How to Improve School Productivity? Caroline Minter Hoxby

Published September 1, 2001

Although it’s only seven years since Caroline Minter Hoxby received her Ph.D. in economics from the Massachusetts Institute of Technology, the brilliance of her subsequent research at Harvard University into the effectiveness and cost of public education has established her as one of the leading economists of the day.

What is particularly striking about Hoxby’s research is her insightful ability to identify natural variations in different public school conditions that are equivalent to the implementation of different public policy proposals for class-size reduction, school district consolidation, and centralization of education funding.

Hoxby’s research conclusions often contradict conventional wisdom and consequently have rankled teacher union leaders and members of the public education establishment.

Among her findings:

  • Competition among public schools improves productivity by lowering costs and increasing student achievement;
  • School choice results in higher pay for more effective teachers;
  • The local property tax is one of the best and most stable methods of financing public schools;
  • Class size has no effect on student achievement; and
  • Teacher unions lead to lower student achievement.

Currently a professor of economics at Harvard University, Hoxby comes from a family of educators. Her mother was a teacher and her father, now involved in the reform of Cleveland’s public schools, was an undersecretary of education in the Carter administration. Hoxby received an A.B. in economics from Harvard University in 1988, a M.Phil. from England’s Oxford University in 1990, and a Ph.D. from MIT in 1994. At all three institutions, the theses she submitted each garnered an award.

The author of numerous research papers on the economics of public education, Hoxby has testified before Congress, state legislatures, and courts on school finance equalization, tax policy for education initiatives, and charter school legislation. Besides teaching graduate and undergraduate classes at Harvard, her other affiliations include Faculty Research Fellow with the National Bureau of Economic Research, senior advisor to the Brookings Institution, and membership in the Hoover Institution’s Koret Task Force on K-12 Education.

During a brief break from her summer research projects, Hoxby spoke with School Reform News Managing Editor George Clowes.

Clowes: Tell me a little bit about your background and how you came to be involved in looking at productivity in the education sector?

Hoxby: In two ways my background is related to education. For an economist like myself, it’s natural to be interested in the reason why the education sector, which is very important in the United States, has low productivity and very poor productivity growth compared to other sectors in the U.S. Economists are increasingly interested in sectors that are neither for-profit nor strictly government-controlled, but are somewhere in between

. Health care and education are two such sectors.

The other reason I am interested in education is that I care about children and the kind of a life they will face if they do become well educated. American children need to be educated so that they can contribute to the future economy of the United States. I think that the education sector is in crisis because it is no longer doing an adequate job of providing the key input–educated workers–for our complex and technology-intensive economy. You simply cannot build an information-oriented economy with workers who are poorly educated.

Clowes: You have said that American schools are in a productivity crisis. What are the symptoms of this, and how did we get into this situation?

Hoxby: The main symptom of the productivity crisis is the fact that productivity has fallen almost 50 percent in the past 30 years. We measure productivity by dividing a measure of student achievement by per-pupil spending in inflation-adjusted dollars. Regardless of which achievement measure we use, we find a decline in productivity of 40 to 50 percent. This is because achievement has been flat or slightly declining, while costs have been escalating rapidly.

Most people favor measuring achievement with standardized tests that are given to a representative sample of students in the nation. The most appropriate test for this purpose is the National Assessment of Educational Progress, because it is designed to trace the achievement of American students over time. However, you would get similar results with other measures, such as the percentage of students who graduate from high school or scores on tests like the Stanford-9 or the Iowa Test of Basic Skills.

Clowes: How are schools able to increase costs so enormously without increasing the quality of the product?

Hoxby: Schools don’t face enough competition. Just imagine competing grocery stores. If one of them decided to increase its prices but offer exactly the same products, people would go to the other grocery stores and the store would quickly go out of business. Unfortunately, parents do not have sufficient opportunity to change schools so that they can say, “My school is more expensive than before and it doesn’t seem to be doing better than other competing schools. Therefore I’m going to send my child to another school.” In other words, competition in the education industry is too weak to be an effective brake on costs.

In addition, some factors that have decreased productivity affect many schools and are, therefore, difficult to combat in a low competition environment.

For instance, many schools go through curricular fads that last only two or three years. The fads have very little effect on achievement, but they are costly to implement, so they lower productivity. However, a fad will hit most schools in an area at the same time, so unless there is strong competition, schools will jump on the latest bandwagon.

Clowes: What has been the effect of teacher unions on productivity?

Hoxby: Unions are another great example. In the United States, there are both private-sector unions and public-sector unions, like the teacher unions.

In the private sector, union demands are constrained by the firm’s need to compete in the marketplace. A union knows perfectly well that it should not negotiate a contract that would cripple the firm vis-a-vis its competition and drive the firm out of business.

In contrast, the teacher unions can be very demanding for their members, who often want contracts that are not best for students. There’s no effective brake on the teacher unions because they are not forced to negotiate contracts that permit their schools to remain competitive. There is just not much of a competitive marketplace out there.

For instance, suppose all the local teacher unions say, “Look, we want teachers to spend fewer hours in the classroom.” The superintendent who opposes that request faces an unpleasant battle. More importantly, the superintendent won’t be rewarded for sticking his or her neck out. Parents cannot come flocking to the superintendent’s school because there isn’t any mechanism that allows them to do that. Thus, there is almost no incentive to be the superintendent who fights a bad contract or an increase in wages that isn’t accompanied by an improvement in teaching. Without competition, a superintendent cannot get rewarded by the marketplace.

If there were competition, teacher unions would not have nearly as bad an effect on productivity as they currently have. Competition, in and of itself, reins in the most excessive of union demands.

Clowes: That leads to the issue of class-size reduction, which both Bob Chase and Sandra Feldman recently claimed in The Washington Post would increase student achievement. Would it?

Hoxby: The evidence suggests that class-size reduction has no effect on student achievement. There are literally hundreds of studies of class size, which have been very effectively reviewed by Eric Hanushek, that suggest that reducing class size does not raise achievement.

I have a study in which I examined every change in class size at every elementary school in Connecticut over a 20-year period. In schools, class size varies from year to year because enrollment varies. Therefore, with 20 years and 800-some schools, there is a tremendous amount of variation in class size to examine.

I found there was no effect of class size on achievement at all, even when children were in small classes for all six years of elementary school.

There really is only one study in which a class-size reduction improved student achievement: the Tennessee STAR experiment. But the effect on achievement was tiny–a 10 percent reduction in class size raised achievement by two one-hundredths of a standard deviation in achievement test scores.

More importantly, in the Tennessee STAR experiment, everyone involved knew that if the class-size reduction didn’t affect achievement, the experimental classes would return to their normal size and a general class-size reduction would not be funded by the legislature. In other words, principals and teachers had strong incentives to make the reduction work. Unfortunately, class-size reductions are never accompanied by such incentives when they are enacted as a policy.

Also, it is worth keeping in mind that class-size reduction is very expensive. A 10 percent reduction in class size typically costs about $850 per year per student. For such a large amount of money, we could fund programs that are much more likely to raise achievement.

Reducing class size is a perfect example of a policy that the teacher unions like, but that lowers school productivity. The unions like it because each teacher has less work and more teachers have to be hired.

Clowes: Chase and Feldman also compare class-size reduction to school vouchers, which they claim don’t have any effect. But I understand your analysis of the Milwaukee voucher program shows they do have a significant effect.

Hoxby: Milwaukee is interesting because it’s really the only voucher program in the United States that is big enough to be a threat to the local public school system.

The Milwaukee voucher program started in 1990, but for the first eight years of its existence it was poorly funded and forced to remain very small. No more than 1,000 children were allowed to receive vouchers, and parents could never be certain the program would exist in the next school year.

But, starting in the 1998-99 school year, the ceiling on the number of vouchers was raised to 15,000 students, and the voucher amount increased substantially, to about $5,000 per student. Also, the voucher program was approved by the courts, so its existence was not so precarious. Thus, it suddenly became a threat to the Milwaukee public schools. More than 65,000 Milwaukee public school students became eligible for vouchers, and there were some schools in Milwaukee where 90 percent of the students were eligible for vouchers.

From 1998-1999 onwards, the schools that faced the most competition from the vouchers improved student achievement radically–by about 0.6 of a standard deviation each year. That is an enormous, almost unheard-of, improvement. Keep in mind the schools in question had had a long history of low achievement. Yet they were able to get their act together quickly. The most threatened schools improved the most, not only compared to other schools in Milwaukee but also compared to other schools in the state of Wisconsin that served poor, urban students.

Milwaukee shows what public school administrators can tell you: Schools can improve if they are under serious competition.

Most superintendents and principals know where the problems are in their schools. Unfortunately, they often cannot make needed changes because they are opposed by constituencies like the teacher unions. Such constituencies back down only when a school’s existence is threatened.

For instance, administrators usually know which teachers aren’t performing, but they hesitate to fire them or pressure them to quit because the union will fight the change. However, when a school is faced with losing students–and, thus, teaching positions–the union will decide that getting rid of under-performing teachers is the lesser of two evils.

Clowes: What kinds of policies would restore school productivity?

Hoxby: Although one might say, “We shouldn’t have teacher unions because they’ve harmed school productivity,” one has to recognize that the unions are here to stay. Thus, it is more reasonable to say that, if we are going to have unions, then we need to have competition. Competition can ensure that union demands are in concert with the goals of improving student achievement and satisfying parents.

In short, I think school choice is probably the policy most likely to restore school productivity.

The other policy that can help is spreading information about school performance. Such policies are known by a variety of terms–school accountability, school report cards–but the key thing is that parents, voters, and legislators get solid information about achievement and school costs.

Parents, voters, and legislators are supposed to control schools. How can we expect them to a good job if they do not have information?

Clowes: Could you comment on the quality of the voucher studies that have been done through Paul Peterson’s group using randomized field trials?

Hoxby: The best way to compare achievement in voucher schools and regular public schools is to do randomized access to the voucher schools.

In a medical experiment, some patients are randomly assigned to be treated and others are randomly assigned to be in the control group. In a voucher experiment, it’s ideal if some students are randomly given access to vouchers and others are randomly assigned to the control group. The “control” students do not get vouchers, so they remain in whatever school they were in.

If there is random assignment in the experiment, it is easy to learn how vouchers affect achievement. Anyone with a calculator can compare data from the group of students who go to the voucher schools and data from the control group. Therefore, a researcher can get high-quality evidence by:

1) Setting up the experiments properly; and,

2) Making a serious effort to collect data from both the voucher and control students.

I’ve read all of the studies by Peterson’s group, and I’ve seen them present a number of their studies. All of their studies use a proper randomized setup, and my sense is that they have tried very hard to gather as much data as possible. Of course, the group has not had perfect success gathering data on every voucher experiment, but some of that is inevitable.

In their most recent studies, where they’ve had more control over the data gathering, they’ve had very good rates of success by normal survey standards. I do not know that anyone could do much better.