But it has become increasingly common for people, even otherwise numerate analysts, to write as if it can.
Consider a recent instance. In the Spring 2014 issue of Regulation, Sam Batkins and Mitch Boynton discuss a case in which an estimate of a regulatory cost fell from $672 million to $89 million. That's a drop of 87 percent.
But that's not what they wrote. They call this "a 750 percent drop in costs." How did they get that? It looks as if they take the drop as a percent, not of the number from which it dropped, but of the number to which it dropped. The drop was $583 million. That's 655 percent of the number it dropped to. Then, for some reason, they add another approximately 100 percentage points.