Beyond subsistence: the aggregate contribution of campesinos to the supply and conservation of native maize across Mexico

Abstract

Mexico is the center of domestication and a center for diversity of maize. Area planted with maize is the country’s largest agricultural land use, mostly planted by smallholder family farmers known as campesinos. They generally plant native varieties, saving and sharing seed by and among themselves, enabling the evolutionary processes that sustain and generate crop genetic diversity to continue today. Campesinos have been viewed as largely subsistence farmers generating limited maize surpluses. Here, we show that subsistence production is insufficient for explaining the quantity of maize they produce and the extent of the area they plant across Mexico. Our hypothesis is that beyond supplying their own consumption needs, campesinos collectively produce maize to respond to the demand of non-maize producing local consumers. We quantify the extent of subsistence versus surplus production among campesinos, showing that they produce more maize than would be needed to feed themselves and generate substantial surpluses. We test statistically the association between the area campesinos plant with maize across the country with socioeconomic variables that link their production to the demand by other consumers, and examine the implications of the results for the supply and conservation of native maize in the country. Our results suggest that maize trading linking campesinos to other consumers may be important and widespread, contributing to create additional incentives beyond self-consumption to plant native varieties from saved seed. We conclude that there are important opportunities for maintaining maize evolution under domestication at large scale by strengthening local maize markets.

B.Supplementary Information

ESM 1

(PDF 1224 kb)

A.Appendix 1. Diagnostic tests

Table 4 presents the value of the Moran I statistic two-sided on the residuals of an OLS regression of the model in Eq. 1. It is highly statistically significant, indicating the presence of spatial autocorrelation and thus the need to address it. Table 4b-c shows the values of different Lagrange Multiplier tests statistics for different specifications of the spatial autocorrelation: error dependence and lagged dependent variable. Since both the LMerr and LMlag are statistically significant, we compared their robust forms RLMerr and RLMlag. Results suggest that the lag model as the likely alternative (RMLag = 22.273, p value = 0.000 versus RLMerr = 0.63505, p value = 0.426).

Table 4 Diagnostic tests for spatial autocorrelation (new)