This morning Atrios had an interesting comment about proposed housing-market "reforms":

I saw Shiller speak awhile back and like Felix I was quite puzzled by his desire to fix things buy adding even more financial "innovation." One proposal involved insurance against falling home prices, which just seemed to be kind of absurd. The whole game seems to involve injecting as much risk into the system and then conning or forcing the suckers to pay to remove it again. Oh, and then repackaging that risk and conning still more suckers into buying it.

And it came to me that with a few exceptions, economists have been drinking the free-market koolade for so long that they're complete newbies when it comes to the idea of designing market institutions. So like most untrained designers (look at the average myspace page, or some kid's doodle of a machine to fly underwater, tunnel underground and take people to Mars) their idea of how to fix problems is to add more bling.

What makes this particularly sad is that people have been doing this kind of job right for 200 years or so. Steam engines didn't all rely on operators to monitor pressure; they had safety valves. Locomotives had (and have) "dead-man" switches so that any accident to the driver would bring the train to a stop. Designers of aircraft in WWII learned to make controls in different shapes for different functions so that a pilot could recognize them by feel. Software designers (some of them, anyway) have figured out how to simplify interfaces to avoid presenting users with an unmanageable array of options. Engineers and designers know how to build stuff so that it won't blow up the first time an inexperienced user touches it.

So instead of starting with a clean slate and thinking they know so much better (that worked out really well the last couple times around) economists should look at human-factors rules and good engineering practice when thinking about how financial markets and other institutions should be structured.

Among the rules:

• Simplify ruthlessly. Parts that aren't there can't fail.
• Make the default behavior safe, or at least no more dangerous than other choices.
• Be willing to accept some loss of performance for reduced change of disaster.
• If part of your performance envelope is potentially unsafe, don't go there.
• Don't present non-experts with expert choices.

The profits of some individual players can be maximized by ignoring such rules, but that's exactly why you need to the rules. The history of early engineering was littered with the fragments of people who thought that they were exceptions to Murphy's Law. Let's not keep littering the economic landscape in the same way.