The government's Cash for Clunkers program made August an extraordinarily difficult month to predict. Analysts' forecasts for the month's annualized sales rate spanned an unprecedented 4 million range - from a low of about 12 million to a high of 16 million. Edmunds.com's forecast proved too low at about 13.2 million.

But now the final numbers are in. The August SAAR tallied up to 14.1 million.

Still, some numbers simply don't add up.

One of the difficulties in forecasting came as a result of a midstream change in the officially named Car Allowance Rebate System (CARS) program. Cash for Clunkers took off so fast that not only was the $1 billion used up in days, vehicle inventories disappeared quickly. As a result, the program's administrators changed the rules allowing dealers to take clunker trades on vehicles that were in transit or on order, instead of just inventory on the lot.

"As a result, we're still seeing clunker deals trickle in and are yet to be counted as sales. We expect that to continue for a few weeks," said Edmunds.com Senior Analyst David Tompkins, PhD. "We expect about 50,000 vehicles will be delivered and counted as clunker sales in September's sales reports."

Government's Own Numbers Don't Figure

What also threw off analysts in some forecasts was the number of voucher applications the government reported it had received from dealers under the program.

The Department of Transportation (DOT), which administers CARS, reported on Aug. 26 it had received 690,114 Cash for Clunkers voucher applications that were for either the $3,500 or $4,500. As of Sept. 2, the DOT had not changed or updated that figure.

However, the government's own data doesn't add up. In its Aug. 26 press release giving the supposed final tally for the program, the DOT gives the total application voucher number as well as a breakout of the number of vehicles purchased by category and the number of vehicles traded in by category.

None of the numbers are the same.

The overall application number of 690,114 is different from the number of clunkers the government said were turned in under the program - 685,201 - which, in turn, is different from the number of vehicles purchased - 684,941.

While the differences aren't huge nor enough to make a difference in forecasting, they should match. Presumably the number of clunkers turned in and the number of vehicles purchased should be identical, which should add up to the number of applications.



Applications Versus Sales

More problematic is that some analysts apparently counted the government's 690,114 figure as sales in their forecasts when the number is not a sales figure at all but the number of applications for vouchers. The government is still sorting through those applications. Only this week the government said it was auditing dealers to ensure the clunker deals were legitimate.

Dealers tell Edmunds.com more clunker deals are pending government approval or rejection than have been approved. Less than 10 percent of the dealers have been reimbursed the nearly $3 billion allocated for the program.

Clearly, if that 690,114 really converted to sales, then the SAAR should have been far higher than the 14.1 million in August.

Since Cash for Clunkers kicked off officially on July 24, Edmunds.com has tracked actual sales transaction data; the months leading up to the program's launch serves as the base line for comparison using the Cash for Clunkers eligibility requirements.

Daily and weekly tracking of clunker trade-ins under the CARS program as a percentage of transactions shows the program never exceeded a third of transactions and averaged 28.5 percent throughout the program. A lag time exists between a dealer filing a Cash for Clunkers application and reporting the sale to the auto manufacturer, which could account for some of the difference and will be updated in Edmunds.com's data, which is not expected to change that dramatically.

"It is mathematically impossible for there to have been nearly 700,000 new car sales during the course of the program, given the actual sales numbers announced by automakers -- who should have no motive to under-report," said Edmunds.com Senior Statistician Zhenwei Zhou, PhD. "The prominence of that number in the media helps explain why some analysts' sales forecasts were way too high."



Why Sorting the Numbers Matters

While the Cash for Clunkers program has ended, sorting the numbers matters for a number of reasons.

First, it's needed to determine if the program was successful and in what way it was successful. Did it boost the economy through higher auto sales as it was intended? Did it improve the environment by putting vehicles on the road that achieve higher fuel economy and lower emissions? If it was successful in one or both of these regards, how successful was it and at what cost to taxpayers?

From an economic standpoint, how many of the clunker trades resulted in sales that wouldn't have occurred otherwise? Or did it merely pull from behind and pull from the future naturally occurring sales? If it did produce incrementally higher sales, what was the cost per vehicle to shareholders?

Even more important, sorting Cash for Clunkers data is important for forecasting going forward. Analysts, including those at Edmunds.com, auto companies and other forecasting firms, are trying to determine what is the true underlying demand for vehicles by retail customers excluding the clunkers program and what is the trajectory of vehicles going forward. For automakers, dealers and suppliers, such information is critical for planning their businesses - for automakers scheduling factory production and ordering parts from suppliers and for dealers managing their inventories.

Discrepancy: What Consumers Bought

Throughout the Cash for Clunkers program, the list of what consumers bought from the government has not meshed with Edmunds.com's list, the first of which was posted in early August.

Edmunds.com's most recent list, representing dealer sales transaction data through Sept. 2, reveals the Ford Focus and Ford Escape continue to top the list. That's in contrast to the government's list that puts Toyota Corolla, Honda Civic and Toyota Camry, respectively, in the top slots, with the Focus ranking fourth and the Ford Escape in the No. 10 spot.

What Edmunds.com analysts discovered was that the government was cutting the data by drive configuration as the Environmental Protection Agency (EPA) does to rate fuel economy, a critical element of the clunker program. For instance, the Ford Escape comes as a front-wheel drive (FWD), all-wheel drive (AWD) and hybrid version. It appears the government counted those individually, whereas Edmunds.com counted all versions of the Ford Escape as one and the same.

However, Edmunds.com sliced its data by drivetrain configurations and it still did not clear up the discrepancy as Edmunds.com's newest data through Sept. 2 indicates.



Less Discrepancy: What Consumers Traded In

The list of top clunker traded-ins under the CARS program shows less discrepancy what they purchased between government and Edmunds.com data.
The discrepancy that exists may due in part to the lag time between a voucher application filed and the deal counted as a sales. It is also due, in part, to the fact that the government breaks down its numbers by drivetrain. For example, the CARS numbers divide the Ford Explorer four-wheel drive and the Ford Explorer rear-wheel drive whereas Edmunds.com combines the Explorer into one model.