Smaller Districts: Closing the Gap for Poor Kids

Craig Howley & Robert Bickel (August 2001, draft article manuscript)

School district size has been researched almost as much as school size, and yet, it's hardly ever mentioned positively in the context of school reform. Our recent work on school and district size is prominently cited in literature reviews, and yet the school size results have gotten all the attention, and the district size results have been completely ignored--even though the findings were nearly the same!
Most recently we published a study that explicitly links district size and school size. That study should make it harder to ignore district size in the future. ASBJ asked us to explain our research, especially the study linking school and district size. Three things actually need explaining here. One: Why do school reformers ignore district size? Two: What did we find in our studies? And Three: What does it all mean anyhow?

Why Do School Reformers Ignore District Size?

This might be the most difficult of the three questions. One answer might be that school reform is all about doing the possible. That means easy and quick, and the seemingly easiest and cheapest things to change have to do with classroom instruction, directly. The theory is that we're already paying teachers (the most costly investment in schooling), and what's wrong is that they're teaching the wrong way. That's the theory. In reality, however, getting teachers to change the way they operate in the classroom is profoundly difficult. It seems largely to be a question of "resistance to change," but it's also a question of justifiable skepticism on the part of experienced teachers. They've seen the fads come and go, and they've resolved to trust their own judgement. (1)

The next cheapest thing to change is the curriculum. The theory here is that we're teaching the wrong things. Curriculum change is more expensive than professional development for two reasons. First, it involves the development and publication of new materials. This takes time, lots of new money, and comparatively rare expertise. Second, it's still more expensive, since to teach the new materials, teachers have to change the way they're teaching. Curriculum change, therefore, absolutely requires the same sort of professional development required simply to encourage teachers to teach differently. (2)

More expensive, and considered much more difficult to change, are the environments of schooling. (3) The theory, in this case, is that context determines the meaning of everything else that happens. A rotten environment makes everything in it rotten. In this view, large schools (or perhaps the toxic cultures for which size is a substantial risk factor), will usually subvert the best teaching and materials. Anyhow, that's the theory. It's the one that makes sense to us. Think of a fine meal. Then think of throwing the meal straight from the kitchen into a dumpster (or compost bin). Or think of business: throwing good money after bad.

To summarize: Class-size and school-size reform tends be viewed as an add-on to (a) teacher training and (b) new materials development. It might be nice, but it's not considered very affordable or certainly not essential. (4) From this vantage on class- and school-size reform, believing that district size might influence student performance is like believing that the alignment of the planets influences human destiny: wacky.

So, the answer to our first question is that (a) educators and policy makers don't think district size has much to do with learning, (b) they consider smaller districts too expensive, and (c) they think it's not possible (too dangerous to existing power relations). Our recent research, however, strongly suggests that (a) and (b) are probably wrong, and extant state systems provide compelling counterexamples to (c). (There are indeed many states that maintain lots of smaller schools and districts--which is why it's possible to study them as we did.) It's time, we think, to create all school districts on a human-friendly scale (but not all the same size). Let's turn to the evidence.


What Did We Find?

The initial series of studies, dubbed "The Matthew Project" (5) looked at school and district size (separately from one another) in Georgia, Ohio, Montana, Texas, and West Virginia. In our five states we found two important patterns for both schools and districts--with one notable exception (the inspiration for our multi-level study).

Two patterns. The first pattern we call the "excellence effects" of size. Basically, schools and districts with lots of poor kids (percentage-wise) are predicted to have higher average test scores if they are smaller: the smaller the school, the higher the predicted score. In scientific terms, "the effect of size is negative in impoverished schools": as size goes up, school and district-level student achievement go down (not a law, just a predictable relationship consistently evident in data in six states). In affluent schools and districts, however, the opposite is true. Larger sizes improve average test scores in affluent communities (up to a certain reasonable limit (6)). In scientific terms this means that "the effect of size is positive in affluent communities." It's important to stress that the very same pattern applied to districts in California, Ohio, Montana, Texas, and West Virginia. Georgia was the notable exception--we found no such "excellence effect" at the district level. Also, in Montana, with many small schools and many small districts, these patterns were evident, but they were much weaker than any of the other states.

The second pattern we call the "equity effects" of size. These effects focus on the strength of the relationship between socioeconomic status (SES) and achievement. Nearly everyone knows that SES exerts the consistently strongest influence on student achievement. The relationship is not perfect, just very, very strong, usually accounting for about one-half to two-thirds of the variance in achievement. In general, the odds of getting high test scores are improved by high SES and worsened by low SES. It's not a law, just a strong tendency.

Thinking about this relationship, we thought that the path to higher test scores in impoverished schools might be related to the weakening of the typically strong bond between SES and achievement in smaller schools. Kind of makes sense, right? Poor kids do better if the if the link between achievement and poverty is disrupted--if unfair disadvantages are removed and they get a more equal opportunity to succeed. It's the American Way (or, for some of us, the supposedly fictionalized American Way).

Anyway, to test this idea, we divided schools and districts in half according to size, creating "smaller" and "larger" halves in each state. Then we simply calculated the correlation of SES and achievement in each half; the range of achievement and SES in the halves was comparable, incidentally.

Our hunch turned out to be very much on target. The relationship between aggregate achievement (student achievement averaged for a school or district) and SES was consistently weaker in smaller schools and districts. Often it was much weaker, even dramatically weaker. In Montana (again, a state maintaining many absolutely small schools and districts), this mitigating effect was so strong that the measured correlation between achievement and SES was sometimes not statistically different from zero. In any case, the "equity effects" were substantial, and remarkably consistent across all states, at 3 to 5 grade levels within state, and at both levels of size (school and district).


--------------------------------------------------------------------------------

Sidebar: Mitigating the Influence of SES

Mitigating the influence of SES might not necessarily strike all readers as a good thing. Certainly, generous people would favor mitigating the influence of poverty, but mitigating the influence of affluence? Does that make sense? Don't we want affluent kids to succeed? The only legitimate answer is "yes." But it's important to remember that the correlation of SES and achievement is what's called a "zero-order" correlation: a relationship without such things as effort, relative difficulty of the curriculum, or ability factored in. One major reason that affluent schools do so well on these tests is that they are far too easy for many of the kids that attend them. In the technical terms of test construction, the test ceiling is low. The tests are constructed to probe grade-level competence. Some schools are filled with kids in the fourth grade who are reading at the 8th or 12th grade levels, but who nonetheless continue to be measured against the prevalent age-grade treatment, despite its inappropriateness. From this vantage, the positive association of SES takes on a quite different meaning. We reward high-SES schools for coasting. The association of SES and achievement is as unfair as the association of poverty and achievement, but it promotes bad habits in high-SES schools.


--------------------------------------------------------------------------------

This is powerfully good news, since educators have been looking nearly in vain for systematic ways to blunt the negative effects of poverty on school performance. If smaller schools and districts represent the American Way of decentralization, it's a Way that evidently works to promote academic achievement among poor kids and impoverished communities. Based on our initial findings, we'd say that it should be policy that poor kids attend smaller schools and that they attend smaller districts. Instead, of course, de facto education policy seems to have been doing the opposite--making sure many, probably most, poor kids attend large schools in large districts! It's proved to be a recipe for academic and probably social disaster. (7)

Larger schools (we recommend upper limits of about 1,000 for 9-12 high schools and 500 or so for K-6 elementary schools) should be reserved for affluent communities only. Unfortunately, schools much larger than this are still being built, as we know personally from a continuous stream of correspondence sent to us by alarmed community members across the nation.

The multi-level study. A remaining question--and the one that prompted our study linking the two levels of size--is this: should we make sure that poor kids attend smaller schools in smaller districts? To address this question, we looked at Georgia, because that's the one state in which we did not find an "excellence effect" for district size (though we did find a strong equity effect of district size). Perhaps, we thought, district size might work indirectly, in combination with school size to influence school-level achievement. Because the single-level studies looked at schools and districts separately, they would not have been able to detect such a "cross-level" relationship.

Now, people who work in schools have a strong sense that cross-level influences exist. District politics and economics are felt at the school level. Results can be disastrous, for instance, when a lousy principal is assigned to a school to satisfy bureaucratic procedures. Furthermore, the community context of a district has an impact at the school level, again according to anecdotal accounts. Drug deals and shootings on the corner render the cultivation of a "learning community" a pretty doubtful effort. The multi-level study merely tested the possibility that cross-level interactions might influence achievement, using a formal analysis of relationships evident in a large set of numerical data. (8)

What did we find in this multilevel study? Substantial cross-level influences most certainly exist. The "excellence effects" are stronger at the 8th grade, and weaker at the 11th grade (the two grade levels studied). Interestingly, district size interacts with school-level SES, and school size interacts with district-level SES to influence achievement at the 8th grade level. At the 11th grade level, school size interacts with district level SES only. The single-level pattern of interaction (school size with school-level SES) was not evident at either the 8th or the 11th grade. A number of other single-level and cross-level interactions, however, also exerted an influence on school-level achievement: size (the two levels of size interacting) and percent African American (again, percentages at the school and district levels interacting). Class size was a weak and inconsistent influence at best, and exhibited no cross-level interactions. It exerted a statistically significant negative influence only in the case of 11th grade "composite" achievement.

The two-level "equity effects," as with all the single-level studies, were very strong. As with the single-level studies, we divided districts and schools in half by size. But here, we further assigned our smaller and larger schools according to whether they were located in smaller or larger districts. This gave us four groups: larger schools in larger districts, larger schools in smaller districts, smaller schools in larger districts, and smaller schools in smaller districts. Then we calculated the correlation of school-level SES and school-level achievement within each group, and for the four comparable tests at the two grade levels. The accompanying figure details the results for mathematics achievement (numbers refer to the proportion of variation in school-level mathematics achievement associated with variation in school-level SES).

The effects are not only strong, but they show a distinctive pattern--and not only in mathematics, but in the other three comparable achievement domains, as well, and at both grade levels. The relationship is strongest (that is, worst) among larger schools in larger districts. At the 8th grade level, 71% of the variance in school-level mathematics achievement is accounted for by school-level SES. The relationship weakens somewhat for larger schools in smaller districts (59%), weakens still more for smaller schools in larger districts (46%), and is weakest among smaller schools located in smaller districts (29%). At both the 8th and 11th grade levels, then, smaller schools in smaller districts reduce the influence of SES by about 60% (e.g., 71%-29%/71%). Across the four comparable domains-- math, composite, reading, and science--reductions in the strength of the SES-achievement bond vary from a low of about 50% (8th grade composite) up to a phenomenal reduction of 80% (11th grade composite). The pattern of equity effects is the same, as well: achievement among...

...larger schools in larger districts shows the strongest relationship with SES,

...larger schools in smaller districts shows a somewhat weaker relationship,

...smaller schools in larger districts shows a still weaker relationship,

...smaller schools in smaller districts shows the weakest relationship with SES.

It's hard to characterize what is happening among larger schools in larger districts as anything less than scandalous--particularly in light of the contrary situation prevailing among smaller schools in smaller districts. Among the former group, SES exerts an unacceptable (and scary) determining influence on school performance, but among the latter group, this prevailing relationship has been substantially and consistently--almost miraculously--disrupted.
In short, then, we found that district size makes a considerable difference in the equity and excellence of school-level achievement, even in Georgia, where the previous (single-level) study had not detected such an influence. We'd surmise, then, that such cross-level relationships would prevail in the other states studied (and probably elsewhere, given the variety of state contexts involved).

It's essential to note that these patterns are not the result of intentional reform. They merely show up when data on ordinary life, with all its warts, is examined. These results are inherent in the way some of us have structured educational systems. The good stuff here is associated with durable structural features of schooling--the very things that have usually disabled the best reformist intentions of the past century!

Of course, a multi-level study sets out to study complexity, and the results it reports are indeed complex--not to mention difficult for most readers to understand, including researchers. This brings us to the last of our three questions.


What Does All This Mean?

The multi-level study suggests something important (at least to the two of us!): a more user-friendly scale supports the equity and excellence of school performance. (9) Others have written about the scale of the schooling enterprise, but it's seldom actually been studied in a formally rigorous way. The multi-level Georgia study is, so far as we know, among the first to link levels of organizational complexity (e.g., such levels as classroom, school, district, state, or nation), inching toward a more empirical outlook on "scale" (see sidebar).

Modern industrial nations pioneered the idea of "economy of scale": bigger production units lower the cost of the items produced. Henry Ford put cars in all the former horse-stables in America with this insight.


--------------------------------------------------------------------------------

Sidebar: Scale

Scale is more than size. Scaling has to do with proportion, and that is a complex idea, as middle-school math teachers trying to explain it to students well know. It's also an idea inherent in fractal geometry and the "new sciences" of complexity and chaos theory. The idea is that at different levels of organization, structures repeat themselves. It's just possible to imagine that the organizational structures inherent in larger districts impose themselves on larger schools in those larger districts, and ultimately, through the organization of classrooms and instructional routines, on the mindsets of students, teachers, administrators, and parents interacting with such systems. This insight is very speculative, and the means of studying it are hardly clear to anyone, but the speculation does illustrate the import of the idea of "scale" as opposed to the far more innocent idea of "size." The upside is this: the kind of mindsets that more humanly scaled educational systems evidently seem to generate do already exist, and rather widely at that. Possibly, more humanly scaled institutions are better at cultivating humanity. It would, at any rate, seem logical!


--------------------------------------------------------------------------------

And true to form, Americans came to agree that what was good for Ford was good for Ford Elementary and, especially, Ford High School. Not coincidentally, given the prerogatives of professional management, the number of school districts has declined since 1900 by an almost unbelievable 90%. This change, incidentally, has also removed a proportionate number of citizens from the governance of public schools. Public schools are therefore less "public" than ever.

Smaller-scale institutions are widely needed in America. This need is especially critical in schooling, because ours is the one public institution that affects everyone, and occupies long stretches of our lives. If schools propagate mindsets, and it seems that instruction ought to do just that, then we surely want institutions where people know and attend carefully to one another. It seems this is not generally the case in larger schools in larger districts, and we should not be surprised to learn that this exacts a toll from student achievement.

Not all schools and districts need to be absolutely small. High schools of 1,000 are plenty large to provide a marvelously comprehensive curriculum, according to David Monk, a prominent school size researcher, now dean at Penn State. Even 400 kids is plenty, he says. Smaller schools can be equally effective with a more focused curriculum; the "need" for a comprehensive curriculum was not, after all, a commandment handed to Moses from the Allmighty.

So far as district size is concerned, those of 500,000 students or even 50,000 students can believably claim they provide responsive treatment for patrons and their children. Even 5,000 might be a bit too large! Perhaps (this will surely infuriate some readers) districts should be no larger than needed to "field" a single high school. That proposition might suggest an upper limit to district size of about 3,000 kids in an affluent community, and a district of 750 or less in impoverished communities. If this is a radical proposition, however, it's already being practiced by a lot of conservative communities nationwide. Just remember that there's evidence that such schools in such districts make more equal educational opportunity a reality for poor kids and families.

If you regard this thinking as nutsy, consider the counterexample of Montana. The average district there operates less than 2 schools! Montana even retains the K-8 district that many American educators have difficulty even imagining (not those in some New England states, however). Note also that states like Montana and Vermont are have smaller populations than many states; arguably, the scale of their whole educational systems are more humane. While we're at it, nations like Singapore and Hong Kong (a nation-within-a-nation) might also claim such an advantage, if comparison of national size is the reference point.

Are we holding Montana, Vermont, Singapore, and Hong Kong up as models? Hardly. They all have lots of warts. Let's be plain: don't emulate Montana or Singapore! But readers should note that Montana, with a large American Indian population, and its small schools and districts, and small overall population, turns in consistently top-drawer performances on successive waves of the National Assessment of Educational Progress (NAEP). Montanans tell us they are sick and tired of hearing about this success; Indian education there, as everywhere, confronts enormous challenges, and kids' aspirations for higher learning aren't robust enough, anti-intellectualism is rampant, school funding is in the cellar, and so on and so forth. This all makes it sound to us like they're taking care of needful business out there, however. Perhaps the scale of the enterprise helps them address the inevitable dilemmas more effectively. It's a theory.



Afterword

We keep asking ourselves the very same questions that we've posed in this article, and the answers, even these answers, are a work-in-progress. They are by no means carved in stone, but neither are they written in sand on tidal flats. And they are not just our opinion.



Resources

Bickel, R., & Howley, C. (2000). The influence of scale on student performance: A multi-level extension of the Matthew principle. Education Policy Analysis Archives [On-line serial], 8(22). Available World Wide Web: http://seamonkey.ed.asu.edu/epaa/v8n22.html.

This is our recent multi-level study.


Guthrie, J. (1979). Organizational scale and school success. Educational Evaluation and Policy Analysis, 1(1), 17-27.
A seminal study on school size (with size taken as symbolic of "scale").


Howley, C. (2001). Research on smaller schools: What education leaders need to know to make better decisions. Arlington, VA: Educational Research Service.
Very practical summary of research, particularly developed to inform thoughtful administrative practice and policy making.


Howley, C. (2000). School district size and school performance. Charleston, WV: AEL, Inc. (ERIC Document Reproduction Service No. ED 448 961)
One of the few summaries on the topic; presents the Matthew Project district-level findings.


Howley, C. (1997). Dumbing down by sizing up: Why smaller schools make more sense--if you want to affect student outcomes. The School Administrator, 54(9), 24-26, 28, 30.

Written for administrators.


Howley, C. (1996). Ongoing dilemmas of school size: A short story (EDO RC-96-6). Charleston, WV: ERIC Clearinghouse on Rural Education and Small Schools. (ERIC Document Reproduction Service No. ED 401 089)


Contrasts rural and urban dilemmas of size.


Howley, C., & Bickel, R. (1999). The Matthew Project: National report. Randolph, VT: Rural Challenge Policy Program. (ERIC Document Reproduction Service No. ED 433 174)


Formal report to the Rural Trust (sponsors of the Matthew Project research). Popular accounts are available at th following URLs:

http://www.ruralchallengepolicy.org /sapss.html (sponsoring organization)

http://www.ael.org/eric/digests/edorc0010.htm (ERIC Digest)


Slater, R. (1989). Education scale. Education and Urban Society, 21(2), 207-217.
A rare overview of the issue of educational scale.



1. We're beginning to wonder ourselves, and we suspect that a justifiable behaviorist backlash might be right around the corner. Even in the domain of "small schools" a "small schools" unfounded rhetoric has begun to subvert careful action, with a predictable result: the status quo of school size.

2. Professional development (in-service training) has not been shown to improve student performance, nor have "improved" instructional materials. Some professional development is useful, and some materials are worthy--but, on average, it's a wash. One size doesn't fit all when it comes to instruction, curriculum--or, in fact, school and district size.

3. Systemic reform, which aims to change everything all at once, also avoids tampering with structural features of the environment like school and district size.

4. Class size limits have been imposed, but most don't reach the levels of reduction advised for substantial gains, and small school size -- though rhetorically acknowledged as having a substantial impact on achievement particularly for poor kids -- isn't materializing: schools aren't getting any smaller. And no one is even talking about district size--except, in rural areas, about how there are "too many" small districts.

5. After the New Testament verse, Matthew 12:13-- "Whoever has will be given more, and he will have an abundance. Whoever does not have, even what he has will be taken from him." The Matthew Project, with support from the Rural School and Community Trust, extended a 1988 study by Noah Friedkin and Juan Necochea, who pioneered a new theory of size and tested it with data from California districts and schools. All six states are very different from one another, of course. The studies included virtually all schools in all these states: urban, rural, and suburban.

6. We're against mega-schools anywhere, however: high schools with more than 250 kids per grade and elementary schools with more than about 125 kids per grade. Schools in impoverished communities need to be much smaller than this. We also think that schools should have broad grade span configurations (e.g., K-12, K-8, K-6, 7-12, 9-12 instead of K-4, K-2, 4-6, and "ninth grade academies"). A forthcoming research article (in the online journal, Education Policy Analysis Archives) by our research team will provide the additional evidence that supports this position.

7. The rural and urban challenges are very different here, with rural community members struggling to fend off district consolidation and school closure, and urban residents struggling to convince power brokers to break up mega-schools and make mega-districts somehow more responsive. (See resource list for more information.)

8. School-level achievement was the dependent variable, and independent variables (assessed at both the school and the district level) included: size (i.e., of school and of district) subsidized meals rate, percent minority students, percent African American students, and class size. The data represented practically all Georgia schools containing an 8th grade or an 11th grade. We tested for effects in all 13 achievement domains available in Georgia SEA data sets (8 at the 8th grade level and five at the 11th grade level--we deemed four comparable across the two grade levels, and used these four to assess "equity effects."

9. Maybe, like so many "conclusions of research," this one too can be accused of being "common sense." Perhaps, but it's not necessarily a widely practiced common sense. It's worth repeating: it is practiced some places.