Indeed, the massive relocation of the descendants of black slaves to northern cities, combined with intermarriage among blacks from source counties with different exposure to sugar planting, ultimately weakened the direct relationship between sugar and the black family structure, but created a novel one which is best captured today by the share of blacks in the population.
According to Wilson (1987), the growing diffusion of factors such as poverty, unemployment, and incarceration have disproportionately affected African American urban males, which has consequently made them less likely to form stable unions. We also test the Wilson hypothesis using contemporary data, and we find that the share of blacks in the jailed population – a proxy for the withdrawal of black males from the marriage market – emerges as a powerful mediator of the black share. This suggests that the demographic and social dysfunctions inherited from sugar plantations, which spread all over the country with the Great Migration, are channelled through contemporary black incarceration.7 We can therefore establish a link running from the patterns of family relations that were especially common among slaves in sugar plantations, including the diffusion of male-absent families and the condition of forced celibacy, and the “tangle of pathology” affecting present-day US society.
Summing up, the evidence we produce provides strong support for Moynihan's belief that the African American family owes its origins to the history of slavery, and more specifically slave life on sugar plantations in the US south. While our focus is on family structure, our results carry deep ramifications for other related facets of the “tangle of pathology”, and in particular for the workings of the US welfare, education, and health care systems.
Baldwin, R (2020), “The COVID-19 upheaval scenario: Inequality and pandemic make an explosive mix”, Voxeu.org, 15 March.
Bertocchi, G, A Dimico (2020), “Bitter Sugar: Slavery and the Black Family”, CEPR Discussion Paper No. 14837.
Coclanis, P A (2010), “The economics of slavery”, In Paquette, R L and M M Smith (eds.), The Oxford Handbook of Slavery in the Americas, New York: Oxford University Press, 489-512.
Du Bois, W, E Burghardt, I Eaton (1899), The Philadelphia Negro: A Social Study, Cambridge: Harvard University Press.
Fogel, RW, S L Engerman (1979), “Recent findings in the study of slave demography and family structure”, Sociology and Social Research, 63: 566-589.
Follett, R J (1997), “The Sugar Masters: Slavery, Economic Development, and Modernization on Louisiana Sugar Plantations, 1820-1860”, LSU Historical Dissertations and Theses 6540.
Frazier, E F (1939), The Negro Family in the United States, New York: Citadel Press.
Gutman, H G (1976), The Black Family in Slavery and Freedom, 1750-1925, New York: Pantheon Books.
Kaye, A E (2007), Joining Places: Slave Neighborhoods in the Old South, Chapel Hill: University of North Carolina Press.
Manson, S, J Schroeder, D Van Riper, S Ruggles (2019), IPUMS National Historical Geographic Information System: Version 14.0 [Database]. Minneapolis: IPUMS.
Moynihan, D P (1965), The Negro Family: The Case for National Action, Washington: Office of Policy Planning and Research, United States Department of Labor.
Rodrigue, J C (2001), Reconstruction in the Cane Field: From Slavery to Free Labor in Louisiana’s Sugar Parishes 1862-1880, Baton Rouge: Louisiana State University Press.
Ruggles, S, S Flood, R Goeken, J Grover, E Meyer, J Pacas, M Sobek (2019), IPUMS USA: Version 9.0 [dataset]. Minneapolis: IPUMS, 2019.
Smith, R T (1982), Family, social change and social policy in the West Indies, New West Indian Guide/Nieuwe West-Indische Gids, 56: 111-142.
Tadman, M (2000), “The demographic cost of sugar: Debates on slave societies and natural increase in the Americas”, American Historical Review, 105: 1534-1575.
Vespa, J, J M Lewis, R M Kreider (2013), “America’s families and living arrangements: 2012”, U.S. Census Bureau Current Population Report No. P20-570.
Wilson, W J (1987), The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy, Chicago: University of Chicago Press.
1 Before Moynihan, the connection between slavery and family structure had been proposed by DuBois et al. (1899) and Frazier (1939). After Moynihan, it had been rejected by Fogel and Engerman (1974) and Gutman (1975).
2 Using 2011 American Community Survey data, Vespa et al. (2013) document that female-headed family households are 29.4 percent among blacks and 20.1 percent among whites.
3 Crop suitability data are from FAO GAEZ. See here.
4 Data are from the 1 percent sample. The source is IPUMS USA. See Ruggles et al. (2019). Official data for the 1890 Census have been lost.
5 Because we face partial compliance, we rely on a fuzzy design that combines county-level data on distance from the border with actual average sugar production in 1850-1860.
6 The source is IPUMS NHGIS. See and Manson et al. (2019).
7 Incarceration data are from the Vera Institute of Justice.