With Governor Bill Haslam’s announcement to give districts flexibility to determine class sizes we thought we would share some of the research on class size. There has been an enormous amount of research into the impact of class size on student learning. However, quantifying this impact is difficult since there are so many other factors that must be taken into account. As explained by Grover J. “Russ” Whitehurst and Matthew M. Chingos in Class Size: What Research Says and What it Means for State Policy (May 2011) of the Brown Center on Education Policy at the Brookings Institution:
The primary difficulty in interpreting this research is that schools with different class sizes likely differ in many other, difficult-to-observe ways. For example, more affluent schools are more likely to have the resources needed to provide smaller classes, which would create the illusion that smaller classes are better when in fact family characteristics were the real reason. Alternatively, a school that serves many students with behavior problems may find it easier to manage these students in smaller classes. A comparison of such schools to other schools might give the appearance that small classes produce less learning when in fact the behavior problems were the main factor.
We will come back to the Brookings paper later, but we will begin with The Evidence on Class Size by Eric A. Hanushek (February 1998) of the University of Rochester Institute of Political Economy. This paper is 40 pages, but roughly a quarter of that is graphs, tables, footnotes, etc. While a lay person can follow along a second reading of the occasional sentence or paragraph may be helpful. We recommend a full reading of the paper to get a full understanding of the impact of class size reduction; however, we are pasting some highlights below.
The first section begins with what aggregate data indicate about the effectiveness of class size policies.
The available evidence and data suggest some uncertainty about the underlying forces related to families, school organization, class size, and achievement. Allowing for changes in family background and in special education, however, it remains difficult to make a case for reduced class sizes from the aggregate data. A natural experiment in class size reduction has been on-going for a long period of time, and overall achievement data do not suggest that it has been a productive policy to pursue. Nonetheless, the aggregate data are quite limited, restricted to a small number of performance observations over time and providing limited information about other fundamental changes that might affect school success.
The second section reviews international data and the third section summarizes the available extensive econometric evidence about the effectiveness of reducing class sizes.
Table 4 Percentage Distribution of Estimated Influence of Teacher-pupil on Student Performance, By Level of Schooling
Table 4 summarizes the available results for estimates of the effects of teacher-pupil ratios on student outcomes. Of the total of 377 available econometric studies of the determinants of student performance, 277 consider teacher-pupil ratios. Ignoring the statistical significance, or the confidence that we have that there is any true relationships, we find that the estimates are almost equally divided between those suggesting that small classes are better and those suggesting that they are worse. This distribution of results is what one would expect if there was no systematic relationship between class size and student performance.
The table also shows the results divided by level of schooling. As Table 4 shows, there is little difference between the estimated effects in elementary and in secondary schools, but, if anything, there is less support for increasing teacher-pupil ratios at the elementary level.
Table 6 Percentage Distribution of Other Estimated Influences of Teacher-pupil ratio on Student Performance, Based on Value-added Models of Individual student Performance
One type of statistical investigation – those employing a value-added specification – is generally regarded as being conceptually superior and likely to provide the most reliable estimates of education production functions. Table 6 provides a summary of value-added results, both for all 78 separate estimates of class-size effects and for the 23 estimates that come from samples in a single state.
There is simply little reason from the results in table 6 to believe that smaller classes systematically lead to improvements in student achievement. Of the best available studies (single-state, value-added studies of individual classroom achievement), only one out of 23 (4 percent) shows smaller classes to have a statistically significant positive effect on student performance.
The econometric evidence is clear. There is little reason to believe that smaller class sizes systematically yield higher student achievement. While some studies point in that direction, an almost equal number point in the opposite direction. Moreover, restricting attention to the best of these studies, including those with the most accurate measurement of individual class sizes, merely strengthens the overall conclusion.
The fourth section turns to the evidence developed in Project STAR, an experiment conducted by the State of Tennessee in the mid 1980s.
The key to interpretation for policy purposes revolves around expectations about student performance over time. Perhaps the most standard interpretation form learning theory begins with the view that education is a cumulative process, building on past achievement. From this view, if students learn certain skills in the first grade, they tend to carry over to later grades, albeit possibly with some depreciation. According to this view, the basic evidence of the STAR study suggests that smaller classes may be important at kindergarten but have no average effect subsequently. Specifically, since the growth in achievement across experimental and control students is the same from first through third grade, that added resources of the smaller classes appear to add nothing to student performance. Early differences simply remain the same over time. If resources had a continuing impact, we should observe a widening of achievement as more and more resources are applied.
Some have argued that the observed pattern could be consistent with small classes making a difference in all grades if students are expected to fall back to a common mean performance each year. This is equivalent to a view that educational performance is not cumulative.
…the differentials in performance found at kindergarten remain essentially unchanged by third grade after class size reductions of one-third were continuously applies and remain largely unchanged by sixth grade after class size returned to its prior levels for another three years. This latter finding leads to rejection of the fall-back model and indicates that class size reductions after kindergarten have little potential effect on achievement.
A third interpretation-which helps to reconcile the basic data-is the small classes, particularly if they occur early in the schooling process, have a one-time effect on student performance that is not linked to the acquisition of cognitive skills per se. This one time effect could reflect early training in the “activity of school.” Students in small classes, by this view, learn the norms, behavior, and learning patterns that are useful in subsequent years, so that they are able to continue achieving at a higher level. In fact, this last interpretation is the one most consistent with the STAR data (ignoring the other possibilities of flaws in the underlying experimental design and data collection). It provides a parsimonious explanation of why there is a one-time but lasting effect of class size reductions in kindergarten. But, this interpretation also has powerful implications for any policy discussions.
The most expansive conclusion that can be reached from Project STAR and the Lasting Benefits Study is that they might support an expectation of positive achievement effects from moving toward small kindergartens, and maybe small first grades. None of the STAR data support a wholesale reduction of class sizes across grades in schools.
Project STAR and related programs do support one aspect of the econometric results from Texas that were mentioned previously: Disadvantaged students appear more sensitive to class size variations than the majority of students. Again, however, disadvantaged students on average are not currently in larger classes than more advantaged students, and the effects appear small relative to costs of programs and alternative policy approaches.
The final section provides possible interpretations for the lack of any results from reducing class size and then relates the evidence to prospective educational policies.
The extensive investigation of the effects of class size on student performance has produced a very consistent picture. There appears to be little systematic gain from general reductions in class size. This story comes through at the aggregate level, where pupil-teacher ratios have fallen dramatically over the past three decades and where student performance has remained virtually unchanged. It also comes through from international data, where extraordinarily large differences in class sizes are found without commensurate differences in student performance. But, since the aggregate analyses could be misleading for a variety of analytical reasons, more weight should be put on school level analyses and on experimental data. From production functions estimates, there is little reason to believe that overall reductions in class size will yield much in the way of positive achievement gains. With several hundred separate estimates of the effects of reduced class size, positive and negative effects almost evenly balance each other, underscoring the ineffectiveness of overall class size policies such as those being currently advocated. Finally, the one major random-assignment experiment – Tennessee’s Project STAR study – provides no support for widespread class size reductions, although it holds out hope for gains from reduced-size kindergartens.
Much of the case for reduced class size rests on common sense arguments. With fewer students, teachers can devote more attention to each child and can tailor the material to the individual child’s needs. But, consider, for example, a movement from class sizes of 26 to class sizes of 23. This represents an increase in teacher costs alone of over ten percent. It is relevant to ask whether teachers would in fact notice such a change and alter their approach. The observations information from teacher and classroom process effects of the one-third reductions in Project STAR suggested no noticeable changes.
…given the strong advocacy of reducing class sizes by teachers and parents, politicians appear frequently to find supporting proposals to reduce class size simply irresistible. There are clear limits to the amount of funds and attention available for education. Squandering the current public and political attention on policies that reinforce existing inefficiencies and that promise little hope of success is likely to have long term consequences, albeit consequences that will not be apparent until some time in the future.
We now return to Class Size: What Research Says and What it Means for State Policy since it is relatively recent, May 2011, and they do a good job of pointing out the practical implications of class size policies. The paper is only 14 pages and very easy to read. Again, we recommend a full reading, but are pasting some highlights below.
The current fiscal environment has forced states and districts to rethink their class-size reduction (CSR) policies given the high cost of maintaining small classes. When school finances are limited, the cost-benefit test any educational policy must pass is not “Does this policy have any positive effect?” but rather “Is this policy the most productive use of these educational dollars?”
The average U.S. pupil/teacher ratio in the public schools is currently 15.3. With an average U.S. teacher salary of approximately $55,000, each student has an individual cost of about $3600 in teacher salary alone. With about 49.3 million public school students enrolled, a one-student decrease in class size from the present average would cost over $12 billion a year in aggregate for the U.S. A one-student increase in class size would generate an equivalent savings. The costs of CSR are not limited to teacher salaries. More classrooms are needed for smaller classes. In our example of one-student reduction in class size across the U.S., more than 225,000 additional classrooms would need to be added to the nation’s stock. In any context $12+ billion a year for any government’s largest single K-12 education program, Title I of the Elementary and Secondary Education Act, involves about the same level of annual expenditure as would a one-student reduction in the nation’s average pupil/teacher ratio.
There is no research from the U.S. that directly compares CSR to specific alternative investments. In other words, the comparison condition for all CSR studies has been business as usual rather than, for example, a comparison of $20 billion invested in smaller classes vs. $20 billion invested in higher teacher salaries. Thus, estimates of effects and costs from different education investments have to be extrapolated and estimated from different studies, and this process is necessarily inexact. Nevertheless, Harris finds short-term rates of return for computer-aided instruction, cross-age tutoring, early childhood programs, and increases in instructional time that are all greater than those for achievement from choosing more effective curriculum; reconstituting the teacher workforce (for example by substituting Teach for America teachers for new teachers from traditional training routes); and enrolling students in popular charter schools in urban areas that are all as large or larger than those obtained from CSR.
By one estimate, an increase in average class size by 5 students would result in an across the board increase of 34 percent in teacher salaries if all the savings were devoted to that purpose. Higher salaries would likely draw more qualified people into the teaching profession, and keep them there.
The potential for negative consequences of larger classes clearly needs to be weighed against the fallout from cutting other programs in order to preserve smaller classes – both academic programs and non-academic offerings such as athletics and the arts.
If the teachers to be laid off were chosen in a way largely unrelated to their effectiveness, such as “last in first out,” then the associated increase in class size could well have a negative effect on student achievement. But if schools choose the least effective teachers to let go, then the effect of increased teacher quality could make up for some or all of any negative effect of increasing class size.