On the Efficiency of Income - Testing in Tax - Transfer Programs

The consensus of economic experts is that income-tested programs are more efficient than non-income-tested tax-transfer programs. Most widely favored as a policy proposal is the negative income tax, (NIT) an income-tested program. This view apparently stems from the widespt:ead use of the· target efficiency measure--a conceptually flawed measure of technical rather than economic efficiency. In this paper we examine the economic efficiency issue within a two-class model which includes taxpayers along with beneficiaries. Our theoretical analysis establishes the possibility that non-income-tested programs are more efficient than income-tested programs. However, no general qualitative conclusion on the efficiency of income-testing can be drawn. Rather, the theory indicates that the marginal efficiency of income-testing depends upon a number of program parameters and empirical magnitudes. For a simple two-class model, the marginal efficiency of income-testing improves as the substitution effect of poor workers declines relative to that of rich workers. The form of the NIT program and the relative sizes of. beneficiary and non-beneficiary groups also affects the marginal efficiency of income-testing. To illustrate some quantitative aspects of the efficiency of incometesting, we calculate several feasable overlapping NITs and their corresponding credit inco~e taxes (CIT)'s. These are based upon empirical estimates·of the labor-supply functions and substitution effects for a national cross-section of U.S. prime-age married males. Some observations· are also made for the fully integrated NIT. We present measures of the marginal efficiency of income-testing and of the relative efficiency in terms of welfare loss for comparable NIT and CIT programs. For relatively generous programs, CITs are superior at the margin and entail less welfare loss than .comparable NITs. For the less generous programs examined here, the CIT retains its superiority over certain NITs but some income-testing becomes desirable at the margin. More important, the difference in welfare loss between the two programs is invariably small-less than one-half of One percent of aggregate earnings. Perh~~~ the major implication of our findings for ~olicy +qtmulation is that any differential eco~omic efficiency costs between CIT an4 NIT schemes maY we+l .be domin~ted by other program differences.