Improved rural connectivity can lower trade costs while reshaping rural food environments. This paper combines Nepal survey rounds from 2013, 2014, and 2016 with a harmonized measure of external market connectivity that holds road geometry, each ward’s baseline gateway, and baseline food-market destination fixed, using official road records for time variation. A 10 percent improvement in local-to-regional food-market connectivity is associated with about a 0.6 percentage-point increase in the stable processed/convenience-food budget share and a 0.07-point decline in child dietary diversity. Estimates are similar in ward and household fixed-effects models for food purchasing and in repeated cross-sections and within-child models for dietary diversity. Total food expenditure changes little, and agricultural production and commercialization show little short-run response. The findings suggest that market integration can reshape rural food demand before local production adjusts and need not improve diet quality, although finite-cluster inference is less precise for purchasing outcomes.

Soil conditions can shape human nutrition through agricultural productivity and the nutrient composition of food, but large-scale evidence is difficult to interpret because soils are multidimensional and strongly correlated with broader geography. We link georeferenced Demographic and Health Survey data for approximately 180,000 children in 31 Sub-Saharan African countries to high-resolution soil nutrient maps. A principal component summarizing the local multivariate soil nutrient environment is positively associated with height-for-age and negatively associated with stunting. In the preferred specification, which compares children within country-specific 1 geographic cells, a one within-country standard-deviation increase in the soil nutrient component is associated with 0.066 standard deviations higher height-for-age and a 1.45-percentage-point reduction in stunting. The association remains similar under finer geographic fixed effects, alternative grid origins, spatially correlated inference, country exclusions, rainfall controls, and alternative constructions of the nutrient index. The evidence documents a robust local association between multivariate soil nutrient conditions and child linear growth, but does not identify the causal effect of any individual mineral.

How much households spend after an adverse shock may conceal substantial adjustment in what they buy. We study food-expenditure responses to monsoon rainfall shocks in rural Nepal using three rounds of household panel data and locally standardized CHIRPS rainfall anomalies. Same-household first differences across 2013–2014 and 2014–2016 show that drier prior monsoons produce little systematic change in reported total common-food expenditure but shift expenditure away from processed and convenience foods and toward staples. The processed-food response replicates across two rainfall measures and both panel transitions and remains robust to balanced-panel restrictions, leave-one-ward-out exercises, Conley spatial covariance estimates, and inference clustered by identical rainfall histories. Annual purchasing data show higher expenditure on purchased staples and lower expenditure on processed foods across all four design cells, while production responses are more heterogeneous. Together, these patterns point to a broader adjustment process rather than a uniform own-production replacement channel. Overall, aggregate food expenditure can appear relatively smooth even when households materially reallocate the food budget.