National Study Links Small Schools and Higher Achievement

Last Updated: May 01, 2004

This article appeared in the May 2004 Rural Policy Matters.

A major study by rural education researchers Craig and Aimee Howley addresses the vexing problem of how individual students of various income levels fare in larger and smaller schools nationwide. It is the strongest evidence to date that small schools are better for low income children.

Importantly, it also refutes other research that concludes a school that is "too small" can harm achievement, and casts doubts on conclusions from yet other studies that students from wealthier families do better in larger schools.

The Poverty and School Size Connection

A series of previous studies (many involving the Howleys) has shown that small schools reduce the harmful effects of poverty on student achievement in nine states (see summary of these studies in April 2004 RPM). Most have also concluded that larger schools produce higher achievement among the most affluent student populations, but that the benefits of smallness for the others is greater than are the benefits of largeness for the affluent.

These earlier studies have used the average test scores of all students in a school or a district as the measure of student achievement and the overall percentage of those students who are poor as the measure of poverty. Using average student test scores means these studies can only measure the performance of schools of various sizes serving communities of various poverty levels. They do not reveal how individual students of various income levels perform in those schools.

One study on that issue that is often cited (V. Lee and J. Smith, High school size: Which works best, and for whom? In Educational Evaluation and Policy Analysis, 19(3), 1997, pp. 205-227) examined high school level performance only. It concluded that while high schools should be smaller than many are, they can be too small too, and the ideal size does not vary by the income level of students who attend the school. The study concluded that the ideal size for a high school is somewhere between 600 and 900 students.

This result has been widely cited both as justification for smaller schools in urban areas where many are over 900, and as justification for school consolidation in rural areas where many are below 600.

But as the Howleys point out, the data set used for the Lee and Smith study-the National Educational Longitude Survey (NELS)-does not reflect the national distribution of school sizes. Even though Lee and Smith used statistical procedures designed to reduce this bias in the data set, the procedure still resulted in a sample in which the larger half of the schools was represented by ten times as many schools as the smaller half. As a result, there is a much greater likelihood of sampling error that invalidates the conclusions about the lower size limit for an "ideal" school than for the upper size limit.

The Findings

The new study focused on all 8th grade students in 22,811 public schools offering 8th grade in 1988, the first year for the NELS data set.

Those schools with fewer than the median number of 8th graders-84-were classified as "smaller" schools. Those with 84 or more 8th graders were "larger" schools. The schools were also grouped into school-size deciles-that is, the smallest 10 percent of schools with an 8th grade were the first decile, the next larger 10% of schools were the second decile, and so on. Within those deciles of schools, students were grouped into poverty quartiles-the one-fourth of the students from lowest-income families, the next lowest-income one-fourth, and so on. Test scores in math, reading, science, and history were analyzed.

The study considered three major issues.

1. How does the average test score for students in each of the four income quartiles compare for those students attending larger versus smaller schools, and are the differences statistically significant?

Results: Students in the lowest income quartile did better in smaller schools than in larger ones in every subject area test. So did those in the next two income groups. Only among the highest income group did students in larger schools perform better. But the positive advantage of larger schools for the highest income group was less than one-tenth as great as the positive advantage of smaller schools for the lowest income group. And in all four subject areas, the advantage of smaller schools for the lowest income groups was large and showed strong statistical significance. But in all four subject areas, the apparent advantage of larger schools to the wealthiest students was small and not statistically significant.

A separate analysis indicated the same results-but even stronger-for students attending rural and small town schools.

2. How do the average test scores of the lowest income quartile compare among each of the ten school size groups?

Results: In each subject area, average scores among the lowest income quartile of students were highest in the smallest decile of schools reported (which was the second decile because there were too few students in the first decile schools to allow statistical reporting under NELS guidelines). The smallest reported decile included schools with between 12 and 23 eighth graders, the largest included schools with more than 296 eighth graders.

In each subject area, scores generally fell in each successively larger decile of school size, with only a few exceptions. In every subject area, the difference between lowest income students' performance was higher in the smallest schools (that is, the second decile) than in the largest schools (the tenth decile) by a very wide margin.

3. Is the relationship between individual student achievement and individual income level stronger in the largest or the smallest school size groups?

Results: The correlation between individual poverty and individual test scores in the smallest reported decile (again, schools with between 12 and 23 8th graders) was 60 percent weaker than in the largest decile in math, 39 percent weaker in reading, 50 percent weaker in science, and 45 percent weaker for history.


These findings affirm most of the conclusions reported in the state studies summarized in RPM (April 2004). Smaller schools benefit student achievement for all but the very highest income groups. They weaken poverty's power over achievement, and the relationship is continuous among schools of all sizes-the smaller, the better for achievement among the poorest populations.

Contrary to the flawed findings in the Lee and Smith study, the evidence in this study shows that there is no lower limit for the "ideal" school size. Schools with as few as 12-29 eighth graders performed better than larger ones. Below that, there is not enough data to reach any conclusion.

Contrary to the Howleys' own findings in earlier studies that looked only at average test scores at the school level, this analysis using individual student-level test scores shows little or no achievement benefit of larger schools even for the most affluent students.

What does it mean? The Howley's rip off ten crisp "practical recommendations" in plain language rare in education research. Among them:

  • Sustain the smallest schools in the poorest communities.
  • In communities that serve all social classes, do not build large schools.
  • In affluent communities, do not build high schools larger than 1,000 students.
  • Provide decent support to smaller schools: small size improves the odds of success, it does not guarantee it.
To read the full report, go online to: