Eight alternative methods of eliciting preferences between money and a
consumption good are identified: two of these are standard willingnes
s-to-accept and willingness-to-pay measures. These methods differ with
respect to the reference point used and the dimension in which respon
ses are expressed. The loss aversion hypothesis of Tversky and Kahnema
n's theory of reference-dependent preferences predicts systematic diff
erences between the preferences elicited by these methods. These predi
ctions are tested by eliciting individuals' preferences for two privat
e consumption goods; the experimental design is incentive-compatible a
nd controls for income and substitution effects. The theory's predicti
ons are broadly confirmed.