Thai nailed it.
Purely mathematical in calculation. No real-world factors, not even suspension stiffness, etc are taken into affect. It's just a bunch of body spec/dimensional stuff crunched together into a 'rollover ratio' and it does not take into account stability control, traction control, electronic brakeforce distribution or other nannies that might help avoid a rollover.
Thus, Volvo and every other mfgr can advertise that their electronics actually make the vehicles even safer.
I've seen a post on Edmunds where the author claimed an automotive publication tested the same SUVs with bigger and smaller overall wheel packages (as compared to stock), as well as bigger/smaller wheel/tire examples (same total circumference). The ratio numbers changed quite drastically, with some vehicles ranging from 2 to 4 stars. Not really surprising, but SUVs with bigger tires (and a higher center of gravity) get lower rollover ratings and SUVs with 20s and low-profile street tires have a far better rollover rating than the same SUV equipped with 15" wheels and big-aspect rubber.
I think this data is pretty much marketing BS. I suggest taking stock OEM models of every SUV out on the same day, get them up to the same speed on the same stretch of road, hang two wheels off the pavement at the same spot and onto a soft shoulder and then crank the wheel and mash the brakes (uh - the only real situation in which SUVs tend to roll). THEN give them ratings *on the number of times* they roll.
But then of course the best cars would have zero or one roll and that's not a powerful, confidence-inspiring marketing figure...
And let's not forget that you'd have to head to Poland to find your testers...
