The data for the study was collected using both primary and secondary data collection techniques.
Primary data related to gambling ballot outcomes (yes/no votes on gambling proposition) were obtained from all 50 state-election offices.
Secondary data was compiled from a variety of sources :
county-level religious data containing information about the number of churches and church members;
other country-level data, such as population estimates, age, personal income, ethnicity, gender, poverty level, and education, were extracted from the U.S. Census Bureau and state data centers.
The dataset included 1,287 records, each representing a county’s voting outcome from a past ballot. The following lists variables to be used in their ANN models :
Ballot Type I (Gambling versus Wagering, a binary variable)