All Models are Wrong, and Some Models are Useless

Yihui Xie 2007-09-23

Poor economists are said to be always searching for a non-exist black cat in a dark room, while poorer econometricians are regularly accused of finding one (see Peter Kennedy, A Guide to ?Econometrics, 3d ed., The MIT Press, Cambridge, Mass., 1992, p.82). The other day I saw another funny description:

An economist is a surgeon with an excellent scalpel and a rough-edged lancet, who operates beautifully on the dead and tortures the living. (by Nicholas Chamfort)

I believe statisticians will certainly be confronted with the same criticism too (at least in China)… Today when I was helping my supervisor review a manuscript, I just laughed to myself because the paper was almost completely useless, and the authors dared to say there was NO previous research on their topic – that’s too ridiculous. In the sense of economics, I cannot find any significance of this paper – it only revealed the relationship between FDI, import & export trade using the Error Correction Model for Jiangsu Province. What’s the use of such a simple relationship? (This variable increases, that decreases, blabla…) Can it still hold when the model is applied to other districts?… In a word, a statstical model is useless if it only describes the past history.