While not hundred percent sure, it is highly possible that finance complexity will magically disappear once we get a proper model for money transactions. I certainly hope so, because a better model means more means to solve economic issues.
Right now economists are like the physicians in the time of Molière, they understand nothing and have only two cures: bloodletting and lavement (change interest rate and remove taxes)
His system still used circular orbits and epicycles. The only real difference was that he put the sun in the middle instead of the Earth. In fact, it required more epicycles / gave more imprecise results.
The real improvement came when circles were replaced with ellipses by Kepler.
That wasn't entirely accurate, either, but replacing the ideal circles with the law of Gravitation (Newton) fixed that.
Economists have many more than just two cures -- which is why some of them keep harping on regulations and overly protective labour laws in Greece (and the rest of the Mediterranean swamp), their very expensive pension system, and the huge size of their public sector. Krugman doesn't, which is why I have absolutely no respect for him.
Can someone smarter than me tackle this argument?
Like a software system, it's extremely naive to think that complexity is a sign that a system is rotten.
Why would such an assumption be naive?
It doesn't seem naive to me at all. I was reading a 1994 paper that I posted recently called "A Uniform Name Service for Spring's UNIX Environment" which made me think how simplified a modern Unix could be if it adopted the name service primacy described therein from its onset.
It is just as naive to assume that all complexity is intrinsic and not incidental or the result of inadequate abstractions.