Chapter 2: Making Policy Choices: Is Class-Size Reduction the Best Alternative?

Doug Harris
Michigan State University and Michigan Department of Education

David N. Plank
Education Policy Center
Michigan State University

Introduction

Everyone wants to improve student performance. The question is, how? States and school districts across the country have chosen class-size reduction as the answer. Anyone who has ever taught a class or read the available research knows that reducing class size does make a difference. Teachers have more time to give personal instruction to students and spend less time dealing with disciplinary issues.

Unfortunately, knowing that class-size reduction makes a difference is not very helpful for making school policy. Yes, lowering class size can help, but so can improving technology, providing better textbooks, expanding professional development opportunities for teachers, and increasing teacher salaries to attract and keep more able educators. All of these initiatives can improve student achievement. They also cost money. Information about the relative gains and costs of different policies should be accounted for in making the tough decisions about education spending.

How should these choices be made? In theory, all programs should be adopted if the benefits exceed the costs. In the real world, however, funding levels are fixed. Only some of the beneficial programs can be adopted. To get the most out of their limited resources, state and local policymakers must instead try to find the most cost-effective mix of programs—the biggest bang for the buck.

The problem that policymakers face is that they rarely have good information available to compare the costs and benefits of multiple programs. These comparisons are essential for sound decisionmaking, yet education research usually focuses on the benefits and only for individual programs. As a result, administrators and policymakers are forced to rely on instinct, intuition, the demands of key constituencies, and imitation of policies adopted in other districts or states. It is difficult to fault them for this approach, since they generally lack better information on which to base their choices.

In this paper we seek to promote the use of cost-effectiveness analysis to support more informed policy decisions. We illustrate the value of this approach by comparing policies aimed at increasing the number of teachers (decreasing class size) with those intended to increase the quality of teachers. Our results call into question the conventional wisdom about class size. Analysis of the relative gains and costs of these two policies suggests that greater improvements in student performance could be achieved through sustained policies aimed at increasing teacher quality rather than through further class-size reductions. The remainder of this paper provides more detail about these conclusions.

National Trends

Policies affecting class size and teacher quality have changed dramatically in recent years. Table 1 below summarizes these changes from 1960 to 1990.

 

Table 1—Changes in School Resources (1960-1990)1
Year	Class Size	Teacher Daily Wage (1990 dollars)	Proportion of Total Spending
			Teacher Salaries and Benefits	Other Instructional	Administration Other Costs
1960	25.8	$124	0.68	0.07	0.25
1970	22.3	$155	0.67	0.11	0.22
1980	18.7	$143	0.60	0.15	0.25
1990	17.2	$183	0.612	0.12	0.27

Table 1 shows that teacher salaries (adjusted for inflation) have increased dramatically in recent years. What the table does not show is that the salaries of other college-educated workers have also increased. In fact, the relative salaries of teachers have gone down even as real salaries have increased.³ This is especially true for women, who have gained much greater access to many professional jobs.

Table 1 also shows a significant downward trend in class sizes. Together with the increase in teacher salaries, these reforms have produced a large increase in total spending on education. Despite the increase in average teacher wages, however, the proportion of funds going to teachers decreased from 68 percent in 1960 to 61 percent in 1990. This decrease reflects even larger percentage increases in spending for special education and support services.

The Effectiveness of Class-Size Reductions

There are many ways to measure student performance. No single measurement will capture all of the things that parents and educators value. For our purposes, "effectiveness" refers to changes in two common measurements of student performance: test scores and future wages. Test scores might reflect student achievement in knowledge, verbal and quantitative ability, critical thinking, and other academic skills. Wages reflect many different aspects of performance after students leave school, such as the likelihood of graduating from high school, motivation and ability to finish college, people skills, and general work ability.

Experimental research consistently shows that class size affects how students perform on standardized tests. The Tennessee Student/Teacher Achievement Ratio (STAR) study of class-size reduction is the best-known piece of research in the field. STAR was a large-scale experiment that showed clear achievement gains from smaller classes in lower grade levels. The experiment took place in the 1980s and included random assignment of 12,000 students to small and large classes for kindergarten through Grade 3 (K-3). The average large class had 24 students, and the average small class had 15 students.⁴

A more recent experiment is Wisconsin's Student Achievement Guarantee in Education (SAGE) program, which included random assignment of 5,000 students to small and large classes with average sizes similar to the STAR program.⁵ Many other smaller-scale experiments have been conducted. Glass and Smith studied 725 separate estimates of class-size effects based on studies completed before 1979. These studies used varying degrees of sample control and randomization.⁶

The results from these studies are summarized in Table 2. The numbers represent the percentile gains for students who start at the 50th percentile in student achievement and who experience an average class-size reduction of five students (e.g., from 25 to 20) over a period of six years (grades).⁷ These findings assume a flexible reform in which class sizes can be reduced for any grades and student types—as long as the average decreases by five.⁸ According to the Glass and Smith results, the effect of such a reform in elementary grades (1-6) would help a student at the 50th percentile to move to the 52nd percentile—an increase of approximately two percentage points.

The other class-size effects reported in Table 2, such as those for math and science, also assume that the change in average class size occurs over six years. This approach allows for direct comparisons across all of the grade levels, subject areas, and student characteristics listed. (A different number of years would imply a different length of treatment, making such comparisons more difficult.)

 

Table 2—Percentile Gains in Test Scores From Class-Size Reductions (Base = 50th percentile. Treatment = reduce classes by 5 students over 6 grades.)
Study Characteristics	Study
	STAR	SAGE	Glass & Smith
Elementary Grades	2.4	1.0	1.9
Secondary Grades	—	—	4.4
Black	—	1.7	—
White	—	0.8	—
Reading	2.0	0.8	—
Language	—	1.2	—
Math	3.4	2.0	—
Science	2.6	—	—

The most important idea to take away from Table 2 is that reduced class sizes do produce gains in students' academic achievement. The data in the table also suggest that the gains are larger in math than in other subjects. Black students appear to gain more than whites, which is consistent with other qualitative evidence showing that disadvantaged students gain more from class-size reductions. In addition, the table suggests that gains are larger for secondary students, which departs from the conventional wisdom about policies to reduce class size.

The Costs of Class-Size Reduction

In this section, we develop cost estimates for the simple class-size reduction policy described above. The main cost of reducing class size is the cost of hiring more teachers. If an individual school district seeks to reduce class sizes, teachers might move into the district from other school districts. Class-size reduction at the level of a single school district is unlikely to require much change in the total number of teachers in a county or state. State-level policies to reduce class size, however, almost certainly will require additional teachers to enter the workforce. The effects of class-size reduction therefore depend on the number of available candidates for new teaching positions and the qualifications of those teachers.

Research evidence on the supply of candidates for teaching jobs suggests that most districts face a surplus of candidates for teaching positions. The overall size of the pool is uncertain and there is great variation both within and across states.⁹ In the cost model used here, we assume that the pool of candidates is 10 percent of the total number of employed teachers. For example, if a district employs 1,000 teachers, we would assume a surplus of 100 teachers. We vary this assumption from 0 to 30 percent.

Regardless of the type of district, it is likely that new teachers hired in response to class-size reduction policies will be of lower quality than those already employed. This is almost certainly true if districts now are selecting the best available teachers in their hiring processes.¹⁰ The benefits from reducing class sizes are therefore likely to be at least partially offset by a decrease in average teacher quality. One way to prevent this would be to offer starting salaries high enough to attract new teachers of higher ability who are now working in other professions or other districts. This change is likely to be costly, though, because most collective bargaining agreements would require that all salaries be increased, not just those of new teachers.¹¹

To illustrate some of the trade-offs involved with these policies, consider California's recent program that provides $1 billion annually to reduce class sizes in the elementary grades. In many California districts, the evidence suggests that test score gains for students in smaller classes were offset by test score losses for students who ended up with less able teachers.¹² Wealthier districts were able to reduce class size by attracting the best teachers from low-income districts, while low-income districts were obliged to hire less qualified (and often unqualified) teachers. The net effect of the policy was to help some students at the expense of others.

Cost estimates require specific assumptions about the relationships between class-size reductions, the pool of candidates for teaching jobs, and average teacher quality. "Teacher supply elasticity" is the percentage change in the number of available teachers divided by the associated change in teacher salaries. The available research suggests that this elasticity ranges from 0 to 2.0, depending on the time frame.¹³ A change in salaries today obviously will have little effect on the number of teachers applying tomorrow. So the short-term elasticity is essentially 0. As information spreads through the workforce, however, more teachers might enter the pool of candidates. The long-term elasticity might be as large as 2.0, meaning that a 1-percent increase in salaries will produce a 2-percent increase in the number of workers willing to teach.¹⁴

Our model also depends on how teacher salaries might affect the quality levels of teachers who apply. Manski (1987) concluded that a 10-percent increase in average teacher salary would increase the average teacher SAT score by 10 points. Ballou and Podgursky (1992, 1994) and Figlio (1997) obtained similar results.

Class-size reduction policies might also require capital expenditures if more classrooms are needed to accommodate students. Two facts suggest that the cost of classroom space is small compared to the cost of teachers, however. First, capital costs represent only 10 percent of total education expenditures. In addition, half of U.S. schools have extra space available that cannot easily be used for other purposes.¹⁵ For these reasons, we initially excluded capital costs from the model.¹⁶ If the results had shown that class-size reductions were more cost-effective, then it would have been necessary to include capital costs in our calculations. However, the results presented below suggest just the opposite. Adding capital costs to the model would make class-size reductions compare even less favorably to increases in teacher salaries.

The cost model described in this section, combined with the effectiveness information in the previous section, suggests that it would cost $198 annually per student to move a student from the 50th to the 52nd percentile by reducing average class size by five students. These costs will vary depending on assumptions about labor supply elasticity, quality elasticity, and capital costs. Changes in these assumptions do not change the main conclusions of the paper, however.

The Cost-Effectiveness of Policies Designed to Increase Teacher Quality

So far, the term "teacher quality" has not been carefully defined. For our purposes, teacher quality refers to the ability of teachers to improve student outcomes. We make no claims about what these characteristics might be. Instead, we focus on research evidence showing the effects that teacher hiring policies have on students.¹⁷

There are two main ways to increase teacher quality. The first is to hire teachers who begin their jobs with high ability. The second is to develop teachers with high ability through training and professional development once teachers have been hired. The focus of attention here is on hiring better teachers, specifically through increases in teacher salaries. Although teacher salary increases could be implemented in many ways, the focus here is on across-the-board changes.¹⁸

The choices that potential applicants must make include whether to take formal teacher training (usually in college), whether to stay in a particular job or switch jobs after entering the profession, whether to stay in the teaching profession or change professions, and whether to continue working or to leave the labor force. Considerable research has been done about how prospective and current teachers make these decisions.¹⁹ Not surprisingly, the reasons are complex and many are not under the control of school districts or state governments.

As suggested earlier, salary is certainly one important factor in determining who teaches. Unfortunately, in contrast to the research on class size, there is no experimental evidence regarding the effects of teacher salaries on student achievement. Instead, most studies on teacher salaries use a statistical tool called regression analysis that tries to imitate experimental conditions. This nonexperimental technique requires that the regressions include control variables that affect student performance, including class size and student characteristics. Regression estimates that exclude these variables suffer from omitted variable bias and can yield misleading estimates of the effects of the teacher salaries.²⁰

Most nonexperimental studies find that salary matters, but the effects appear to be small.²¹ Most of these studies suffer from the omitted variables problem, however, which means that their results might be misleading. To identify unbiased estimates of how teachers' salaries affect student achievement, we first selected regression estimates that included both class size and teacher salaries in the analysis. This is important because we do have experimental estimates of the class-size effect. We then restricted our attention to regression estimates in which the class-size effect was close to the experimental estimates found in STAR, SAGE, and other studies. If one of the effect estimates is unbiased, it is more likely that the other effect estimates are unbiased as well.²²

Twenty regression estimates from seven different studies include both a class size and a teacher-salary variable. Eight of these estimates were left after restricting this sample based on the class-size evidence. Taking the average of these teacher salary effects suggests that a statewide increase of 10 percent in teacher salaries will raise a student from the 50th to the 52nd percentile on a norm-referenced test.

Multiplying the change in teacher salaries by the number of teachers yields the total costs of the reform. As with the class-size reduction policies discussed above, policies that seek to increase teacher quality by increasing salaries will affect the number of teachers available. The final cost will depend again on teacher supply elasticity, teacher quality elasticity, and other factors. We make the same assumptions here as we did regarding class sizes. Our results suggest that raising a student from the 50th to the 52nd percentile using teacher salary increases would cost about $100 per student per year.

Cost-Effectiveness and Simple Policy Reforms

Most research provides information about benefits or costs of individual programs. The purpose of analyzing cost effectiveness is to combine information on costs and benefits for various programs to obtain the highest possible level of student achievement. Policymakers need more than just information about benefits—they need to know how they can get the biggest bang for the buck.

The costs and benefits described above for class-size reductions and teacher salary changes are presented in a way that makes these comparisons easy. The cost of increasing student test scores from the 50th to the 52nd percentile is estimated to be $198 per student for class size, but only $100 per student for teacher salaries. These results vary somewhat based on assumptions about labor supply elasticity, labor quality elasticity, and other factors. We considered ranges of values for these factors, but even values at the extreme ends of these ranges did not alter the main conclusion.

The results above suggest that sustained teacher salary increases would be more cost-effective than class-size reduction in raising student achievement. The short-term effect of a change in class size would almost certainly be greater, but the long-term effect would be smaller. As teachers leave the profession, higher salaries would attract better candidates for teaching positions. Over time these new teachers would have a greater impact on student performance at a lower cost.

The discussion thus far has focused on improving student test scores as the main goal of education. An additional purpose of education is to prepare students for their adult lives, including their careers. Different types of analysis are possible when the focus is on increasing student wages, since both costs and benefits are expressed in dollars. This contrasts with an analysis in which test scores were the outcome of interest. Test scores may be related to personal success in adult life, but they are also related to our general desire for knowledge and good citizenship. It is difficult to place a dollar value on these goals and values.

As with test scores, it is important to have precise estimates of the benefits that reducing class size or increasing teacher salaries provide for future student wages. Our review of the evidence suggests that a $100 increase in teacher salaries would increase students' future wages by 0.8 percent. A decrease in class size by one student in all grades would increase students' future wages by 1.6 percent.

Again, it is necessary to combine the above results about policy benefits with information about policy costs. One simple and useful way to approach the analysis with future wages is simply to ask, "At what point do the extra costs of education reforms exceed the extra future wages for students due to the reforms?"²³ The results of our analysis are similar to those in the earlier sections. The absolute size of the gain in student wages is larger when we reduce class size than when we increase teacher salaries, but the cost of the change is also higher. This suggests that it would be more cost-effective to raise teacher salaries than to reduce class size further.²⁴

Cost Effectiveness: The Details Matter

Real-world policies are rarely as simple as those considered in the previous sections. Instead, policies almost always are filled with fine print, restrictions, and rules that add complexity. This section focuses on how the earlier results might change based on differences in policies and different cost assumptions.

Possible restrictions on class-size policies include maximum class sizes, restrictions by grade, and unique rules for special education and other student populations. All of these restrictions would add to the cost of reform. For instance, suppose a school imposed a maximum class size of 25. This would require splitting a class of 26 into two very small classes (13 and 13) and hiring an additional teacher, because a class of 26 would violate the rule.²⁵ Imposing this rule would make class-size reduction far more costly; more flexible reforms are generally cheaper.

Similarly, there are many different ways to increase teacher salaries, ranging from signing bonuses for new teachers to across-the-board increases for all teachers to merit pay based on professional development, experience, peer evaluations, or other factors. These different strategies can produce similar increases in average teacher salaries, while producing very different increases for particular teachers and very different consequences for the pool of candidates for teaching jobs. The details of policy design are therefore likely to make a big difference in how salary increases affect student achievement.

The available evidence suggests that increasing the average quality of teachers by increasing teacher salaries will improve student achievement, but our research does not provide guidance for choices among alternative strategies. As in the case of class-size reductions, however, complicating salary policies by adding rules might increase the cost of reform, sometimes in unforeseen ways. For example, offering signing bonuses or increased salaries for beginning teachers might be a good strategy for attracting talented young people to the teaching profession. Such policies might produce dissatisfaction or resentment among veteran teachers, however, leading to increased turnover or reduced retention. It clearly would be desirable to design policies that targeted salary increases in ways that would reward or retain more able or more talented teachers, but it is far from clear that administrators or policymakers are able to make the fine distinctions among teachers that would make such policies effective.

Policies to reduce class size or increase teachers' salaries could be implemented at various levels of the education system, from individual school districts to the federal government. In our analysis, we assume that these policies are part of state or federal reforms. Action by an individual district might make local students better off, without having much impact on the larger education system. Any single school district that reduces class sizes or increase salaries can probably attract the new teachers that it needs from nearby districts or private schools. State or federal policies that affect all school districts are likely to bring about major changes in the market for teachers, however. As the California experience suggests, these changes might benefit some schools and school districts at the expense of others unless policies are designed with care.

Conclusion

The results presented in this paper run against the grain of many recent policy initiatives in education, at both the state and federal level. Our research suggests that sustained increases in teacher salaries will be more cost-effective than class-size reductions in increasing student achievement. This is true regardless of whether we focus on academic achievement or students' future wages.

If research suggests that salary increases and increasing teacher quality are so cost-effective, why might actual policy decisions differ? One possible explanation is that class-size reductions have a much more immediate and identifiable impact on student performance. Teachers like class-size reduction because it has demonstrable effects on student achievement and allows them to focus more closely on each of their students. Parents like the reductions because they want their kids to have personal attention. This puts a great deal of pressure on school leaders to reduce class sizes, even if it is not the most cost-effective use of resources. In contrast, changes in teacher quality are hard to observe, and they might affect student achievement only over an extended period of time. Teachers might appreciate the importance of higher salaries, but making the case to parents and taxpayers is harder. Hiring fewer teachers and paying them more in order to raise teacher quality simply lacks the intuitive political appeal of hiring more teachers and reducing class size.

How can we improve student performance? Past research is useful, but it falls far short of providing a definitive answer to this question. Researchers and educators, devoted to finding a solution to pressing educational needs, search hard for programs that work. Through innovation and creativity, many successful programs have been developed in the process. These successes will remain meaningless if we cannot successfully choose among them, however. The best apple in the barrel is no better than the worst if we cannot find it. Instead of reaching out and picking the first good one we see, it is probably wise to dig deeper, compare several possibilities, and choose the best of the bunch. These choices are not merely a matter of economics—of mundane calculations of benefits and costs. Rather, the ability of educators to make these tough choices will determine how much we can contribute to the learning of our students.

Endnotes

1. Source: Digest of Education Statistics (1999).

2. This number is estimated because information about teacher benefits was not available for 1990.

3. See Hanushek and Rivkin (1996).

4. Nye, Hedges, and Konstantopoulos (1999) describe the study in greater detail. Other papers focus on the weaknesses of the project and previous interpretations of results, including attrition and some violations of random assignment.

5. The SAGE program, which started in 1996, also included staff development, after-school programs, and a new curriculum for kids in the treatment group. While this might appear to complicate the analysis, Molnar et al. (1999) found that these other programs had no significant effect after controlling for class size.

6. Hedges and Stock (1983) reanalyzed the Glass and Smith sample with some modifications, but found that these changes did not affect their results.

7. All calculations based on test scores assume that student performance is normally distributed.

8. More precisely, the policy experiment we propose assumes that schools will reduce class sizes in the same manner as those schools used in the studies. This approach ignores possible nonlinear effects. For instance, the effect of decreasing class sizes from 30 to 25 might be different from decreasing class sizes from 25 to 20.

9. See, for example, Ballou (1996).

10. Supporting evidence for this conclusion is presented below for California's recent class-size reduction. There are certainly some circumstances in which this would not be reasonable. For instance, the assumption might be false if the new teachers are expected to provide instruction on computers or other skills with which newer, younger teachers might be more familiar. Other reasons might also prevent this from being the case and these are discussed in later sections. However, these reasons do not appear to be strong enough to contradict the general conclusion that teachers on waiting lists are less able on average.

11. One way around this rule is the signing bonus, which provides one-time payments only to new teachers.

12. Class Size Reduction Research Consortium (2000) [Online]. Available: http://www.classize.org.

13. It is extremely difficult to estimate elasticity for particular time frames—e.g., short term versus long term. However, the distinction is important.

14. See Manski (1987), Currie (1991), Ballou and Podgursky (1994).

15. See The Condition of America's Public School Facilities: 1999 from the U.S. Department of Education. Also note that if extra space could be rented out, then the cost would include this rental value.

16. Excluding capital costs certainly does not affect the main conclusion of the paper. If capital costs were included, class-size reductions would fare even worse when compared with teacher salary increases.

17. There is substantial evidence that certain teacher characteristics are associated with student success, including teacher test scores and years of experience. Again, no assumptions are made about these characteristics in this paper.

18. This is an important assumption given the substantial evidence that the way in which salaries are now paid is far from cost-effective. See Manski (1987).

19. These exams may also exclude teachers who lack verbal and quantitative skills, but who have other valuable abilities.

20. Some econometric estimates are based on "panel data," which can produce unbiased estimates of effect sizes. Unfortunately, such data are rare, especially regarding the effects of school resources, such as class size and teacher salaries.

21. See, for example, Hanushek and Pace (1997), Ballou (1992), Ballou and Podgursky (1994).

22. To check the validity of these estimates, we also estimated the effect that teacher salaries have on average teacher test scores (e.g., SAT scores from college or certification exams). We then reviewed the evidence regarding the effect of teacher quality on student outcomes, where teacher quality is measured by teacher test scores. The results were similar to those obtained using the procedure described in the text. See Harris (2000) for details.

23. Economists use the terminology, "marginal costs and marginal benefits." The point at which these are equal is called the "optimal" level.

24. This approach requires a discount rate that reflects the common assumption that future benefits and costs should receive less weight than current benefits and costs. The base discount rate in this model is 3 percent with a range of 0 to 6 percent. In addition, this is a partial equilibrium model in which the interrelationships between some markets (e.g., the teacher labor market and the market for other workers) are ignored.

25. The costs of class-size reductions are interrelated with two other polices of many schools and states. First, schools-of-choice policies make it increasingly difficult for schools to shift student populations across schools to avoid these class splits. In middle and high schools, an additional problem is that students request specific classes, and teachers specialize in ways that prevent splitting classes.

References

Ballou, D. (1996). Do public schools hire the best applicants? Quarterly Journal of Economics, 111(1), 97-133.

Ballou, D., & Podgursky, M. (1994). Recruiting smarter teachers. Journal of Human Resources, 30(2), 326-38.

Ballou, D., & Podgursky, M. (1992). Buying quality: Will higher pay improve the teaching workforce? Working paper, University of Massachusetts at Amherst.

Brewer, D. J., Krop, C., Gill, B. P., & Reichardt, R. (1999). Estimating the cost of national class size reductions under different policy alternatives. Educational Evaluation and Policy Analysis, 21(2), 179-92.

Currie, J. (1991). Employment determination in a unionized public-sector labor market: The case of Ontario's school teachers. Journal of Labor Economics, 9(1), 45-66.

Ehrenberg, R. G., & Brewer, D. J. (1994). Do school and teacher characteristics matter? Evidence from high school and beyond. Economics of Education Review, 13(1), 1-17.

Figlio, D. (1997). Teacher salaries and teacher quality. Economics Letters, 55, 267-71.

Goldhaber, D., & Brewer, D. (1997). Why don't schools and teachers seem to matter? Journal of Human Resources, 32(3), 504-23.

Hanushek, E. (1986). The economics of schooling. Journal of Economic Literature, 24, 1141-77.

Hanushek, E., & Pace, R. R. (1995). Who chooses to teach (and why)? Economics of Education Review, 14(2), 101-17.

Hanushek, E., & Rivkin, S. (1996). Understanding the twentieth-century growth in U.S. school spending. Journal of Human Resources, 32(1), 35-68.

Harris, D. (2000). Different methods, different results: New approaches to meta-analysis with applications to education production functions. Doctoral dissertation, Michigan State University.

Hedges, L., & Stock, W. (1983). The effects of class size: An examination of rival hypotheses. American Educational Research Journal, 20(1), 63-85.

Levin, H. (1983). Cost Effectiveness: A Primer. Sage: Beverly Hills.

Manski, C. F. (1987). Academic quality, earnings, and the decision to become a teacher: Evidence from the national longitudinal study of the high school class of 1972. In Public Sector Payrolls, David A. Wise, Ed., University of Chicago Press: Chicago.

Molnar, A., Smith, P., Zahorik, J., Palmer, A., Halbach, A., & Ehrle, K. (1999). Evaluating the SAGE program: A pilot program in targeted pupil-teacher reduction in Wisconsin. Educational Evaluation and Policy Analysis, 21(2), 165-77.

National Center for Education Statistics (1999). Total and current expenditure per pupil in public elementary and secondary schools: 1919-20 to 1997-98. Digest of Education Statistics, Washington, DC.

National Center for Education Statistics (1999). Teacher salaries in public elementary and secondary schools. Digest of Education Statistics, Washington, DC.

National Center for Education Statistics (1999). Pupil-teacher ratios in public elementary and secondary schools. Digest of Education Statistics, Washington, DC.

National Center for Education Statistics (1999). Condition of America's public schools. Digest of Education Statistics, Washington, DC.

Norris, G., & Richburg, R. (1997). Hiring the best. American School Board Journal, 184(11), 46, 48, 55.

Nye, B., Hedges, L., and Konstantopoulos, S. (1999). The long-term effects of small classes: A five-year follow-up of the Tennessee class size experiment. Educational Evaluation and Policy Analysis, 21(2), 127-42.

Previous | Table of Contents | Next