Cross-national Public Opinion Surveys

The Economic Vote:
How Political and Economic Institutions Condition Election Results (Cambridge University Press 2008). In this book Duch and Stevenson propose a selection model for explaining cross-national variation in economic voting: Rational voters condition the economic vote on whether incumbents are responsible for economic outcomes, because this is the optimal way to identify and elect competent economic managers under conditions of uncertainty. This model explores how political and economic institutions alter the quality of the signal that the previous economy provides about the competence of candidates. The rational economic voter is also attentive to strategic cues regarding the responsibility of parties for economic outcomes and their electoral competitiveness. Theoretical propositions are derived, linking variation in economic and political institutions to variability in economic voting. The authors demonstrate that there is economic voting, and that it varies significantly across political contexts, and then test explanations for this variation derived from their theory. The data consist of 165 election studies conducted in 19 different countries over a 20-year time period.
Data collection for the project was unprecedented: it involved identifying, translating and analysing over 300 public opinions surveys from over 20 countries. It developed a novel statistical estimation technique for the thousands of complex regression models estimated. The project also developed a formal model to explain how political and economic institutions condition the economic vote. These data collection, statistical estimation and formal model efforts took eight years.

The information used to produce measures of economic voting (the change in vote probabilities due to changing perceptions) comes from the surveys of 165 different national populations in seventeen western democracies from 1980-2001. Each survey randomly samples the national population of adults at the time of the survey. In some cases the sample had to be weighted to account for non-random over-sampling of some groups in the population. The principal concern was to build measures of general economic voting in each population that are comparable across the different populations.

The measure of the economic vote is a casual effect – it is the impact of one variable on another. The effect must be estimated from the data on the two variables that make up the relationship (vote choice and economic perceptions). Since each of the 165 surveys is based on a large probability sample, this effect can be measured along with measures of uncertainty associated with it. Considerable efforts were made to ensure consistent estimates of the economic vote in each population, i.e., estimates of economic voting that reflect its true value in the population. In particular this meant according careful attention to model specification and correctly specified functional relationships in each of the 165 surveys. The specification of these models was guided by the vast literature on voting behavior in general, the country-specific literatures on voting, and a growing body of work concerned with the particular statistical problems associated with estimating vote choice models. Logical constraints were employed to rule out many possible specifications (e.g., a variable that can only have positive values can not be generated by a random process that sometimes results in negative values). Knowledge of the substantive problem was used (e.g., which explanatory variables are likely to be important?); and efforts were employed to ensure that the model produced estimates that conform to what we is widely known about real political world. The 756 estimates obtained from these models (one for each party in the 165 populations) are the numbers used to characterize the strength of economic voting in each population.