Women's subsistence strategies predict fertility across cultures, but context matters

Abstract

While it is commonly assumed that farmers have higher, and foragers lower, fertility compared to populations practicing other forms of subsistence, robust supportive evidence is lacking. We tested whether subsistence activities-incorporating market integration-are associated with fertility in 10,250 women from 27 small-scale societies and found considerable variation in fertility. This variation did not align with group-level subsistence typologies. Societies labeled as "farmers" did not have higher fertility than others, while "foragers" did not have lower fertility. However, at the individual level, we found strong evidence that fertility was positively associated with farming and moderate evidence of a negative relationship between foraging and fertility. Markers of market integration were strongly negatively correlated with fertility. Despite strong cross-cultural evidence, these relationships were not consistent in all populations, highlighting the importance of the socioecological context, which likely influences the diverse mechanisms driving the relationship between fertility and subsistence.

Keywords: anthropological demography; cross-cultural analysis; demographic transition; fertility; subsistence-based populations.

Fig. 1.

Descriptive statistics for four key subsistence metrics—(A) proportion of activities in different modes of subsistence, (B) proportion of diet from different sources, (C) self-reported occupation, and (D) type of residence. Plots are at the population level (Y axis), but not all subsistence measures are available for all populations, so those listed vary. Plots are structured by degree of market integration, running from Top to Bottom: “low,” “medium,” to “high.” The colors of the curves/bars reflect the investment in the subsistence strategy. Note that in plots (A) and (B), “farm” represents the combined influence of agricultural and horticultural production. The sample size varies by measure; please see SI Appendix, Table S1 for further descriptives.

Fig. 2.

(A) Live births as a function of age in each study population. Within populations, each line is the posterior median age function for an individual woman. The dotted intervals denote 90% credible intervals of the posterior predictive distribution. (B) Points represent posterior median predicted cumulative fertility ($\stackrel{\sim }{\mathrm{CF}}$). Bars represent 90% CI. Vertical facets on the right denote level of market integration (low/medium/high). In both plots, the main subsistence type is color coded.

Fig. 3.

Relationship between five measures of individual subsistence strategy and mean cumulative fertility. (A) Foraging activities as a proportion of all economic activities (n = 8), (B) cultivation activities as a proportion of all economic activities (n = 13), (C) proportion of diet from farmed sources (n = 12), (D) log count of livestock (n = 18), and (E) log (acres) of land owned (n = 13). Points represent posterior median expected cumulative family size ($\stackrel{\sim }{\mathrm{CFS}}$). Bars represent 90% CI. Vertical facets on the right denote level of market integration (low/medium/high) in all plots except A, where all populations are categorized as “low” market integration. The main subsistence type is color coded. Effect sizes in plots (1+SD) are relative to each individual population.

Fig. 4.

Relationship between six measures of market integration and mean cumulative fertility. (A) Wage labor activities as a proportion of all economic activities (n = 15), (B) waged versus subsistence occupation (n = 13), (C) proportion of diet from market sources (n = 13), (D) log years in education (n = 24), (E) urban versus rural residence (n = 13), and (F) log income (n = 15). Points represent posterior median expected cumulative family size. Bars represent 90% CI. Vertical facets on the right denote level of market integration (low/medium/high) in all plots. The main subsistence type is color coded. Effect sizes in plots (1+SD) are relative to each individual population.

Fig. 5.

DAG representing our assumptions about the data-generating process. The dashed box contains our main variables, with threats to identification [birth year, selection (S) due to missing records of deceased individuals] shown outside the box alongside of unobserved confounding (U).