Tyler Cowen notes some evidence that astronomers have been underestimating the frequency of catastrophic asteroid impact events here on the planet earth. Depending on how you interpret a certain set of deposits, it's possible that super-huge impacts occur once every few thousand years instead of once every 500,000 years or so.

At any rate, it sounds dumb, but I really do think the world's major governments should pony up the money that would be required to better track the paths of asteroids and the like. Right now, we don't really have a good sense of where everything is all the time. Building the necessary monitoring capacity would be pretty cheap if you put it aside other kinds of national defense expenditures, and it would be something all the big players and wannabe players (USA, EU, Japan, Russia, Brazil, India, China, whomever else) could do cooperatively. Whatever the exact frequency, the question of an extremely destructive collision with an object in outer space is very literally a "when not if" sort of thing. Given ample warning, though, people could probably figure out some countermeasures. Keep in mind that beyond the truly catastrophic impact events, there are things like the Tunguska Event that could kill hundreds of thousands -- if not millions -- if they happened in a big city.