The question behind the question
Ask someone how much they would pay for a shorter hospital wait and they will give you a number. Ask the same person the same thing a week later, worded slightly differently, and you will often get a different number.
The instinct is to call this unreliability. Somebody is lying, or careless, or does not know their own mind.
I think that reading is wrong, and it took me a long time to let go of it.
Most of us have never priced a hospital wait. There is no figure filed away waiting to be reported. So when you ask, the person builds an answer, right there, out of whatever is nearby: the last bill they paid, the number you mentioned in the previous question, whether the option was on the left or the right. Change what is nearby and you change the construction.
This is uncomfortable if you want survey answers to be measurements, like temperature. It is much less uncomfortable if you treat them as what they are: a decision made under specific conditions, which you designed.
Which puts the responsibility on the person asking. If the ordering of your attributes moves the answer, then the ordering is part of your result whether you report it or not. If half your respondents never looked at the cost row, your willingness-to-pay number is partly an artefact of your layout.
None of that makes the exercise pointless. Real decisions get constructed on the spot too, in a consulting room, under fluorescent light, with someone waiting. The survey is not failing to capture a stable preference. It is capturing a constructed one, which is the only kind there is.
The useful move is to stop asking whether the number is true and start asking what it was built from.