As development practitioners, we face the challenge of ensuring food security in an increasingly shock-prone world. Poor and vulnerable households are unable to smooth their food consumption in times of drought, flood or when pests destroy their crops. This dissertation draws on the poverty literature and posits resilience as a latent capacity allowing households to recover from the effects of shocks. It presents several definitions and measurements across multiple contexts. In particular, this thesis explores the tension between focusing on a shock-specific response and emphasizing household well-being in a stochastic context. It analyzes both short and long term measurements of food security in Malawi and Ethiopia, investigating the effects of internal interventions as well as household-level characteristics that may allow households to better manage risk. The first chapter motivates the investigation in the context of eliminating hunger and expanding our understanding of socio-ecological systems. The second chapter investigates the causal impact of the Productive Safety Net Program (PSNP) in Ethiopia, finding that it mitigates the effect of drought on long-term food insecurity. The third chapter uses a novel 12 month high frequency dataset in Malawi to track the incidence and persistence of subjective shocks. It finds that households living in the flood plain and those with fields far from home are more resilient to the effects of drought, while female-headed households are less resilient to the effects of illness. It also illustrates the use of machine learning algorithms to identify predictors of short-term food insecurity. The fourth and final chapter picks up on the insight that households with spatially dispersed parcels may better manage risk. Using a natural experiment from Ethiopia, it shows that land fragmentation reduces both short and long-term food insecurity. Endowed with a diversified set of parcel characteristics, households grow a more varied set of crops, mitigating the effect of drought. Together, these chapters argue that reducing food insecurity and improving resilience is possible. In order to avoid doing more harm than good, external interventions must take into account households' and communities' existing ability to mitigate shocks.