That is how it works: it's regression to the mean. If you expect a major hurricane every n years and you didn't have one this year, the expected likelihood of one next year (1/n) is greater than what you actually got this year (0).
If the non-hurricane damage is independent, then you expect $(1/n * cost of major hurricane) additional costs next year. In a major hurricane year, you expect (1-n)/n fewer such hurricanes next year. And therefore you expect less expensive damage.
One could debate whether "likely to be worse" is correct. If there's a a major hurricane every 10 years, there's only a 10% chance of it happening next year. The expected value of the damage is higher, but it's still "unlikely" (as in, less than 50%) to happen at all. So it's expected to be worse, but unless there's over 50% chance of a major hurricane per year, it's not likely to be worse.
If year-to-year hurricanes are also independent, then knowing that there was or was not a hurricane this year, it's also no more or less likely that a major hurricane occurs next year then it was looking forward to this year, as of this time last year. But, I don't know if that is an independent thing.
> That is how it works: it's regression to the mean. If you expect a major hurricane every n years and you didn't have one this year, the expected likelihood of one next year (1/n) is greater than what you actually got this year (0).
Not if hurricanes are independently and identically distributed. Which I think you acknowledge could be the case in your last paragraph, but I have not ever come across anything linking hurricane to one another.
No, it's not a gambler's fallacy. You know you got zero this year, it already happened. You know it's expected 1/n next year (assuming independence, etc), just like every year. So the expected amount of hurricanes next year is more than what you did get this year, which was 0. In total you now (as of now, year 1 end) expect only 0 + 1/n total hurricanes over the two years (this year and next). That is 1/2n per year: less that you would have predicted at the start of this year, which would have been 2/n over two years for 1/n per year.
A gambler's fallacy is thinking getting 0 this year means it's more likely than 1/n next year to act to retroactively "correct" 0 this year.
If the non-hurricane damage is independent, then you expect $(1/n * cost of major hurricane) additional costs next year. In a major hurricane year, you expect (1-n)/n fewer such hurricanes next year. And therefore you expect less expensive damage.
One could debate whether "likely to be worse" is correct. If there's a a major hurricane every 10 years, there's only a 10% chance of it happening next year. The expected value of the damage is higher, but it's still "unlikely" (as in, less than 50%) to happen at all. So it's expected to be worse, but unless there's over 50% chance of a major hurricane per year, it's not likely to be worse.
If year-to-year hurricanes are also independent, then knowing that there was or was not a hurricane this year, it's also no more or less likely that a major hurricane occurs next year then it was looking forward to this year, as of this time last year. But, I don't know if that is an independent thing.