I'm reminded of statistics showing X website doesn't support mobile because only Y% of their traffic comes from mobile. Guess what? If the site doesn't work on a given browser, you're biasing you sample size. I think it was in Coupland's Microserfs that there was the story of corporate waiting until all the soda was gone to restock the fridge -- how can you know what people actually want if you articifically restrict their choices.

Now, the premise seems plausible, but you can't back up that premise with data that I know a priori is skewed to support that premise.