There are some caveats to this study which mean isn't as spectacularly powerful as Ars makes it. - They only have subscriber data from 1987-1995, when almost n

There are some caveats to this study which mean isn't as spectacularly powerful as Ars makes it. - They only have subscriber data from 1987-1995, when almost nobody had mobile phones (7%), and they were so expensive to use that they were used rarely (mostly outdoors and in cars). - But since then, 86% of the population has started to use mobile phones, and because they don't have tracking data for them, all of these mobile phone users are in the control group. So they're comparing exposed users to only-slightly-more exposed users (as I mention above, cell phones were rare and had quite expensive per-minute fees before 1995). They only know that most in the control group didn't have a mobile before 1995. Since then, they don't know how long any user in either group has used their mobile. - During the period for which they have data (pre-1996) 30% of users were on corporate plans and couldn't be tracked--so the heaviest users were lumped in with the control group too. - The 1987-1995 group mostly used analog phones, not today's GSM phones. Still, earlier phone users have used their phones slightly more than non-users in aggregate, so this study is noteworthy despite the lumping of the 30% heaviest users with the controls and the massive cell phone use by both groups since 1995. You'd expect cancer rate to correlate with mobile phone exposure, and both groups are very large.