The Quantitative Analysis of Electorates
This appendix discusses the sources and methodology for the analyses in Chapters 3 and 4: the data, multinomial logit techniques, and the LERS machine learning algorithm. The data on public demands and priorities comes from the “Party Systems and Electoral Alignments in East Central Europe” database of public opinion surveys. Available in machinereadable data form, the database comprises public opinion polls in Poland, the Czech Republic, Slovakia, and Hungary regarding sociodemographics, evaluations of parties and the market transformation, and political identities. The project is supervised by G bor T ka of the Central European University and employed several polling agencies: CBOS in Poland, STEM in the Czech Republic and Slovakia, and Median in Hungary. The samples were representative of the adult population in each country. In Poland, clustered random sampling was used, with sample sizes of 149, 1,188, 1,468, 1,209, 1,162, 1,173, and 1,173. In Czech Republic and Slovakia, random clustered sampling was used in 1992, and quota sampling thereafter. The Czech sampling sizes were 815, 939, 1,117, 1,562, 1,515, 1,291, 1,569, 1,443, and 1,595. The Slovak samples were 712, 920, 871, 845, 757, 1,213, and 1,027. The Hungarian samples used clustered random sampling, and the sample size was 1,200 across the years, except for 1,196 in June 1995.
These population surveys were conducted annually in 1992–6 in the region. Since they started in 1992, we cannot trace the crucial transformations that took place in the party images in 1989–91, but can judge the evolution of their outcomes. Another limitation is that that respondents had to choose from among several categories rather than providing their own answers and priorities. Therefore, to supplement these findings, I used public opinion data gathered by several polling agencies in each