However I'd be concerned that the law would work differently with a subset than the whole population. For instance take a law banning cellphones while driving. If you try to apply this to half the population, then the police aren't going to try to apply it because at least half the time they would be pulling over people who are allowed to use cellphones. But if you pass the law for all of the people, then the police could enforce it. So you try to solve this by saying that it will be selectively enforced according to location. But now you've gotten rid of the random selection that is the heart of evaluating the statistics.
For another example, look at the tax cut example they were discussing. The challenge here is that the touted economic benefit of tax cutting is that people with lower taxes spend their money, increasing circulation, and general prosperity. (And eventually improving tax revenue.) However there is no way to economically separate out the people you're giving the cut to from those you aren't. Therefore any economic change cannot be attributed to the tax cut, and you're unlikely to get a clear economic difference. (Other than that the less taxed have more disposable income.)
That depends on whether or not license plate numbers are assigned by county, or by driver last name, or some other nonrandom criterion. And in some states, significant numbers of drivers choose their own customized license plate numbers.
However I'd be concerned that the law would work differently with a subset than the whole population. For instance take a law banning cellphones while driving. If you try to apply this to half the population, then the police aren't going to try to apply it because at least half the time they would be pulling over people who are allowed to use cellphones. But if you pass the law for all of the people, then the police could enforce it. So you try to solve this by saying that it will be selectively enforced according to location. But now you've gotten rid of the random selection that is the heart of evaluating the statistics.
For another example, look at the tax cut example they were discussing. The challenge here is that the touted economic benefit of tax cutting is that people with lower taxes spend their money, increasing circulation, and general prosperity. (And eventually improving tax revenue.) However there is no way to economically separate out the people you're giving the cut to from those you aren't. Therefore any economic change cannot be attributed to the tax cut, and you're unlikely to get a clear economic difference. (Other than that the less taxed have more disposable income.)