This paper uses a Computable General Equilibrium (CGE) model to simulate the short-run effects of alternative food- subsidy scenarios. Savings from reduced subsidy spending are used to reduce direct taxes uniformly for all household types. The model uses a 1996/97 database with detailed household information. The simulated impact of targeting or eliminating oil and sugar subsidies is small: disaggregated real household consumption changes by ±0.3 percent. It is progressive if the subsidy is targeted to the needy (the bottom two quintiles in rural and urban areas) and regressive if it is eliminated. The targeting of all food subsidies is pro-needy, in part due to important indirect effects. It raises the consumption of the needy by 0.5 percent with, on average, little change for the nonneedy. The strongest gains are recorded for the rural needy (consumption increase by 1.0 percent). Food subsidy elimination is regressive: the needy suffer a consumption loss of 1.1 percent. If the government savings instead are transferred to the needy, the impact is reversed: consumption increases by 4.2 percent for needy households while the nonneedy register a small loss. The overall policy implication of the paper is that there is scope for reducing food subsidy spending without hurting the low-income groups.