e d m a n * Smog c h e c k data o n 307 of these vehicles.Of these, 126 (41%) showed deliberate tampering, and another 77 (25%) had defective or ~nissi~lg equiptnent that tnay not have been the result of tampering (for ex-I n response to continuing pollution prob-lected 91,679 records with valid H C and ample, missing air pump belts). Overall, 282 lems, in 1990 Congress passed the Clean CO tneasurements (8). T h e etnission distri-(92%) failed the inspection even though all Air A c t Amendments, many parts of which bution is highly skewed; the half of the fleet had valid registrations. Thus, less than 8% deal with motor vehicles. Motor vehicles are with the lowest emissio~ls contributed less of the vehicles identified as gross polluters the ~r i m a r v source of urban carbon tnonox-than 10% of the C O and HC. while a few bv remote sensine nassed a n immediate " L ide (CO) and are an important source of the high-emission vehicles dominated the rdadside test. W h e n r a~l d o~n pullover studhvdrocarbons I H C ) and oxides of nitrogen mean values. I n this instance. 7% of the ies were carried out bv the California BLI--(NO,) that are responsible for the fortnation of photochemical stnog and ground-level ozone (1). Cost estimates for itnplenlenting the act's mobile source provisions range up to $1 2 billion annuallv 12). Thus it is im-,~, portant to analyze the scientific basis for these legislative programs (1 ).For any region violating air quality standards, a State Itnolenle~ltation Plan (SIP) , , must be submitted to and approved by the U.S. Environmental Protection Agency (EPA). In approving the SIP, EPA grants credit for each portion of the plan, including new vehicle emission standards, new fuels, vehicle scrappage, and inspection and maintenance (IM) programs, o n the basis of predictions from a spreadsheet computer tnodel 13). This EPA model treats all cars of ~, a given tnodel year as having the same odometer reading, the same annual mileage accutnulation, and an equal likelihood of emission control systetn probletns. T h e model has had little success in predicting urban on-road vehicle etnissions (4), and the use of unverified computer models as the sole guide for public policy decisions is controversial (5). Calvert et al. (6) criticized the tnodel and recommended in-use vehicles accounted for 50% of'the on-road C O emissions, and 10% of the vehicles accou~lted for 50% of the on-road HC etnissions. These vehicles we call gross polluters. About 5% of the vehicles were gross polluters for both HC and CO (9).It is often assumed that gross poll~tters are sitnply old vehicles. In fact, all model vears of vehicles we have measured o n the road include some proportion of gross polluters, as shown in Fig. l. W e found that the highest emitting 20% of the newest cars were worse polluters than the lowest emitting 40% of vehicles from any tnodel year, even those from model years before the advent of catalytic converters (1970 and earlier). These data are typical of CO and H C results across the United States an...