This is getting old. I find myself correcting false claims regarding the scientific evidence on private school choice all too often. For example, using only one correlational study that did not detect any statistically significant effects, Valerie Strauss recently concluded that “private schools aren’t better at educating kids than public schools” in the Washington Post. As I have pointed out many times before, the preponderance of the causal evidence indicates that school voucher programs in the U.S. improve student test scores and more important outcomes such as high school graduation, college enrollment, and tolerance of others.

But science shouldn’t determine whether families are allowed to pick the schools they want for their kids.

Just imagine if we used the scientific evidence to decide whether people should be residentially assigned to their nearest government grocery store. What if we randomly assigned thousands of families to government-run grocery stores and found that, on average, those families consumed fewer calories than the families with the freedom to shop for groceries on their own? Such an empirical finding certainly wouldn’t mean that the government should compel all individual families to accept the grocery basket deemed ideal by the experts. After all, a crude metric such as caloric intake can only tell us so much about how well an individual’s nutritional needs are being met. And, of course, some people may simply value eating appetizing foods more than the benefits of having a lower BMI. It would be a disgraceful limitation of freedom to compel people to consume—and pay for—a basket of groceries they did not want.

Yet this is precisely the type argument often used against freedom in education—that parents are somehow unfit to choose schools for their own kids.

But it’s worse than that with education because most of the random-assignment evaluations find that students are better off when their parents are allowed to choose their schools. And even though the most rigorous scientific evidence available says families should have school choice, less than one percent of the school-aged population in the U.S. uses a private school choice program.

But why shouldn’t society use science to force other people to do the “right” things?

Of course, science can tell us a lot about the world around us. And random assignment (the “gold standard” of scientific research) is the best thing we have available to determine how certain treatments (or policies) affect groups of people. However, even the best methodology has important flaws that do not allow researchers to give central planners enough information to make good decisions for individual families.

For example, since the internal validity of experimental design relies on something called the law of large numbers, the methodology only allows researchers to determine the average effects of policies for large groups of people. In other words, even the best scientific methodology that exists cannot determine the effect of policies on specific individuals. And then there is the external validity problem—even if a school choice program is found to have large positive (or negative) effects on one group of students, on average, it is unlikely that the effects will be exactly the same for the other cohorts of students or in different settings. Similarly, the effectiveness of school choice programs—and the supply of private schools—can change over time.

And education technocrats frequently use faulty measures of success—standardized test scores—because it is extremely difficult for researchers to get their hands on more important long-term outcomes such as earnings, crime, and character skills. The main problem is that effects on math and reading test scores often do not predict effects on more important long-term outcomes. For instance, a recent American Enterprise Institute review of the evidence found that 61 percent of school choice programs’ effects on math test scores—and 50 percent of effects on reading test scores—did not predict effects on high school graduation. And at least 11 other studies have found divergences between private schools’ effects on test scores and their effects on more important long-term outcomes. For example, a peer-reviewed evaluation found that private school vouchers in Ohio had no effect on test scores but increased students’ charitable donations by 23 percent. In other words, focusing too much on standardized test scores could compromise the character skills necessary for true lifelong success.

And that’s just the tip of the iceberg. Many evaluations do not use random assignment—and the problems only get much worse when weaker empirical methods are used. Put simply, central planners face a severe knowledge problem with education today—just as central planners faced severe knowledge problems with five-year plans in the Soviet Union.

The fact is that families have more information about what their individual children need than education bureaucrats and scientists sitting in offices—hundreds of miles away. Families are also more interested in their own children’s lifelong success than anyone else.

The strongest scientific evidence we have on the subject suggests that private school choice works. But that really shouldn’t even matter. Just as people have the right to pick their own groceries, people should have the right to pick the schools that they believe will work best for their own kids. And just as government officials cannot force families to eat at particular restaurants, government officials shouldn’t be able to force families to send their kids to failing government schools.