The Additional Member System - New Zealand's experience

Yesterday New Zealand had an election. In itself it wasn't too dramatic - the main governing party slightly advanced, the opposition fell back rather more and there was some shuffling round amongst the smaller parties. The most prominent story is the return to parliament of the populist New Zealand First. But of other interest is that New Zealand also held a referendum on its voting system and opted to retain its current Additional Member System (known as Mixed Member Proportional over there, but I'll translate the terminology for a British audience).

Now here in the UK changing the voting system was dealt a huge blow at the AV referendum in May and many, myself included, are willing to take the losing campaign's cries of "this is the only chance to change the system" at face value. However in politics few proposals ever truly die and circumstances can change, bringing once seemingly dead issues back to the table. (The New Zealand referendum came 18 years after two previous ones that adopted the system, and six elections in the meantime. Historically this is slow for New Zealand which until 1989 had referendums on prohibition alongside every election.) And despite the referendum we still have other voting systems in use - the Additional Member System is used for the Scottish Parliament, Welsh Assembly and London Assembly; the Single Transferable Vote is used for the Northern Ireland Assembly and local government in Northern Ireland and Scotland; party lists are used for elections to the European Parliament (apart from Northern Ireland which still uses STV); and the Supplementary Vote is used for directly elected Mayors and will also be used for Police Commissioners. So one way or another the voting system question hasn't gone away completely.

New Zealand's implementation of the AMS system is relatively straight forward. There are seventy constituencies ("electorates") which each elect a single member of parliament by First Past The Post. Then there is a top-up list vote, operating on a New Zealand-wide basis, which elects a further 50 members. Crucially the 70 constituency members are factored into the overall proportionality. There are, however, some additional key details:

• There are separate constituencies for Māori voters. When registering voters can choose to go either on the main ("general") electoral register or on the separate Māori register (but cannot appear on both). The seven Māori constituencies are organised on a geographic basis, overlapping the sixty three general constituencies. All voters vote for the same list.
• In order to qualify for list seats, a party has to get either at least 5% of the list vote or win at least one constituency. (The German implementation includes a further qualification that parties that represent specific minorities - e.g. the South Schleswig Voter Federation representing ethnic Danes in Schleswig-Holstein - are exempt from the threshold requirements. Such a rule was considered for New Zealand but the final implementation opted instead for keeping the separate Māori constituencies.)
• Proportionality is calculated only amongst list votes for parties that qualify for list seats. Thus if 10% of the vote goes to parties that don't qualify, the 120 seats are apportioned on the basis of 90% of the votes.
• If a party wins more constituency seats than its list vote entitles it to then there is an "overhang". This is partially corrected by automatically creating additional list seats equal to the number of overhang constituencies. For example in 2008 the Māori Party won five constituencies but a list entitlement of only three, so two extra list seats were created.

(The British uses of AMS have all voters on a single "general" register and maintains the 5% rule, which in London has caught out variously the Christian People's Alliance, RESPECT and the BNP in some elections. In Scotland and Wales the use of separate AMS regions means there are fewer seats available overall. Also AMS here uses a different mechanism for allocating the list seats though in practice to much the same effect. There is no mechanism to counter an overhang.)

AMS can throw up a number of anomalies and confusions because of this - for those interested in the more technical details have a read of Antony Green's Election Blog: The MMP Sealed Section: How the Intricacies of MMP Could Prevent a National Majority which is written for an audience with no direct experience of the system. By far the most interesting, and potentially controversial for any British consideration, are the rules on qualifying for list seats and the potential for gaming the system.

Preliminary results for 2011 suggest that this isn't an issue - every party that has won list seats is over the 5% threshold and every party below it with seats has only constituency seats. But in 2008 New Zealand First won 95,356 list votes and no seats whilst ACT New Zealand (a free market, free people, small government party) won 85,496 and was entitled to five seats overall. Were the UK to have a referendum on AMS I wouldn't be surprised if this occurrence got fished out at some stage. The difference was that ACT had successfully targeted the Epsom constituency, albeit with help from the Nationals, whereas New Zealand First won no constituencies at all. These things are swings and roundabouts - New Zealand First had itself relied on constituency success in 1999 when leader Winston Peters narrowly retained Tauranga by just 63 votes but by 2005 they lost there and were at the mercy of the list.

(Other parties who've qualified for list seats this way include: the Progressive Party in 2002, United Future in 2005 and ACT also in 2005. However it was rare in any of these elections for such parties to be outpolled by others who ended up with no seats.)

And the result partially reflected a gaming of the system by the larger National Party - they gave a discrete message to their list voters in Epsom to give their constituency vote to ACT so that the latter would qualify for list seats and boost the chances of a National government with ACT support. The constituency and list votes are cast on the same ballot paper and the detailed count includes the precise combinations - in 2008 nationally 86.3% of National Party list voters also voted National for the constituency vote, but in Epsom this figure was just 26.2%, with 69.9% of National list voters voting ACT in the constituency. (Antony Green's Election Blog: 2011 New Zealand Election: How MMP Works) Other stats from that breakdown show a lot of Green list voters vote for other parties for constituencies and indeed there are some signs of Labour and Green list voters in Epsom trying to counteract the National/ACT effect by voting for the National constituency candidate. It is a bizarre situation when a party's opponents are keener for it to win a particular contest than the party is itself!

In the past I've blogged about possible ways to game AMS - How the London Assembly could have been so different details how "decoy lists" would have worked in London in 2008. New Zealand's overhang correction mechanism could throw up interesting effects - either many more list seats would be created to counter the effect of lots of constituencies being won by parties with next to no list votes (something similar happens in some German lander parliaments, resulting in wild fluctuations of the number of politicians) or the seats might be completely disregarded in the list calculations. But a system where small parties can survive or die depending not directly on the votes but on whether they can get tactical help from big parties will not be to everyone's taste, especially if the small parties are felt to wield disproportionate influence in government.

Is AMS a "simple" system for the lay voter to understand? Well that's very hard to measure as one person's simplicity is another's convolution. But some factors are there - on the one hand AMS lacks the surpluses and fractions of STV and it never sees the most popular constituency candidates (as measured by who gets the most votes) denied victory. However details like overhangs and multiple thresholds can be complicated to explain and would expand an official summary beyond conciseness. Furthermore the prospect of a party winning more than 50% of seats with less than 50% of votes is a real one under AMS - it happened in Scotland this year due to factors such as the use of multiple regions but also because of "wasted votes" for parties that secured no seats - and again this sort of detail that works counter to expectations can confuse. Now all systems have their anomalies but given the tendency of PR campaigners to make so much of victories on less than 50% of the vote, if the alternative is shown to not live up to their own standards then it's natural to be sceptical.

Could AMS win a future referendum on voting systems? That's very hard to say as much of it would come down to a variety of factors including events in advance, how well campaigners prepared the ground before the announcement of a vote, the effectiveness of the campaigns themselves and so forth. I do hope that in any such referendum better arguments would be put forward than "vote yes for utopian politics" or "vote no 'cos the Germans use it".

However for the moment such a referendum seems a long way off. But time will tell...