18 May 2011

Brin's Pareto Pipe Dream

Daily Kos | David Brin | The "No-Losers" Tax Simplification Proposal

There is nothing on Earth like the US tax code. It is an extremely complex system that nobody understands well. But it is unique among all the complex things in the world, in that it's complexity is perfectly replicated by the MATHEMATICAL MODEL of the system. Because the mathematical model is the system.

Hence, one could put the entire US tax code into a spare computer somewhere, try a myriad inputs, outputs... and tweak every parameter to see how outputs change. There are agencies who already do this, daily, in response to congressional queries. Alterations of the model must be tested under a wide range of boundary conditions (sample taxpayers.) But if you are thorough, the results of the model will be the results of the system.

Now. I'm told (by some people who know about such things) that it should be easy enough to create a program that will take the tax code and cybernetically experiment with zeroing-out dozens, hundreds of provisions while sliding others upward and then showing, on a spreadsheet, how these simplifications would affect, say, one-hundred representative types of taxpayers.
No. This is not possible. It is an extremely appealing vision — Pareto improvements! provability! the frailty, irrationality and narcissism of human lawmakers replaced by impassionate math! no difficult sacrifices or hard decisions or political bargaining! — but it is a pipe dream.

I do not have the patience to explain why. I will simply point you to the Wikipedia article on The Curse of Dimensionality and leave you to it.

That is the only nail you need to close the coffin on this proposal, but I will give you some more.

This is very, very wrong:
But [the tax code] is unique among all the complex things in the world, in that it's complexity is perfectly replicated by the MATHEMATICAL MODEL of the system. Because the mathematical model is the system.
Many things are equivalent to their mathematical models, and yet doing complex calculations on them is extremely tricky and often impossible. Computer programs are equivalent to their models, and yet analysis of them is maddeningly complex. Despite the perfect congruence between programs and their models some questions are provably impossible to answer. (See, for instance, The Halting Problem.)

The tax code is not equivalent to a mathematical model of the tax code, because so much of it is vague or outright undefined. Those aren't provisions that lend themselves to mathematical analysis.

Brin is succumbing to the comforting illusion — quite popular now — that we can avoid raising taxes by instead "closing loopholes." The bad news for these folks is that all those "loopholes" were put there on purpose. Maybe for selfish purposes, but commonly enough for altruistic ones. They aren't accidents. Very vocal groups care very much about them, and the law of concentrated benefits and diffuse costs makes them devilish to get rid of. Then there is the meta-problem: assuming you can find a significant number of loopholes and eliminate them, how will you keep them from re-arising?

Now I would rather like to eliminate all such special provisions, raze the personal income tax code to the ground, sow it with salt and start over with a flat consumption tax, but Brin dismisses that as "utopian fantasy." Quite the reverse, I say: the goal of building a plan around identifying "loopholes" that no one cares about is the real illusion.

We ought to aim for a tax system which is machine operable, but let's not pretend we're close now.

(Via io9)

Edited to addWarren Meyer points out another crippling error in Brin's proposal that I completely missed: taxing human activity isn't a static operation.  People's behavior will change in response to changes in the tax code.  Brin is assuming that you can twist any of the numerous knobs on the tax side of things and you won't see any changes in the activities being taxed.

No comments:

Post a Comment