I also think that fully accounting for multi-generational consequences is murky/questionable and not really something we do even in much more obvious cases: Eligible people deciding against having children are not punished for depriving future society of centuries of expected workyears, and neither are mothers/fathers rewarded for the reverse.
But even if you accounted for losing 3 full generations and some change (for biodiversity loss), that still leaves you in the ~$200M/year range.
Currently we don't have reliable asteroid deflection capability at any price (but it would be technically somewhat in reach), but just imagine a future NASA budget discussion that goes "we're gonna have to mothball our asteroid deflector 3000 because it eats 5% of yearly NASA budget and thats just not worth it"-- that could be the mathematically correct choice, which confounds me.
All the math assumes that the probabilities will follow historic trends and is relatively static. With single digit events, we really have no way in knowing what the actual likelihood of impact is. It could be 1 in 100 million, it could actually be 1 in 1 million and we've been rolling a bunch of nat 20s.
Before we build out the asteroid blaster 9000, the first step is detection. With that in place then we get actual good risk and probability calculations. If the detector tells us "There's no object that will strike earth in the next 1000 years" we can safely not put any budget into asteroid defense. If, on the other hand, the detector shows "Chicxulub 2.0 will hit in the next 100 years" then your probability of an impact is 1 and the actual budget worth it is going to be much closer to that $8e+16 number calculated earlier.
I also believe the approximate bounds we have on impact probability are good enough for this estimate and quite unlikely to be off by a factor of 100, because we can guess at both size distribution and impact likelihood from craters (on earth and moon), and if the >10km object impact likelihood was over 1/million years we would expect to see a hundred times more craters of the corresponding size...
We already have 10s of years of certainty with the current observations. Most of the uncertainty comes from the interactions of unknown objects. As the mappings of objects increase, our predictions will become much better.
The other thing to consider is that large objects will have much better certainty. A 10km asteroid won't be influenced (much) by colliding with 100 1m asteroids. It will only be impacted if it hits or swings by something like a 1km asteroid.
Rubin should in a pretty short timeframe (a few years) give us an orbital mapping of all the >1km asteroids, which is pretty exciting.
We can already say that we have very high completion of cataloguing near-Earth objects that are anywhere near extinction-event / Chicxulub-sized (~10km), and have a majority of catastrophic / country-killer (~1km), and are digging deeper and deeper into regional / city-killer (~100m) bodies.
What we don't have is comets. Comets on long period orbits just aren't readily detectable with this sort of survey unless they're quite close in to the Sun, and I don't think we have great statistics on frequency vs size, size being something that requires very specific radar cross-checking to establish with any confidence. A long-period comet or hyperbolic body has a potential impact velocity much higher than inner system asteroids, and impact energy scales with impact velocity squared.
You can fight this a bit by working in the thermal infrared, which you really need a specific sort of space telescope for. But long-period comets and hyperbolic impactors will be a probabilistic threat for the foreseeable future. I would say "Be thankful that they're so rare", but the data from observatories like Rubin on these bodies during points of their orbit where they're close enough to the sun to actually detect, is necessary to statistically characterize their existence with any confidence.
To be honest, I think the 1km diameter range might still be a major fraction of the actual risk, because the estimates around "human exctinction every 100Ma" are probably much too pessimistic.
So you could actually make an argument that to a country like the US, full 100% reliable asteroid protection is only worth like $50M/year (even if an impact means full extinction)?
So if upkeep for a detection/deflection system costs more than that we'd be "better off" just risking it?! Thats insane. I would have expected this number to be much higher than $50M/year.