WTP is just a marginal rate of substitution. In the typical case of models with linear in parameters utility functions the subjective value of time, for instance, is the ratio between the travel time coefficient and the cost coefficient. Remember that even for the MNL model estimation produces just an estimator of the true values of the parameters (with a given probability distribution). In other words the computed WTP is itself an estimator with another probability distribution. So you have to include in your analysis a method to incorporate this randomness of the estimated WTP measure. You have to calculate confidence intervals. The most popular method adopted (with heroic hypothesis made) is the so called Delta Method (approximation). Popular softwares such as , for instance, Nlogit, estimate both the delta method as well as the krinsky robb (a parametric simulation). A good reference for al this is Armstrong et al (2001) Transportation research part E "Confidence intervals to bound the value of time".
regards
Edoardo
I read about Krinsky Robb. Aren't WTP is dividing the coefficient of attribute and coefficient of price?
Sydney, a WTP (if the attribute you are considering is something you like , e.g. reduction of travel time) or a willingness to accept (WTA , increase in the level of an attribute you do not like), is always a ratio between an attribute's coefficient and a coefficient of a monetary attribute (not necessarily a price it could be a cost or a subsidy). The important thing to remember is that the ratio of two random variables as the estimated coefficients always are is itself a random variable with (most often) a difficult to treat distribution (unless you are willing to make strong assumptions a s the delta method does i.e. the distribution of the ratio is asymptotically normal). regards Edoardo
Sydney, you could also look into the book by Hensher, Rose and Green - Applied choice Analysis (2nd Edition) for further notes on the same. Best wishes, Hilary.
- Badges
- Science topic
- Similar topics
- Vectorization