Based on a nationwide survey administered before the House of Representatives election in Japan, which will be held on October 31, 2021, we attempt to understand voters’ preferences for policies and parties. The broad objective is to measure voters’ multidimensional preferences using conjoint analysis. A specific aim is to understand how or why the Liberal Democratic Party (LDP) dominates elections in Japan, despite proposing policies that seem to be unpopular with voters.
Each respondent will be asked to complete three sets of seven exercises. Each exercise contains two profiles (policy bundles) and five attributes (policy issues). The seventh exercise is the same as the first exercise (meaning that there are six unique comparisons). For our confirmatory analysis, we drop the seventh exercise in each set. We use the repeated exercise to measure the average accuracy of survey responses, which we use in exploratory analysis.
The levels of each attribute are based on summaries of the main six political parties’ actual policy proposals put forward in the campaign. The order of attributes is randomized (across respondents) and the levels for the five attributes are randomly assigned from the predetermined sets of levels.
In the first set of exercises (named the “Without Label” exercises), a respondent compares two hypothetical political parties’ policy proposals (or “manifestos”) and answers the following question: “Suppose that the following two parties were nominating candidates in this general election. Which party would you support? Even if you are not entirely sure, please indicate which of the two you would consider supporting.” Respondents choose between “Party A” and “Party B” without any specific labels.
The order of the next two sets of exercises is randomized for each respondent.
In the second (third) set (named the “With Label” exercises), the same pairs of manifestos are presented to each respondent. This time, we also randomly assign party names (“labels”) instead of using “Party A” and “Party B.” In each pair of profiles, one is always LDP and another is always one of the other five parties. We then ask the same question.
In the third (second) set (named the “Ownership” exercises), the same pairs of manifestos (without the party label) are presented to each respondent. We then ask: “Assuming that the parties putting forward the following proposals were fielding candidates in this general election, please select the party whose policies are more similar to those of the Liberal Democratic Party.” Respondents choose between “Party A” and “Party B” without any specific labels.
These experimental questions are preceded by basic demographic questions (pre-treatment) and followed by questions about political attitudes and vote intention in the actual election (post-treatment).