This paper provides an empirical and theoretical analysis of theft from opaque food subsidy schemes in developing economies. Some 48% of the subsidized rice distributed in the Philippines in 2006 does not show up in household surveys. Pilferage is suspected, and we introduce several new robustness tests that rule out alternative interpretations of this …nding. We argue that …finding pilferage in an opaque system does not simply imply that redistribution is costly, as the literature has recognized. It also implies that reducing the amount of product pushed into the distribution system should reduce pilferage one-for-one, and that such reductions and transparency improvements should unambiguously improve consumer and taxpayer welfare. In contrast, reductions in the price subsidy may not be welfare enhancing.

We argue that parastatal food-distribution organizations are less effective when their agents vary the quality of subsidized food and retail service (henceforth "quality") in response to local market conditions. These actions make assessing need from market signals difficult, hindering the effective allocation of subsidized food. In theory, policies and market conditions can affect outcomes (quality, demand, rationing and pilferage) differently, depending upon how markets for subsidized rice clear. We provide tests to reveal which market clearing mechanism is relevant. Using administrative and household survey data, we …nd evidence of quality variations, pilferage and market clearing through ad-hoc rationing of subsidized rice in the Philippines. Poorer markets and those facing higher rice prices do not receive more subsidized rice, but do appear to receive lower quality service and smaller rations. This may explain why the nationwide incidence of hunger rose with rice prices in poorer areas during 2003-2007 despite costly subsidies. We recommend explicit rationing rules, increased transparency regarding rice allotments and measured reductions in the amount of subsidized rice distributed.