That simple regression does not indicate anything...the author does not show that the data follows an exponential; he just imposes it. What about a non-parametric regression? What about structural breaks in the time series?
Indeed, fitting a line would have demonstrated the main point just as well (that Usain Bolt is way outside the normal progression of the record).
Of course, it would have also implied that by the year 3000 some guy will be finishing before he starts, but from the data given that's no sillier a conclusion than the one given.
PS. Seriously though, well done Mr Bolt. I'm not usually impressed by athletes, but this guy really seems to be way out of the norm.
That would be the "structural breaks" idea. The process in the last 80 years that generated the downward trend in times has changed recently due to technology, etc.
He plainly justified an asymptote: "there must be some limit where humans can not run any faster". The exponential has this property and is simple. There also looks to be a good r^2.
