The basic assumptions of the theory of economic voting are very simple:
- voters care about unemployment, inflation, and growth
- voters blame the government for adverse economic conditions
- voters use the ballot to punish the government.
Unfortunately, the impact of this effect is not constant over time and across countries, which is slightly embarrassing. In their recent book, van der Brug et al. do not claim that they have solved this puzzle, but they maintain that they have taken the discussion one step further. According to them, previous research has looked at the wrong variable, i.e. (dichotomous or multinomial) vote intentions. This is hardly surprising. For the last decade or so, these authors and their associates have campaigned for an alternative measure, namely the subjective probability to vote for each single party. However, their measure (which has been implemented in the European Election Studies) is not uncontroversial. First, analysts must account for the clustering of these ratings (while we might look at 4,000 or 6,000 ratings, we still have only 1,000 truly independent cases, i.e. persons). Second, if a respondent does not rate a party, is that a missing value or a zero probability? Third, comparisons across political systems (especially comparisons of two-/multiparty systems) are at least as dodgy as comparisons of the traditional variable. And finally, while counting votes/vote intentions obviously discards valuable information about the individual calculus that leads to this decision, subjective probabilities are closer to party sympathies than to real thing. Nonetheless, an interesting read.