I'm not sure that I see the ethics in using New Car reliability statistics to make recommendations on Used cars.

For many years, cars like software, have followed the bath-tub curve for costs - typically the first year or 2 are expensive due to teething issues, followed by ~7 years of low cost maintenance, and then increasing costs as all the original components hit the end of their expected life. (That 7 years figure has probably increased now, I'm quoting (from memory) an early 90s study).

The teething issues costs are covered by warranty - that is the entire reason for the warranty. Some people may get the bad tranny, whilst others will get the bad spark plug lead. Since one is exhorbitantly more expensive to replace than the other, and you have no way of telling at the time that you buy the car, it would not be fair on consumers to _not_ have a warranty. Effectively, the warranty averages out the costs of all teething issues and puts them into the price of the car.

So some cars will have more teething issues than others. Some will spend more time in the shop during their first 2 years than others, and this will be reflected in the statistics. But once these teething issues have been dealt with, do those statistics have any meaning?
Both VW and Audi have been hit extremely hard by coilpack issues on their 1.8T engines. The coilpacks were designed and manufactured by a third party supply company. It turns out that 2001 and 2002 models had a bad design that was prone to failure, especially in extreme cold weather conditions. _Every_ 2001/2002 VW/Audi 1.8T car had this issue, and due to the large percentage of failures, those coilpacks are being replaced as a routine operation. Now the statistics reflect that failure, even though by the time you buy a second hand 2001 car in late 2003 it's probably not got those coilpacks any more.