Drivers who track their own fuel economy have long known that their results seldom match the gas mileage claimed by the Environmental Protection Agency (EPA) on new-car stickers.

Our study of gas mileage in hundreds of vehicles for model-years 2000 to 2006 – in which we compared EPA fuel economy with the miles per gallon (mpg) we measured in our real-world tests – quantifies that problem across a wide swath of makes and models. Among our findings:

¢ Shortfalls in mpg – those instances in which our numbers lagged EPA’s – occurred in 90 percent of vehicles we tested.

¢ The largest discrepancy between the EPA’s claim and our actual mpg involved city driving. (Although its results were “adjusted” downward in 1984, the EPA city-driving test still uses 1970s-era protocols. Our test recognizes that motorists drive more miles in dense traffic these days, resulting in slower speeds and a greater percentage of time spent idling.) Some models we tested fell short of EPA claims by 35 to 50 percent. As for highway mpg, our numbers more closely reflected EPA’s, on average.

¢ Hybrid vehicles, whose mix of gasoline- and battery-powered propulsion is touted for its fuel thriftiness in city driving, had some of the biggest disparities there, with real fuel economy ranging from 11 to 25 mpg below EPA ratings. Still, hybrids won three of the best five spots in our tests or overall mpg, which is calculated from city, highway and mixed-driving tests.

¢ The disparity between our numbers and EPA’s is increasing. For gasoline-powered vehicles, the shortfall was 6 percent for 2000 model cars that we tested, but about 12 percent for 2005 and 2006 model cars.

Their use as a predictor of real miles per gallon aside, consumers still can trust EPA’s figures to the extent that the fuel economy of, say, a Honda Civic EX (33 mpg claimed by EPA; 29 actually tested by us) is considerably better than that of the BMW X5 (18 claimed; 17 actual). But until the EPA ratings are made more realistic (in the past two years, we’ve urged the agency to overhaul its testing protocols) you should discount the EPA sticker numbers for city travel in conventional cars and trucks by 30 percent, in larger hybrids by 35 percent, in diesels by 36 percent and in smaller hybrids by 42 percent.

Or you can go by our findings. Following are some vehicles – still sold – that provided the most fuel economy in their class, as determined by our recent testing. (Figures are for overall mileage. The Toyota Prius appropriately appears in two categories.)

¢ Small cars. Honda Insight (51 mpg, with a manual transmission); Toyota Prius (44 mpg, with an automatic transmission); Toyota Echo, (38 mpg, with a manual transmission.)

¢ Family sedans. Toyota Prius (44 mpg); Volkswagen Passat GLS TDI (28 mpg); Honda Accord Hybrid (25 mpg).

¢ Large sedans. Toyota Avalon (22 mpg); Mercury Montego FWD (21 mpg); Ford Five Hundred (21 mpg).

¢ Small SUVs. Ford Escape Hybrid (26 mpg); Honda CR-V EX (21 mpg); Subaru Forester 2.5 X (21 mpg); Toyota RAV 4 (21 mpg).

¢ Midsized SUVs. Toyota Highlander Ltd. (19 mpg, with the V6 engine); Nissan Murano (19 mpg); Acura MDX (17 mpg); Honda Pilot (17 mpg).

¢ 4WD 4-door Crew Cab Pickups. Subaru Baja (20 mpg); Toyota Tacoma (17 mpg).

You can improve the fuel economy of any car by keeping it in top shape. A poorly-maintained engine can cut gas mileage by 10 to 20 percent. A clogged air filter can cause up to a 10 percent increase in fuel consumption. And underinflated tires, because they require more energy to roll, can reduce fuel economy by 5 percent.