this is a well-known issue in market research, and survey design in particular. stated preferences are nearly useless in understanding customer segments and behavior. respondents have hundreds of different reasons for not answering with their true preferences (in this case, to not to pay more/separately).
instead, you want to reveal preferences using techniques like conjoint analysis (not always feasible), which better simulates the actual choices respondents might make in real situations.
It’s pretty obvious when you look at things from an awkward level of first principles.
“Based on this forced social interaction of questioning strangers as they left the park, we have forecasted how different people entering the park will react to a commercial interaction.”
instead, you want to reveal preferences using techniques like conjoint analysis (not always feasible), which better simulates the actual choices respondents might make in real situations.