Abstract
This paper evaluates income variable specification and functional form issues in contingent-valuation studies using the utility-theoretic approach. Applying data from a study of willingness to pay for irradiated beef, results suggest that functional form does affect the estimates, but variable specification does not introduce statistically significant differences.
Acknowledgements
The authors thank Linda Crenwelge for editing the manuscript and the Institute of Food Science and Engineering, Texas A&M University, for project funding.
Notes
For example, if the family income is under 10 000 USD, income = 1; if income falls between 10 001–25 000 USD, coded income = 2, and so on.
This is different from the Cameron (Citation1988) approach, which has been shown to provide a statistically efficient use of referendum survey information.
If income categories are uniform, then using c or y is equivalent as y is simply a rescaling of c.
From , and , each E(WTP) value is a nonlinear function of estimated coefficients. If WTP is formulated as a function of estimated coefficients vector, b, and call it g(b) function, then the asymptotic variance of g(b) = Asy.Var[g(b)] =[∂g(β)/∂β′] Asy · Var(b)[∂g(β)/∂β′]′ which is referred to as the delta method, see Greene (Citation2000).