The phenomenon that prior gains may increase people’s willingness toaccept risky gambles is named as the house money effect (Thalerand Johnson, 1990). Many studies have shown that the “house moneyeffect” is a robust phenomenon but few scholars explain the mechanism ofit well. We suppose the reason for the house money effect is that the ante(starting amount) is from the prior gambling profits, and its potential loss hasrelatively low psychological value. To test this hypothesis, we designed aseries of studies using two-stage gambles. A total of 915 university studentsparticipated. In Study 1, in addition to a standard condition (which replicatedthe basic effect), we test how people respond to “prospect theory, withmemory” frame, a “concreteness” frame and“quasi-hedonic” editing. None of these types of frames result in asignificant house money effect. In Study 2, we certify the reference point shiftto 100 Yuan in the second-stage gamble, thus the house money effect can beregarded as the absence of loss aversion; Study 3, consisting of 3sub-experiments, indicated that gambling profits and normal income will opendifferent mental accounts which are spent quite differently. The pain of losing100 Yuan allowance is more serious than that of losing 100 Yuan gambling wins.People will typically reject the gamble of 50/50 chance to gain or lose 100 Yuanif the ante is from the “normal income account”, but accept if theante is from the “windfall account”. The results of the series ofexperiments prove the accuracy of our hypothesis mostly.