Models, Maps, and Simplifying Assumptions

by PM

14 December 2012, 1216 EST

I don’t want to say too much about the Rathbun-Arena smackdown (smackduck?) taking place but I will say this: We often rely on maps that are not “true.” The Mercator projection is “wrong” if our criterion is “showing countries according to their relative size” but “useful” if our criterion is “helping to plot sea voyages.” Moreover, we sometimes explicitly distort maps to make them more useful:

As Wikipedia notes, the Tube map is considered a huge advance on earlier maps despite being flatly wrong about distance, train routes, and everything else.

When we consider theories, it may be that depicting actors’ behaviors as following rationalist behaviors are deeply false, but nevertheless useful. We shouldn’t be deluded into mistaking assumptions for reality but we should also not forego the power of simplification.

(And I have to say that every empirical researcher who’s ever used a parametric estimation method like OLS has used untested, and often blatantly wrong, assumptions in their analysis, unless they have a signed note from God that their errors follow a known distribution.)

Late Update: Similar and longer argument (read after posting).

this is text because there’s a bug in the template