I saw the 2007 Annual Consumer Reports Auto issue. They rate all models of used cars. One of the things that bothered me was the way some categories jump around from far better to far worse in a single year, when there were no major changes. So I started to poke around to see if I could reach any of my own conclusions about Consumer Reports. I don't think they're biased, but I don't think that their conclusions are statistically valid for low volume models.
Here is some basic math. There are about 10k Boxsters sold in the US each year, give or take a thousand or two. There are about 16.5 million cars sold in the US each year also. So the Boxster makes up about .06% of the market. T(hat's not six percent, it's six one-thousands.)
There are about 1.3 million cars in the Consumer Reports survey, but that's not 1.3 million from each model year. It's just 1.3 million. Let's cut them some slack and say that they had 200,000 cars from each of the last five model years, then it tapers off.
If we assume that their 1.3 million cars has the same mix of models as the population as a whole, then we can determine how many Boxsters are in their sample. So we need to multiply, 200,000 times .06% and you get 120 cars. So Consumer Reports is coming to the conclusions they are reaching based on about 120 cars of the 10,000 produced each year.
They listed the category of Major Engine problems for the 2002 as 'Much Worse than Average'. I don't know the answer to this, but if a 2% failure rate is average and you get one more with a bad engine, then what happens to your score. My guess is that you end up in 'da crapper'. They do get to track problems over time, but unless they are reporting on a different set of 120 cars from each model year, there is no improvement on the statistical validity of their sample.
I don't want to suggest that Porsches are without issues - they aren't, but I don't think Consumer Reports presents reliable data in their reliability analysis. Sorry, I couldn't help myself. Interesting, JD Powers used car data states 'Not enough data'. Sounds like a more responsible assessment.
Interested in comments...