In the framework of the standard model, the estimated response of labor supply to the tax rate, holding income constant, reveals information about the utility function (in particular S, the rate of change of the slope of the indifference curve). The general model of behavioral response makes clear that in fact this response reveals a mixture of information about individual preferences and the avoidance technology.
There are several reasons why it is important to separately identify the structure of preferences and the structure of the avoidance technology. First of all, it is likely that the avoidance technology varies across individuals in systematic and observable ways. This implies that the behavioral response of labor supply to taxation will also vary, and the differential response can potentially be related to observable characteristics. Second, while in the standard model the labor supply response to a change in the pre-tax wage rate is identical (except for a multiplicative constant) to the response to a change in a proportional tax rate, that connection breaks down in the general model. Thus, estimates of the wage elasticity of labor supply do not necessarily carry over to estimates of the tax elasticity of labor supply, and vice versa. Separate identification of the structure of preferences and of the avoidance technology is required to nail down the relationship between these two elasticities. Thus, this model forces a reinterpretation of the results of, for example, Rosen (1976), who stimates a labor supply function of the formw6582-9
and interprets the estimated value of (qi-q2) as a measure of “tax illusion,” or the tendency of taxpayers to be more aware of the pre-tax wage rate than the appropriate marginal tax rate. In the general theory of behavioral response presented here, the value of qf- q2 has a structural
interpretation that is unrelated to tax illusion.

Note that the interpretation of an estimated wage or tax elasticity depends on the data one has available. Some data sources have evidence directly about L. Others have only information on (wL). Data from tax returns contain no information about L, nor about (wL), but at best have some information about (wL – A).
The key to identifying separately the structure of preferences and the avoidance technology is to specify observable influences on the latter. What are appropriate variables depends on the nature of the avoidance behavior being studied. One response to the taxation of labor income is to report less, or no, income to the tax authorities; thus, evasion is, in the parlance of this paper, a form of avoidance. The cost of this avoidance strategy includes the expected penalty for such behavior, which in turns depends on the probability of the evasion being detected and the expected penalty from detected evasion, including both financial penalties and any stigma (positive or negative) associated with detected evasion. add comment