The survey sample we use was constructed by first drawing a target population sample of 1,800 individuals. This sample is based on the 2005–06 American Community Study, November 2008 Current Population Survey, and the 2007 Pew Religious Life Survey... Polimetrix invited a sample of their opt-in panel of 1.4 million survey respondents to participate in the study. Invitations were stratified based on age, race, gender, education and by simple random sampling within strata. Half of the 2,494 respondents who completed our module (1,247 individuals) were randomly assigned to complete the questions about ballot secrecy that we use here. A subset of the 2,494 respondents were matched to the target sample of 1,800 cases based on gender, age, race, region, metropolitan statistical area, education, news interest, marital status, party identification, ideology, religious affiliation, frequency of religious services attendance, income and voter registration status. Approximately 70 per cent (893) of the respondents who completed the secret ballot items were matched to cases in the target sample. Of these, 804 respondents provided responses to all of the items we use in our analysis and, thus, are included in our sample.