Edmunds has released this analysis explaining their approach on determining the cost of Cash for Clunkers.

Their approach differed from mine, but provides insight into one of the assumptions. I had assumed that 50% of the assumed gross margin derived from the Cash for Clunkers units were “marginal.” Edmunds’ PhD’s have corrected that, stating that they believe 125,000 units were marginal, which is only 18% of the 690,114 units traded in. This is a big variance that hurts the marginal return to taxpayers on their investment in the auto manufacturers.

Their approach takes the nearly $3 billion in claims filed and divides it by the 125,000 marginal units. The implied statement is that the other 565,114 units were unnecessarily subsidized by the government. Those costs, therefore, become attributable to the necessarily subsidized units, or the 125,000.

While the targeted endpoint isn’t different from my analysis, by discussing this research I realized I left out a consideration, and that is government overhead. This is something that Edmunds shares with me - neither of us considered the cost of administration. This is a material omission, as the claims were not processed and paid at no expense.

In short, by assuming only 18% of the sales were marginal, and by considering marginal costs of administration, the -34% return for taxpayers that I originally calculated is significantly understated, with the real return being much, much worse.