Monday, 24 December 2012

Second-Chance elections

South Africa has a purely proportional representation (PR) election system at all levels, except for town councils, which have a mix of PR and ward councillors. A ward is essentially the same thing as a geographic electoral district (US terminology), constituency (UK and old South Africa) or electorate (Australia). To standardise the terminology, I choose the shortest term for the concept of voting for a representative in a given region: district, and MP for representative, since “member of parliament” is very common terminology.

A pure PR system as we have in South Africa, in which each voter chooses a party, and the party decides how to order its party list, puts a disproportionate amount of power in the hands of the party. How high you are on the party list determines your chances of election: the higher you are on the list, the fewer votes the party needs to win for you to make it. Yet your position on the party list is not in any way influenced by how you perform in the election. A party can reward a stalwart, or worse, corruption, by placing a candidate high on the list, while making sure they do no public campaigning that might reveal their weaknesses. Also, you have no concept of your own MP, so the line of accountability is very indirect.

It would be desirable to fix these problems without losing the benefits of PR system. But first, let’s examine the disadvantages of the alternatives.

A pure district system has the problem that it can skew representation relative to the popular vote. If a party has concentrated regional support, for example, it may score a low fraction of the vote nationally yet win several seats. A party with wide appeal but not quite enough to win anywhere may score significantly more votes than this regionally-based party. In Australia, we see this effect with the Greens, who win typically 10–15% nationally, yet only hold one seat in the lower house of parliament. By contrast, the Nationals poll about a third of the votes the Greens win, and have 7 seats. The Australian system, even so, is not as unfair as the lower house systems in the US and UK, which use a first past the post system (FPP).

The Australian system requires that voters rank candidates in order of preference. If your first-preference candidate has the lowest number of votes in a first-round count where no one has over 50% of the vote, that candidate drops out, and your ballot is reallocated to your second-preference candidate, and this continues until one candidate has over 50% of the vote. The advantage of the Australian system is that it doesn’t “waste” the vote of someone who votes for a small party or an independent. The disadvantage is it’s complicated to understand and difficult to audit results in a close race. And it can still result in unfair outcomes, as I explain above with the skewed representation of the Nationals versus the Greens in Australia.

To illustrate the point, I constructed a scenario for a 400-seat parliament, with each seat (to keep things simple) contested by one each of a small left party, a bigger left party, a bigger right party, a small right party and an independent. I didn’t do the distribution of preferences accurately, but the results represent a plausible outcome:
systemsmall Lbig Lbig Rsmall RInd
Australian 6276901414
FPP 15143392210

What we have is three very different outcomes (based on the way I chose to allocate votes geographically). In a PR election, the two left parties have 192 seats, enough to form a coalition if they can win over a bit over 20% of the independents. The right parties have 173, and need almost all the independents to form a ruling coalition. In the Australian system (based on plausible assumptions about preferences), the bigger left party has enough seats to rule in its own right. In the FPP system, the bigger right party has enough votes to rule in its own right.

Of these variations, arguably the PR outcome most accurately reflects voters’ intentions: they are reasonably evenly divided between left and right, with a slight preference for the left, but with a significant minority disenchanted with the parties, and who want an independent voice in government. In practice, in an FPP system, small parties lose because a bigger party tries to frighten voters of a smaller party of their side of politics into supporting them (think Ralph Nader’s Green votes being blamed for Al Gore losing in Florida), so an outcome like this FPP example is not too likely – but the stifling of the growth of smaller parties is bad for democracy.

How, then could you achieve the fairness of a PR system without the disconnect between voters and MPs resulting from a party-list system as in South Africa? There are various hybrid systems including the municipal election system in South Africa, where some candidates are elected from a party list, and others directly (on a first past the post district system). However that hybrid system still puts significant power in the hands of the party, and still has the risk that the party becomes a source of unaccountable patronage.

My proposal, which I call a second-chance election (2CV for second-chance vote, not the venerable Citroën car, illustrated), is to have two MPs per constituency. The first is elected on a first past the post basis, and the second on a party list system with the wrinkle that the party list is formed out of the candidates who ran and didn’t win in the first round, ranked within party by how well they did. If for example a party wins 70% of the vote in a constituency, 20% of their vote there is in effect wasted since they only need 50%+1 to win, and their second candidate should probably be in parliament. If on the other hand the first past the post candidate won with 30% in a highly split race, the second candidate from that party should end up a lot lower down the party list. The tricky part is working out exactly how to do the voting; I work through one approach here, and don’t prescribe it as the only option.

Going back to our example, if we only elect 200 MPs on a first past the post basis, assuming the votes split the same way as in the above example, on the first pass, the elected members will look as follows (with deficit indicating how many seats each party is short of its entitlement):

2CV stepssmall Lbig Lbig Rsmall RInd
FPP half87169115
error vs. PR-13-1945-7-6

We can’t completely fix the deficit, because one party, the bigger right party (big R), already has more seats than it’s entitled to in a pure PR system, so it gets no more seats. The other parties’ deficits have to be scaled to a correction that adds up to 200 (each proportionate to the number of seats they are short of the PR total), so we get 400 MPs in total. The bottom row of the table shows the error versus a pure PR election (a minus value means a party has too few seats).

Compare the revised chart with the original. The 2CV system ranks the parties in the same order as PR, which is an improvement, though the result is not exactly the same. The two right parties can form a ruling coalition, which is not an exact reflection of the electorate’s intent, but the result is not as skewed as the FPP result, or as inaccurate as the Australian system, for this example. This is however a fairly extreme outcome, as it requires that one party be able to win nearly 85% of the seats in an FPP election, with only 31% of the national vote.

You could reduce the probability of an inaccurate outcome such as this by having more MPs per district, or by converting the entire election into a party list system after voting in districts. Working through such detail is a secondary concern once the principle is accepted.

Finally, two issues are worth considering: how independents are handled, and how switching parties is handled.

Independents may be of extremely diverse views but treating them collectively as a party list after voting in districts ensures that those with strong enough support are elected. If their support base is diverse, this will be reflected in a diverse independent “party list” being elected.

Since representatives are directly elected, unlike with a party list, they should have a right to switch parties, and this creates a clear division of accountability between the party (which can provide resources for re-election) and the voter (who may reject a candidate who abandons their principles or fails to deliver). In a pure party list, switching parties (or “crossing the floor” as it is described in South Africa, though that technically only applies to joining or abandoning the ruling party) is less defensible because representatives are not elected in their own right – though they can argue that their party has deviated from its principles or platform).

Where does this all take us?

What I have here is a basis for discussion. There are many details to work out. To create a direct line of accountability, if there are two representatives per district, that implies you have 2 votes. Do you have a different ballot for a different slate of candidates, or are all candidates on one ballot, that has to be marked twice? If a party exceeds its quota in the FPP stage, is there a good way to correct for this? For example, you could have an upper house that absorbs some of the excess MPs – though typically an upper house is elected on a different basis to provide some sort of checks and balances.

There has to be a better way: pure PR has served South Africa badly. The Australian system has too many imperfections to be a good alternative, and a pure FPP system is too unfair and locks out small parties. I don’t like a mixed-member PR and representative system like the South African municipal system (with variations in other countries like Germany and New Zealand) because it retains a party list as a component. The party role should end at nominating and supporting the candidate; if the party chooses badly, they deserve to lose.