Friday, July 9, 2010

What the class size research REALLY says.

As school system budgets tighten, more journalists find themselves writing about—and misinterpreting the research on—class size. Nearly every education writer knows about Project STAR, the only large-scale, random-assignment experiment that has been conducted on class size. Over four years in the late 1980s in Tennessee, researchers assigned children in 79 schools to classrooms ranging from 13 to 25 students. They found significant academic advantages in reading and math for students in small kindergarten and first-grade classes, and the effects diminished in second and third grades. We do not know much about is what kind of difference class size makes outside the parameters of that experiment, at least not with the certainty that comes with the methodological rigor of an experiment like STAR.

To nail down “what the research says” about class size, I contacted three authorities on the topic: Eric Hanushek at Stanford, Thomas Dee at Swarthmore and Beth Graue at University of Wisconsin-Madison. This issue, like nearly any in education, divides people somewhat; I think it’s fair to say that Graue and Dee think reducing class size has more merit than Hanushek does. But they all agree that it is not accurate to extrapolate and say, as some do, that because of STAR we know that class size “only matters” in kindergarten and first grade. Some quasi-experimental studies show benefits of smaller classes; some do not. But there isn’t high-quality evidence to confirm one way or another. There has not been another study the likes of STAR, and in this climate—small class size is out of favor among national reformers, and it is very expensive—I am not sure we will see one, though some researchers would like to make that happen.

In my gut, I cannot help but think the number of kids in a room matters in a variety of ways (what sort of assignments teachers choose to give, how connected students feel to their teachers) when we are talking about big differences in size. To take an extreme example: Just because nobody has studied what would happen if we increased the size of an eighth-grade math class from 28 to 58 doesn’t mean the outcome would be pretty. Of course, on average across a big district, class sizes usually increase only a couple of students at a time. But in individual schools and classrooms, big disparities in class size can happen. We do not have much research on this kind of thing; Dee recommends a study from the early 1990s by Joshua Angrist in Israel, where a maximum class size of 40 made for interesting comparisons. (Forty kids in a grade? One class of 40. Forty-one kids? A class of 20 and another of 21.) For those interested in looking at larger classes than evaluated in STAR, Graue recommends the work of Peter Blatchford in the U.K., who compared classes in the mid-20s to those above 30 students. 

Hanushek points out that an issue of key importance is how teachers are hired (in the case of class-size reduction) or fired (in case of increases). He suggests that changes in teacher quality borne of such decisions matter as much as—probably more than—the number of children in the room. If a district reduces class size without a pool of good teachers to draw from, why would you expect improvement? If a school increases class size and has to get rid of teachers with no regard to their effectiveness, the big classes might matter less than losing good educators.

This is all to say, be careful about how you represent the research—only STAR represents the so-called gold standard of research, and even that has its limitations. I suppose “We don’t know if it matters” is not a great line to stick in your story when the school board is debating whether to enlarge classes. If this is an issue in your area, do some shoe-leather reporting comparing classrooms and give qualitative examples. Did a teacher whose classes got bigger stop assigning essays when she felt like she had too many to grade? To what degree do teachers attribute differences in classroom climate to the number of students versus the other myriad factors? I could go on; there is a lot to look at. While doing so,  remember that there is a difference between a teacher who says he would leave if his class were significantly larger and one who actually does so. 


  1. To play devil's advocate, I should point out that Montessori education actually encourages larger class sizes. Because of the mixed age groups and emphasis on independence, it works well. I don't know if there have been any studies on this. It would be an interesting comparison if there were.

  2. Some of the best teaching I have seen has been in China and Japan where class sizes are often double or more of those in the U.S. But teaching loads are also about half of what they are here. Isolating class size as the lone variable as in STAR makes for interesting research, but what we really care about is how to improve learning in the most economical way possible. So the real question is: What combination of class size and teacher prep time gives the best results for the least resources? Posing the question this way also has the practical effect of depoliticizing the class-size issue that clouds the issue. Unions can abandon their knee-jerk opposition to large class sizes as just another teacher-bashing tactic. Cash strapped school boards would have harder time cherry-picking data to justify increasing class size with no adjustment to teacher work load.

  3. @Carl, learning IS the issue here, I totally agree with you. I also agree that this is one of the biggest issues for the teacher unions, and one that - at least i in Sweden - has been left out in the discussions about school and teachers for a very long time. As I see it there has to be a limit somewhere, a limit for what is reasonable in respect to what tasks you are expected to performe in addition to the definition of the teacher job in itself. There has to be some kind of correlation between the amount of time teachers are able to spend preparing and evalutaing their students learning, hasn´t there?