Dr Geoffrey Pritchard and Dr Mark C. Wilson, members of the Centre for Mathematical Social Science at the University of Auckland.
In November 2011 a referendum will be held in New Zealand to determine whether the electoral system for parliamentary elections should be changed. Full background can be found at the Electoral Commission's referendum website. We assume that anyone reading this FAQ is familar with the basic information on that site. We aim to allow voters to compare the proposed referendum options in a quantitative way, by allowing them to compute quickly, for a given polling scenario, the party seat distribution in Parliament under each system. More generally, we aim to promote the use of quantitative/mathematical analysis in making important public decisions.
All parties that won at least one seat in the 2008 general election, plus the New Zealand First Party. There is room in the input form to add up to 2 new parties if the user wishes.
It is written in Javascript and the source code is publicly available.
For information about the simulator, see this seminar presentation, or contact the authors above. For information about the 2011 referendum, see the Electoral Commission's referendum website or the Wikipedia article.
The Electoral Referendum Act stipulates different numbers of electorates for the various options. The size of Parliament is 120 MPs under each option (except that MMP may also produce overhang MPs). The existing rules for allocating Maori seats and South Island electorates must be unchanged.
The party support is translated into votes in different ways, depending on the electoral system:
In electorate-based systems, we need to know how each party's support is distributed among the electorates. For this purpose, we use the distribution of the party vote from the 2008 general election. This is a particularly important assumption in situations where the most popular candidate in an electorate does not belong to the most popular party (e.g. Maori electorates in 2008, or many of the electorates where National Party candidates won the electorate vote in 2002).
Some of the voting systems require voters to rank candidates in order of preference. Voters with a given first-preference vote may have a variety of rank orders for their second and subsequent preferences. In order to infer the frequency of the various rank orders, we used data from the New Zealand Election Study 2008 as detailed here.
A real-life electorate design would probably fudge the numbers a little to avoid doing this. But if New Zealand is to be evenly divided into 120 electorates without any fudging, then each new electorate must correspond to 7/12 of an existing electorate. This is not possible with a whole number of electorates in each island - there must be at least one electorate that spans Cook Strait. There is (and always has been) a similar problem with the Maori electorates.
We considered, but rejected (for now): drawing new electorates by aggregating the polling places; more complicated ways of mapping national support for a party to the electorate vote; more user options.
The MMP electorate vote is a very different animal from the single FPP vote, because voters need not consider party affiliation in casting it. This was crucially important for National's 21 successful candidates in 2002, most of whom were standing in electorates where their party was deeply unpopular (as indicated by the party vote). It seems likely that under FPP, some of them at least would have lost their seats.
National's overall party vote in 2002 was 20.93%, and it is this number that the "2002" button fills in on the form. But arguably, National would have got more of the popular vote under FPP than it did of the party vote under MMP; you can use this simulator to explore the consequences of different scenarios along these lines. The NZES has a question on hypothetical voting behaviour under FPP, which may be helpful.
In general, though, it is quite realistic to expect that a party with only about 20% of the popular vote (e.g. Social Credit in 1981) will win few seats under FPP. For a rather startling real-world example, see here.
This is a similar problem to National's in 2002. Maori Party candidates won the electorate vote in 5 of the 7 Maori electorates, but the Labour Party won the party vote in all seven. This leaves it a little unclear what the voters would do if they had only one vote each, instead of two. We have chosen to stick with the same assumption we use in the general electorates: party support is distributed in the same manner as the 2008 party vote. The implication is that if the 2008 general election were re-run under a single-vote, single-seat-electorate-based system (i.e. FPP, PV, or the electorate-seat part of SM), the Labour Party would win all of the Maori seats. Under STV, the Maori Party gets 5 of the 12 Maori seats, and under SM it gets one of the 30 list seats.
The simulator doesn't use candidate-specific information, so it cannot capture these outcomes.
It is probably true that a change in the electoral system would eventually lead to different behaviour by parties (and mergers of existing parties) which would change the assumptions behind the simulator so much that results may be less reliable. Since the purpose of this simulator is to compare electoral systems on as straightforward a basis as possible, we have not tried to anticipate such behavioural changes.
The spatial distribution of party support around the country is a crucial point for all the systems except MMP. For example, if we distribute support uniformly, a party with 51% national support will win all 120 seats under FPP. So assumptions must be made about spatial variation, and for a referendum affecting NZ we thought it much better to base the model on real data about NZ than to make yet another (possibly very ad hoc) assumption.