From Nathaniel Hendren:
This paper applies basic price theory to study the marginal welfare impact of government policy changes. In contrast to the canonical marginal excess burden framework, the framework does not require a decomposition of behavioral responses to the policy into income and substitution effects. The causal effects of the policy are sufficient. Moreover, in the broad class of models where the government is the only distortion, the causal impact of the behavioral response to the policy on the government budget is sufficient for all behavioral responses. Because these behavioral responses vary with the policy in question and are, in general, neither pure Hicksian nor Marshallian elasticities, I term them policy elasticities. The model provides formal justification for a simple benefit/cost ratio measure for non-budget neutral policies: the welfare impact on beneficiaries per dollar of government expenditure. I calculate this ratio using existing causal effects from five policy changes: the top marginal income tax rate, EITC generosity, food stamps, job training, and housing vouchers. Comparisons across beneficiaries of such policies is accomplished using social marginal utilities of income. For example, the mid-range of existing causal estimates suggest increasing EITC generosity financed by an increase in the top marginal income tax rate is desirable if and only if one prefers giving an additional $0.44-0.66 to an EITC-eligible single mother (earning less than $40,000) relative to an additional $1 to a person subject to the top marginal tax rate (earning more than $400,000).