The author presents some data in the article. Also, the absence of hard quantitative data doesn't necessarily make it a complete guess. (At least not any more than starting with the assumption of a Gaussian distribution.)