Election 2026: Has vote splitting saved a battered coalition?

Analysis: Pretty early on election night, it became obvious that a handful of electorate races may well decide who has the best shot at forming a government.In an MMP system where the party vote has always decided the make-up of Parliament, this is remarkable. It has thrust a handful of electorate races previously of passing national interest into a potentially decisive role for deciding the government.The coalition probably lives or dies by whether or not National can cling on to - or claw back - a handful of tight seats.On the election night numbers, it is very vulnerable in Mount Roskill, quite vulnerable in Hamilton East, and vaguely vulnerable in Whangārei. It has a good chance in Wairarapa and some dim hope in Hutt South.This all adds up to a pretty reasonable chance at retaining the Beehive for a party that hemorrhaged about 9 percentage points of party vote support.While it will shed five list MPs, it will likely end up losing only a net one or two, or even no electorate seats. Labour gains one, none or two electorate seats. ACT and Te Pāti Māori have lost one each, the Greens unchanged with three.How did this happen? An unpopular Prime Minister, an election night result in the 20s and four of the party's most senior MPs - including its deputy leader - dumped out of Parliament. And yet, its electorate tally is practically unscathed.Two of those four banished senior MPs alluded to part of the reason: a vote splitting campaign National did not endorse, and still does not like but that ultimately appears to have saved its bacon.National voters were urged by some prominent people to split their ticket by giving their party vote to ACT or NZ First and candidate vote to National.The aim was to grow National's overhang and boost its coalition partners' seats. It appears to have worked.All across the country, in seats it's on track to win, National Party candidates outperformed the party vote by large margins - 17.2 points was the median gap.These numbers don't lie: a large number of people ticked National for their candidate vote and a different party with their other vote. ACT and NZ First added roughly six percentage points to their tallies from 2023 between them, so you don't have to look far to see where most of National's support has gone.This will really hurt National and its leader Christopher Luxon. If special votes don't take it out of his hands, Luxon will enter coalition talks with unquestionably the weakest hand a would-be prime minister has ever had. He and his party are severely diminished, whatever happens from here.No wonder the outgoing Speaker and National Party grandee Gerry Brownlee was unhappy."It's like kids playing with matches," Brownlee said of the vote splitting campaigners.There was "no doubt" that ACT had "benefited a little bit from that vote splitting campaign," he said.Paul Goldsmith, beaten by ACT leader David Seymour in Epsom and now out of Parliament, was even more direct. "A lot of confusion was sowed" by the vote splitting campaign "which didn't help," he said.There were no such quibbles for Te Pāti Māori who nailed their electorate-only campaign extremely successfully, adding another four seats to a huge 11 seat overhang.Paul Goldsmith is among National's casualties.RNZ / Samuel RillstoneAnother kind of vote splittingThe suddenly outsized significance of the electorate seats puts into sharp relief a completely different - and unrelated - kind of vote splitting.This happens in every election: a National or Labour Party candidate loses a winnable race where there are a good number of votes going to candidates from parties aligned with their own. It can be really annoying for those candidates and their supporters but it doesn't matter to the overall make-up of Parliament because the party vote has always decided that.This time, the 'what-ifs' could have been - and may yet still be - much louder.Mount Roskill is the worst example of this problem for Labour. National's Carlos Cheung had only a weak ACT candidate nibbling away a few votes from his column in this race. Labour's Michael Wood had a much more meaningful bite taken out of his vote by the Green candidate, and he probably lost a few to the Opportunity one as well. If even a quarter of the votes that went to the Green candidate had gone to Wood, he'd be much more comfortable this morning. As it stands, he could still lose.Labour's Michael WoodNelson is interesting too. You can bet the 2952 people who voted for the Green candidate wanted a change of government. If they'd given those votes to Labour's Rachel Boyack instead, she would have gone ahead of National's Blair Cameron and we'd be looking at a hung Parliament.National suffered this affliction too, albeit to a lesser extent. In Whangārei, there was no Green candidate so almost all of the left-leaning vote went to Labour's candidate Gary Payinda. There were, however, ACT and NZ First candidates who got a lot of support between them, leaving National's Lloyd Budd in a tight race with Payinda for the seat.And Wairarapa is too close to call. If Ron Mark hadn't been on the ballot for NZ First, maybe it wouldn't have been.The MMP features of an old school electionThis unquestionably was, as National's defeated Chris Bishop said, a very weird election. The importance of the electorate races gave it the look of an old-school FPP race. But in other ways, it was very MMP.The parties formerly known as 'minor' had their most successful election ever. Opportunity came from way back to surge over the 5 percent threshold and make it into Parliament. ACT, Green and NZ First all grew their vote.NZ First and the Greens have both achieved more than 10 percent before, but it's the presence of four smaller parties over the 5 per cent threshold that stands out.This was only possible because Labour and National both scored in the 20s, and there was very little vote siphoned off to parties that didn't make 5 percent.How we compared electorates:Electorate boundaries were redrawn in 2025, so some 2026 seats don't match a 2023 seat. We used 2023 Census population counts for small neighbourhood areas (Stats NZ statistical area 1) to work out where each 2026 seat's residents lived under the old boundaries.If at least two-thirds of a seat's residents were in one 2023 electorate, we compare it with that electorate. This includes five renamed seats:- East Cape (was East Coast)- Mount Maunganui (was Bay of Plenty)- Ōtāhuhu (was Panmure-Ōtāhuhu)- Wellington Bays (was Rongotai)- Wellington North (was Wellington Central)Six seats didn't meet that test, so we treat them as new and don't compare them with 2023: Glendene, Henderson, Kapiti, Kenepuru, Rangitīkei and Waitākere. Rangitīkei kept its name, but 39 percent of its residents were previously in Ōtaki.Four seats were close to our test's two-thirds. In Mt Albert, Mt Roskill, Botany and Wellington North, 70 to 72 percent of residents were in the same electorate in 2023. We still compare them, but some of the change there may come from the new boundaries.Port Waikato has no 2023 comparison. A candidate died during the 2023 campaign, so the seat was decided at a by-election, not on election day.