Evaluating Cash for Clunkers: The Case of the Missing Denominator

by Tom Bozzo

An AP story (via Felix Salmon) based on an analysis of “Cash for Clunkers” transactions is circulating with the non-news that a number of the transactions involved trades of gas-guzzling trucks for only modestely less gas-guzzling trucks. Here’s the lede:

The most common deals under the government’s $3 billion Cash for Clunkers program, aimed at putting more fuel-efficient cars on the road, replaced old Ford or Chevrolet pickups with new ones that got only marginally better gas mileage, according to an analysis of new federal data by The Associated Press…

The single most common swap — which occurred more than 8,200 times — involved Ford F-150 pickup owners who… trade[d] their old trucks for new Ford F-150s. They were 17 times more likely to buy a new F-150 than, say, a Toyota Prius….

Whoa now. It isn’t until the 6th paragraph that readers are told the analysis involves more than 677,000 transactions, so the most common swap accounts for just 1.2 percent of cash-for-clunkers transactions. The seventh paragraph finally deigns to mention the average fuel economy statistics: 15.8 mpg for the traded clunkers, 24.9 mpg for the purchased vehicles — if you invert the mpgs, that’s a 36.5% reduction in fuel consumption at the means.

Salmon for whatever reason reckons this makes cash-for-clunkers the administration’s worst policy initiative yet, which brings a lolcritter to mind. Even William Buiter granted in an otherwise cranky post that, under conditions clearly met by the program (i.e. a temporary incentive that’s big enough), the programs are “bound to work” as Keynesian stimulus which arguably the administration hasn’t pursued in sufficient quantity. [*]

More below the jump.

None of this is news because the ability to trade old trucks for slightly more fuel-efficient new trucks was a widely reported feature of the enacted program design, which was biased towards the stimulus end of things, given the truck-heavy product lines of the domestic three. Participants could get $3,500 for a 1 mpg improvement in a light truck-for-light truck trade. Still, program participants generally used the credits for something better than 2 mpg improvements. Here’s the full distribution of differences between the old- and new-vehicle MPGs as reported in the CARS public transaction database:

mileage_diff | Freq. Percent Cum.
————+———————————–
-3 | 4 0.00 0.00
-2 | 5 0.00 0.00
-1 | 20 0.00 0.00
0 | 116 0.02 0.02
1 | 9,961 1.49 1.51
2 | 37,486 5.61 7.12
3 | 31,130 4.66 11.78
4 | 36,584 5.47 17.25
5 | 50,526 7.56 24.81
6 | 56,730 8.49 33.30
7 | 50,187 7.51 40.80
8 | 49,397 7.39 48.19
9 | 41,127 6.15 54.35
10 | 62,020 9.28 63.62
11 | 71,757 10.74 74.36
12 | 47,101 7.05 81.41
13 | 38,482 5.76 87.16
14 | 26,377 3.95 91.11
15 | 16,205 2.42 93.53
16 | 10,801 1.62 95.15
17 | 5,433 0.81 95.96
18 | 3,019 0.45 96.41
19 | 1,478 0.22 96.63
20 | 896 0.13 96.77
21 | 1,530 0.23 97.00
22 | 475 0.07 97.07
23 | 1,576 0.24 97.30
24 | 1,016 0.15 97.46
25 | 870 0.13 97.59
26 | 717 0.11 97.69
27 | 361 0.05 97.75
28 | 333 0.05 97.80
29 | 141 0.02 97.82
30 | 95 0.01 97.83
31 | 64 0.01 97.84
32 | 5,504 0.82 98.67
33 | 1,965 0.29 98.96
34 | 2,401 0.36 99.32
35 | 2,276 0.34 99.66
36 | 1,223 0.18 99.84
37 | 680 0.10 99.94
38 | 303 0.05 99.99
39 | 60 0.01 100.00
40 | 7 0.00 100.00
————+———————————–
Total | 668,439 100.00

The distribution is a bit bimodal, with peaks at 6 and 11 mpg improvements; the latter is the most common swap in mpg improvement terms, or to borrow the AP article’s reporting style, that’s around 9 times more common than F-150 for F-150 trades. Fuel economy improvements of more than 30 mpg over the trade-in vehicle (roughly, trading a very large truck for a Prius) are more common than F-150 for F-150 trades. The 1 mpg minimum improvement likewise was not overly common at 1.51 percent of transactions.

In addition, fuel consumption reductions and stimulus aren’t the only potential benefits from the program. While current emissions regulations hold light trucks to the same standard as cars for regulated pollutants (as of 2009, in fact, though notably not [yet] including carbon dioxide), trucks previously were allowed to meet lower emissions standards. All cars and light trucks have been held to higher standards since the mid-1990s as stricter emissions regulations have phased in. Indeed, as the following table shows, fully 30 percent of trades predate the 1994 start of the phase-in for the Tier 1 emissions regulations. New Tier 2 vehicles are much cleaner than pre-MY2003 vehicles, which account for 99 percent of Clunkers trades.

Trade_in_Year | Freq. Percent Cum.
————+———————————–
1984 | 1,835 0.27 0.27
1985 | 9,703 1.43 1.70
1986 | 10,771 1.59 3.29
1987 | 12,171 1.80 5.09
1988 | 18,537 2.74 7.83
1989 | 23,826 3.52 11.35
1990 | 23,993 3.54 14.89
1991 | 28,036 4.14 19.03
1992 | 32,248 4.76 23.80
1993 | 41,652 6.15 29.95
1994 | 57,145 8.44 38.39
1995 | 66,363 9.80 48.19
1996 | 61,430 9.07 57.26
1997 | 65,955 9.74 67.00
1998 | 65,876 9.73 76.73
1999 | 62,441 9.22 85.95
2000 | 43,758 6.46 92.42
2001 | 29,183 4.31 96.73
2002 | 16,114 2.38 99.11
2003 | 4,472 0.66 99.77
2004 | 1,226 0.18 99.95
2005 | 254 0.04 99.99
2006 | 52 0.01 99.99
2007 | 23 0.00 100.00
2008 | 17 0.00 100.00
————+———————————–
Total | 677,081 100.00

There’s also a cautionary lesson in here investigating large databases. In a 677,000-record dataset, there will be hundreds or thousands of errors unless the data are meticulously cleaned. The AP report takes note of the 145 transactions listed above where the mileage of the trade was no worse than that of the new vehicle, which could indicate frauds against the program if correct. Among the putative horror stories:

A driver in Negaunee, Mich., traded a 1987 Suburban that got 18 mpg for $3,500 toward a new Silverado pickup that got only 15 mpg. An Indianapolis driver traded a 1985 Mercedes 190 that got 27 mpg for $3,500 toward a new Volkswagen Rabbit that got only 24 mpg.

The article is correct based on the mpg for the trades as coded in the database. However, other information in the database suggests that the transactions actually may have involved valid fuel economy improvements. The database includes the vehicle identification number (VIN) of the trade, which provides important eligibility information on the cars when decoded.

In the case of the 1987 Suburban, to get 18 mpg it would need to be a RWD model equipped with a 6.2-liter diesel engine. That’s what the database record says, but according to the VIN, the Suburban was a 13 mpg gas-engined model. Likewise, the VIN of the 1985 Mercedes is consistent with the car being an eligible 18 mpg gas-engined 190 rather than a 27 mpg diesel 190. In both cases, the database is ambiguous as to the validity of the transactions; if the VINs are correct, then they were allowable.

The Cash for Clunkers program pretty clearly was not designed as an efficient environmental program, and there have been questions as to its efficiency as stimulus, too. However, as its costs are spread over aims of stimulus, reducing fuel consumption (which has costs external to car owners), and reducing emissions pollutants other than fuel consumption-related GHGs, it doesn’t have to be. To call it the Worst Obama Program Ever is to go overboard.

[*] Buiter, who takes the line that it’s crazy per se to destroy a portion of the capital stock, doesn’t carefully consider that automobiles don’t just magically produce transportation services; they consume other inputs and emit various wastes doing so.