Willingness To Pay is out of ranges of bids asked in survey

I'm trying to estimate the willingness to pay (WTP) for whale watching by using a probit model (the results are similiar with a logit). But the WTP estimated is out of the bid levels asked in the survey, which were 10, 20, 30, 40, 50 and 60 USD, but my estimate WTP results in 100 USD. So I want to know why this happen and how to correct this?

By the way, I have included and dismissed different variables but the problem persists. I'm using Stata for my regressions.

Best regards,


No cake for spunky
I am not sure why the model would generate predicted results that matched your survey. Commonly predicted and observed data is not the same, that suggests there are problems with your predictors.
Surveys usually "cluster" corner solutions.

Example Question:

Write the donation amount given to charity last year ( from $0-infinity)

In this kinds of questions you will typically have a normal distribution over the positive values, but a big cluster of observations in the $0 donations, because it is completely normal than most people don't decide to donate.

Since you have clustered obervations around and specific value your OLS estimates will be inconsistent and biased.

In this cases a Tobit model is appropiate.

I always thinked that willingness-to-pay is an amount rather than a probability (as you were using probit).

I'd use something like this:

Y*=B0+BX +u

Where Y=max(0,Y*)