One would expect smaller classes to produce better results. But education policy must be informed by evidence and not just our intuitions. Small classes have a big price tag, but the empirical evidence shows that they don’t produce results.

One of the oldest issues in education policy is class sizes. It’s often used as an example of why schools don’t perform as well as we would expect, or how we could improve their performance. It would be better to use it as an example of why it’s so important for education policy to be informed by evidence and not just our intuitions.

We expect smaller classes to produce better results. Smaller classes must mean more individualized attention for each student, which in turn must mean a better education. That makes sense to me as much as it does to everyone else.

The basis of this intuition is sound. Children are not widgets and education is not the same process for any two human beings. More individual attention is exactly the direction we should be moving in. We have had far too much one-size-fits-all standardization in the classroom.

It does not follow, however, that smaller classes must necessarily be a good policy reform. In fact, the empirical evidence shows that it doesn’t produce results.

Let’s start with the big picture. The ratio of students to teachers in the public school system plummeted in the second half of the 20th century and has declined slowly since then. In 1955, when the U.S. Department of Education began tracking the ratio, there were 27 students per teacher in the public school system. There were 20 in 1975, 17 in 1995, and 16 in 2015. The department projects 15 by 2025.

Across that time period, as the number of students per teacher in the system has plunged, outcomes have been flat. Scores on the National Assessment of Education Progress are flat, and so are graduation rates. You will often hear that education spending per student has been going up rapidly for generations, even on an inflation-adjusted basis, and we’ve had nothing to show for it; that’s true, and one of the most important ways we’ve spent that money without any results has been to hire a lot more teachers per student.

However, the student-to-teacher ratio may not tell the whole story. Some teachers don’t actually have their own classrooms; they have the title “teacher” but work in support capacities. Unfortunately, the Department of Education doesn’t collect data on actual class sizes, as opposed to the student-to-teacher ratio. So, it’s worth looking at further evidence.

California enacted a big class-size reduction policy in 1996. It sounded easy when it was pitched to voters, but it ended up costing the state billions of dollars. And it produced no measurable improvement in any education outcomes—not test scores, not graduation rates, nothing.

Alas, the lesson was not learned. Florida enacted an even more ambitious class-size reduction policy in 2002. It cost the state $20 billion to implement as it was scaled up over eight years, and costs between $4 billion and $5 billion to maintain every year. And it produced no positive effect on education outcomes.

Unfortunately, most “studies” examining class size are not scientifically rigorous. Among the few that are, the most common result is that when class size reductions are very substantial—such as between 22 and 15 students—they produce modest positive effects on students in elementary grades. Unfortunately, the effects don’t last. A few years later, the differences disappear, and the students who were in smaller classes have the same outcomes as others. A major review of the evidence just published by Campbell Systematic Reviews reported disappointing results.

What explains the failure of class-size reduction policy? For one thing, we can only change the size of classes within a limited range. We can’t reduce classes to a size of one student per class.

Small classes have a big price tag, as California and Florida found out. My wife and I looked at private schools before ultimately enrolling our daughter in public school, and we looked at one school where she would have been in a class of six students. The price tag was impressive. And even at that price, the school was doing everything else on a shoestring budget in order to have enough staff to keep classes that small. It closed its doors not long after we looked it over.

Another factor to consider is the teaching labor pool. To make classes smaller, you have to hire more teachers. To hire more teachers, you have to hire the teachers who didn’t make the cut before. In other words, you have to lower your standards. Large-scale reductions in class-size imply large-scale reductions in teacher standards. And teacher quality is one of the factors with the strongest empirical evidence supporting its importance to education outcomes.

Of course, education special interests have their own reasons for advocating smaller classes. Teachers’ unions profit when there are more teachers, and they don’t care about taxpayer costs. But the rest of us can look at other possible ways of improving education.

Intriguingly, smaller schools are a more promising reform than smaller classes. The Bill and Melinda Gates Foundation funded some initial experiments in the 2000s with the idea that smaller schools might create more personal overall learning environments and improve school governance. The positive effects on student outcomes were moderate, but promising. It’s too early to say much more, but further experiments in this direction would be welcome.

And, of course, the ultimate form of individualized attention is school choice. When families direct a child’s education, that’s a class that really does have a size of one. Parents with choice have the power to choose the school that is the right school for their child and hold their schools accountable for success.