I agree with Becker that it's a silly program.
Like the bailout of the auto companies, the program had dual environmental and economic-recovery goals. The environmental goal, to reduce carbon emissions, was trivial; the aggregate improvement in gas mileage from the program is certain to be minuscule. The contribution to economic recovery was probably very small as well--possibly negligible. The program was one of transfer payments, not government investment. The distinction is important to Keynesian deficit spending (what is now referred to as "stimulus") as a method of fighting a depression. The idea behind such programs is to replace deficient private investment with public investment, for example, the construction of a new highway. The government hires a contractor who hires workers and by doing so increases employment, which raises incomes and therefore spending. A transfer payment does not do that, at least immediately.
It is true that people who participated in the "cash for clunkers" program couldn't pocket rather than spend the money they received from the government, as they could with the other transfer payments included in the stimulus program; they had to use it to help them buy a new car. But that is different from paying a road contractor to build a new highway. The contractor as I said has to go out and hire people to build it, so unemployment falls (on the assumption, correct with regard to construction, that there is a high rate of unemployment in the industry). The purchase of a new car merely reduces a dealer's inventory, and whether the reduction leads to new production will depend on estimates of future demand. Those estimates are likely to be inverse to the success of the "cash for clunkers" program. For, as Becker notes, the program may to a large extent merely have caused people to accelerate a previously determined intention to trade in their old car.
Timing is important; had the program been put into effect in the winter, the buying spurt that it induced might have had a bracing effect on consumer confidence. But by August the economy had sufficiently improved that the need for confidence-boosting measures that had no other effect on economic activity had waned.
Unlike Becker, I do not conclude from this unhappy episode that the Keynesian approach to fighting depression is misconceived. The problem with the $787 stimulus package that Congress enacted in February, to which the "cash for clunkers" program was a belated addition, was that it was poorly designed and has been lackadaisically executed. Roughly two-thirds of the program consists of transfer payments rather than public works, and because the Administration has failed to push the public-works components (it should have appointed an expediter to try to cut the red tape that smothers public projects), virtually all the stimulus disbursements to date have consisted of transfer payments (including, what are not really transfer payments, tax reductions that don't put cash in people's pockets until they are reflected in reductions in withholding or estimated tax payments, or in increased rebates when one files one's year-end return on April 15).
Keynesians recognize that timing is key to the success of a stimulus program in fighting an economic collapse. The stimulus program should have been enacted last fall and heavily weighted in favor of public works concentrated in areas and industries of high unemployment, with provisions for cutting red tape even at the risk of a higher incidence of fraud and waste, which are constants in government programs.