IPOP: A Dataset of Individual Preferences over Parties throughout Democracies

• In Event: Lightning Round: Research Design

Abstract

Theories about comparative elections and party systems often hinge on one universal feature about which our data are exceptionally sparse: peoples' sincere preferences over their country's political parties. We propose to create the dataset of Individual Preferences Over Parties (IPOP) by gathering pre-existing public opinion data and conducting original surveys, beginning in Brazil, Canada, India, and Spain.

There are two components to this proposal. First, we will compile the data from other sources that have published individuals' rankings of the parties; for example, the Canadian Election Study has asked Canadians to rank their political parties for many years. The second challenge, however, is that many important democracies do not have an existing survey that includes rankings of political parties. In these cases, we will need to conduct our own original surveys, and we are seeking feedback on how these surveys should be designed to create the most useful dataset.

There are three reasons why survey data is needed on this point. The first major theoretical question that these data would help to address is on the types and prevalence of strategic voting around the world. Strategic voters may cast a vote for a candidate other than their most-preferred option, but this cannot be directly measured, and it cannot be measured at all without knowing the prevalence of different preferences in the population.

The second reason why information on individual preferences over parties are needed is because vote counts alone are usually not enough to recover voters' full rankings of the alternatives. In single-vote contexts, we can only observe how many people placed each candidate as their top choice candidate, after any strategic considerations they may have undergone.

Finally, there is also a distinction between the candidates in a particular election and the political parties that they belong to. A candidate may be more or less likeable than their political party for any number of reasons. To make more general statements that will have validity when applied to different elections, we need to know how voters think about enduring political parties, not just the specific candidates on offer.

The IPOP aspires to collect data on preferences over parties from every country where such data are feasible and useful to collect, but we propose to begin with Brazil, Canada, India, and Spain, for the sake of initial validity checks and institutional variety. By including cases which have extremely high-quality survey data on party preferences, such as in the Canadian Election Study, we can compare our survey results to established results. The countries also vary in the important ways: two are presidential while two have prime ministers, they vary between pure Westminster systems and closed list PR, and the list includes countries that are accustomed to majoritarian rule by one party as well as countries where coalitions are the rule.

We are looking for two major pieces of feedback on research design. First, we want to focus the dataset very narrowly on party preferences specifically, so we want to consult other researchers on how sparse a dataset is too sparse for it to be useful to other researchers. Second, we wish to discuss the survey design techniques that would elicit the most accurate responses about peoples' true preferences over political parties. We hope that this dataset will be useful for formal modelers and empirical scholars, in the widely varied applications that require knowing peoples' political preferences.

Authors

• Samuel Baltz, Massachusetts Institute of Technology

• Fabricio Vasselai, University of Michigan, Ann Arbor