This study investigates the influence of social preferences on voting decisions using a new Random Price Voting Mechanism (RPVM), which is best thought of as a public goods voting extension of the Becker-DeGroot-Marshack mechanism for private goods. In particular, this mechanism is used to investigate experimentally whether voting decisions are affected by the distribution of net benefits associated with a proposed public program. Recent papers have shown that, in additional to selfishness, factors such as inequality aversion, maximin preferences, and efficiency may influence individual decisions. However, the effect of social preferences on voting, the predominant funding mechanism for public goods by legislatures and public referenda, has not been thoroughly examined. We first establish the presence of anomalous behavior in dichotomous voting, and introduce the RPVM as a more efficient mechanism to examine such anomalies. We show that it is demand revealing in the presence of social preferences and empirically consistent with dichotomous choice voting. Laboratory experiments involving 440 subjects show that when net benefits are homogeneously distributed, the new RPVM is demand-revealing in both willingness-to-pay (WTP) and willingness-to-accept (WTA) settings, for both gains and losses. When the voting outcome potentially results in a heterogeneous distribution of (net) benefits, a systematic wedge appears between individuals' controlled induced values and their revealed WTP or WTA. With induced gains, the best-off subjects under-report their WTP and WTA in comparison to their induced value. Worst-off subjects express WTP and WTA that exceed their induced value. With induced losses a mirror image is evident. Best-off subjects over-report their induced value while the worst-off subjects under-report. Theoretical and econometric results presented in the paper suggest that these differences are caused by a concern for social efficiency.