SECTION: Vol. 77 ; No. 3 ; Pg. 210; ISSN: 0031-7217
LENGTH: 4080 words
HEADLINE: Detracking America's schools: the reform without cost?
BYLINE: Brewer, Dominic J. ; Rees, Daniel I. ; Argys, Laura M.
ABILITY GROUPING or tracking has long been used in public schools as an important method for organizing students, particularly at the high school level. It is seen as a way to narrow the range of performance and motivation in a group of students, thereby making teaching easier and preventing less able students from "holding back" those with greater academic talent. However, as concern over test scores, dropout rates, and related social ills has grown, tracking has become increasingly controversial. Opponents of the practice have argued that tracking stigmatizes students who are consigned to low-track classes with less experienced teachers, fewer resources, and lower expectations. Moreover, the detrackers maintain that average and even above-average students do not derive substantial academic benefits from being grouped together.(1)
The detracking bandwagon has gathered some steam. In her 1992 book, Crossing the Tracks: How 'Untracking' Can Save America's Schools, Anne Wheelock argued that detracking is a necessary component of successful school reform and detailed several experiments.(2) A recent Kappan article described the case against existing tracking practices as "compelling" and offered examples of practical ways in which detracking might actually be carried out.(3) The National Education Association has recommended that schools discontinue "conventional tracking practices"; the National Governors' Association has also endorsed detracking.(4) The conventional wisdom" seems to have evolved into a belief that tracking is universally bad for low-ability students and neutral for other students. If this is true, the policy prescription is very clear: detrack. Everyone will gain, and no one will lose.
Our purpose here is not to pour cold water on detracking efforts or to argue that detracking is necessarily bad policy. Rather, we want to challenge the view that tracking can be ended with little or no cost. First, we revisit the previous research on the effects of tracking on student academic achievement and find remarkably little support for detracking efforts. Second, we present some findings of our own, based on analyses of nationally representative data and using current statistical techniques. Our estimates suggest that, while public high school students in low-track math classes do worse on standardized tests than they would have done had they been in an untracked class, students in high-track classes actually perform better academically. This suggests a less clear-cut policy prescription.
How Widespread Is Tracking?
Tracking is a pervasive phenomenon in American secondary schools. The National Educational Longitudinal Study of 1988 (NELS), a nationally representative student survey conducted by the National Center for Education Statistics (NCES), provides perhaps the best available picture of tracking practices.(5) In 1988 NELS sampled more than 20,000 eighth-grade students in more than 100 schools. Most of these students were resurveyed in the 10th and 12th grades in 1990 and 1992. The survey contains a range of questions dealing with student academic performance, family background, attitudes, and school experiences. Teachers, administrators, and parents were also surveyed. Indeed, NELS is unique in that it allows researchers to link a student with a particular class and teacher in a given subject area (mathematics, science, English, or social studies). Eighth- and 10th-grade teachers were asked about a number of classroom characteristics, including whether the class each student attended was composed of students of above-average, average, below-average, or widely differing (heterogeneous) achievement levels relative to other students in the school.
Using the descriptive statistics from these responses and adjusting them for the particular composition of the NELS sample enables us to obtain nationally representative estimates of the extent of tracking.(6) Clearly, tracking is a widespread phenomenon.(7) Only 14.4% of eighth-grade students were enrolled in math classes that their teachers characterized as heterogeneous; 38.8% of students were in math classes that their teachers considered to be about average, 25.8% were in classes considered above average, and 20.9% were in classes considered below average. The percentages of students who were in classes in other subjects that teachers considered to be heterogeneously grouped were 15.7% for English and approximately 20% for science and social studies. Similarly, only 10.8% of 10th-grade math students in pubIic school s were judged by their teachers to be in mixed-ability classes; 39.4% of students were in classes that their teachers considered to be about average, 24.6% were in classes considered above average, and 25.1% were in classes considered below average. The figures for heterogeneous grouping in other subjects were 11.6% for science, 17.7% for social studies, and 14.7% for English.
One of the major reasons that tracking has become unpopular has less to do with the outcomes the practice generates than with the types of students who tend to be assigned to the different tracks. A major concern is that tracking is used to segregate gate students on the basis of class and race, as well as ability. A great deal of research search has been devoted to investigating whether family background is an important determinant of track placement.(8) Although evidence from past studies (which typically control for student ability and other factors) has been mixed, it is clear from the NELS data that, if one does not control for ability, there is a strong correlation between socioeconomic status and track and between race/ethnicity and track. For example, for 10th-grade math classes, only 14% of children in the lowest socioeconomic quartile are in classes judged to be above average, while almost 38% of those from the highest socioeconomic quartile may be found enrolled in such classes (see Figure 1).(9) Blacks and Hispanics panics are more likely to be placed in a class judged to be low track or in a mixed-ability class.(10)
The Conventional Wisdom
Perhaps the most influential review of tracking research was done by Robert Slavin.(11) After summarizing 29 separate studies at the secondary level, Slavin concluded that the effect of tracking on students of any ability was essentially zero."(12) After reexamining this work, we have come to believe that such strong conclusions are unwarranted, for the following reasons. First, of the studies examined by Slavin, many (13) were unpublished dissertations, which, of course, were not subjected to independent peer review. Second, of the experimental studies, most used small samples, often taken from a single school. Third, of the nonexperimental studies, none used nationally representative data. Finally, with only one exception, all the studies examined were done prior to 1978.
Furthermore, the conventional wisdom that tracking does not have beneficial effects on student achievement has been undermined in more recent nonexperimental research that was based on large-scale data sets and that used more sophisticated statistical models. For example, Adam Gamoran and Robert Mare estimated that, after controlling for prior test scores, race/ ethnicity, and socioeconomic status, high school seniors in the college-preparatory track scored approximately 8% higher on their mathematics exams than seniors in the non-college-preparatory track.(13) Similarly, Thomas Hoffer found that placement in an upper-track (rather than a heterogeneous) class for eighth- and ninth-grade math and science was associated with an increase in students, scores.(14) Placement in a lower-track class (rather than a heterogeneous class) was associated with an even larger decrease in students, scores.(15)
Of course, evaluating the effects of tracking on student performance is problematic in both experimental and nonexperimental settings. Typically, experimental studies are poorly designed, there is a desperate need for systematic, independent evaluations of detracking experiments, but rarely are these carried out. Nonexperimental studies have, until recently, suffered from inadequate data. Student performance is related to a host of observable (and many unobservable and perhaps unmeasurable) individual, family, teacher, school, and community characteristics. If statistical models ignore some of these important factors, then biased results are likely. Early quantitative studies tended to suffer from this problem.
Two issues are worth highlighting. First, all previous research has failed to take account of the fact that observable teacher characteristics and other educational inputs, such as class size, vary across tracks. There is evidence to suggest that lower-track classes tend to be assigned to the least qualified teachers and, in general, tend to receive less than their share of educational resources. Similarly, upper-level classes seem to receive a disproportionately large share of resources." If this pattern of allocation indeed exists, then it is obviously important to control for it in any effort to examine die effect of ability grouping on learning.
Second, since assignment to tracks is made at least in part on the basis of prior ability, any real attempt to measure the effect of tracking itself must disentangle the influence of tracking from the process of assignment. This is obviously difficult to do. It is likely that factors that researchers cannot observe (e.g., student motivation that are correlated with student performancee also help determine the track to which a student is assigned. While statistical techniques exist to get around this problem, only the most recent studies have used them.
Some New Evidence
We recently undertook a study of the effects of tracking using a sample of students taken from the NELS data. The research was funded by the National Center for Education Statistics(17) and focused on public school students, academic achievement as measured by standardized test scores in mathematics.(18)
Specifically, we examined the effect of 10th-grade tracking on end-of-year 10th-grade achievement in mathematics for a sample of more than 3,900 public school students. Class-specific information was provided by each student's mathematics teacher.(19) It should be emphasized that these measures are not based on student self-reports and refer to a specific math class. One drawback to much of the previous work in this area is the use of student self-reports in order to measure track placement.(20)
We reexamined the impact of tracking on high school student achievement through the estimation of standard education production functions,, for students in each type of class- above average, average, below average, add mixed ability.(21) This involved using regression analyses to explain variation in student test scores on the basis of a set of explanatory factors. Variables at the individual level included the student's sex, race/ethnicity, socioeconomic status, and eighth-grade mathematics test score.(22) Classroom-level characteristics included class size, teacher experience, teacher certification, teacher absenteeism, and teacher education - all of which are likely to be correlated with both student achievement and class track. In fact, our data confirmed the proposition that high-track classes received more educational resources than low-track classes. For example, students in above-average math courses tended to be taught by more experienced teachers than students in either below-average or heterogeneous classes and were more likely to be taught by a teacher with a master's degree.
Our statistical models took into account what statisticians call "sample selection bias" If there are unobserved student or school characteristics that affect both achievement and track placement, then any association between achievement and tracking may stem from these characteristics. Failure to explicitly model the process through which students are assigned to a particular track may therefore lead to erroneous conclusions about the true relationship ship between tracking and student achievement Following previous research,(23) we modeled track placement as a function of prior (eighth-grade) achievement and such student characteristics as race/ethnicity and socioeconomic status.
Finally, we used the estimates of our achievement models to calculate the predicted test score for each individual in our sample had he or she been placed in each of the four tracks, and we found that tracking was an important determinant of student achievement. For example, if our entire sample had been placed in heterogeneous classes, the average test score was predicted to be 63.36 on a 100-point scale. The average 10th-grade mathematics score associated with the placement of all students in average classes was predicted to be 65.30; in below-average classes, and in above-average classes,
By comparing the predictions for the various tracks with those for heterogeneous grouping, we were able to assess the impact of tracking. Placement in a below-average math class, as compared to a heterogeneous one, was associated with a decrease crease in achievement of approximately five percentage points. Placement in an above-average math class was associated with an achievement increase of roughly the same magnitude. And placement in an average class was associated with an increase of somewhat less than two percentage points.(24)
These results suggest that detracking would create winners and losers. Although students in lower tracks would realize achievement gains by being placed in a heterogeneous class, this gain would be at the expense of students placed in higher - level tracks. Our estimates imply that detracking all students currently enrolled in homogenous classes would produce a net 1.7% drop in the average mathematics test score.(25)
The conventional view that detracking has few costs in terms of student performance may be too optimistic. Our study, which, used nationally representative survey data and statistical models that control for both track assignment and classroom characteristics, clearly suggests that this is not the case.
However, our analysis should not be interpreted as an argument against detracking as part of an education reform program. For example, tracking may affect such educational outcomes as selfesteem, dropout rate, and the likelihood of going to college, and our results need to be considered with this limitation in mind. In addition, "detracking" is not a monolithic strategy, the way in which detracking is carried out may well be as important as the policy itself.(26) We merely wish to highlight the fact that the conventional wisdom on which detracking policy is often based - that students in low-track classes (who are drawn disproportionately from poor families and from minority groups) are hurt by tracking while others are largely unaffected - is simply not supported by very strong evidence. Furthermore, it is worth stressing again that it is not only our own work that suggests a more complex picture, several other recent analyses have come to similar conclusions.
While our research, which we believe uses better data and improved statistical techniques than previous efforts, suggests that there may be a small overall gain in efficiency associated with ability grouping, it also raises an equity issue- tracking clearly exacerbates the achievement gap between low- and high-ability students. It would seem, then, that policy makers are left with a difficult choice. There is clearly a case for detracking on equity grounds, however, as a result, students currently in upper-track classes may suffer major losses in achievement test scores.
1. Jomills H. Braddock and Robert E. Slavin, "Why Ability Grouping Must End- Achieving Excellence and Equity in American Education" Journal of Intergroup Relations, vol. 20, 1993, pp. 51-64. 2. Anne Wheelock, Crossing the Tracks.. How Untracking, Can Save Americas Schools New York New Press, 1992). 3. Richard S. Marsh and Mary Anne Raywid, "How to Make Detracking Work," Phi Delta Kappan, December 1994, pp. 314-17. 4. Jeannie Oakes, "Can Tracking Research Inform Practice? Technical, Normative, and Political Considerations Educational Researcher, May 1992,p. and Laura mansnerus, "should Tracking Be Derailed," Education Life, special supplement, New York Times, 1 November 1992,p. 15. 5. National Center for Education Statistics, First Follow-Up: Student Component Data File Users Manual (Washington, D.C.: U.S. Department of Education, 1992). 6. In particular, these descriptive statistics are adjusted using NCES population weights that are specifically designed to enable researchers to generalize from the NELS sample to the U.S. population as a whole. More detailed descriptive statistics on the extent of tracking and detailed breakdowns by race, ethnicity, and socioeconomic status can be found in Daniel I. Rees, Laura M. Argys, and Dominic J. Brewer, "Tracking in the United States: Descriptive Evidence from NELS," Economics of education Review, in press. 7. Jeannie Oakes estimated that approximately 80% of secondary school math and science classes (as opposed to students) were tracked in the mid-1980s. See Jeannie Oakes, Multiplying Inequalities: The Effects of Race, Social Class, and Tracking on Opportunities to Learn Math and Science (Santa Monica, Calif.: RAND Corporation, 1990), p. 20. 8. Recent work in this area includes Aage B. Sorensen, The "Organizational Differentiation of Students in Schools as an Opportunity Structure," in Maureen T. Hallinan, ed., The Social Organization of Schools: New Conceptualizations of the Learning Process (New York: Plenum Press, 1987), pp. 103-29; Adam Gamoran, "The Variable Effects of High School Tracking," American Sociological Review, December 1992, pp. 812-28; and Adam Gamoran and Robert G. Mare, "Secondary School Tracking and Educational Inequality: Compensation, Reinforcement, or Neutrality?," American Journal of Sociology, March 1989, pp. 1146-83. 9. For similar data on eighth-graders, see Jomills H. Braddock and Marvin P. Dawkins, "Ability Grouping, Aspirations, and Attainments: Evidence from the National Educational Longitudinal Study of 1188," Journal of Negro Education, vol. 62, 1993, pp. 324-36. 10. In our own statistical analyses, in which we hold student ability and other factors constant, we find that socioeconomic status continues to play a major role in track placement. However, race is generally not a statistically significant predictor of track assignment when other factors are controlled. 11. Robert E. Slavin, "Achievement Effects of Ability Grouping in Secondary Schools- A Best-Evidence Synthesis," Review of Educational Research, vol. 60, 1990, pp. 471-99. 12. Ibid., p. 484. 13. Gamoran and Mare, op. cit. 14. Thomas B. Hoffer, "Middle School Ability Grouping and Student Achievement in Science and Mathematics," Educational Evaluation and Policy Analysis, Fall 1992, 1992, pp. 205-27. 15. Placement in an upper-track class was associated with an increase in ninth-grade mathematics test scores of as much as 5% over placement in a heterogeneous class; placement in a lower-track class was associated with a decrease in scores of as much as 6.4%. 16. Oakes, "Can Tracking Research Inform Practice?," pp. 15-16. 17. The technical details of the study and its finding may be found in Laura M. Argys, Daniel I. Rees, and Dominic J. Brewer, "The Impact of Ability Grouping on High School Student Achievement: Evidence from NELS," National Center for Education Statistics, U.S. Department of Education, Washington, D.C., 1995. 18. Although test scores in other subject areas are available in NELS, much of the recent work examining track assignment and the effect of tracking on student achievement has been focused on mathematics scores. In addition, it has been argued that family background is the primary determinant of verbal ability, whereas schools play a greater role in such areas as math and science. See, for example, George F. Madaus et al. "The Sensitivity of Measures of School Effectiveness," Harvard Educational Review, May 1979,pp 207-30. 19. Specifically, teachers were asked, "Which of the following best describes the achievement level of the eighth- 10th- graders in this class compared with the average eighth- 10th- grade student in the school- Higher achievement levels, average achievement levels, lower achievement levels, or widely differing achievement levels-,, Of course, one cannot assume that every teacher in the sample interpreted this question in exactly the same manner. However, this was the only method of distinguishing between heterogeneous and tracked classes, and other research has relied on similar tracking measures (see Oakes, Multiplying Inequalities-. There is also evidence that teacher perceptions with regard to the homogeneity of classes closely correspond to information provided by school administrators (see Robert E. Slavin et al., Alternatives to Ability Grouping Baltimore: Center for Research on Effective Schooling for Disadvantaged Students, 198!). Tenth-grade teachers were also asked, "Which of the following best describes the 'track' this class is considered to be? Academic, advanced or honors, general, vocational/technical/business, or other?" By combining the vocational and other responses to this latter question, two sets of track measures, each with four categories, were constructed. The results for this measure may be found in Argys, Rees, and Brewer, op. cit. 20. See, for example, Gamoran and Mare, op cit.; and Gamoran, op cit. Student perceptions of track placement have been shown to be unreliable. See James E. Rosenbaum, "Track Misperceptions and Frustrated College Plans: An Analysis of the Effects of Tracks and Track Perceptions in the National Longitudinal Survey," Sociology of Education, April 1980, pp. 74-88 Samuel R. Lucas and Adam Gamoran replicated the track assignment analysis of Gamoran and Mare's study, using information from actual school transcripts. The results changed substantially. See Samuel R. Lucas and Adam Gamoran, "Race and Track Assignment: A Reconsideration with Course-based Indicators of Track Locations," working paper, University of Wisconsin, Madison, 1993. 21. For a discussion of the conceptual and methodological problems associated with this approach, see Eric Hanushek, "Conceptual and Empirical Issues in the Estimation of Education Production Functions," Journal of Human Resources, vol. 14, 1979, pp. 351-88; and Ronald G. Ehrenberg and Dominic J. Brewer, "Do School and Teacher Characteristics Matter? Evidence from High School and Beyond," Economics of Education Review, vol. 13, 1994, pp. 1-17. A summary of "production function" research can be found in Eric Hanushek, "The Impact of Differential Expenditures on School Performance," Educational Researcher, May 1989, pp. 45-62. A recent example of this approach using NELS data can be found in Ronald G. Ehrenberg, Dan D. Goldhaber, and Dominic J. Brewer, "Do Teachers' Race, Gender, and Ethnicity Matter? Evidence from NELS:88," Industrial and Labor Relations Review, April 1995, pp. 547-61. 22. Students in the NELS data were tested in the spring of their eighth- and 10th-grade years. Unfortunately, no ninth-grade information is available. We assume, therefore, that the effect of eighth-grade inputs on the educational process are captured through the inclusion of eighth-grade achievement in our models. 23. Sorensen, op. cit.; Gamoran and Mare, op. cit.; Lucas and Gamoran, op. cit.; and Michael S. Garet and Brian Delaney, "Students, Courses, and Stratification, "Sociology of Education, April 1988, pp. 61-77. 24. These estimates are statistically significant at conventional levels. That is, we can be at least 95% certain that the differences are not a matter of chance. 25. This 1.7%, figure is marginally statistically insignificant (p less than .13). 26. For example, detracking need not mean students are taught entirely in heterogeneous classes. Rather, there could be a mix between grouped and ungrouped classes, allowing high-ability children to interact with their peers and to assist lower-ability students in cooperative learning situations.
GRAPHIC: Cartoon; Graph
IAC-NUMBER: IAC 17774615
LOAD-DATE: June 24, 1996