Peking University constitutes three precincts of Beijing, and voters are made of up faculty, student and staff of the university. The election results for one of the precincts are used in this experiment. A list of all registered voters in the precinct was collected and the names of non-students were removed, leaving a sample of 4,034 undergraduate and master’s students. Typically four to six students share a dorm room in Peking University. Treatment was clustered at the dorm room level since a canvassing message delivered to one student in a room would likely be passed on to the non-treated students as well. Thus all students in a single room are exposed to the same treatment. Rooms were randomly assigned to one of three groups: control or one of two treatments. The canvassing was done only two days before the election to minimize infiltration of the message to control voters.