|
Clearinghouse on Educational Policy and Management
|
| College of Education · University of Oregon |
|
|
Class-Size Reduction: Effects and Relative Costs By Lawrence O. Picus This article is adapted from a forthcoming book on Resource Allocation by Dr. Picus. The book will be published by the Clearinghouse in fall 2000.
Perhaps the hottest state educational policy initiative in the nation today is the move to reduce class size, particularly in the primary (K-3) grades. In recent years, a number of states have passed legislation either mandating smaller classes in elementary grades or establishing incentive programs to finance smaller classes. Few public policy proposals are more popular than class-size reduction. In March 1997, a Wall Street Journal poll found that 70 percent of adults believe reducing class size would lead to big improvements for public schools. A 1997 Education Week survey found that 83 percent of teachers and 60 percent of principals believed classes should not exceed 17 students (Bell 1998). Parents say their children are happier and learn more in smaller classes. Teachers report they have fewer discipline problems, are able to give students more individual help, and can cover material faster. Many states have enacted class-size reduction measures in recent years. Perhaps best known is California’s effort to reduce the size of all K-3 classrooms to no more than 20 students. Tennessee has had a program in place since 1990 to reduce class size, while Texas mandates that all K-4 classrooms in a school average no more than 22 students. Most states that implement class-size reduction seem to set average K-3 class size at around 20 students. Nevada has the lowest mandated size, requiring no more than 15 students per class. Washington was a leader in this effort. Beginning in 1987-88 the legislature provided funding through the state basic-aid-distribution formula to increase the number of certificated instructional staff members per 1,000 students. Over the next several years, the state incrementally changed the allocation of resources for K-3 students. Prior to the change, each 1,000 students in a school district generated 46 certificated instructional staff members (1 per 21.74 students). In 1987-88 this ratio was increased to 48 instructors per 1,000 students for grades K-3, and the next year to 49 per 1,000 (1 per 20.41). In addition, beginning with 1989-91 the state provided districts the opportunity to increase the ratio to as much as 52.3 certificated instructional staff members per 1,000 students (1 per 19.12) provided the districts could document that the funds were spent on K-3 programs. This incentive for smaller classes was strengthened further in the 1991-93 biennium when the ratio was increased to 54.3 certificated instructional staff members per 1,000 students (a ratio of 1 per 18.42). The continued and growing popularity of this reform was evident in 1998 when President Clinton called for hiring 100,000 new teachers to reduce class size to an average of 18 students in grades 1-3. He also proposed a construction tax to help build and modernize schools to help pay for the estimated $12 billion it would cost to provide enough classrooms. Dramatic class-size reduction is expensive. California’s program provides an additional $800 per student for children in K-3 classrooms with 20 or fewer students. It also provided funds for school and classroom construction. To reduce the class size from an average of approximately 29 to 20 or fewer students, the first-year costs of the program were some $1.1 billion. By the end of the program’s third year (1998-99), the state will have spent more than $4 billion on class-size reduction. This observation is in line with general estimates offered by Brewer, Krop, Gill, and Reichardt 1999). Other states have made similar investments. Tennessee spent about $600 million between 1991 and 1996 to implement its program. In Philadelphia, Superintendent David Hornbeck has unveiled plans to reduce class size in kindergarten through third grade from an average of 27 students to 20 students by the year 2002. He estimates that the program will require 1,000 new teachers at a cost of $50 million a year, as well as 35 new schools at a construction cost of $470 million. Philadelphia school district’s annual budget is approximately $1.2 billion. In Washington, smaller class size is encouraged through a higher staff/student ratio for K-3 in the funding formula. The annual cost of this program, which is part mandate and part incentive, is an estimated $250 to $300 million. Some $90-$100 million of this amount funds the mandate portion of the program, which increased the staff ratio from 46 to 49 per 1,000 students, and the balance pays for the incentive, which can bring the staff ratio to as many as 54.3 certificated instructional staff members per 1,000 students. Class-size reduction efforts become progressively more expensive as class size decreases. For example, a hypothetical district with 10,000 students would need to add about 22 teachers (and classroom space) to move from 22 to 21 students per teacher (a 4.5 percent reduction). However, it would take about 42 more teachers to move from 16 to 15 students per teacher (a 6.3 percent reduction). Aside from new teachers, new spaces are needed to teach these students. Brewer and colleagues (1999)estimated that, to reach a required class size of 20 students, at the current rate we would need 41,574 new classroom spaces. That number increases as class-size limits decrease. If class sizes were set at 15 as they were in Nevada and the Tennessee Star Project, we would need an estimated 221,612 new classrooms to teach these students in reduced classes. This recent estimate reaffirms the billion-dollar price tag that accompanies class-size reduction. Although current research supports the notion that smaller class size can lead to improved student performance, that view is not universally held among researchers. More important, research shows that alternative reforms may be considerably more costeffective in improving student performance. In particular, many have argued that investments in additional teacher training and professional development will lead to even greater gains in student performance for each dollar spent. It is important to understand both the policy and research context of the class-size reduction issue. The next section establishes the policy context for the discussion of class-size reduction. The second section briefly reviews the research literature on the effectiveness of smaller classes on student performance. In the third section, alternative policy options are discussed and compared with class–size-reduction programs. Class-Size Reduction: The Policy Context Reducing class sizes to improve education is not a new idea. Data from the federal government show that the average pupil-teacher ratio in the United States has declined dramatically in the last 40 years (NCES 1997). The pupil-teacher ratio in the United States has declined from nearly 27:1 in 1955 to approximately 17:1 in 1997. Some of this reduction can be accounted for by the increased availability of special programs (Title I and special education) for poor and mentally and physically challenged children; these programs utilize very small classes or rely on "pullout" programs that require a teacher to work with children individually or in small groups. Nevertheless, there have been real declines in the average number of children in most classrooms across the United States. Nationally, as per-pupil spending has increased, pupil-teacher ratios have declined. Research by Barro (1992) found that, on average, when a school district received an additional dollar of revenue, half of that dollar was spent on teachers. Of those 50 cents, 40 cents were spent on reducing class size and 10 cents on increasing salaries. Barro’s findings help confirm the apparent priority educators place on smaller classes, and their willingness to trade increases in salary for smaller classes. As noted in the opening section of this chapter, the cost of implementing smaller class size is high. Brewer and others (1999) go on to note that policy goals and foundation-level policy requirements play a major role in dictating the cost of implementing and maintaining smaller class sizes. Depending on the baseline policy option (size of class) adhered to, the cost of class-size reduction could range anywhere from $2 billion to $11 billion per year, say Brewer and colleagues. They estimated the cost of implementing class sizes at varying levels: 20 students as is the case in California, 18 as proposed by the federal bill, and 15 as carried out in the Tennessee’s project STAR. Further, Brewer and others’framework assumed that various policy issues--grade level, eligibility, phase-in period, measurement level, and flexibility--significantly influence the cost of implementating CSR. Also, Brewer and others show that the cost of maintaining smaller class size rises yearly, from total operational costs of $5.049 billion dollars and $448 per pupil in operational cost in the 1998-99 school year to an estimated $6.028 billion and $562 per pupil in operating cost in 2007-2008. Despite these current and future high costs, legislative efforts to reduce class size are common. Today in Washington, districts generate 3 more certificated instructional staff positions per 1,000 students in grades K-3 than they do for grades 4-12. Moreover, additional incentives--not part of a district’s basic aid--allow that staffing ratio to be as much as 8.3 certificated instructional staff members per 1,000 students higher in grades K-3. The K-3 pupil-teacher ratio can therefore be as low as 18.42, or some 3.32 pupils per teacher fewer than the pupil-teacher ratio of 21.74 generated in grades 4-12, where the formula provides 46 certificated instructional staff members per 1,000 pupils. One of the first states to enact CSR was Texas, which began mandating limited class sizes with the educational reforms of 1984. Today, K-4 programs must average no more than 22 students per classroom. Of the19 states that have some form of class-size reduction, 10 rely on incentives to encourage school districts to reduce class size, whereas 8 use mandates (Education Commission of the States 1998). Washington is unique in that it relies on both a mandate and an incentive if districts spend the funds on certificated instructional staff members who work with students in grades K-3. The focus of state programs is almost entirely on the primary grades, generally K-3. North Carolina’s program is aimed at grades K-2, while Oklahoma’s program focuses on grades K-6, and the program in Texas on grades K-4. In Utah, grades K-2 are the primary focus, and funds can be devoted to reducing class size in grades 3 and 4 only if K-2 classes already are all reduced to 18 or lower. Washington’s program differs to some extent from the others in that the law does not require classes of 18 or 20 or some other number, only that the funds be spent on staff members who work with children in grades K-3. Theoretically this allows for alternative staffing structures as determined by schools and their respective districts. There is no question that class-size reductions are an important educational policy issue. They can also be very expensive as the data above suggest. The general belief of most educators and policymakers is that smaller classes are effective in improving student performance. However, it is difficult to ascertain the "right" class size and to determine whether the positive effects of being in a small class in grades K-3 stay with students into later grades. Essentially, the investment is hardly worthwhile if student outcomes do not improve over the long run. The next section considers the research on class size, particularly its impact on student achievement. Reducing Class Size: A Brief Synthesis of the Literature Today, it is hard to find anyone not in favor of reducing class size. Even those who are not convinced there is a strong research base to show that smaller classes lead to improved student performance are willing to concede that smaller classes can lead to more individualized instruction, higher morale among teachers, and more opportunities for teachers to implement instructional programs that research shows work well. Among those who are convinced that smaller classes lead to better student performance, there is only limited consensus on what the "ideal" class size might be. By looking at those studies that appear to be the most methodologically sound, this section attempts to provide answers to three questions:
The section begins by describing the early meta-analyses on class size and then discusses recent studies that have attempted to resolve some of the methodological issues identified with earlier studies. The Early Meta-Analyses Meta-analysis (Glass, McGaw, and Smith 1981) is a technique for looking at a wide variety of studies on a specific topic and determining if the results of those studies support a conclusion about that topic. The first step is to identify high-quality studies on the subject. This is done by searching for all the documents dealing with the topic and establishing decision rules whether to include each study in the meta-analysis. These decision rules usually pertain to the quality of the study (that is, published in a referred journal or high-quality book) and the relevance of the actual analysis to the topic of the meta-analysis. Once the findings have been identified, researchers need to compare the findings from each of the studies. This is difficult since the studies use different data sets, have different sample sizes, and analyze different variables. To compare studies, the results are standardized and the outcomes compared in terms of these standardized values. Glass and Smith (1979) conducted an early and comprehensive meta-analysis of the class-size literature. They identified more than 300 studies going back as far as 1895 on the topic. Of those 300, 77 met their decision rules for inclusion in the meta-analysis. They calculated a total of 725 effects from the 77 studies. Based on their analysis of those studies, Glass and Smith concluded:
These are strong and important conclusions, and many have used them to support calls for reducing class size to less than 20. Unfortunately, not everyone in the research community found this work to be convincing. Slavin (1984) criticized meta-analysis, arguing that the technique gives equal weight to all study findings, regardless of the quality of the study design. He argued that only 14 of the 77 studies in the Glass and Smith meta-analysis were methodologically sound. He also criticized meta-analysis generally, suggesting that the technique combines studies that are on different topics while claiming to address the same topic. For example, one of the methodologically sound studies with large effects in the Glass and Smith sample had to do with learning how to play tennis. When Slavin (1989) reanalyzed the methodologically sound studies from the Glass and Smith work, he pointed out that there were relatively few studies with fewer than 20 students in a class, and that there were no classes with between 4 and 14 students. He argued that the Glass and Smith findings were thus based on statistical interpolations of the findings in the 14 sound studies. He also concluded that the effects of reduced class size on student achievement were considerably smaller than Glass and Smith has determined. Using these data from earlier meta-analyses, Odden (1990) suggested that the research on class size supports "dramatic--and only dramatic--class size reductions." While he did not necessarily put a figure on what class size should be, Odden argued that reducing class size from 28 to 26, or from 24 to 22, would not be effective. He argued that class size needed to be reduced substantially more--to something like 15 to 17 students per class. This line of reasoning has major implications for policymakers interested in reducing class size. States with large class sizes will need to spend substantial sums of money to make those "dramatic" reductions if the policy is to succeed. Recent Studies In recent years a number of studies have analyzed the impact of class size on student learning. In general, they show that smaller class size leads to greater gains in student test scores. One exception to this is the work of Eric Hanushek, who argues that to date we have not found a systematic relationship between resources and student outcomes. Hanushek (1989) reviewed 152 studies that used the pupil-teacher ratio as an independent variable in estimating the impact of spending and resources on student outcomes. Hanushek found only 27 studies with statistically significant findings, and only 14 of those found that reducing the number of pupils per teacher was positively correlated to student outcomes, whereas 13 found the opposite. Among the other 125, Hanushek found 34 with a positive effect, 46 with a negative effect, and 45 with an undetermined effect. More recently, Hedges, Laine, and Greenwald (1994) and Greenwald, Hedges, and Laine (1996), after reviewing the same studies, came to the opposite conclusion. Relying on newer and more sophisticated statistical techniques, they argued that smaller classes did matter. Their analysis found substantial gains in student performance when more money was spent on education, and smaller class size was related to performance gains as well. Others have reached that conclusion as well. Ferguson (1991) analyzed the effect of class size and teacher preparation on student achievement in Texas, concluding that in elementary grades lower pupil-teacher ratios contributed to increases in student achievement. In a recent study in Alabama, Ferguson and Ladd (1996) attempted to address some of the weaknesses of earlier studies in this area. They used larger samples of students, employed better model specification, and had access to better data than in the past. They concluded that teacher test scores, teacher education, and class size "appear to affect student learning" (Ferguson and Ladd 1996). They also attempted to ascertain the threshold below which further reductions in class size would no longer lead to systematic achievement gains for students. They believe that if such a threshold exists, it is in the range of 23 to 25 students per teacher. This number seems somewhat high compared to other results, but could be a result of the relatively low per-pupil spending in Alabama and the generally larger class size in that state during their study. More important, Ferguson and Ladd sought to measure actual class size, rather than the district’s or school’s pupil-teacher ratio. Consequently, their work may reflect a more accurate picture of the number of students in a classroom at any time. One of the problems with this line of research has been the lack of a true experimental design. In fact, only one study with such a design has been undertaken. The Tennessee Student-Teacher Achievement Ratio Experiment (STAR) relied on an experiment in which children were randomly assigned to classes with low pupil-teacher ratios and high pupil-teacher ratios. The study design placed students into one of three groups: an experimental group where the average class size was 15.1 students and two control groups (a regular-size class with an average of 22.4 students and a regular-size class with a teacher’s aide and an average class size of 22.8 students). Under the study plan, each student was to stay in the original class size assignment until the third grade. Following third grade, the experiment was concluded and all students assigned to regular-size classrooms. Standardized tests were given each school year to measure student achievement. While there are some methodological and data problems in any study of this magnitude, two respected researchers have argued that the Tennessee STAR project is the best-designed experimental study on this topic to date (Mosteller 1995; Kruger 1998). Kruger (1998) summarized the major findings of the Tennessee STAR project as follows:
A number of important policy issues are highlighted by the findings from Tennessee STAR. First, the results of the evaluation suggest that smaller classes do lead to improved student performance, and that those performance gains are maintained at least through the seventh grade. Recently, Nye, Hedges, and Konstantopoulos, in a five-year followup study, wrote that although some students "dropped out," they dropped out having attained a higher level of achievement than their peers in larger classes. Moreover, the results suggest that alternative models that rely on the use of teacher aides to reduce the "effective class size" may be ineffective. The research also suggests that simply reducing class size without changing how teachers of smaller classes deliver instruction is unlikely to improve student performance. It is important that teachers take advantage of the smaller classes to offer material in new and challenging ways identified through research. Absent that effort and the training needed to accompany such a change, expenditures for class-size reduction may be relatively ineffective. Alternatives to class-size reduction The research reviewed above shows that reducing class size can, and probably does, lead to improved student performance. It is, however, a very expensive option: In addition to hiring more teachers, schools need additional classroom space. Before embarking on a substantial CSR program, policymakers may want to consider whether more cost-effective alternatives exist. Current research suggests that such alternatives are available and should be considered, either instead of--or in addition to--class–size reduction. One range of options deals with teacher knowledge and skills, while others relate to the structure of the education program offered at individual schools. Each is discussed below. Teacher Knowledge and Skills Reducing class size gives students greater access to teacher resources. There is evidence this will help students learn. However, what the teacher knows and is able to do is at least as important in helping students learn. Darling-Hammond (1998) argues that "teacher expertise is one of the most important factors in determining student achievement." She quotes Greenwald, Hedges, and Laine’s work, which demonstrated the relative impact of spending $500 more per pupil on increased teacher education, increased teacher experience, and increased teacher salaries. All three of these appear to have a greater impact on student test scores than does lowering the pupil-teacher ratio. For an expenditure of $500, the greatest gains in student test scores (measured in standard deviation units from a range of tests in 60 studies) were achieved through increasing teacher education. Lowering the "pupil-teacher ratio was the least cost effective of the four methods. Increasing teacher salaries and experience fell between lower pupil-teacher ratios and teacher education in terms of cost effectiveness." Ferguson (1991) found that the effects of teacher expertise in Texas were so great that after controlling for socioeconomic status, disparities in achievement between black and white students were virtually entirely explained by differences in teacher qualifications. He found that teacher qualifications explained 43 percent of the variation among the factors affecting math score test gains, whereas small classes and schools only accounted for 8 percent of the gain. Home and family factors were identified as explaining the remaining 49 percent of the variance. Darling-Hammond (1998) summarizes these findings by stating that "teachers who know a lot about teaching and learning and who work in settings that allow them to know their students well are the critical elements of successful learning." Smaller classes are clearly desirable in her view, but given limited funds to invest, her work suggests policymakers should at least take a close look at improving access to high-quality professional development first. Professional development is frequently poorly funded in school districts and often the first item to be cut when finances become tight. Darling-Hammond’s research suggests this may be a mistake, and in fact, more resources should be put into professional development. Even if class size is reduced, professional development still may be essential to help teachers maximize their skills and capitalize on the benefits of having a reduced number of children for whom they are responsible. Certainly investments in professional development would be complementary to class-size reduction programs. Reducing class size and providing greater training opportunities for teachers are not the only options available for improving student learning. There are many things school board members and site leaders themselves can do to restructure their schools for improved learning. Several of these are briefly discussed below. Reorganizing Schools Many of today’s educational reforms are restructuring how educational resources are used. A number of the reform designs supported by the New American Schools (NAS), for example, rely on using teaching resources differently, rather than purchasing more. While seven designs supported by NAS require some investment on the part of a school or school district, most are less expensive than dramatic reductions in class size or pupil-teacher ratios. Most also come with substantial teacher-training components. Odden and Busch (1998) found substantial gains in student performance, often as high as one-third of a standard deviation, at NAS design schools. These schools reach these performance levels with relatively little additional expenditures, generally averaging around $50,000 to $250,000 a year for a school of 500 students (an extra $100 to $500 per pupil each year). Odden and Busch argue that any school can reorganize itself into one of the NAS designs by looking closely at its current allocation of teachers and aides and reassigning them as needed to meet the design specifications. In many instances this calls for eliminating aides in favor of more teachers. Given the results of the Tennessee STAR project reported above, spending for teacher aides may not be productive anyway. Another option schools can consider is restructuring the use of time. The National Commission on Time and Learning (1994) reported on a number of successful schools and school districts that had improved student performance through different ways of organizing the school day to give students more access to, and time with, teachers. Models that provide more access to learning resources, particularly teachers, may also be substantially more cost-effective than class-size reduction. Conclusion Class-size reduction is currently one of the most popular--and most expensive--educational reforms today. At least 19 states have enacted mandatory or voluntary policies aimed at reducing class size in the primary grades, and one (California) has even created an incentive to reduce the number of students in ninth-grade English and math classes. The question facing state policymakers is this: Should substantial investments in smaller classes be made? The research shows that such investments will lead to improved student outcomes. However, the research also shows that attention to teacher training and expertise may have a bigger payoff per dollar spent. Moreover, as California’s experience shows, states that jump into a major CSR program quickly may find they have a shortage of qualified teachers. Given the importance of high-quality teaching to student learning, investment in the quality of the teaching force first might be a better way to maximize the potential of the dollars that are used to reduce class size. In short, few appear to oppose class-size reduction. However, there are a number of things states and school districts can do to ensure that the substantial investment made in teachers and classrooms pays off to the maximum extent possible. Virtually all of them revolve around ensuring that the state has the highest quality teaching force possible. |