Certain types of expenditure--e.g. mortgage interest and medical insurance- receive favorable tax treatment and are effectively subsidized relative to other (non-tax-favored) expenditures. Labor taxes (e.g. income taxes) can therefore produce efficiency losses by distorting the allocation of consumption, in addition to distorting the labor market. Using evidence on the responsiveness of taxable income to changes in tax rates, a seminal study by Feldstein (1999) estimates that the marginal excess burden of taxation (MEB) could exceed unity, when the effects of tax deductions are taken into account. This is several times larger than in previous studies of the MEB that focus exclusively on labor market effects. This paper develops a "disaggregated" approach to estimating the MEB that decomposes welfare impacts in the market for labor and tax-favored consumption goods, and uses micro evidence on labor supply elasticities, the demand elasticity for mortgage interest, medical insurance, and so on. Based on Monte Carlo simulations, we find a 68 percent probability that the MEB lies between .31 and .48 for government transfer spending and between .21 and .35 for public goods. These estimates are below Feldstein's, but are still considerably higher (70 percent or more) than when we ignore tax deductions.