Tuesday, October 27, 2009

Stephenson and Tanguay on electoral reform. Or why you need multiple methods.

Yesterday afternoon, Dan Rubenson forwarded on to me Laura Stephenson and Brian Tanguay's IRPP research paper on electoral reform in Canada. It's a nice piece in a lot of ways. It charts the path to electoral reform very comprehensively. It draws the link between the decline in turnout and calls for electoral reform (even though the latter will do next to nothing to change the former). And then it presents a series of results on why citizens of Ontario voted against reform 3 to 2.

This is where their piece goes off the rails a little bit.

S and T take a very conventional approach. They survey 1000 Ontarians on a battery of items, including attitudes on things they may have never thought about. In particular, they ask lots of questions about whether citizens value fairness and proportionality, and whether they feel that election results under the current FPTP system are unfair. This is fine as far as it goes. I don't have too much trouble with the idea that citizens can hold more or less sensible opinions on things about which they know nothing and have thought very little. The more damning objection is that they miss a central fact about electoral reform.

Here it is: electoral reform, especially away from a FPTP system, is not about being for proportionality or against proportionality. On principle, I can't imagine many people are against fairness. It's about trade-offs between two largely irreconcilable goals. In the case of PR, the goals are responsiveness and policy that reflects the preferences of as wide a swatch of the population as possible. For plurality systems, like FPTP, it's about effective governance and accountability. It's about being able to effectively identify the rascals and show them the door if you dislike their actions in the years prior. This argument has been made by people a lot smarter than me. Powell, for instance, makes this argument very clearly. Yet mention of him is entirely missing from this piece. Asking about only one aspect of electoral reform largely misses these contours. As a result, S and T's study sheds dim and incomplete light on why electoral reform failed. It doesn't make it wrong, just not as right as it could be.

Now, the shameless self-promotion. Daniel and I ran a couple of studies during the same referendum. We ran them a lot differently than most of our colleagues who rely solely on survey based methods and lots of post-hoc modelling assumptions and simulations. First, we conducted an actual field experiment in which we exposed voters to varying levels of campaign materials from either or both sides of the campaign. You can read a paper about it here. (Of course, you can also send comments). What do we find? So far, we've completed failed to find a persuasion effect. Neither side was much good at convincing voters that their system was better; those who received mail from either side did not hold opinions any different from those who received no mail at all. If you want an exogenous and uncorrupted measure of information, this is a lot better than a quiz on a survey. Second, we found a small mobilization effect, such that those who received more mail from either side were more likely to vote in the referendum. But the effect is very small. Those in the heaviest treatment condition only voted in rates about 1.4 percentage points higher (Table 4). These effects are much smaller than any that would be estimated according to some information coefficient in a regression model. It's just hard to believe that increased information in the referendum would have increased turnout or support for MMP.

Why is it hard to believe? In another paper with Arthur Spirling, Dan and I investigated the power of arguments for and against reform (the paper is under R and R, so it's not public. But I am happy to send it on request). By power, we mean the ability of an argument to illicit agreement with its preferred position in the face of an argument from the other side. To measure the power of arguments, we gave respondents in a survey experiment one of six arguments for FPTP and one of six for MMP. We then asked them which system they supported. We ignored all individual level aspects, and instead focussed on the charactersitics of arguments. Through something called a Bradley-Terry model, we are able to test which arguments are most powerful (or convincing) and why. The picture below demonstrates our key results.

Here's the main story: there is a lot that is convincing about MMP, namely that it provides for much greater proportionality. But, there are two things that hurt it. First, it weakens local representation.* Second, it gives a central role to political parties. Voters like proportionality, but they don't like the other two factors. Unfortunately, you can't argue for MMP without mentioning these other factors. Likewise, you can't invoke proportionality when arguing for FPTP.

The figure below shows the effects of an argument invoking some feature of electoral systems on its probability of winning an argument in our experiment. As can be seen, an argument that invokes fairness experiences an almost 14 point advantage. But as soon as political parties are mentioned, a hit of about 7 points is incurred, cutting the fairness advantage in half.

And what of FPTP? It's advantage came from a small built-in bias for the status quo. All FPTP arguments enjoy about a 7 point kick. And then arguments which invoke local control get a 9-10 point kick. So, FPTP has small but consistent advantages over MMP.

The real story is that both systems have elements which make them attractive but which also draw away from them. If you want to explain why a referendum issue falls, you should take a full account of the dimensions of the choice at hand. And one should think about more than just conducting a survey.

*I know, I can hear howls in some departments that it doesn't, but I think the consensus is that it does. I've got a clever experiment in the works to test this proposition.


Anonymous said...

There is, in fact, another survey where a bunch of trade-off questions are put to Ontarians in an effort to account for their voting behaviour. It was conducted by Fred Cutler (UBC) and Patrick Fournier (U de Montreal). The trade-off questions come from their study of the 2005 BC referendum (with Richard Johnston, UBC and Ken Carty (UBC)).

Anonymous said...

"Asking about only one aspect of electoral reform largely misses these contours. As a result, S and T's study sheds dim and incomplete light on why electoral reform failed. It doesn't make it wrong, just not as right as it could be." Maybe - but maybe these alternative contours are also not relevant.

This debate is illustrative of what's wrong with this type of political science. Where's the theory? Theory is what should be guiding the research design but frequently in these types of debates, theory is absent.

Peter Loewen said...

I like to think that the place where I talk about previous work on the nature of electoral systems qualifies as theory, but I could be wrong. Perhaps you want something more formalized.

Anonymous said...

That's not theory. That's a lit review.

No, no need for formal theory.

What I mean by theory in this case would be something(s) that addresses the question of why citizens vote during referendums.

Anonymous said...

Or how they make choices in the context of referendums.

Peter Loewen said...

Agreed, it would be nice to have theory on what motivates choice in a referendum, particularly in a paper. I generally don't think blog posts have to theoretically motivated. For the time being, I (and Daniel) just like to point out that the choice presented to voters is not one-dimensional.

Anonymous said...

Of course - I think you are right that if one chooses to focus on one part of the justification for MMP vs. FTPT, then you should also consider that other parts of the debate may matter (like accountability vs. efficiency, etc.).

But I guess my point would be that maybe the goal should be to generate or at least use an existing theory that tells us why:

a) citizens, when faced with two choices in a referendum, focus on things like proportionality vs. not and accountability and efficiency;

b) or more broadly, tells us how citizens evaluate such choices (media; partisanship; etc.).

But I agree with your criticism of S and T. At a minimum, if you choose to focus only on one aspect, then you need to either provide a theory that supports that choice, or take into account the other issues you identified.