AB Testing Laws?

Recently I was thinking about test-driven web development (AB testing and the like) and how you could apply that to governments.  One way it could work is you would pass a law, but then build in some metric that would make the law repeal itself if it didn't work out well.  Let's consider a law that would lower the drinking age to 18.  Well, some would argue that this would increase deaths from alcohol from people aged 18-21.  That sounds reasonable.  Others would say the current drinking age of 21 encourages high school students to drink surreptitiously, which leads to binge drinking and deaths.  I don't know how I feel about it, so I'm not sure how I would vote on a law that lowers the drinking age to 18.  

What if instead the law was "Lower the drinking age to 18 for 2 years.  If the drinking-related fatalities increase in these two years compared to when the law wasn't in effect, increase the drinking age back to 21."  

While I was on the fence for the prior law, I can get behind this conditional law.  You would need some objective third party to measure the data, and you would have to make sure you are measuring the right things.  That's pretty hard, but it seems a lot better than just passing the law and not measuring it.  While many laws have effects too complex to measure this way (like lowering tax rates) there seem to be a decent number of laws that could work (for example, do carpool lanes really help traffic?  how about banning sugary drinks?)

It seems like a good principle for laws to be built with explicit goals in mind.  I'm not a legal expert, so I don't know if this has been done before, and a quick Google search didn't reveal too much.  What do you guys think?