A r c h i v e d  I n f o r m a t i o n

Class Size and Students At Risk: What is Known?...What is Next? - April 1998

Assessing the Costs and Benefits
of Smaller Classes

Without exception, the greatest obstacle to widespread implementation of smaller classes is the expense of additional teachers and classrooms. The cost issue is raised by researchers (Tomlinson, 1990) and by state and local policy makers who control the purse strings. They are to be commended for being cautious with tax dollars until the expenditure is of proven worth. At the same time, we do not have any widely-accepted procedures for determining the dollar value of particular increments in school achievement. And most economic analyses of class size to date have been severely limited. However, several approaches to the problem have been taken.

Educational Production Functions

The production function approach relies heavily on multiple regression analysis to relate a series of inputs (such as cost factors) to an output (such as student achievement). Hanushek (1986) reviewed 112 studies that used educational production functions to examine the effects of instructional expenditures on student achievement, using indicators such as teacher experience, teacher education, and pupil/teacher ratio.15 Pupil/teacher ratio was statistically significant in only 23 of the 112 studies, only 9 of which were significant in the expected direction. This "vote counting" procedure led Hanushek to conclude that pupil/teacher ratio is not an important correlate of student performance. More sophisticated analyses of the same data, however, have led others to conclude that low pupil/teacher ratios (and other cost-related inputs) are associated with increased pupil performance (Hedges, Laine, & Greenwald, 1994; Laine, Greenwald, & Hedges, 1995).

Unfortunately, the production-function approach often fails to consider findings of earlier research on class size. For example, most production-function analyses do not focus on the elementary grades, although two recent exceptions are noteworthy. In an analysis of national survey data at the district level, Wenglinsky (1997) concluded that expenditures to reduce pupil/teacher ratios impact positively on academic achievement at grade 4 but not at grade 8. Ferguson and Ladd (1996) analyzed achievement scores for students in grades 4, 8 and 9 of 131 districts in Alabama. These researchers used average class size in their multi-level regression models instead pupil/teacher ratio, concluding that class size does matter in both the earlier and later grades.

Other important differences remain. Most production function analyses include schools and districts with classes within "normal" ranges--22 to 40 students or so--and the results do not answer the question of what the impact would be if classes were reduced substantially.16

Of greater concern, most production function analyses focus on school-wide or district-wide pupil/teacher ratios rather than actual class size.17 For a host of reasons, pupil/teacher ratios do not indicate how many students are enrolled in any given class or interacting with the assigned teachers (see Boozer & Rouse, 1995, for a comparison). Project STAR demonstrated the benefits of a small- class setting and provided some insight into why they occurred. It did not demonstrate that reducing the pupil/teacher ratio for a school or district would have the same impact, unless actual class sizes decreased at the same time.

Cost Analyses

Cost-effectiveness analysis examines both costs and consequences in considering alternatives for decision making. In educational applications, outcomes are typically assessed in terms of school achievement. Levin (1988) illustrated this approach to compare four strategies for educational improvement: cross-age tutoring, computer-assisted instruction, lengthening the school day, and reducing class size. Data on class size were taken from 14 evaluations collected in previous research; effect sizes were expressed as "estimated months of achievement gain" in reading and mathematics. Costs were estimated using an "ingredients approach" which involved the identification of ingredients of each intervention and their respective values, and determination of the overall cost of implementation. For example, the ingredients needed to reduce class size include personnel, facilities, and equipment.

Although the projected annual cost per student of reducing class size by five students was not found to be as great as either lengthening the school day or use of computer-assisted instruction, larger reductions in class size become quite expensive:

The principles of cost-effectiveness analysis are sound, if fraught with methodological difficulties. The cost of an intervention can often be determined with some degree of accuracy, but the effectiveness side of the equation is more complex. Even in the simplistic applications.given by Levin (1988) a small change in an effect size can have a large impact on the cost-effect ratio. When an intervention has numerous or diverse outcomes (only as different as mathematics and reading), or effects that differ from one population to another, the method provides no clear-cut way to determine cost effectiveness in toto. Introducing small classes into a school or district is at least this complex, precluding any easy answers to the cost-effectiveness question.

One analysis of costs is noteworthy even though it did not consider small classes directly. King (1994) compared costs--time and money--associated with three educational interventions: Henry Levin's Accelerated Schools, Robert Slavin's Success for All, and James Comer's School Development Program. Although the Accelerated Schools and the School Development Program have costs that are similar, Success for All is more expensive to implement. The major expense of Success for All--with demonstrated efficacy--is the cost of additional staff members, particularly tutors. It would be useful to compare the costs, benefits, and feasibility of implementing this program with those of reducing class size. The main effective ingredient of Success for All may be the smaller number of students working with a particular teacher or tutor, that is, a small-class arrangement.

Further Work

The question posed by cost-effectiveness analysis is entirely appropriate, namely: What benefits are associated with what levels of investment? The current state of knowledge dictates that we evaluate the effectiveness of small classes more completely by documenting the full spectrum of outcomes that are realized, and ask whether the extra investment can be put to best use by directing it to schools where it is needed most, for example, those serving students with poor educational prognoses.

The database created for STAR and the LBS can provide a fuller picture of short- and long-term outcomes. There is a real possibility that attending a small class in the primary grades can begin students on a path that reduces the need for special education, grade retentions or disciplinary measures, and increases the likelihood of high school graduation. Even if there is no further payoff after a student graduates, the cost savings would be appreciable.

Economist Alan Odden (1990) explored whether the effects of reducing class size on student achievement could be achieved with other lower-cost interventions, or whether larger effects could be obtained through other interventions at the same cost. He concluded that particular uses of small classes are worthwhile, especially in kindergarten through grade 3. Odden recommended reducing class size for students achieving below grade level and combining individual tutoring with classes reduced to 15 students for language arts-reading instruction. He also proposed that small classes be coupled with a "larger comprehensive set of strategies" shown to be effective for low-income, ethnic and language minority students. Early childhood education is one example.

Unfortunately, at this point in time there are no well-established procedures for summarizing diverse effects of any major intervention or, further, for comparing one intervention with another. The effect sizes in Table 1 only begin to indicate the range of outcomes, and even these are not well represented by one or two figures. If outcomes are attained that are conceptually different (e.g., improved behavior) the problem of comparison becomes even more complex.

To obtain valid comparisons with other specific instructional strategies, the duration of the intervention also needs to be considered. For example, individualized instruction (tutoring and computer-assisted instruction) and cooperative learning (see Slavin & Madden, 1989; Wasik & Slavin, 1990) are often utilized for a portion of the day to provide support in one or a few school subjects. To compare, the costs and effects should be prorated to ask what the cost-benefit ratio would be if the strategy were implemented all day for all school subjects. Making comparisons with full-scale intervention programs in which small classes, small groups, or tutors are a component (e.g., Success for All) is a slightly different matter. It may be feasible to estimate the effect of this component alone and compare that to overall program effectiveness. In either case, a number of methodological issues need to be resolved before meaningful cost comparisons can be obtained.

15 Note that "indicators" are not the same as actual expenditures, and the relationship between the two may be complex.

16 The step cannot even be taken "in theory" since reductions in class size would change the values of other important inputs as well.

17 Ferguson and Ladd (1996) is an exception.

-###-

[Research on the Academic Effects of Small Class Size] [Table of Contents] [Instructional Practice and Student Behavior]