I'm not arguing the validity of the COLA measurement, but I know that fuel costs being 40% lower this year has contributed more than a 2-3% difference in my personal budget. So it's reflective of reality for me.
My understanding is that whenever the COLA count is negative they do not provide an increase, but neither do they reduce the payout. So when you have multiple years of negative COLAs as we have had recently, you are actually getting a bit of an increase (compared to the general economy) just by remaining at the same level.
Then, when the COLA count goes positive again (with respect to the now lowered level) you get paid a second time for those increases. At least this is the way the retirement expert at FEDWeek.com explains it.
Now, weather or not you agree with the method they use to create the COLA count, is another issue. I was just stating that gas is a non-negligible factor in the economy, and it doesn't seem unreasonable that a significant reduction in the cost of gas would have an effect on the count.
The price of fuel does, as you've suggested, have an indirect effect on the CPIs through its effect on the prices of other things we buy. More importantly though, the price of fuel is factored into the relevant CPIs directly. So though the prices of other things have generally gone up (even including those other things which are affected to varying degrees by fuel prices), the index as a whole is a shade below flat YOY because energy costs have fallen so much.
Also, a small clarification: Yes, if the change in the CPI-W (i.e. the average for the 3 months used - July, August, and September) is negative YOY, there just isn't a COLA for that year (really it's for the next year). Benefits aren't reduced because the change is negative, they just aren't increased. However, the base used for the following year is not adjusted to that reduced CPI-W. The base is the CPI for the last year in which a COLA was triggered. That's why in 2010 (effective at the end of 2010, though really for 2011) there wasn't an automatic COLA even though the CPI had gone up from 2009 to 2010. It was still below where it had been in 2008, which was the base for 2010 since there had been no COLA triggered in 2009.
So... CPI in year 1 is 100. CPI goes to 90 in year 2. No COLA - no increase and no decrease. CPI goes to 105 in year 3. The COLA is not 17% ( 15/90 ), instead it is 5% ( 5/100 ).
That's how the automatic COLA works anyway. Congress can, of course, provide for something different in a given year if it wants to.