What the economists found (free but complicated download for researchers) was:
With per vehicle environmental benefits at $596 and the costs at $2,600 per vehicle, the clunker program is a net drain on society of roughly $2,000 per vehicle. Given the approximately 700,000 vehicles in the program, we estimate the total welfare loss to be about $1.4 billion. The welfare loss would be even greater if we fine tuned our estimate of the social cost per gallon to account for the spatial mix of clunkers. Clunkers, especially the trucks that comprise a large percentage of the traded-in vehicles, may have been retired disproportionately from rural locations where the social costs of pollutants are significantly lower. Also, if the average value of clunkers exceeds our conservative figure of $1000, then cost of the program would be higher. Even if the environmental gains were double our estimate, the net drain would still be close to $1 billion. While a more rigorous analysis would no doubt adjust these figures, we doubt that the basic conclusion would change.
It is plausible that a much smaller, very well-designed program might have been a net benefit, but this verdict seems unassailable to me. I imagine that the costs to society in terms of damage to the used-car and spare-parts industries caused by the destruction of the vehicles concerned may result in the actual value of clunkers being more than $1,000.