Christopher R. Knittel of UC Davis has a paper (pdf) looking at likely CO2 reductions from cash-for-clunkers under a variety of assumptions. The $416 figure per ton of CO2 avoided may actually be low, as it does not include the well-documented rebound effect of people with higher MPG cars driving more miles**. Also, he admittedly assumes that cars being turned in will have average future driving miles for a car of similar age, though there is anecdotal evidence that in fact the cars being turned in are driven less than average. Under these assumptions, the cost may be as high as $600-$1000 per ton.
The analysis looks pretty thoughtful, with the proviso (which the author is the first to make) that data on the program and cars bought/turned-in is still sketchy. The interesting part was that there were no reasonable assumptions that even got the price within an order of magnitude of the $28 per ton clearing price the CBO estimates under cap-and-trade.
As a CO2 reduction program, this is the equivalent of the military's $700 toilet seats. But of course we all know that no one ever really considered this an environmental or even stimulus bill. This was always first and foremost 1) another Easter egg subsidy for the middle class and 2) a back door way to subsidize GM and Chrysler to try to make the Administration's investment in them look better.
** This is straight supply and demand -- reduce the cost of miles driven, and people will drive more miles.