In short no, you can't really calculate the mean of a categorical variable. Sure you could assign a single value for each of the  intervals but how to do that well especially for the first and last interval is not really trivial and it will require knowledge about the actual research question. However  if you are comparing groups there are methods to use for comparing a categorical variable like this (non-parametric tests) 

For the variable like it is now you can say what the median answer is like "the median willingness to pay was $1-$3 dollar" I think the question you should ask yourself is "do I need the average, and why do I need it" most likely you do not actually need it.

Thank you so much Kristain.

Kristian, do you know any paper that reported similarly to what you recommended me about the median willingness to pay? If yes, could you please give me the name of the papers? Thank you so much in advance.

You can calculate the mean of grouped data. If you Google mean of grouped data I am sure you will be guided.

Thank you Job.

You'll see the difficulty of the problem with a hypothetical example.  Imagine in your survey there's one person that would pay $2 million for the information.  They'd be in the > $5 category but would distort the arithmetic mean value enormously.  The median may well be the better way to look at an 'average' in these circumstances.  You could assume mid-points for your other groups ($1 -$3 = $2; for example), but this could still be grossly distorted by anything 'atypical' or an outlier in the > $5 category. In simple terms what would be the 'midpoint' for "> $5"?!

Is it logical to ask open-ended willingness to pay question if we would like to calculate mean WTP?

Let me be more direct. I suggest you prepare a tabulation of responses, and then graph that. Use a range for the values at the extremes, and include them in the graph as well.. Use some mean for those two ranges.. Now, given your ranges for the 4 classes, and their corresponding class mean, compute the overall mean, given the number of responses for each..

if open ended questions asked... just take the average....if closed-end question DBDC. ... Mean WTP=-alpha/ Beta

For the double bounded CVM, if you used the interval regression, use the variable coefficients for your final model and multiply them with the weighted means of the corresponding variables. This will give you the Mean Willingness to Pay.

Lemma Zemedu Seifu

There are many ‘averages’. Which one do you mean? (Bad pun...)

I am working in R with the package DCchoice, I am using the double bounded dichotomous choice method. However, I don't know how to analyse the result from the logit model.

For example, the coefficient of income is 0.002, then if income increase 1, the log(odd) will increase 0.002? or the probability of saying yes is increasing 0.002?

Also in double bounded , it 's more difficult because it's YES YES, NO No, No Yes, Yes No...

ANyone help me?