In this thesis we develop a method for selecting households to monitor for efficiently measuring shocks to food security at the village-level. We find that the optimal subset of households to monitor is identified by a variance criterion which intuitively selects households with the lowest variance of their idiosyncratic shocks. In other words, those chosen for monitoring should be the households for whom shocks to their food security are driven by variation in village level shocks, with the variance of household-level shocks being minimized. We compare this method of selection to selecting households at random and selecting the poorest households to monitor and find that the method developed in this thesis greatly outperforms both in terms of root mean squared error and total absolute error. However, we are unable to identify observable characteristics to distinguish the monitored from the non-monitored households, suggesting that the variance criterion for selection is capturing certain unobservable variables and thus we are not able to detect monitored households based on observables in our data.