While the Single Transferable Vote has some nice qualities, it is possible that it may not work that nicely for electing the members of the Constituency Assembly.  The number of candidates and the possibility that voters will rank (far) less than 25 candidates may mean that some of the members of the Constitutional Assembly will be elected with only a few votes.

The electoral system used to elect the members of the Constitutional Assembly is the Single Transferable Vote. While I remain unconvinced about the wisdom of electing the Constitutional Assembly, I think using the Single Transferable Vote (STV) may not have been such a bad idea – in principle at least. On the face of it, STV might seem an odd choice as the system is usually classified as a proportional representations system, which, as a class of electoral system, aims at achieve the proportional representation of parties. However, unlike most other proportional representation systems, the party identity of candidates doesn’t actually figure at all in the allocation of seats. In other words, only provides for the proportional representation of parties when voters vote along party lines. So the proportional representation aspect can be seen as applying to voters’ non-party preferences, i.e., we would expect the system to do a reasonably good job of reflecting voters’ preferences over whether to change the electoral system in a particular way (if that is what the voters care about). In theory, at least, that sounds kinda nice. But perhaps the best case for STV is that it is no worse than the alternatives. Other proportional representation systems are list based systems, which rules them out. So what is left are the variety of majoritarian systems. These are probably not very good options. Take plurality rule for example (actually, the single non-transferable vote). Each voter casts a single vote and the 25 candidates with most votes win a seat – this makes it very likely that the candidate winning the 25th seat would have very few votes behind him/her. There are other majoritarian systems, e.g., the block vote, that might be slightly better suited to the task but might invite strong incentives for strategic behavior (which people tend to see as a bad thing – I tend to think the opposite). In contrast, STV is relatively complicated and the outcome is fairly unpredictable – and there are good reasons to belief that all but the top few seats are for all practical purposes chosen at random. Which I think is better than having the elected candidates think they have a mandate to pursue the particular amendments that they campaigned on – however, the members are likely to disagree with me on that point.

In practical terms, I am starting to think that STV may do a fairly poor job in terms of creating representative outcomes.  Consider the following example – it is unrealistic but it illustrates the problem in the simplest way possible. Suppose there are 104000 voters and 500 candidates. With 25 seats to fill, the quota to be elected is 4001 votes. Each voter is free to rank up to 25 candidates. Now suppose that each voter only chooses to rank a single candidate and that each candidate happens to between 200  and 210 votes (the average would have to be 208). No candidate meets the quota so the candidate with the fewest votes is eliminated (there are no transfers since each voter only votes for a single candidate) until 25 candidates remain.  So, at least theoretically, we could end up in a situation where each elected merely has the support of just over 200 voters- or .2%.  Yes, that is one-fifth of a percentage.  We are bound to worry about the representativeness of the assembly if none of its members is supported by more than one in every five hundred voters.

Now this example is clearly unrealistic – most voters are likely to vote for more than one candidate – but the basic problem remains. If every voter votes for only two candidates and the second preferences are equally distributed among all the candidates the problem remains the same, i.e., once the bottom of each voters’ ranking is reached their vote is dead and doesn’t contribute to any candidates’ vote share. Now, is this a real problem? It depends on a couple of thing. First, how many candidates will voters chose to rank. The election to the Constitutional Assembly is a low information election in the sense that most of the candidates are relatively unknown and it is not clear how voters will choose to cast their votes. Overall, it seems likely that the votes will be distributed fairly equally among the candidates with the exception of a few candidates who are more familiar faces. More importantly, the fact that most of the candidates are fairly unknown makes it likely that most voters will not rank all the candidates – on the social network sites I have seen several people say that they plan to rank about 10 candidates. Will that be enough? It is difficult to say but my guess is that it is not enough. If none of the 10 candidates ranked by the voter are elected then these votes are dead. And with 500 candidates it is probably not that unlikely that a fair number of voters will find themselves in that situation. That would even be the case if some of the ranked candidates include one of the more ‘popular’ candidates.

Second, how the voters’ preferences over the candidates are distributed is also important. I’ve already hinted at an equal distribution of the votes across candidates being a potential problem, that is, if voters tend not to rank very many candidates. But the probably may also occur if voters focus their attention on a small set of candidates. Imagine, e.g., that all the voters only rank 20 candidates and that these happen to be the same candidates. In that case, those twenty lucky candidates are elected, but the remaining seats would actually not be filled… but a single vote for a candidate would be enough for the next candidate to get in.

In sum, it is relatively easy to imagine scenarios where at least some proportion of the assembly members will be elected with the support of only a few voters.  It then becomes more difficult the assembly as a representative and it would make it easy for those opposed to the assembly’s recommendations to claim that the assembly lacks a clear mandate.  If the process of adopting constitutional amendments involved a referendum this wouldn’t be a major problem but since constitutional amendments need to be accepted by two consecutively elected parliaments it doesn’t appear unreasonable to think that parliament may consider itself a better representative than the Constitutional Assembly and, therefore, simply ignore the Constitutional Assembly’s suggestions.

As I am really in favor of random selection of members the fact that each member is potentially only supported by a few voters doesn’t bother me too much – it would mean that who is elected has more to do with the candidates’ luck than anything else, i.e., a handful of votes may be all that distinguishes candidates.  The good thing is that it removes the candidates’ ability to claim that they have a mandate to implement the amendments that they campaigned on, which hopefully will lead to a more open-minded examination of the different alternatives.

A couple of external links:

  • Stefán Ingi Valdimarsson has a nice post (in Icelandic) on the opportunities for strategic voting in the context of the Constitutional Assembly
  • Nicholas Miller has short paper about the ‘chaotic character’ of STV in Electoral Studies that demonstrates that the outcome can change in surprising ways only a few voters change their ranking.  That STV doesn’t satisfy monotonicity (a condition stating that a candidate should do better, or no worse, if some voter decides to rank him/her higher) is well know but this is still a nice example.  [Subscription may be needed]