counting crowds

From an AP snippet quoted in the Omaha World Herald: "Journalists estimated there were about 20,000 protesters, but pro-Chávez lawmakers said there were far fewer." I've always wondered how people estimate crowds. Sentences like the one I've just quoted seem pretty common in newspaper articles involving crowds, and particularly about protest crowds, since, of course, the number of people attending is itself a political statement. A little web research reveals there's really no good way to estimate them. There are various methods, but none without significant detractors, or at least significant margins of error (e.g., 20%). The major problem seems to be density--there's no way to evaluate density accurately in all parts of the crowd (unless they're seated in an auditorium or stadium or similar arena). This article from the Seattle Times outlines the problem pretty well.

. . .