Thursday, 10 November 2011

Democratic Elections (Part 3)

And finally we get to the method New Zealanders know [?] and love [??], the Mixed-Member Proportional system.  I noted at the start some surprise that NZ had enacted such a major reform as the adoption of a completely different electoral system; but it was motivated by possessing possibly the worst system in the democratic world: First Past the Post election of a unicameral legislature.  In 1993, Kiwis changed to a very different beast.

Mixed-Member Proportional (MMP)
The most important feature of MMP is that it is proportional: seats in the parliament are assigned to parties in proportion to the percentage of the national vote that each party receives.  The actual way that is worked out is by the Sainte-LaguĂ« formula, and the final result is, indeed, pretty proportional.  There are, however, some twists and turns.

There are approximately (see below) 120 Members of Parliament, which works out to 0.83% of the vote per MP.  The first twist is that there is a threshold of 5%, i.e., in order for a party to get any Members of Parliament, it must get at least five percent of the vote.  This stops the parliament from being fragmented into many small parties, Italian style, but whether it's a fair barrier to participation is questionable, especially as preferential voting is not used.  As we'll see, this barrier exists for some, but not for others.

The next twist is that rather than being just a party-centred proportional system, there is also a level of regional representation built into MMP.  Seventy MPs are, in fact, "Electorate" MPs: they are elected to represent a particular electorate by a first past the post vote.  This doesn't mess with the proportionality of the system (mostly: see below!), because each Electorate MP counts against the party's total allocation.  The rest is made up from "List" MPs.

So at the ballot box, each voter has two choices to make: the party they want to vote for and the local electorate candidate they want to vote for; one voter, two votes.  To first approximation, the electorate vote is irrelevant: the total number of MPs the party gets is determined by the party vote, exactly which electorates they represent -- or whether they come in on the List -- is a matter of shuffling the chairs.  But for small minor parties the electorate vote can be very important: winning an Electorate seat means dodging the 5% threshold.

In the 2008 election, for example, the ACT party received 3.65% of the party vote, less than the threshold; but one of its candidates, Rodney Hide, won the seat of Epsom, thus defeating the threshold.  ACT's party vote was enough to elect 4 additional MPs.  Another party, New Zealand First, received 4.07% of the party vote, but won no Electorates, and therefore received no MPs.  The Maori Party only got 2.39% of the party vote, but won 5 Electorates, so its representation in the parliament was much larger than its party vote required (3 seats)!  This, by the way, is why the number of MPs is not fixed.  The Maori Party excess means there are currently 122 MPs; an "overhang" of two.  So once again we see that concentrated regional support trumps broad, thin support.

The main problem with this is not so much the discrimination between parties (though I'm sure Winston Peters would disagree), as the discrimination between voters.  As we've seen, the "one voter, two votes" of MMP is really just "two ticks, one vote" for most voters.  But some voters -- those in the Maori Seats and the Good People of Epsom -- actually do get two votes!

Renowned Aussie psephologist Malcolm MacKerras explained this a worthy rant some time ago, and this election this exploit is very much in the public mind.  However, the complaint being made by the SM proponents is very different.  What upsets them about MMP is that it is difficult to get rid of a particular parliamentarian.  If they lose their Electorate, because they're out of favour with their constituents, but are still in favour with their party then they come into parliament as a List MP.  This is a consequence of the party-centred philosophy of MMP, as opposed to the candidate-centred philosophy of STV.

MMP is designed with the expectation that voters vote for a party, whilst STV allows voters to vote for an individual.  While it is true that modern politics in much of the world, and certainly in New Zealand, is based on the party, objections to particular members are not that uncommon.  The example that springs to mind is Larry Anthony, turfed from the seat held by his father and his grand-father despite the overall swing to his party.  This would be almost impossible in MMP.  The voters' inability to rid themselves of unpopular politicians is a fair complaint, however the correct solution is to opt for STV, not SM.

First Past the Post 
  • Pro: Voters feel they understand it.
  • Con: They don't.
Preferential Voting
  • Pro: Can vote for the candidate you want.
  • Con: Overwhelmingly favours the largest parties.
Supplementary Member
  • Pro: Better than FPP.
  • Con: Not better enough.
Single Transferable Vote
  • Pro: Preferential and (approximately) Proportional.
  • Con: Complicated.
Mixed Member Proportional
  • Pro: Proportional representation.
  • Con: Not preferential, favours voters in some electorates.

Tuesday, 8 November 2011

Democratic Elections (part 2)

Continuing our psephological journey we look at Single Transferable Vote (STV) and Supplementary Member (SM).  But first I'd like to point out one of the challenges of this discussion.  The proposals blend two things: the voting method, and the thing you're voting for.  In the two proposals I discussed last time, First Past the Post (FPP) and Preferential Voting (PV), the only difference between them was the voting method; both proposals are for single member electorates.  The next two systems -- STV and SM -- include multi-member electorates.  One of them uses preferential voting (note the lack of caps) and the other uses first past the post (ditto).

Single Tranferable Vote (STM)
Unlike FPP and PV (but like MMP), Single Transferable Vote is a proportional system, i.e. it is constructed so that representation in the parliament is (roughly) proportional to the number of votes a team gets.  So if a third of people vote Green, a third vote Labour and a third vote National, that's how the seats in the Beehive will be handed out.  Of course, all the proposals look reasonable at first glance, so let's look at the details.

Suppose New Zealand gets divided into 24 electorates, each returning 5 Members of Parliament (it looks like it won't actually be quite that neat, but anyway).  That means that each electorate will have around 100,000 voters.  In order to be elected each candidate will need just over one sixth (for five members) of those votes, i.e. roughly 17,000 votes.  That number is called the "quota".  So if you like a candidate, and so do 16,999 of your fellows in that in electorate, then s/he can win a seat.  Note that this is a much easier challenge than trying to convince half of the voters in an electorate.

So far it's pretty simple.  The main complexity of this system comes from the fact that it is preferential as well as being proportional.  In part, that's because there has to be a way to deal with surplus votes.  For example, suppose a candidate gets 20,000 votes.  That's more than quota, so that candidate gets elected, but what do you do with the extra 3,000 votes?  The STV solution is this: the voter numbers every candidate from 1 (first preference) up till every candidate has a number.  Each candidate's "1"s are added up.  Any who have quota are elected, and the surplus distributed in order of preference, i.e. to whichever candidate is marked "2" on each ballot.  The process continues till 5 MPs (in this example) have been found.  The counting and distribution of preferences in a little technical, a more detailed description is provided by the AEC.

So the advantages of STV are that it is both proportional -- so highly democratic -- and preferential, so voters can vote for the candidate they actually want.  It also has the advantage of not suffering from what seems to be the major complaint against MMP: because you can vote for individuals, you can vote individuals out.  There's no coming back in on a list!

The most obvious disadvantage of STV is its complexity.  In addition to the complexity of the system itself (and few people really understand any of these systems) the voter is often confronted by a ballot containing many candidates.  At the 2010 Australian Federal election, South Australian voters faced a ballot paper containing 42 candidates (to elect 6 Senate members).  Rather than insist that the voter number every one of the 42 candidates, Australian voters have the option of voting "above the line", outsourcing their preferences to one of the parties (or other ticket).  So in NZ, it would be possible to vote National, for example, saying you're voting the way the National Party would like you to vote.  In effect, it would be much like the Party vote in MMP.  This desire to offer a simpler voting option does, of course, complicate the ballot: two ways to vote on the one piece of paper!

A particularly dramatic corner case for STV occurred in the 2004 Australian Federal Election.  This election saw a turf war between the Labor [sic] party, the Democrats and the Greens.  In an effort to slow the rise of the Greens, Labor and the Democrats both directed Senate preferences to Family First, a new conservative party.  In the state of Victoria, the trickle of preferences was actually enough to elect Family First candidate Steve Fielding with just 2.63% of the first preference votes!  The election of Steve Fielding was a clear failure of the system (and I'm talking purely about voter intention, not his merits as an MP), but one that required a very special collection of circumstances, including the use of above the line voting.

Supplementary Member (SM)
We could just slander SM by pointing out that it's the system used in Japan, possibly the most politically unstable of the developed nations (Japan has chewed through five Prime Ministers in the last four years, and 12 in the last 20).  But that would be childish, so let's give SM a fair trial before, umm...

As I alluded to above, there are basically two competing psephological ideas: single member constituencies and multiple member constituencies.  FPP and PV adhere to the first, while STV and MMP use the second.  Supplementary Member is an attempt to blend the two ideas (in a way, so is the combined use of PV and STV in Australia).  In New Zealand, there would be 90 electorates where the MPs would be chosen by first past the post -- "Electorate MPs" -- and then 30 members chosen proportionally (also by first past the post, but in multi-member electorates it's called Single Non-Transferable Vote), called "Supplementary MPs".

In practice, voting would be much like in MMP, with an electorate vote and a party vote, but the relative importance of these would be swapped.  The Electorate seats would behave just like they would in FPP, and the major parties would also get the lion's share of the Supplementary seats, relegating the minor parties to fighting over the scraps.  The Supplementary MPs would presumably be allocated using the Sainte-LaguĂ« formula, but the Electoral Commission's website doesn't explicitly say so.

There are some nice examples comparing SM to MMP in actual elections at Kiwiblog, but the usual effect is to amplify the influence of the strongest single party and diminish that of everyone else.  Some would see that as a positive, in that it tends to generate decisive results, but then, so does a coin toss.  SM is neither proportional nor even preferential.

Partial Conclusion
At first sight, Supplementary Member presents an elegant compromise between the ideologies of single vs multi member electorates, however the rump of 30 Supplementary MPs will have little effect on the body of 90 Electorate MPs, especially as most of them will belong to the major parties.  Voters might seek insurance by voting for a major party in the Electorate, and a minor party in the Supplementary seats, much as many Australians do by voting for different parties in the House of Representatives and in the Senate; however under SM this strategy will not be as effective, as all the MPs sit in the one house.  In short, this is a dollied-up version of FPP, and has the same flaws.

Single Transferable Vote is much more democratic.  It is not strictly proportional in the way that MMP is, but does allow the voter to pick and choose from the available candidates, instead of leaving it to the parties.  And under STV everyone gets one vote, while in MMP some get two (more on that next time).  STV incorporates preferential voting, which means voters can vote for the candidate they want without being "tactical";  but the inclusion of "above the line" voting (or "party" voting, Kiwis might call it) encourages some devious strategising by the parties.  Of course, most of the time such details have little effect anyway.

Next time: Mixed-Member Proportional and Summary

Friday, 4 November 2011

Democratic Elections (part 1)

I had been under the impression that significant political reform generally required somebody getting shot; but in New Zealand, not only can the electoral system change with a vote, it actually does.  In 1993, NZ adopted Mixed-Member Proportional representation (MMP) -- the system used by the Germans -- and at this year's general election Kiwis will be asked if they would like to change it again.  The referendum will have two questions, firstly, should New Zealand keep MMP? and secondly, if NZ were to change, which of a list of four other systems should it use?

I decided to look at each of the alternatives on offer:

First Past the Post (FPP)
Initially, I wasn't planning to spend any time on this one as it's an incredibly silly system, but given that the Poms actually managed to vote for it earlier this year, and polls show it's the second most popular option in NZ, it would seem to be worth a few words.

In this proposal the country would be divided into a number of single member electorates (120 of them), and the winner of each is the candidate who gets the most votes.  Simple.  But if there's more than two candidates, there is a real risk of returning the wrong result, because no one needs fifty percent of the vote.  A candidate despised by sixty percent of the electorate can win.  While two party politics dominates many (most?) democracies, three way contests are not that uncommon, either at the national level -- consider the UK (Tories vs Liberal Democrats vs Labour), or Canada (Conservatives vs Liberals vs New Democrats) -- or at the electorate level, and we'll see a new example of that in New Zealand this year (Maori vs Labour vs Mana).  And then there is always the Greens.

One of the more famous demonstrations of the flaw in this method of voting comes from the 1992 US presidential election where Ross Perot split the conservative vote.  Bill Clinton won with 43% of the vote, when George Bush Senior was almost certainly the more widely preferred candidate.

Actually, the most insidious aspect of FPP is not the contests it gets wrong, but the contests it prevents from happening at all.  There is little point in voting for a candidate that few others vote for - the vote is "wasted".  Instead, the voter is encouraged to second guess the intentions of other voters, throwing support behind a candidate he expects to be competitive with whomever it is he really doesn't want.  Hence the controversy surrounding any of Ralph Nader's electoral tilts.  But lousy though FPP is at dealing with individual competitions, it actually has a much bigger problem, which it shares with the next system to be discussed.

Preferential Voting (PV)
Question: How do you spot an Australian?  Answer: They're the one trying to explain preferential voting to the person next to them.
As an Australian, I regard preferential voting as my birthright, and it's certainly a practical corrective to the problem described above.  Given a list of candidates, the voter marks their preferred candidate with a "1", the next with a "2" and so on, saving the highest number for the most despised candidate on the list.  Or, if you prefer, decide whom you dislike the most and work backwards.  Counting is performed by determining which candidate has the least number of "1"s, eliminating that candidate and distributing those votes to all the "2"s.  This process continues until one candidate has more than 50% of the vote.  The voter can actually support the person they want, without it directly serving the person they least want.

However, PV shares the biggest weakness of FPP: single member electorates.  Consider the following thought experiment: Party A gets a uniform, nation-wide vote of 50.1% (either first preference or two-party preferred, it doesn't matter).  This would lead to that party winning 100% of the seats!  While the example is extreme, the phenomenon of individual parties dominating a parliament on slender majorities -- or even considerable minorities -- is not.  Consider the 2005 UK election in which Labour won 355 seats in the House of Commons with a mere 35.2% of the vote.  The Tories and Liberal Democrats, with a combined 54.4% of the vote managed just 260 seats between them.  Minor parties can thrive in such a system if they have strong regional support, but not if they have broad support.  So, for example, the Democratic Unionists (Northern Ireland) were able to convert their 0.9% of the vote into 9 seats, while the Greens were unable to gain any seats despite 1% of the vote.  So in New Zealand, PV will suit the Maori Party just fine, but at considerable cost to the Greens.

A similar disparity exists between the Nationals and the Greens in Australia, although there the excess of single-member electorates is partly mitigated by having a bicameral legislature, with the Upper House selected using a system of proportional representation, Single Transferable Vote, which I shall discuss next time.

Partial Conclusion
The perfect electoral system accurately reflects the intentions of the voters, but if that could be perfectly defined then the hunt for a system would be over.  All the systems so far devised have their quirks, and can fail at corner cases.  But the fact that there is no one right answer does not mean that there are no wrong answers, and both FPP and PV are, in this context, wrong answers.  Preferential Voting is a good way to resolve individual contests, but by itself cannot overcome the flaw of single member electorates (although it does push things in the right direction, which is why the outcome of the UK referendum was gobsmacking).  The remarkable thing about FPP is not that it fails sometimes, but that sometimes it doesn't fail.

Next time: Single Transferable Vote and Supplementary Member

Thursday, 3 November 2011


On the 23rd of September, the OPERA collaboration announced that they had measured the speed of neutrinos travelling from CERN to their detector at Gran Sasso, and that it exceeded the speed of light by 0.00002%.  The announcement was met with excitement and enthusiasm, but the dominant view amongst physicists seems to be that it is due to some -- as yet unidentified -- systematic error.  Some have sought to draw a parallel between this scepticism, and the opposition to action over climate change.

Four ways in which scepticism about superluminal neutrinos is unlike scepticism about climate change.
  1. It's new:  All new, as yet unverified results should be treated with a healthy dose of scepticism.  OPERA released its preprint just over a month ago.  The current understanding of global climate change and its possible causes is the result of decades of work.
  2. It's weird: Neutrinos travelling at more than the speed of light is difficult to reconcile with a wealth of well-established science.  On the other hand, carbon dioxide is known to act as a green-house gas, and this behaviour is well understood.  While it might be surprising that net human emissions have been great enough to make a difference, it does not pose a problem for fundamental science.  Not all claims are equal.
  3. It's the scepticism of peers: The scrutiny that the OPERA result is receiving is led by physicists, people who understand something of the strengths and weaknesses of such experiments and can put the result in context.  Climate change scepticism is led by commentators, particularly in print and radio.  Of course, there is also one particularly committed comedic performance artist, and a Catholic cardinal!
  4. It's not a threat: the confirmation of superluminal neutrinos would be truly monumental, and might lead to revolutionary discoveries and technologies.  However, it would have little impact on the way people live for the foreseeable future (short though that might be).  The possibility of anthropogenic climate change, however, presents a direct challenge.  Even modest changes in weather patterns will affect land use, agriculture, the prevalence and spread of certain diseases, and the way many of us live.  Studies have repeatedly shown that the cost of mitigating action is modest compared to the potential cost of inaction - a clear case for taking out insurance.  The case for action on climate change doesn't require "proof", just the existence of a credible threat.

Believing Nonsense

In an article in Crikey (paywall), Noel Turnbull asks why we believe so much that’s wrong: "One of the great conundrums of modern life is how so many people have come to believe so much that is just plain wrong."  He goes through some fun facts, such as the research that showed that the more you watch Fox News, the more ignorant you become.  This reminded me of a claim about human cognition I encountered back when I was a student: it's easier to accept new facts than to discard old ones.

I don't remember who wrote the book, but it was someone famous and I think it was about the development of the GSW electroweak theory.  The above claim was followed with this example: when Chadwick discovered the neutron, nuclear physicists readily accepted the existence of the new particle, but had trouble disposing of the notion that the nucleus contained electrons (because they were observed in beta decay).  In other words, whoever tells their story first has a big advantage.

We see this phenomenon all the time in those emails that circulate with amazing stories, the kind of thing that Snopes is devoted to debunking.  A typical exchange might be, for example, that Alice forwards to Bob an email claiming that (US President) Barack Obama was born in Kenya, and therefore ineligible for the Presidency.  Bob expresses some doubt over the quality of the email's contents, to which Alice's inevitable reply is "How do you know it's not true?".  Bob then resorts to plausibility arguments and/or scouring the web for contradictory evidence, when the real question is, why did Alice ever think it was true in the first place?

So long as a new fact doesn't obviously contradict something we already think we know, we accept it easily and incorporate it into our worldview.  When shaping public opinion, the first to act therefore has a big advantage.  This might be why, when we have unprecedented access to information, it is so easy to be mis-informed.  The human brain makes it impossible for de-bunkers to keep pace with the bunkum.

Take 2

Well, I tried to set-up a blog on my google sites page, but got frustrated, so here I am.  To start off, I'm re-posting those posts here.