RODNEY HIDE: Thirty-one thousand votes

Tomorrow's election comes down to about 31,000 votes.

That's how many Opportunity voters would need to stay home or vote for someone else to push the party under five per cent. Do that and the centre-right governs. Don't, and Opportunity decides who does.

Thirty-one thousand is about one party vote in every 94.

That's not a hunch. It's what my model says this morning. Here's how it works, and why the overhang matters more than anyone's letting on.

How our poll works

I take every published party-vote poll since the 2023 election, 134 of them, from 1 News-Verian, RNZ-Reid Research, Curia, Roy Morgan, Freshwater, Talbot Mills and the rest. Nothing is made up.

Recent polls count for more. A poll's weight halves every four weeks. Bigger samples count a bit more. Every pollster has a lean, and I measure it and take most of it out.

The result: Labour 28.6 per cent, National 28.0, Greens 14.8, NZ First 10.2, ACT 10.1, Opportunity 6.1, the Māori Party 1.3.

Then the electorates. I take each seat's 2023 candidate vote and apply each party's swing since, the way Stuff's Flip Forecaster does. Where there's a published electorate poll, as in the Māori seats, I use it instead.

Then I hand out seats exactly as the Electoral Commission does. I checked it against 2023. It reproduces that Parliament seat for seat.

The overhang

Under MMP your party vote sets each party's share of 120 seats. Electorate wins come out of that share. But if a party wins more electorates than its party vote earns, it keeps them all. The extras are the overhang, and Parliament grows to fit them.

In my numbers National wins 35 electorates but its 28 per cent earns only 34 seats. That's one overhang seat.

The Māori Party wins four Māori seats on a party vote that earns two. That's two more. One of those four, Tāmaki Makaurau, rests on a two-point lead in a poll taken before Labour's candidate quit.

So Parliament has 123 seats and a majority is 62.

Here's the count. National 35, ACT 12, NZ First 12: 59. Labour 35, Greens 18, Māori Party 4: 57. Opportunity 7.

Neither side gets to 62. Opportunity picks. Add it to the left and you have 64.

The overhang cuts against the right. Every extra seat lifts the bar. The Māori Party's two hand the left seats its party vote never earned, and the right has to clear a higher line because of them.

Thirty-one thousand

Opportunity is on 6.06 per cent. The line is five. The gap is 1.06 points. On a turnout like 2023's, call it 2.9 million party votes, that's about 31,000 people.

Fall under five per cent without an electorate and you get nothing. Opportunity's only shot at a seat is Mt Albert, and Labour holds it on our numbers. Its votes would be wasted and its seven seats shared among everyone else.

Parliament drops to 122. The majority stays at 62. And the right gets there.

If those 31,000 voters stay home, the right has 62. If they go to National, ACT or NZ First, it's 63. Only if every one of them votes Labour does it end in a 61-all tie.

The overhang shrinks too. With Opportunity gone, National's party vote earns 36 seats, more than its 35 electorates. Its overhang disappears.

Will it happen? Opportunity's last five polls: 6, 7.2, 6.2, 5.1 and 5.8. Freshwater had it a whisker above the line. My model's own dice give it a 96 per cent chance of getting in. But a poll of 1,000 has a margin of error of about one and a half points on a party at six per cent. That's bigger than the gap.

What the right needs if Opportunity survives

Then National, ACT and NZ First need three more seats. How many votes that takes depends on where they come from.

Fresh votes, from people who'd otherwise not vote, are dearest. Extra party votes for NZ First or ACT buy list seats: about 59,000 for NZ First, or about 63,000 for ACT. National is already in overhang, so more National party votes don't unlock list MPs. Spread fresh party votes across NZ First and ACT in proportion to their present support, it's about 61,000; National's path is separate and seat by seat.

Here's the overhang trap. Give ACT 63,000 more votes and it wins three more seats. But that also cuts National's party-vote share, so its entitlement drops to 33 while it keeps 37 electorates. The overhang grows to six. Parliament swells to 126 and the majority line climbs to 64.

The right gains five seats and the bar goes up two. It scrapes in with 64, exactly the number needed.

National's own extra votes don't win it list seats. They win it marginal electorates. Northland is the prime example: our model has Labour taking it by 0.2 points. National's gains come seat by seat, not off the list.

Switchers are cheaper, because each one takes a seat from the other side as well as adding one. About 28,000 Labour voters switching to NZ First does it. About 31,000 switching to ACT. About 57,000 switching to National in the key electorates.

One in a hundred

So there it is. About 31,000 Opportunity voters finding something better to do. Or about 28,000 Labour voters giving Winston a go. Either way, roughly one party vote in a hundred decides who governs.

The polls can't see that far. Their error is bigger than the gap. Nobody knows how this ends. Not the pundits, not the pollsters, not me.

They always say every vote counts. It's usually rubbish.

Tomorrow it's true.

Rodney Hide is a former Minister and ACT Party leader