What we are reading: Corruption performance voting
Do voters punish government parties for high levels of corruption?
Performance voting is a generalisation of economic voting: the idea that voters governments punish/reward for good/bad, well, performance. Low levels of systemic corruption are both an aspect and a precondition for a polity’s performance, so studying how voters’ perceptions of corruption affect their voting behaviour is kind of straightforward – only that few people have done it. This article uses data from round 2 of the CSES (2001-2006) to do just that.
One of the main findings (in my view) is that party ID moderates the effect of corruption perceptions. For non-partisans, perceptions of widespread corruption have a strong negative effect on the likelihood of a pro-government vote, as they should. However, government supporters will vote for the government, and supporters of opposition parties will not vote for the government, no matter what either group thinks about levels of corruption in their country. More technically speaking, partisanship emerges once more as one hell of a drug.
Ecker, A., Glinitzer, K., & Meyer, T. M. (2016). Corruption performance voting and the electoral context. European Political Science Review, 8(3), 333–354. http://dx.doi.org/10.1017/S1755773915000053
What we liked
Students liked the clarity of the theory section, the elegance of the economic/performance voting link, and the interaction plots. They also thought that the topic was super interesting.
What we did not like so much
Students may have been a bit grouchy, but they came up with a long list of potential improvements. Here are the most important points:
Students said that the definition of corruption given in the text was very broad, whereas the item in the survey was highly specific. Table 1 in the text has tiny entries (even given their youthful eyes) and did not show the random effects or the number of cases on the upper level.
Contents-wise, students asked about the policy lessons that could be learned from this research (ouch!), demanded additional system-level variables including measures for freedom of the press and even suggested a longitudinal perspective.
After reading a couple of useful primers on multi-level modelling last week, students also noted the absence of an empty model as well as a far bigger problem: the number of contexts (national elections) is really low (about 20, simply because this CSES round covers a limited number of countries), whereas there are three macro level variables and two cross-level interactions. This is asking a lot of relatively few data points. Having realised this, the finally stopped banging on about more interesting system-level variables that could be included.