Why not provide separate results for each powertrain?

People often ask me why results for each model are not broken out by powertrain. For example, the VW Jetta, Rabbit, and GTI can be purchased with either a five-cylinder gas engine, a four-cylinder turbocharged gas engine, a four-cylinder turbodiesel, or a V6 gas engine. Transmissions include a manual, an automatic, and a DSG. Shouldn’t repair rates be expected to vary based on which of these a car has?

It helps to look at the actual data. Repairs reported for the 2005.5-2006 Jetta (57 cars):

  • satellite radio quit (two reports, now a TSB)
  • power windows failed
  • missing bolt on door skin
  • broken rear seat vent deflector
  • both headlights burned out
  • electrical outlets in center console
  • loose door panel causing rattle
  • windshield washer hose broke (two cases)
  • lumbar support broke
  • more rattles
  • steering box cracking sound

Repairs for the 2006.5-2007 Jetta, Rabbit, and GTI (31 cars):

  • airbag light
  • window seal
  • seatbelt retractor
  • interior hatch trim piece misaligned
  • rattle
  • compass stopped working
  • visor light
  • sunglass holder woudln’t stay closed
  • bad coilpack (the only powertrain-related issue in the bunch)
  • A/C compressor 

Out of both lists, we have a single powertrain repair. All of the other repairs involved components that are common across all variants of these models.

In this case as in most cases these days, powertrain problems have been rare. And the more rare something is, the larger the sample size you need to accurately measure it. To measure problems as rare as these seem to be, you’d need a very large sample. And how meaningful is it, anyway, to report failure rates well under ten percent? If one engine requires repairs two percent of the time, and another requires repairs three percent of the time, it is a difference of 50 percent. But we’re still talking about one extra powertrain repair per 100 cars. Is this really important to know when buying a single car?

If a particular powertrain ever is actually trouble-prone, it’ll show up even with TrueDelta’s current sample sizes, and I’ll note it on the results page. There are a couple cases where I’m considering this currently.