Monday, September 15, 2003

Anecdotes and Antidotes

Radley linked my defense of him versus Al Giordano, saying (among other things) that “empirical evidence is always preferable to anecdotes.” Giordano responds in the comments box, saying (also among other things) that “[Radley] based [his] claim exclusively on polling data (which, after all, is no more or less than the science of collecting anecdotal information from many people at once).”

At this juncture, I think it would be useful to consider the meaning of “anecdotal.” The definition is “Based on casual observations or indications rather than rigorous or scientific analysis.” The very nature of an anecdote is that it is a particular instance observed in an idiosyncratic fashion. When you start collecting a large number of anecdotes in a systematic fashion, they cease to be anecdotes and become a data set.

Now, it’s true (as Giordano emphasizes) that polls and surveys can be very misleading. So perhaps something is wrong with the surveys Radley cites, which indicate that poor people in the developing world actually like globalization. Maybe the wording of the questions was biased. Maybe the sampling method didn’t generate representative samples of the populations. But it seems awfully strange to dismiss the results of such surveys, which at least attempt to be scientific, on the basis of a handful of observations with no consistent sampling method. Even if we accepted Giordano’s strange notion that a poll is nothing more than a collection of anecdotes, does it make sense to reject the clear results of a large sample of anecdotes (38,000, in fact) in favor of the unclear results of a smaller sample of anecdotes generated largely via undocumented collection methods?

No comments: