Discussion - comment on John Bishop, in this issue - Special Issue: Earnings Inequality

This is the right line of research at the right time. The fortieth anniversary of Brown v. Board of Education just passed and the thirtieth anniversary of the Coleman Report (Coleman et al. 1966) is arriving; the time has come for a serious reconsideration of race, segregation, and schooling. Over the past decades, a wide variety of desegregation and compensatory programs have been introduced, so that their effects should now be evident. Additionally, there is a new willingness (perhaps overwillingness) to consider major restructuring and even elimination of programs. Thus, it would be nice to have evidence about what is and is not important in student achievement. Toward this end, John F. Kain and Kraig Singleton are creating a truly unique data set that will permit investigation of some of the key questions that have almost completely eluded educational researchers. And of course Kain, an early interpreter of the Coleman Report and one of the nation's premier researchers into the nexus of race and space, is uniquely prepared to undertake this investigation.

Given the local basis of education and the patterns of local control of educational decisions, a discussion of education is inherently a discussion of the spatial distribution of opportunities. In terms of this conference, the spatial structure of schooling provides clear linkages between today and the future. So it is of some importance to understand how schooling opportunities interact with school attendance patterns and racial disparities in educational quality.

Race and Schooling

The motivation for the Coleman Report, a study mandated in the Civil Rights Act of 1964, was to investigate the "lack of availability of equality of educational opportunity for individuals by reason of race, color, religion, or national origin." This report and the follow-on by the U.S. Commission on Civil Rights (1967), entitled Racial Isolation in the Public Schools, focused attention on one of the most obvious characteristics of the schools of the mid-1960s, their separation by race of the students. While not in their direct charge, these studies also began to provide information that could be used to evaluate the achievement effects of what is one of the largest and most long-running social programs in our nation's history - the effort to desegregate the schools of both the Old Confederacy and the rest of the Union. Given this backdrop, it is useful to begin with a quick summary of what we know about race and schooling from these original studies and intervening studies.

My overall summary is as follows:

1. Large disparities by race exist in school performance (measured, say, by the National Assessment of Educational Progress), although the gaps have closed some over the past 10 to 15 years.

2. The racial composition of schools has changed in fairly complicated ways related to the imposition of desegregation policies (voluntary or otherwise), to the development of housing patterns within cities, and to the general decentralization of the population. Nonetheless, while the patterns vary across regions, the amount of racial contact in the schools has increased over the past decades (Welch and Light 1987).

3. The racial composition of the schools has minimal effects on student test performance, other things being equal.

4. Quality of the schools may be correlated with racial composition, although this is not particularly well documented.

5. Limited progress has been made in addressing the important issues of how school policies interact with racial disparities in performance, largely because the data available for analysis have not been at all adequate.

Kain and Singleton have embarked on a data construction effort in the state of Texas that directly addresses point 5 and holds promise for filling in the details on points 3 and 4. Their data set, which is still under construction, could become the richest data set ever compiled to address central issues of educational policy, particularly as related to race and space. Until now, the largest and most comprehensive data base has been the one for the original Coleman Report, even though it has a number of serious flaws for the investigations of interest here. The Kain and Singleton data set will clearly leapfrog that data base. Without repeating their description, the key features include the extraordinarily large samples, the ability to follow individual students over time, and the ability to link school resources rather closely to individual students.

The Kain-Singleton Analysis

The analysis in this paper largely concentrates on a series of very important descriptive issues. While this analysis considers only a small part of what they can eventually exploit, the authors begin to provide important insights that motivate analyses yet to come.

The basic starting point is a finding that clear and systematic differences exist in student test performance by race and ethnic background. While not surprising in light of other data, this finding sets the scene for the central analysis. An important point, however, is that the differences are larger for low-income blacks and Hispanics. This interaction between income and race is less well known or documented in past work.

Most of the new analytical efforts within this paper are devoted to understanding the distribution of school resources across schools in Texas. Before doing this, however, they present what I believe is the key table for interpreting all of the results - their Table 6. Table 6 presents the only estimates in the paper of the determinants of student performance. These 6th-grade results are clearly preliminary and subject to modification with further refinements. Nonetheless, they are rather remarkable. The first column presents estimates of achievement models that employ just income-race interactions (plus student gender and age). The fifth column presents estimates of this same model with individual school fixed effects, that is, a dummy variable for each of the about 2,000 separate campuses. At least at the visual level, the estimated differences in performance by race appear independent of school level inputs. In other words, the racial differences are not affected by differences in school level resources.

This finding does not particularly surprise me, because I have long held that school quality is not closely related to expenditure or conventional school inputs (Hanushek 1986). But it does provide a somewhat different interpretation of much of the Kain-Singleton analysis.

The focus of attention of their study is how school resources vary by race of the school. They examine scores on teachers' tests (TECAT), master's degrees, teaching experience (novice or old), and class size. The analysis is very clever and demonstrates the power that comes from their data set. They investigate how resources differ by race, holding constant overall district factors through the use of district fixed effects. The general form of the regressions calls for regressing each of the school resource measures on percent black, percent Hispanic, and interactions with income along with a district fixed-effect term.

Several aspects of these analyses stand out. First, and most important, these resources consistently are distributed such that more resources go to schools with low minority populations. Schools with high proportions of blacks and Hispanics simply get less of each of these resources.

Second, and somewhat unexpected, the pattern and the magnitude of these race effects are very similar across districts. Large MSAs as a group or individual large districts look quite similar to all districts in the state. (Again, these conclusions are not based on formal statistical tests but instead on qualitative summaries of the estimated models.) The apparent uniformity belies conventional views that such race effects are larger and more intense in the big urban centers.

Third, their careful consideration of the measurement of inputs is admirable. They work hard at constructing solid estimates of teacher test scores. They also provide an interesting supplemental analysis of how class size varies widely by type of instruction and grade level, adding a real caution about inherent conceptual difficulties in measuring class sizes for districts. Average class size for a district, for example, will be a very poor measure of potential performance effects if there are nonlinearities in how class sizes affect performance.(1)

As mentioned, the interpretation by many people of these resource variations is that they indicate disparities in the quality of schooling received by students. My interpretation is different, because the evidence on resources indicates that master's degrees and class size are not closely related to student performance. For example, in 277 separate estimates of the effects of teacher-pupil ratios on student outcomes, 15 percent find statistically significant positive effects while 13 percent find statistically significant negative effects (Hanushek, Rivkin, and Taylor 1995). The remaining 72 percent are statistically insignificant; that is, we are not very confident that student outcomes are affected by teacher-pupil ratios. Teacher experience shows somewhat stronger effects but, as Kain and Singleton point out, causality is not well sorted out. The evidence on test scores tends to be stronger: 26 of the 36 studies with estimated effects on student achievement are positive and 15 of those are statistically significant.(2) Thus, past work might suggest taking the TECAT variations more seriously in terms of potential effects on student outcomes.

But remember Table 6. That table suggests that school-level differences do not affect racial differences in student performance. By implication this supports a finding of "no effect" of these factors, because we know that these factors are themselves distributed in a systematic manner by race and ethnicity.

The overall patterns of resource variations remain inherently interesting. If these hold up to further refinement of the data and analyses, they suggest systematic discrimination in the operation of schools. Resources that are conventionally thought to affect student achievement are systematically distributed toward the majority whites in Texas and away from the blacks and Hispanics. We can thus be thankful that these resources in reality do not appear to have much to do with student performance.

Finally, I must conclude with a statement of anticipation. Kain and Singleton have constructed a data base that is likely to become the source of much new knowledge about schooling. Issues ranging from the effects of school desegregation to the impacts of student migration to the effects of special education and other distinct programs all can be brought under the spotlight of their data. They should be encouraged to work faster, so we can have the answers sooner.

1 Some people, beginning with Glass and Smith (1979), argue that class sizes above some level have little effect on performance but have significant effects below a cut-off - roughly 15 students per teacher in the Glass and Smith analysis. Ferguson (1991) argues from Texas data that class size effects become more important when pupil-teacher ratios rise above a threshold. Specifically, "the number of 'students per teacher' is important only when it exceeds eighteen" (p. 477). Both of these studies imply nonlinear responses to variations in class size, and suggest that aggregation across grades and schools within districts will lead to significant biases.

2 This summary omits the five studies that report statistically insignificant effects but do not report the sign of the estimated relationship.

References

Coleman, James S., Ernest Q. Campbell, Carol J. Hobson, James McPartland, Alexander M. Mood, Frederic D. Weinfeld, and Robert L. York. 1966. Equality of Educational Opportunity. Washington, D.C.: Government Printing Office.

Ferguson, Ronald F. 1991. "Paying for Public Education: New Evidence on How and Why Money Matters." Harvard Journal on Legislation, vol. 28, no. 2 (Summer), pp. 465-98.

Glass, Gene V. and Mary Lee Smith. 1979. "Meta-analysis of Research on Class Size and Achievement." Educational Evaluation and Policy Analysis, vol. 1, no. 1, pp. 2-16.

Hanushek, Eric A. 1986. "The Economics of Schooling: Production and Efficiency in Public Schools." Journal of Economic Literature, vol. 24, no. 3 (September), pp. 1141-77.

Hanushek, Eric A., Steven G. Rivkin, and Lori L. Taylor. 1995. "Aggregation and the Estimated Effects of School Resources." Working Paper No. 397, Rochester Center for Economic Research, University of Rochester, February.

U.S. Commission on Civil Rights. 1967. Racial Isolation in the Public Schools. Washington, D.C.: Government Printing Office.

Welch, Finis, and Audrey Light. 1987. New Evidence on School Desegregation. Washington, D.C.: U.S. Commission on Civil Rights.

Eric A. Hanushek, Professor of Economics and of Public Policy at the University of Rochester.

COPYRIGHT 1996 Federal Reserve Bank of Boston
COPYRIGHT 2004 Gale Group