Projection of special votes for the 2014 election

On my projection for the 2014 special votes, based on the 2011 impact, the Greens will miss out on a 14th seat by a whisker and National will retain its outright majority. Maybe.

I crunched some numbers to project (I didn’t say “predict”) the impact of special votes. I suspect most of you will just want to know the answer, so I’ll cut to the chase first, then give a bit of analysis, then give a bit more details, and I’ll talk about my method fourth.

Special votes include, among other things, overseas votes and votes cast for a different electorate to the polling place location. They aren’t counted on election night; they’re just set aside for the full count released two weeks later. And they’re not normally representative of the vote as a whole. Historically, special votes have favoured the Greens significantly—they have often picked up an extra seat from it, as they did in 2011.

The most fundamental assumption I make is that you can use the 2011 impact of the special votes as a guide to the 2014 impact. Some other minor assumptions will become apparent in the method description at the end.

Summary
Basically, on my projection, the results would stay the same. The Greens will get closer to a 14th seat, but because on the preliminary results they only just got their 13th one, special votes won’t propel them far enough for a 14th. However, they will be tantalisingly close, moving their 14th seat to the 121st quotient.

Party	Preliminary results			Projected final results			Gain/loss
	Votes	%	Seats	Vote	%	Seats
National Party	1,010,464	48.06%	61	1,136,155	47.37%	61	-0.69%
Labour Party	519,146	24.69%	32	602,304	25.11%	32	+0.42%
Green Party	210,764	10.02%	13	252,394	10.52%	13	+0.50%
New Zealand First Party	186,031	8.85%	11	204,919	8.54%	11	-0.30%
Māori Party	27,074	1.29%	2	33,134	1.38%	2	+0.09%
ACT New Zealand	14,510	0.69%	1	16,685	0.70%	1	+0.01%
United Future	4,533	0.22%	1	5,098	0.21%	1	-0.00%
Conservative	86,616	4.12%	0	94,357	3.93%	0	-0.19%
Internet MANA	26,539	1.26%	0	32,826	1.37%	0	+0.11%
Aotearoa Legalise Cannabis Party	8,539	0.41%	0	10,894	0.45%	0	+0.05%
Ban1080	4,368	0.21%	0	4,992	0.21%	0	+0.00%
Democrats for Social Credit	1,609	0.08%	0	1,983	0.08%	0	+0.01%
The Civilian Party	906	0.04%	0	1,035	0.04%	0	+0.00%
NZ Independent Coalition	895	0.04%	0	1,023	0.04%	0	+0.00%
Focus New Zealand	677	0.03%	0	774	0.03%	0	+0.00%

I project the Green gain to be 0.50%, and the National loss to be 0.69%. In a lot of cases, this can be enough to win and lose a seat, respectively: one seat is roughly 0.8% of the vote (0.83% if you ignore “wasted” votes), so it’s easy to imagine that another 0.5% might push you over.

But in this election, it wouldn’t be enough. The reason is that, in the preliminary results, the Greens only just made their 13th seat: it was the 118th quotient, i.e., the 118th seat to be allocated. National, on the other hand, has some room to slack before losing a seat—in fact, it would have picked up the 121st quotient, if there was one. In effect, the Greens would need to pick up to nearly a “whole seat” worth of votes to pick up seat number 14.

In my projection, the Greens nearly get there, but not quite. They move their 14th seat from the 127th quotient to the 121st. So if there was one more seat in Parliament, it would be theirs. And they’re very close: with 1,129 votes more (all other vote counts staying the same), they would steal the 120th quotient from National. Conversely, if National had had 5,058 votes less (all other counts staying the same), they would give up the 120th quotient to the Greens.

What does this mean?
In practice, what this tell us is not necessarily that Steffan Browning won’t make back in at Maureen Pugh’s expense. It’s that he might—but it’ll be very close. Certainly, the Greens shouldn’t be expecting another seat from specials, like they could in 2011.

There are, however, reasons to believe the wind might blow in the Greens’ favour. Firstly, advance votes were way up on previous elections. This was partly due to a concerted effort from parties to promote advance voting this election, and (I think) more so from the left than the right. Now, ordinary advance votes are counted on election night. But special advance votes—which include votes cast in advance outside the voter’s electorate—are not. And, as @annagrammatiste pointed out to me on Twitter, a lot of those special advance votes will have been cast at universities, some of which had advance voting booths on campus this year. University students, on average, lean left.

Secondly, estimated overseas votes doubled in this election, from 19,500 in 2011 to 38,500 in 2014. Overseas voters are known for being particularly Green-heavy.

Without those reasons, I would have said that it’s more likely that the seat allocations will stay put, but not enough to put money on it. But these weaken that likelihood. My instinct is that it probably won’t be enough, but objectively, I’m not really sure. It’ll be super close.

That’s basically the gist of this post. If you like voting systems, read on.

~

Some more detail
To dive in a little deeper, here’s a quick primer on how the Sainte-Laguë method works. This is the method used in New Zealand to proportionally allocate seats. The easiest way to think of it (in my opinion) is that you allocate the seats one by one, to the party with the highest “quotient” at the time. Every party’s initial quotient is their total number of votes, so the first seat effectively goes to the party with the most votes. Then, every time you allocate a seat, you divide that party’s total votes by their next divisor to get their new quotient (leaving the rest untouched). The first divisor for each party is 1, then it goes 3, 5, 7, and so on. You repeat till you’ve allocated all 120 seats.

How are quotients relevant? Well, here are the quotients near 120 for the preliminary results:

Quotient	National	Labour	Greens	NZ First	Māori	ACT	UF
118	8350.9	8240.4	8430.6	8088.3	5414.8	4836.7	4533.0
119	8350.9	8240.4	7806.1	8088.3	5414.8	4836.7	4533.0
120	8215.2	8240.4	7806.1	8088.3	5414.8	4836.7	4533.0
121	8215.2	7986.9	7806.1	8088.3	5414.8	4836.7	4533.0
122	8083.7	7986.9	7806.1	8088.3	5414.8	4836.7	4533.0
123	8083.7	7986.9	7806.1	7441.2	5414.8	4836.7	4533.0

As I said, the Greens got the 118th quotient, and won’t pick up another one until 127. National’s getting quotients more often because it has more votes, and the gap between dividing by 121 and 123 is “smaller” than between 25 and 27 (Greens), so to speak.

Here is the same for my projected results:

Quotient	National	Labour	Greens	NZ First	Māori	ACT	UF
118	9547.5	9560.4	9347.9	8909.5	6626.7	5561.7	5098.4
119	9547.5	9266.2	9347.9	8909.5	6626.7	5561.7	5098.4
120	9389.7	9266.2	9347.9	8909.5	6626.7	5561.7	5098.4
121	9237.0	9266.2	9347.9	8909.5	6626.7	5561.7	5098.4
122	9237.0	9266.2	8703.2	8909.5	6626.7	5561.7	5098.4
123	9237.0	8989.6	8703.2	8909.5	6626.7	5561.7	5098.4

The Green’s 14th quotient is just 41.8 away from National’s 61st. So we find a rudimentary “margin” for the Greens by multiplying this difference by the 14th divisor: 41.8 × 27 = 1,129. For National, we use the 61st divisor: 41.8 × 121 = 5,058.

Method
Okay, now for the exciting part. (Heh.) There are lots of ways to do this projection; Graeme Edgeler has one that basically gives the same outcome. This is how mine works.

I took the preliminary results for 2011 and subtracted them from the final results to get the “special vote addition”. I use quote marks because it’s not actually all special votes: it also includes votes cast at polling places with fewer than 6 votes (these aren’t counted on election night) and any corrections. But anyway, I compare the preliminary percentages to the “special vote addition” percentages, by division, to get a “multiplier” for each party. This multiplier represents a relationship between the preliminary votes and the special votes for each party.

I apply this multiplier to the preliminary vote percentages for each party in 2014. If a party didn’t contest the 2011 election, I just use 1 (i.e., no adjustment). Now, after doing this, the percentages won’t add up to 100, so I scale them so that they do.

The next part is the complication. The preliminary total vote count includes informals (ballots where you can’t tell who they voted for); obviously the sum of parties’ votes don’t. And the preliminary special vote estimate (I take the figure that includes overseas votes and fewer-than-6 places) includes ballots that might later be ruled invalid, because the statutory declaration wasn’t completed correctly or something like that. So I use 2011 ratios of preliminary to final counts to project how many valid special votes there will be in 2014. It doesn’t end up making much of a difference (0.4%), but hey, I wouldn’t know unless I tried.

The last step is to use the projected special vote percentages and the projected valid special vote count, to project the “special vote addition” for each party, add those to the preliminary count and that’s the projected result.

If you get excited by numbers like I do, here’s the spreadsheet I used to figure this out: Special vote impact projection 2014. The actual Sainte-Laguë stuff is in a different spreadsheet, but I’ve pasted the borderline quotients into that spreadsheet.