CONTEXTThe challenge of food security has become more relevant in global agendas given growing food demands and the persistence of hunger and undernutrition. Due to the needs for addressing food security, there is considerable interest in identifying major drivers, especially in the case of emerging economies with asymmetric agricultural systems involving smallholder and industrial agriculture. OBJECTIVEThe present study proposes a conceptual model to identify how the structure of the agricultural systems in Mexico relates to food availability, accessibility, and utilization. METHODSOur rural food security conceptual model was tested using structural equation modeling at the national and ecoregion scales. RESULTS AND CONCLUSIONSFood availability and food accessibility showed a stronger influence than food utilization on the modeled food security concept at both scales. Among a total of 81 statistically significant path relationships between agricultural attributes and food security, the presence of smallholder agriculture accounted for 52 (primarily through small farm size, high crop diversity, low agricultural expansion and economic diversification) while commercial agriculture accounted for 29, primarily through market commercialization of the produces and irrigation practice. Our findings suggest that agroecologically efficient cultivation systems, including the milpa system, are widespread, highly productive, and substantially contribute to food security in the Temperate Sierras. Small-scale irrigation support programs for smallholder multi-cropping farms have great potential in drier ecoregions to solve for the current food security challenges in Mexico. To fully value food security assets of the milpa cultivation system, the authors urge the inclusion of more agroecological parameters in the agricultural census. SIGNIFICANCEThe modeling and analytical approaches employed in this study are proposed to identify and geographically differentiate influences of the agricultural systems on food security in the Global South and can adapt any spatial unit, geographic disaggregation, and data type.