This article uses the design science approach and a semi-field experiment to explore the differences between a laboratory and an on-line version of an electronic voting mechanism that allows citizens to express their preference intensities as well as to be compensated in case their preferred alternative is not chosen. The authors find that participants in the online version of the mechanism vote less frequently for their preferred alternative than participants in the laboratory version. Even though this difference negatively affects the participants' income distribution, it does not have a major effect on total social value. The article proposes changes that can be made to the online version of the mechanism in order to achieve results as good as the ones achieved in the laboratory.