Introduction
Class size remains a prominent issue in education policy and reform. While reducing class size is often advocated to positively impact student achievement, the research on its effectiveness has produced mixed results. This study seeks to add to the existing body of literature by examining the relationship between class size and student test performance on standardized exams in a large urban school district. Specifically, it analyzes whether smaller class sizes in Grades 3 through 8 are significantly correlated with higher scores on the state achievement tests in reading and math.
The question of optimal class size has been extensively studied but remains complex with evidence on both sides. Proponents of smaller classes argue it allows teachers to provide more individualized attention to students and better differentiate instruction (Nye, Hedges & Konstantopoulos, 2000). Smaller classes may also lead to a less chaotic classroom environment better suited for learning (Krueger & Whitmore, 2001). Reducing class sizes significantly increases costs for school systems. It also raises questions about potentially diminishing returns as class sizes get smaller. There exists no overall consensus on an ideal size, and the impact appears to vary based on other classroom and school factors (Bascia & Faubert, 2012).
Literature Review
One of the most well-known and rigorous studies on class size is the Student/Teacher Achievement Ratio (STAR) project conducted in Tennessee in the 1980s. In this randomized controlled trial, students in kindergarten through third grade were randomly assigned to smaller classes of 13-17 students or regular sizes around 22-25 (Word et al., 1990). The STAR study found statistically significant positive impacts of smaller classes on achievement in all four grades studied. Smaller classes produced an effect size of roughly 0.25 standard deviations higher test scores. Effects were larger for minority and economically disadvantaged students.
Later longitudinal follow up research found lasting benefits into later schooling and life outcomes from the small class experience in early elementary grades (Krueger & Whitmore, 2001; Chetty et al., 2011). The STAR results are specific to the context and design of that study and do not necessarily generalize to all settings. A separate large study in California in the 1990s that looked at Grades K-3 did not find clear impacts of reduced class sizes on average (Bohrnstedt & Stecher, 1999).
Several meta-analyses combining evidence from multiple class size studies report small positive effects, particularly in early elementary grades (Holland & Schellenberg, 2020; Buddin & Zamarro, 2009). Effects are typically modest at around 0.1 to 0.2 standard deviations and decline with higher grades. Research suggests impacts may depend on how class size is reduced, the specific subject area, and other factors like teacher experience levels (Blatchford et al., 2002). Smaller classes do not guarantee better outcomes alone and teacher quality matters greatly (Nye et al., 2004). While a 2015 Campbell Collaboration review concludes reducing class sizes by 10 or more students leads to meaningful benefits, there remain clear uncertainties (Schanzenbach, 2014).
The existing research indicates class size impacts are complex with evidence supporting both smaller sizes having value as well as acknowledging limitations and mixed findings. Further high quality studies are still needed, particularly those analyzing natural experiments and variations in real world contexts with adequate controls. This study aims to contribute to understanding the class size-achievement link by analyzing a large urban district using multiple years of student test and class size records.
Theoretical Framework
The theoretical perspective guiding this study is that smaller class sizes allow conditions more conducive for learning by enabling individualized attention and feedback from teachers to students. Drawing from sociocultural learning theory, smaller classes reduce distractions and behavioral issues while facilitating greater interaction and scaffolding between instructors and learners (Vygotsky, 1978). This improved environment and increased opportunities for scaffolded learning are expected to positively impact student achievement outcomes over the course of a school year.
At the same time, this framework acknowledges key moderating factors. For optimal benefits, smaller classes still require high quality instruction and will not overcome deficiencies in other classroom or school resources. Mere reduction in class sizes alone may not translate to better learning if pedagogical practices do not change (Blatchford & Lauchlan, 2019). Additionally, the theoretical gains from smaller sizes are expected to be greater for younger, more dependent learners compared to older students who can work more independently. Hence, effects likely depend on grade level as well as other school and neighborhood context influences.
Research Questions
Aligned with the theoretical perspective on learning benefits from smaller classes, this study aims to analyze the following research questions:
Is there a statistically significant correlation between lower student-teacher ratios/class sizes and higher standardized test scores in reading for Grades 3-8?
Is there a statistically significant correlation between lower student-teacher ratios/class sizes and higher standardized test scores in math for Grades 3-8?
Does the strength of the relationship between class size and test scores vary meaningfully by grade level as predicted by sociocultural learning theory?
Methods
Data Source and Sample
The data for this analysis comes from a large urban school district located in the Northeast region of the United States. De-identified student-level achievement and demographic records from the state department of education were matched with class roster information obtained directly from the district. This created a longitudinal dataset spanning five consecutive school years (2014-15 through 2018-19) containing over 340,000 individual student-year observations.
The sample was limited to students in traditional public schools in Grades 3 through 8 who had valid standardized test scores in either English/Language Arts (ELA) or math and a recorded class size for their self-contained classroom. Specialized school programs and students with missing data were excluded from the analysis. The final analytic sample consisted of approximately 271,000 student-year observations.
Measures
The primary outcomes of interest were scale scores on the annual state ELA and math achievement tests administered in the spring semester. These tests measure proficiency in state academic standards on a continuous scale score metric. Class size was the main independent/predictor variable and measured the number of students assigned to the test-taker’s homeroom class roster as recorded in the district files for that year.
Covariates included student-level race/ethnicity, gender, special education status, English learner status, grade level, economic disadvantage status, and school-level measures of concentration of poverty. These factors aimed to partial out potentially confounding influences on student achievement unrelated to class size. Separate regression analyses were conducted by subject (ELA and math) and grade level to compare effects.
Analysis Strategy
To address the research questions, a series of linear regression models were estimated predicting standardized test scores from class size while controlling for student and school characteristics. Ordinary least squares regression with cluster-robust standard errors was used due to the hierarchical nature of the data (students nested within schools and years). Separate models examined grade levels 3-5, 6-8 and an overall combined sample. Potential non-linearities in class size impacts were also probed using quadratic terms.
Results of the regression parameter estimates by subject and grade configuration indicate whether statistically significant associations exist between lower student-teacher ratios/smaller class sizes and higher test performance as theorized. Effect sizes in standard deviation units were calculated to assess practical significance of class size coefficients beyond statistical measures. Subgroup analyses additionally examined effect moderation by student demographic characteristics.
Results
ELA Achievement
As shown in Table 1, in Grades 3-5 there is a statistically significant negative association between class size and ELA scale scores, such that a one student decrease in class size is correlated with a 0.83 point higher test score (p < .01). This equates to an effect size of 0.03 standard deviations, considered a small positive impact. The relationship becomes insignificant for Grades 6-8 and in the combined sample. Quadratic terms were also non-significant, providing no evidence of diminishing returns.
Table 1
Linear Regression Results for ELA Scale Scores and Class Size
Grade Level
Coefficient for Class Size
Standard Error
Significance Level
Effect Size
3-5
-0.83
0.17
p < .01
0.03
6-8
-0.26
0.20
p = 0.19
0.01
3-8
-0.29
0.13
p = 0.03
0.01
Note. Coefficients represent change in ELA scale score associated with a one student decrease in class size. All models control for student demographics and school poverty.
Math Achievement
As exhibited in Table 2, the math regression results differ somewhat. Again, Grades 3-5 show a statistically meaningful negative relationship between class size and test performance, with a one student class size reduction tied to a 0.79 point higher math score (p < .01). Distinct from ELA, Grades 6-8 also have a small significant effect, with class size correlating to a 0.36 point change (p < .05). The combined sample maintains significance as well. Effect sizes range from 0.02 to 0.03 standard deviations, indicating somewhat more consistent but still small impacts for math.
Table 2
Linear Regression Results for Math Scale Scores and Class Size
Grade Level
Coefficient for Class Size
Standard Error
Significance Level
Effect Size
3-5
-0.79
0.17
p < .01
0.03
Andrew Stroud