Honda Insight Forum banner

1 - 1 of 1 Posts

156 Posts
Discussion Starter #1 (Edited)
This should probably be elsewhere but it is sort of a technical issue, albeit a statistical one, so...

My HCH1 is on the way out and I have been researching "new" (ie used) cars to replace it with. One resource is the annual Consumer Reports car issue, which has a large table giving the reliability of various components for a range of model years based on their large survey. Today I noticed something interesting, this is for the 2012 Chevy Volt, with CU's ++ + = - -- scale converted to 5 to 1.

Column A is CU 4/2018 car issue
Column B is CU 4/2019 car issue

A B  What
5 5 Engine Major
5 5 Engine Minor
5 5 Engine Cooling
5 5 Transmission Major
5 5 Transmission Minor 
2 1 Fuel System
4 4 Drive System
5 5 Electrical
5 5 Climate System
5 5 Suspension
4 4 Brakes
5 5 Exhaust
4 4 Paint/trim
5 5 Noises/leaks
4 4 Body Hardware
1 5 Power Equipment 
1 4 In car Electronics
3 4 overall
Most of this is pretty consistent, but note the bottom 3 values. It makes no sense physically that as all cars aged a year the quality of those components vastly improved with respect to those in other cars. Yet this change bumped up the overall rating from average to good. The Chrysler 300, which is the next table over in both cases, had a change half as large, but in the opposite direction:
2018: 5 5 4
2019: 3 3 5

What I think is going on here is that, for cars that are not super common there is a very large noise component in these measurements, the rarer the car, the greater the swings caused by random events. CU surveys each year, and as far as I can tell, presents only that year's result, not any sort of running average. (Edit: correction, it is the sum of the previous 3 years.) In other words, the values represent failures per THIS year and not the sum or average of all failures to the current year. So a couple of failures in a given year can cause wild fluctuations in rarer cars. Look at the tables for the 2012 Honda Civic (a very common car) and there are only a couple of changes of +- 1 between the two issues (not shown).

Among hybrids and electric vehicles the only one which is as common as the Civic is the Prius. Like the Civic, the differences between the 2011 Prius in the 2018 and 2019 CU issues are all +- 1, and the difference is not in a consistent direction. That too is probably noise.

Since we don't have access to the raw data, what is one to do? Well, it seems to me that if one can obtain all the CU car issues since a particular model was made, one could make the conversion to the 5 - 1 scale, sum the values for each year over all those issues and then have a total failure rate number. Rather a PITA to do though, and to compare N different cars one would have to do it for all N of them.

Or just buy the Prius if reliability is really a factor, since by all indications it really is more reliable than any other hybrid!
1 - 1 of 1 Posts