One thing that strikes me about the arguments I've been having with voucher opponents is just how little they seem to understand how markets work. Markets don't work because they get it right the first time; they succeed because if at first they don't succeed, they try, try again.

A public school, by and large, cannot fail. If it screws up, no matter how badly, we will continue pouring money into it. This is particularly true because most of the employees of most systems can't fail either. They can be atrocious at their jobs, but provided that they are not actually molesting the students, it's nearly impossible to get rid of them.

Failure, to put it bluntly, works. Failure is nature's way of telling you "Hey, that doesn't work!" The American economy is vastly strengthened by the fact that companies are allowed to fail--and also by the fact that our crazy culture encourages us to try things that don't work.

In the first few iterations, this often looks inferior to a centralized system. Look, the critics say, they sat down and planned it all! Compare that to our messy, fragmented market where half the stuff doesn't work!

It can take a decade or more before the cracks in the planning appear. The planners, it turns out, didn't foresee that the world would change, and now the giant, planned system can't cope. One no longer hears so many complaints about how American cell services suck compared to Europe--not since their 3G debacle. I am fairly optimistic that in ten years, the current whinging about America's high-speed internet networks will be quieted when the decentralized model produces some unforeseen improvement.

At a conference last year, I saw an incredibly compelling presentation from the guy who does usability for Treo. He talked about design philosophy, and showed slides of a project he does where he goes into various institutions, divides people into groups, gives them spaghetti and some tape, and asks them to build the tallest self-supporting structure they can. The worst-performing group, you'll be unsurprised to hear, was MBA students; they spend all their time arguing about who will be boss. Engineers do okay. But the best performing group? Kindergarten students.

The students don't plan anything. They just try stuff, and if it doesn't work, they try something else. The presenter's argument was that if you want to do something quickly, and well, you need to have a lot of failure. Failure is the quickest way to learn.

But the way public schools are set up, they can't really fail--and so they don't succeed at the hardest task we've given them. The schools are not set up to learn; they're set up to follow the rules, and to serve their customer base, who are not in the case of poor schools the parents, but the various people who work for the system.

Ezra wrote a post criticizing my position on vouchers because there are, you know, really serious experts who care about education, and have all these awesome plans, and why the hell would we listen to some ideological libertarian whack job who just worships the market? I'd argue that first, all those plans suffer the fatal flaw of having to assume away all the poisonous interactions between the various constituents that have so far doomed school reform; second, they suffer the fatal flaw that the educational experts never fail either, because they just claim their plans weren't tried; and third, that serious planning may not be the right way to go about this. The right way may be to let a lot of people try stuff to see what works . . . and let a lot of other people copy what works because they're afraid of losing their jobs. While it's true that I think vouchers are good even if they just let a few kids exit an awful system, I also think it's true that they are the best shot we have at improving the system.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.