Lies and Statistics

Manuel Quezon III, writing in the Phillipine Daily Inquirer, has some comments about statistics that are worth examining (even if you don't live in the Phillipines.)

The article begins:

There is, apparently, a debate going on about the origin of the famous aphorism that there are three kinds of lies: lies, damned lies, and statistics. One source noted in a University of York Department of Mathematics website quotes Sir Robert Giffen (1837-1910).

In the Economic Journal in 1892, Giffen wrote, “An old jest runs to the effect that there are three kinds of comparison among liars. There are liars, there are outrageous liars, and there are scientific experts. This has lately been adapted to throw dirt upon statistics. There are three degrees of comparisons, it is said, in lying. There are lies, there are outrageous lies, and there are statistics. Statisticians can afford to laugh at and profit by jokes at their expense. There is so much knowledge which is unobtainable except by statistics, especially the knowledge of the condition and growth of communities and growth of communities in the mass, that, even if the blunders in using statistics were greater and more frequent than they are, the study would still be indispensable. But just because we can afford to laugh at such jests we should be ready to turn them to account, and it is not difficult to discover one of the principal occasions for the jest I have quoted, and profit by the lesson.”

We've the joke floating around the internet that 47.6% of all statistics are made up on the spot. And we've seen some pretty outrageous claims made using data, and we've all winced at seeing someone use perfectly good data to draw the exact opposite conclusion.

Quezon uses the anecdotes above to launch into an examination of official government statistics in the Phillipines. The same kind of examination is worth paying attention to wherever we live, and that's one reason why metadata information is so critical.

So the next time you see someone calling for massive increases in funding because half the performance is below the median, just smile and nod. We'd all like to see the place where all the children are above average. In the meantime, however, we document our sources and scrutinize the data we collect before we report it, because if we want data to make a difference, we need people to receive the data differently.

(Hat tip: Uniffors)