I did find this while browsing around some people's blogs:
http://wtfnoway.blogspot.com/
Search for Variance, there's an entry about the differences between sit-n-gos and cash games.
As far as I could tell, it basically boils down to "variance at SNGs depends only on your winrate, while variance at cash depends on a lot of other factors, because you can lose partial stacks".
While it is true that variance at SNGs depends purely on winrate, that winrate depends on all of the same things that your winrate in a cash game does.
A tournament is a series of heads-up hands, just like a cash session. And since there are no chipEV/EV differences in a heads-up tournament (and people almost always overestimate those differences even in situations where they are fairly significant), plays that are correct on a cash table should also be correct in a tournament.
The only differences are that you can't rebuy, and that the blinds increase over time, but both of these things mean you're making higher variance plays with smaller effective stacks, which would tend to lead me to the conclusion that HUCASH is *lower* variance than HUSNGs, not higher.
Maybe the difference is that there's sort of an auto step-down mechanism built into tournament play, because you can't rebuy?
On a cash table, if you lose half your chips (or take half the other player's chips), there's probably going to be a rebuy, and further hands are again risking a full stack instead of just half a stack. In a tournament, you're basically now playing the equivalent of a lower buyin where you have double the max buyin.
Only not really...because of the blinds, it's more like playing a higher buyin but buying in short (while your opponent buys in even shorter), which is, again, a higher variance situation than having 100BBs at lower stakes. So it's not really an auto-stepdown mechanism, but an auto-stepup into a small effective stack situation.
Obviously I'm either very confused, or conventional wisdom really isn't all that wise.