What’s Going to Happen With The Special Votes?

In an attempt to stave off boredom, I ran a few scenarios of where the 399ish thousand Special Votes might end up.

1. Starting position

The Electoral Commission reports:

There are an estimated 399,365 special votes outstanding, with official results due on 27 November. Elections NZ

For this exercise, I’ve assumed all 399,365 become valid party votes. In reality, some will be disallowed or informal.

I’ve also held all 71 electorate winners unchanged, allowing us to isolate the party-vote effects.

2. Results: all six scenarios

Here are the resulting parliamentary seat allocations.

Historical scenarios use approximations, as explained below.

ADVERTISEMENT

The polling scenario uses the illustrative percentages discussed earlier, rather than a newly verified polling average.

What happens to the parliamentary majority?

A margin of zero means the bloc holds exactly the number of seats required for a majority.

These are conditional parliamentary configurations, not predictions about the official result.

3. The 2017 scenarios

Your suggestion to include 2017 adds an interesting comparison.

In 2017, special votes represented approximately 17% of votes cast. National lost two seats between the preliminary and official counts, while Labour and the Greens each gained one. No electorate winners changed. Elections NZ

I’ve modelled two different interpretations.

2017 distribution: Use the approximate party shares among votes added between the preliminary and official counts.

2017 bias: Apply the approximate difference between special-vote and ordinary-vote preferences to the 2026 preliminary party shares.

These produce quite different results.

The distribution scenario gives Labour 38 seats and the governing bloc 65 in a 130-seat Parliament.

The bias scenario instead gives the Greens 17 seats and the governing bloc 67 in a 132-seat Parliament.

This demonstrates why the distinction matters. Repeating the absolute distribution of votes from an election in which National received 44.4% overall is very different from repeating the relative behaviour of special voters.

For comparing historical voting behaviour with 2026, the bias approach isolates a more relevant feature, although neither approach establishes what this year’s special voters will do.

4. The 2023 scenarios

The same distinction appears in 2023.

That election had 603,257 special votes, representing 20.9% of votes cast. National lost two seats between the preliminary and official results, while four electorate winners changed. Elections NZ

Using the approximate 2023 distribution produces:

The existing coalition holds 66 seats, with 65 required for a majority.

Applying the 2023 bias instead gives the Greens an additional seat, with NZ First retaining 12.

That produces 131 MPs and a 67-seat governing bloc.

An important limitation is that the 2023 bias model must be normalised after adjusting each party’s share. The provisional calculation does that, but the precise normalisation will need revisiting once we have the actual special-vote distribution.

5. The split-vote strategy

This is particularly interesting because National’s electorate overhang changes the effect of party votes moving between coalition partners.

I modelled three hypothetical transfers from National’s preliminary special-vote share, divided equally between ACT and NZ First.

All other parties retain approximately their preliminary shares.

Resulting seat allocations

Notice what happens in the strong scenario.

National remains at 42 electorate MPs despite losing proportional entitlement. ACT gains one seat, NZ First gains one, and Labour loses one.

The total Parliament is 132, with the coalition holding 69 seats.

This is a concrete illustration of the overhang effect we discussed earlier.

It also demonstrates that a shift of party votes from National to ACT and NZ First can increase the coalition’s actual parliamentary representation under these conditions.

However, the model doesn’t establish whether such a transfer is occurring among special voters.

6. What about Opportunity?

Across these particular scenarios, Opportunity remains at seven seats.

That’s significant for the arithmetic because the party sits outside both of the three-party blocs we’ve been comparing.

In the provisional 2017-distribution scenario, for example:

• National + ACT + NZ First: 65 seats.
• Labour + Greens + TPM: 58 seats.
• Opportunity: 7 seats.
• Parliament: 130 seats.
• Majority: 66 seats.

Neither three-party bloc has a majority independently.

Opportunity’s seven MPs could therefore be relevant to confidence-and-supply arrangements, although the arithmetic doesn’t establish which agreements would be politically acceptable.

7. Two corrections before we treat this as a finished model

I want to distinguish what we’ve established mathematically from what still needs verification.

First, the historical special-vote distributions. The Electoral Commission provides separate special-declaration party-vote data for both elections:

• 2017 election statistics
• 2023 election statistics

Those are the datasets we should ultimately use. The preliminary-to-official differences are not exact substitutes, particularly for 2017.

Second, the polling scenario. The 28.2% National, 28.5% Labour and other polling figures are assumptions carried forward from our earlier discussion. I haven’t independently established that they represent a valid average of the latest 2026 polls.

Consequently, the seat allocations above are reproducible results of the specified inputs, but the historical and polling inputs still need validation.

8. What stands out

Three mathematical features are worth carrying forward into our next round of modelling.

National’s 42 electorate MPs are the dominant constraint. In every scenario tested, National remains above its proportional entitlement. Consequently, changes in its party-vote allocation affect its overhang rather than its actual number of MPs.

The split-vote scenarios change the relationship between proportional seats and actual parliamentary seats. This is why ACT and NZ First can gain representation without an equivalent reduction in National’s electorate representation.

Electorate changes could materially alter these results. So far, we’ve assumed National retains all 42 electorates and TPM retains all five. Changing either assumption could alter both the number of MPs and the majority threshold.

Then we have some scenarios where the Special Vote is a bit more Green-heavy, due to the call from TPM for a Split Vote.

2026 Election: Green-heavy special-vote scenarios

I’ve calculated the three left-leaning scenarios using the 399,365 outstanding special votes, the preliminary party-vote totals and the modified Sainte-Laguë formula.

The results illustrate an important feature of National’s overhang: the existing coalition can lose a list seat while Parliament expands, rather than another bloc necessarily gaining that seat directly.

All calculations assume:

• All 399,365 special votes are valid party votes.
• The preliminary ordinary-vote totals remain unchanged.
• All electorate winners remain unchanged.
• Parties below 5% without electorate seats receive no proportional seats.

The starting figures are from the Electoral Commission’s 2026 preliminary results.

1. Special-vote assumptions

These are hypothetical distributions, not polling estimates.

2. Resulting parliamentary seats

Here are the actual results of the Sainte-Laguë calculations, with electorate overhangs applied.

Parliamentary majority arithmetic

The Green-heavy and Very Green-heavy scenarios both produce a Parliament in which the existing three-party coalition has fewer seats than the majority threshold.

That doesn’t automatically determine who can form a government. It establishes the parliamentary arithmetic under those assumptions.

3. What’s happening to the overhang?

This is the interesting part.

National’s proportional entitlement changes, even though its actual representation stays at 42 electorate MPs.

There is a subtlety here: in the moderate scenario, Parliament grows to 132 even though National’s overhang remains seven.

That’s because TPM’s proportional entitlement falls from one to zero in the calculation, increasing its overhang from four to five.

Correction to the table above: TPM’s overhang is five, not four, in all three hypothetical scenarios. That produces the parliamentary totals shown.

The complete overhang calculation is therefore:

This is precisely why we need to calculate proportional entitlements before applying electorate seats.

4. What about the Greens + Opportunity + NZ First combination?

Andrew Geddis identified a particularly interesting combination: the Greens and Opportunity each gaining a seat while NZ First loses one.

His analysis identifies approximate final nationwide party-vote thresholds of:

These are approximate thresholds calculated with other parties’ votes held in a particular distribution. They aren’t independent fixed thresholds that apply to every possible scenario.

Our Green-heavy scenario produces two of those changes:

• Greens gain two seats.
• NZ First loses one.
• Opportunity remains at seven.

So let’s modify the distribution to test Opportunity gaining a seat too.

Scenario H — Greens and Opportunity both gain

For this test, I’ve transferred one percentage point of special votes from the Greens to Opportunity, relative to our Green-heavy scenario.

Result:

The parliamentary arithmetic becomes:

Neither three-party bloc has a majority independently.

Opportunity’s eight MPs would therefore be relevant to any confidence-and-supply arrangement.

5. What this tells us about the special votes

The Electoral Commission estimates that 399,365 special votes remain to be counted, including 96,947 overseas and dictation votes. Elections NZ

There is a particularly important difference between 2026 and previous elections: changes to enrolment rules mean the composition of special voters is different from 2023.

Geddis notes that approximately 200,000 fewer special votes are expected this time, while overseas votes have increased.

That matters when comparing our historical scenarios.

A repeat of the 2017 or 2023 special-vote distribution isn’t necessarily equivalent to a repeat of the political preferences of the people who would have cast special votes under the old rules.