Simplicity -- a screed in two paragraphs

I have been reading a recent short book by Timothy Gowers, on mathematics and models and such (Mathematics: a very short introduction). In it he makes a point I agree with: "When devising a model, one tries to ignore as much as possible about the phenomenon under consideration, abstracting from it only those features that are essential to understanding it..." This is from a mathematician effectively explaining that there is a special art in good science, since theoretical models are by definition simplifications and abstractions. I made a similar point in a 1997 book, when discussing partial equilibrium models: "By definition partial equilibrium models do not take into account many of the factors emphasized in general equilibrium trade theory. While this is the root of the practical limitations of applied partial equilibrium modeling, it is also the source of its basic advantage... It may be difficult to justify... more complex and less transparent models, when they may yield only margin extensions of the basic insights drawn from simpler approaches." The critical thing is that good models (theoreical or empirical) offer useful simplifications. Musicians (Charles Mingus: "Making the simple complicated is commonplace; making the complicated simple, awesomely simple, that's creativity." ) and super-mega-geniuses (Albert Einstein: "Everything should be made as simple as possible, but not simpler.") would agree. But apparently not all members of the tribe of trade theorists. Some would, though. Some of my favorite theory papers are by Krugman. His papers can display a brilliant ability to boil down potentially complex and difficult issues to sharp, clear expositional models. His 1980 AER paper on gains from trade, in a world with identical two countries with identical tastes, endowments, and technologies and only one sector, for example, is brilliant. So is his explanation of linkages between productivity growth and the process of job creation and construction, using the example of hot dogs and buns. (from an editorial in Slate). Yes you can make the model in his AER paper more complicated, and subsequent literature has. But the basic point is best seen in a simple structure.

So what do you say when referees, and journal editors who should know better, send you letters and reports that basically say "the point you make is clear, and a relevant one, and the model is simple and clear. However, the referees feel you should use a more complicated model...because it will make the theory more interesting"? What? Interesting as in more complicated, harder to follow, with a more circuitous mathematical path to the same basic results? This is definitely where we step out of the realm of math and science, and into the realm of fashion trends and complexity for the sake of complexity. Complexity with the purpose of making what we are doing look too hard and complicated for others to grasp. In terms of models, minimalism (where possible) should be the rule. If it is not necessary, don't build it in. And if your model is too complicated to explain anything with, well... I told you so. Enough said.