Ok. Recall that the value of a statistical life for policy purposes in New Zealand, or at least the one used by MoT for transport planning and subsequently adopted elsewhere in policy, is $3.85 million per fatality.
Let's work out whether the policy here makes sense.
The median 4-bedroom house in Redcliffs/Sumner, where most of this kind of risk obtains, is $690 per week. Let's take that as the value of housing services over and above the value of services that would be provided by a park in the same spot: the rental costs understate the value of housing services of owner-occupied properties, but parks provide some value too. Let's be safe and call it $500/week extra value. For a year, let's round that down to $25,000 per year. We want an annual figure because the 1/10,000 risk is annualised. Alternatively, we could take the value of the house and the lifetime risk of landslide for that house's life.
If a house has an intolerable risk at a 1/10,000 risk of landslide death, and if that risk is sufficient for buying out the home-owner and taking that property out of housing use, then we're willing to forego $25,000 per year in housing services to avoid a 1/10,000 risk of landslide death.
Now let's suppose that a 4-bedroom house, our valuation basis here, has 5 people in it.
$25,000 * 10,000 = $250,000,000.
$250,000,000 / 5 = $50,000,000
The Ministry of Transport is willing to spend up to about $3.85 million to save a life by roading improvements that prevent deaths.
The government here is spending at least $50 million per (ballpark) statistical life saved.
I'm not sure that this makes a lot of sense where there are other projects that, for the same total cost, could save more expected statistical lives.
Unless we think that dying in a landslide is about thirteen times worse than dying in a car accident. I'm really rather sure I'd rather die in a car accident than in a landslide. Suppose a genie came to me and said, "Eric, you and your family, I know with certainty, have a 10% chance of dying in a bad car accident next year. It'll be quick though. Would you like to trade that for a 0.8% chance of your family dying in a landslide? It'll be pretty terrible."
If you were given that choice, and you'd take the deal to get the landslide instead, then the government's buyout doesn't make sense. If you prefer the car accident, then the buyout can make sense.
Update: Note too that there's an important difference between houses and roading investments. The government, in the latter case, makes investments to mitigate risk of death and accident for anyone using that road. While there may be some roads that risk-averse drivers avoid because they're too terrifying, we all have reasonable expectations of safety on government-owned roads. If I choose to buy a house at the bottom of a very unstable hillside, I have demonstrated that I'm comfortable with that risk. While it's true that the earthquakes reveal more about the actual risk and that some owners may have erred, that could be an argument for an insurance payout for the amount of capital loss, not for buying the owners out and barring future residential use. All we then need is the one-time compensation, plus a great big highlighted section at the front of the property's LIM report noting the substantial landslide risk present at the property. If some prefer taking that risk for a lower-cost house, why should that be illegal?