APPENDIX A

Summary

Research Design

The data for this article is a result of a survey on public opinion and voting behaviour that was done on three cities (Medan, Surabaya, Samarinda) shortly after the 9 December 2015 elections. To determine the population samples in the location, a multi-stage random sampling strategy was used, with villages as the primary sampling unit (PSU). First, the population of each city based on total population at the subdistrict (kecamatan) level was stratified to obtain proportional samples in each of the sub-districts. Then the population was stratified further based on the area of domicile (rural vs. urban) to ensure a representative proportion between urban and rural residents in each sub-district. For this article, the respondents are completely urban residents as Surabaya is overwhelmingly urban. Finally, gender was stratified to have an equal number of female and male respondents.

For each municipality, samples were drawn in proportion to the population size of each sub-district, and the randomization process followed these main steps. First, kelurahan (the urban equivalent of villages) were selected in each sub-district according to the subdistrict's respective proportion of population. Second, all neighbourhood organizations (rukun tetangga or RT) in each PSU were listed and five of them were selected at random for each unit. Third, two households were selected at random in each RT. Finally, in each selected household, all household members aged 17 years or older were listed, and one person selected with the aid of the Kish Grid. If a female respondent was selected from one household, a male respondent would be selected from another household. In case the selected respondent could not be interviewed (not available after two visits during interview time in the village, refused to be interviewed, etc.), the respondent was substituted by repeating stages three and four above.