Work has been a little hectic lately, so the blogging has been light of late. I do plan on resuming a more consistent schedule soon, however. Before diving back in the Egypt numbers, however, I wanted to highlight two very good papers I saw presented at APSA several months ago.
When advising on electoral system design, election assistance practitioners often recommend that developing countries implement simple systems that their voters will be able to understand. While the theory behind this is sound, I’ve always wondered if we were overestimating the impact of ballot complexity. After all, I’m sure many Americans aren’t aware that many of their city council races are MNTV, while their congressional race is FPTP, yet our system still functions. In two papers that actually attempt to answer this question, Aina Gallego of Stanford University and Saul Cunow of UCSD both looked at what impact various electoral variables have on voter turnout. Aina argued that increases in ballot complexity, while not impacting educated voters, has a negative effect on turnout rates of less educated citizens. Aina used two strategies to test her hypothesis. For the first, she conducted a field experiment where she sent several fake ballot questions to a random group of Spanish citizens. The control group was given a straightforward ballot containing descriptions of candidates and which party they belonged to, and then asked to vote for one. The second group was given the same list, but was asked to vote for up to five. The second group had a 12 percent drop in responses from low-educated citizens. For her second strategy, she ran a cross-national regression of electoral system design by turnout among various demographics and found that controlling for other factors, increased preference votes were associated with a decrease in turnout.
I had two problems with Aina’s methodology. The first is her field experiment failed to simulate a realistic ballot structure that any voter would encounter. The second is her classification of voting system type was binary (ability to cast preference votes or not) and did not account for the many rules that would significantly impact the level of ballot complexity within preference voting systems. This could include the option of voting across multiple parties, the ability to rank candidates, and the ability to punish candidates on a given party list. Also, her theory is based on the cognitive perceptions of voters before they decide to vote, not on what they actually experience in the voting booth. Invalid ballots are evidence of voters who were unable to handle the ballot, but showed up anyway. Therefore, I would suggest a more realistic measure would be district magnitude, as an increase in candidate choice would, by her theory, intimidate the voter and make them less likely to turnout. Despite these shortcomings, her paper provided a new look at a concept that has been assumed, but never rigorously tested.
Related, Saul Cunow conducted a field experiment in Brazil and found that there is a curvilinear relationship between the number of candidates on a ballot and turnout. That is, at low levels of candidate choice, turnout is low as voters feel they have little options. As the number of choices increases, turnout increases, due to the more choices. After the number of potential candidates increases past a certain level, however, turnout decreases as people are confused by the number of choices and have a harder time distinguishing between them. Saul also finds that the presence of party labels does not reduce the probability of abstention among higher numbers of candidates.
Both are good papers and you should read them.
One of the takeaway points for me is that voters have consistently voiced strong support for systems that are more proportional, but that support quickly evaporates once it is described to them what a proportional representation system is. This isn’t that surprising; traditional tradeoffs between fair and effective governance, although overstated, are easier to make when they are abstract. When you start to think about how they will impact your preferred party, however, things might be different.
This made me curious to see if there has been anything written about vote choice and direct democracy, with a particular interest in the impact of elite signaling. The best thing I found was this paper by Simon Hug and Pascal Sciarini, which discuss how different variables about a referendum affect vote choice. Examining data from fourteen European integration referendums, Hug and Sciarini essentially claim that issue saliency determines voting behavior. In important “first-tier” elections, voters make a decision by weighing the actual issue. On less important “second-tier” elections, voters may base their decision on their assessment of the ruling party. This comes in the form of voting against the wishes of the government if one is dissatisfied, and voting in favor if one is supportive. This makes sense but it only explains voting behavior using a rational choice/retrospective model, where voters retroactively form their opinions of parties after evaluating their performance in office.
I’m not against looking at things this way, but I don’t think it’s appropriate for every type of referendum. The UK AV referendum, in particular, doesn’t fit any existing model. I’m not sure if the election would count as “first or second tier” in importance – the 40 percent turnout leads me to believe second – but I don’t think it matters. Even if it was first tier, voters would not be able to punish the ruling government as it was made up of a coalition divided on the issue. To me, it’s very difficult for a voter to not weigh the issue through a partisan filter because the referendum is essentially a vote on future partisan performance. Yet, as we’ve previously discussed here, there still seemed – at least according to one earlier survey – that a decent amount of partisans were going against their party. I would like to see more written about this by people who know more and have more data to play with. I’m guessing that committed Tory and LibDem partisans took cues from their party leaders. Labour partisans, having been sent such mixed signals from thier elites, would be interesting to examine. Did Labour voters have a clear idea over whether AV would help or hurt them? I think you could make arguments either way but I’m not sure what they heard. Also, how did those with weak identification vote, if they even turned out?