Hoisted from Comments: Frank:
Grasping Reality with Both Hands: The Semi-Daily Journal Economist Brad DeLong: impact rates based on the cratering record are pretty much in line with estimates based on the population of Earth-crossing asteroids. Size distributions roughly fit an inverse square law distribution (i.e., 2 x diameter = 1/4 probability; 1/2 diameter = 4 x probability). Using this one can calculate impact rates based on a probability of a 500 m impactor every 100,000 years. So on average, 50 m projectiles should impact every 10,000 years or so (10 x 10 times more frequent). Once you get down to about 50 m the probability of the projectile exploding in the atmosphere (like Tungusaka) is quite high. Much smaller events might be like nuclear explosions, but they are rare and very unlikely to hit populated areas.
More to the point of the Easterbrook article, I've searched the Web of Science publication database for papers by the chief protagonist in his article. DH Abbot has not published a significant paper in 5 years and has never published anything other than unreviewed abstracts on this subject that I can find. Looks like a squeeky wheel getting Easterbrook's attention, but no follow-up to credible experts in this field.
And then there is the cross-section problem: we have what? 6000 cities each of roughly 100 square miles = 600,000 square miles of devastating impact cross section in a world of 200M square miles. That means only 1 out of 400 50M impacts will be "devastating" if we say that a 50 meter meteorite--2 megatons, Barringer crater-sized--hitting a city is our threshold for "devastating."
So we are down to one devastating every 4,000,000 years--not the one every thousand years of the Atlantic Monthly's lead to Gregg Easterbrook's article.