I can't wait for Bryan's The Case Against Education: every semester, my beliefs move in favor of the signaling model and against the human capital model of schooling. This isn't to say there aren't a lot of students who are there for the human capital. Far from it: one of the great pleasures of my job is being around students who really want to wrestle with great ideas. As a matter of public policy, though, I think a lot of education subsidies are subsidies for wasteful arms races.
With respect to primary and secondary education, the Carden children are being home schooled (I write "are being" because while they're not of school age yet, we take advantage of lots of opportunities to learn cool new stuff). We're letting them follow their developing interests and then using those as contexts in which to explore the natural and social world, at least as far as we can go with kids who aren't yet five, three, or one.
This raises a lot of interesting questions from people who send their kids to traditional schools, and with respect to public policy there are a lot of questions about the effects of homeschooling. It's relatively easy to do in some states and relatively difficult in others, which raises potentially testable questions: what are the effects of homeschooling?
I think the burden of proof in public policy should fall on those who seek power rather than those who seek liberty; in short, there should be a presumption in favor of liberty. I offer the following $100 bet to those who do not think there should be a presumption of liberty with respect to schooling, to be settled on April 1, 2033, which will be when our youngest is finishing college (you might recognize this from Facebook if you're in my network):
Over the next two decades (around the time our kids are finishing college), scholarly research will show that the treatment effect of homeschooling--after controlling for selection effects, ability bias, etc--is either neutral or positive relative to government-operated schools on a majority of academic and social outcomes (test scores like GRE and ACT, # of friends, # of close friends, behavioral problems, parent/child relationships, etc). "Negative and statistically significant" can still count as neutral if the negative effect is small enough to still pass a cost/benefit analysis that we can agree is reasonable. To offer an extreme example, I think we would agree that (say) a 0.0000001 point reduction in ACT score, even if statistically significant, might be worth the trade-off that comes with a better parent/child relationship. The bet will be settled in nominal dollars to keep things easy.