Tyler Cowen put up a post on his blog saying that Malcolm Gladwell asked him to put this question to his readers: “What’s the best example of an experiment or trial that could be scientifically useful and informative but which can’t be done for legal, ethical or logistical reasons?”
The responses I found interesting fell into two categories: child rearing and education. Here is one on child rearing:
Parenting Styles. Compare strict v. permissive. Do a group with physical punishment and yelling etc. to find out just how bad and harmful these practices are. And also do a hippie-dippy group where the kids can do whatever they want. Seems like so much of our knowledge on this is hopelessly confounded.
Others specifically called out breast feeding, having your infant sleep in your bed or not, and parental separation. Because everything we think we know about this stuff is indeed confounded by the problem that people who raise their kids in different ways are different from each other in a million other ways. The same might be true for people who get divorced, although nobody really knows that, either.
One of the education experiments people asked for involved taking 10,000 children and randomly assigning them to different kinds of school systems, or to home schooling, and seeing what happens. Another was to randomly send half a group of 18-year-olds to college and prevent the other half from doing so, then checking on how they were doing ten years later.
It seems that doubt over whether formal education does any good is rampant in America these days.