The frightening thing is that, although I didn't doublecheck every angle, I think I see how all of these statistics can be truth simultaneously. There are the old adages that "90% of statistics are made up" and "80% of statistics are lies," but I tend to believe most statistics that come from reasonably authoritative-looking sources. The catch is that statistics can be phrased in ways so as to suggest nearly anything.

On this issue, gabby and Ellie presented a number of statistics, and there are more outstanding. Accidents per mile travelled in cars. Accidents per mile travelled in SUVs. Accidents per mile travelled in the Ford Explorer. Deaths due to accidents per mile travelled in cars. Deaths due to accidents per mile travelled in SUVs. Deaths due to accidents per mile travelled in the Ford Explorer. Rollovers per mile travelled in cars/SUVs/the Ford Explorer. Percentage of deaths per accident (not calculable from previously mentioned stats, since more than one death can happen per accident) in cars/SUVs/the Ford Explorer. Chance of surviving an accident (not calculable from previously mentioned stats, since cars and SUVs probably carry a different average number of people per mile on the road) in cars/SUVs/the Ford Explorer.

Each one of the stats above has a companion stat when you restrict "accidents" to "single vehicle accidents" and another when you make it "multiple vehicle accidents." For every stat involving deaths of occupants of the vehicle, there is a stat about total deaths in accidents in which the vehicle was involved -- these stats can be broken down further: deaths in accidents in which the vehicle was the one at fault; deaths in accidents in which the vehicle was involved but not at fault. Accompanying the "per mile on the road" stats are "per vehicle" stats, which would differ if SUVs are driven for a different average number of miles than cars. There's also a "per hour on the road" stat, which would differ if SUVs are driven at a different rate of speed than cars. Then, of course, I haven't specified the time period that these statistics can be gathered in. Over several years? Over a single year? Which year? (Different models of vehicles make that important.)

Even objective reports will happen upon bias simply because it's so difficult to avoid. And if you ARE biased, look at all the opportunities you have. If you don't like the accidents per mile rate, you use the rollovers per mile rate. If you don't like some per mile statistic, use the per vehicle statistic. If you don't like the deaths per accident or the deaths per rollover rate or the percent chance of survival in an accident rate, use the percent chance of survival in a rollover. If you don't like some figure about all accidents, just talk about single vehicle accidents. If you don't like how some figure for the Ford Explorer compares with that same figure for cars, compare it with other SUVs. If you don't like the stats for 1995, use the stats for 1996, or 1997, or for the whole decade.

Until you find an article that gives you all of the above, plus potentially crucial variations I haven't thought about, then it's important to look for the loopholes. I'm not saying all stat-based reports should be suspect. I'm not so cynical that I think purportedly unbiased publications always or even often have some secret agenda. I'm just saying that statistics can be made to suggest almost anything even if they are all accurate statistics. Whenever you read any statistics, don't automatically become cynical and disbelieve what you are reading, but do ask yourself what statistics are missing.