Student Segregation and Achievement Tracking in Year-Round Schools
by Ross E. Mitchell & Douglas E. Mitchell - 2005
Twenty-five percent of California's elementary schoolchildren attend schools operating on nontraditional, staggered, overlapping attendance calendars collectively referred to as multitrack year-round education (MT-YRE). This case study reveals substantial differences in the characteristics of students and teachers across the four attendance tracks of eight MT-YRE schools in one large California school district. Analyses of Stanford Achievement Test data, controlling for student and teacher characteristics, reveal strong association of achievement with student demographic, programmatic, and teacher segregation within these MT-YRE schools. These findings suggest that MT-YRE readily (re)segregates students within schools and thereby inhibits access to equal educational opportunity relative to traditional and nontraditional single-track school calendars.
Twenty-five percent of Californias elementary schoolchildren attend schools operating on nontraditional, staggered, overlapping attendance calendars collectively referred to as multitrack year-round education (MT-YRE). This case study reveals substantial differences in the characteristics of students and teachers across the four attendance tracks of eight MT-YRE schools in one large California school district. Analyses of Stanford Achievement Test data, controlling for student and teacher characteristics, reveal strong association of achievement with student demographic, programmatic, and teacher segregation within these MT-YRE schools. These findings suggest that MT-YRE readily (re)segregates students within schools and thereby inhibits access to equal educational opportunity relative to traditional and nontraditional single-track school calendars.
Year-round (modified-calendar) schools are an important, but largely unstudied, component of the American public school system.1 More than 4 percent of the nations 47 million public schoolchildren attend a year-round school.2 Over 60 percent of the nations year-round-school students are enrolled in the California public school system alone. One million schoolchildren (2 percent of the national total and more than 15 percent of the California total) attend a California public school operating on a modified-calendar system known as multitrack year-round education (MT-YRE).3 To gauge the scale of MT-YRE in California (the most prevalent form of year-round schooling in that state), less than one-third of the remaining 49 states have total public school enrollments as large as that in Californias MT-YRE schools.
The prevalence of MT-YRE in California is not the only reason to emphasize its study, though any system of education that affects one million schoolchildren is worthy of attention in its own right. What is more important about Californias MT-YRE schools is that they are a striking example of how a state and its local school districts may administratively respond to population growth under fiscal constraint, a response seen or actively contemplated in other states as well (e.g., Florida, Nevada, North Carolina, and Utah);4 and this response has potentially important, but generally unintended, educational consequences. In particular, MT-YRE is a system that differentiates school attendance groups with the potential for creating both social and academic segregation comparable to other curriculum-tracking practices that have received a great deal of scholarly attention in recent years.5 Also, enrollment and staffing patterns within MT-YRE schools may be subject to the dynamics of family choice, choice in a context that has no transportation costs and relatively low information and transaction costs.
MT-YRE AS ADMINISTRATIVE RESPONSE
Fiscal and political constraints on school construction in California have encouraged the widespread adoption of MT-YRE calendars because such attendance scheduling allows schools to serve more children with the same physical building space.6 This is accomplished by creating multiple, staggered attendance calendars (tracks with differing vacation schedules) such that at any given time, some fraction of the students (and their teachers) are not is session.7 The prevalence of MT-YRE in California has essentially two causes: student population growth exceeding school capacity and state policy encouraging MT-YRE implementation.
First, California has experienced an unabated influx of poorer, often immigrant families into older urban and suburban neighborhoods since the 1980s, which has increased the population densities of those neighborhoods.8 This increased population density has been a major factor in why MT-YRE has been used to accommodate overcrowding and, since 1996, to find classrooms to implement the California class size reduction initiative.9 These conditions also help to explain why Californias MT-YRE schools are more frequently low-performing schools compared to those operating on traditional or other single-track calendars.10
Second, throughout the 1990s, Californias Year-Round School Grant Program provided an incentive for districts to continue or newly adopt MT-YRE operation in order to qualify for state building funds for new school construction regardless of community demographics.11 From San Diego in the south to the Sacramento Valley in the north and in roughly half of the coastal and inland counties in between, over 1,000 schools in more than 100 urban, suburban, and rural districts operate on some form of MT-YRE calendar. And of greatest significance to the present study, at the end of the last decade, about one in every four of the states elementary school students was attending an MT-YRE school.12
HOW MT-YRE SCHOOLS ARE ORGANIZED
The most common multitrack calendar is a rotating, four-track system with roughly one-fourth of the student body not in attendance at any given time. The most prevalent rotation cycle is the 60/20 model, where students are on track for three months (60 days) and off track for one month (20 days).13 Thus, one-fourth of the students and their families are on vacation in any given month. The typical process for making track assignments involves setting a sign-up date for parents (often in the spring months of May or June) when new students and their families express their track preferences.
In addition to calendar preference, student assignment is likely to be influenced by several rules and practices governing track enrollment. Returning children are nearly always guaranteed the right to remain on their current track if they wish. Families with siblings on a specific track are typically given preferential access to placement on that track. And special-circumstance appeals are sometimes allowed (e.g., to facilitate parental visitation for children with divorced parents). Quite often schools also designate specific tracks for special programs (such as athletic teams, band or other music programs, and bilingual education programs) in order to avoid duplicating costs or to accommodate community preferences. In order to participate in these programs, students are typically assigned to specific tracks.14 Once assigned to an attendance track, students typically have little or no exposure to children in other tracks during the instructional portion of their day. As discussed more fully below, the sign-up system contributes to strikingly differentiated enrollment patterns.15
The MT-YRE calendar adds a layer of complexity to the assignment of students and staff. Three mechanisms for distribution among attendance tracks have been identified in the research literature: attendance boundary division, program differentiation, and preferential choice. First, attendance boundary division subdivides the schools neighborhood catchment area, effectively creating multiple schools within the school. This appears to occur relatively infrequently; only one case was found in the literature.16
Second, program differentiation concentrates specific types of students, personnel, and resources on particular attendance tracks. There is no adequate empirical work on exactly how programs are assigned to tracks. There are, however, anecdotes about this in the literature.17 As noted in our findings, there is some confirming evidence from this study.
Third, preferential choice separates students in accordance with family preferences for particular attendance tracks and allows teachers to seek their preferred work schedules. These opportunities for choice are structured by district policies. Family and staff choice is unique to MT-YRE schools. There are no choice opportunities in traditional-calendar schools,18 but schools operating on an MT-YRE calendar are subject to the dynamics of parental choice through sign-up queues employed to allocate children to preferred schedules. Parents exercise their choices within the neighborhood school, a circumstance with information-gathering and transportation costs much lower than those typically associated with interschool or interdistrict choice options.19 Additionally, the teacher labor market sometimes provides opportunities for staff to seek assignment to preferred tracks. As shown below, significant consequences follow these choice dynamics, which, in combination with program differentiation, yields significant student and staff segregation.
Two points are critical for understanding the potential social and educational consequences of differentially distributing students and teachers among multiple attendance tracks. First, by creating typically four distinct schools within a school, the MT-YRE calendar offers a particularly powerful mechanism for separating and, as Shields and Oberg assert, potentially ghettoizing groups within a school site.20 Even though students are enrolled at the same school site, the staggered attendance pattern changes their schoolmates every month. As a consequence, they come to see members of their attendance track as their primary classmates. Regardless of the student assignment mechanism, classroom groups generally are separated by attendance track for all instructional activities. Not only do students stay within a particular track for the entire school year, they typically remain in the same track from one year to the next. Moreover, MT-YRE is more commonly an elementary than secondary school phenomenon, and thus cohort separation begins with the first day of kindergarten.
The second point to emphasize is that separating student track groups creates opportunities for the development of significant biases in the distribution of educational resources and opportunities. The MT-YRE calendar separates teachers, as well as students, into groups by attendance track. As a consequence, teachers are not equally available to all students, other teachers, or even their site administrators. When students go off track, so do their teachers. Frequently as a result of resource limitations, teachers who work with special populations (English-language learners and special education students) or who are curriculum specialists (e.g., music, physical education, and reading) are available only on specific attendance tracks. This leads in turn to redistributing the students according to the available instructional resources. This realignment or redistribution may be the intended consequence of explicit policies or merely the unintended consequence of an effort to use resources efficiently. Regardless, significant track-to-track differences in the distribution of educational resources and opportunities are produced.
These issues of group separation and resource allocation are two major themes in the well-established literature on curriculum tracking. That research helps to inform the data analysis presented in this study by emphasizing a third themeacademic-achievement differenceswhich is seen as dependent upon group and resource segregation. The curriculum-tracking research literature draws attention to the fact that tracked programs do more to create resource and opportunity differences for students than effectively respond to preexisting student performance differences.21 Moreover, inaccurately placed students tend to stay in their initial track placements.22 For example, the lower tracks, where poor and minority students are found in higher concentration, all too often receive the least adequate teaching resources and display stagnated student achievement growth.23 This research concludes that track assignments do more to determine student outcomes than to respond to individual differences.24 Further, since academic-performance advantages tend to be aligned with social-class differences among children, schools curriculum-tracking policies abet the reproduction of social and cultural advantages for certain groups.25 Three issues from curriculum-tracking research are attended to in this study of MT-YRE attendance tracking: (a) the biased distribution of teaching talent, (b) the sorting of students by demographic and programmatic characteristics, and (c) the differential outcomes of schooling.26 As our study finds, MT-YRE schools are characterized by sharp differentiation on each of these three dimensions.
PRIOR RESEARCH ON TRACK-TO-TRACK DIFFERENCES IN MT-YRE SCHOOLS
Prior research on the character and impact of MT-YRE school policies is quite limited. The literature that does exist supports three conclusions:27
1. Attendance tracks that most resemble the traditional calendar are the most popular.
2. Student demographics differ markedly from track to track.
3. The track-to-track student achievement gap can be quite large.
As shown in Table 1, the tracks differ not only in the months during which students are on vacation, but also in family preference, student poverty levels, limited English-language proficiency and nonwhite enrollment, and overall achievement levels. The rows in the table identify the four different tracks of typical MT-YRE calendars. The columns identify inter-track differences. As to the first point, tracks most like the traditional calendar (e.g., those with summer vacation months in July or August) are the most popular and are always the first to fully enroll. Late enrollees are generally assigned to the less popular tracks with more open slots.28
Second, the tracks are demographically differentiated.29 The most popular track has more students, and these students are more often from wealthier, English-speaking, white families than the other tracks. The least popular track has fewer students, relatively more of whom are from poorer, non-English-speaking, and nonwhite families.
In many cases, demographic segmentation reflects de facto segregation resulting from parental choice and the resulting alignment of school programs with the differentiated enrollment groups. Personnel and other resource constraints may accelerate the convergence of preferences with programs, as when shortages of bilingual or special education (and sometimes music) teachers cause school officials to limit some educational programs to one or two of the MT-YRE tracks in order to control costs and allocate limited resources or services.30 However, two other mechanisms leading to demographic segmentation have been observed. First, school catchment areas have been subdivided into smaller neighborhood zones to fill tracks, reproducing the differences already known to be associated with the de facto segregation in family housing patterns.31 Second, de jure segregation has been observed in one case from the 1980s. That is, in response to the preference of Mexican agricultural laborers for extended vacations to Mexico in January, the Oxnard School District had a policy calling for school officials to actively encourage the enrollment of migrant workers children on B-Track, which is off in January, to limit absenteeism for this group.32
To restate the third conclusion, mean achievement also differs sharply across attendance tracks. The most popular tracks have the highest mean achievement, while the least popular tracks have the lowest mean achievement.33 Achievement stratification can occur as multiple strata. That is, each track can have successively higher mean achievement levels regardless of the number of tracks (three, four, or more), or there can be a high track, a low track, and the remaining tracks roughly at the same mean achievement level somewhere between the top and bottom tracks.
While the three conclusions are consistent across the case studies, we note that none of these studies extensively explored the track segregation patterns for systematic covariation among demographic and achievement variables. In particular, there were no attempts to simultaneously consider all or even some of the factors identified in Table 1 when accounting for track-to-track achievement differences. Further, neither the contributions of unequally distributed teaching talent nor the dynamics of choice were identified in previous studies. In what follows, we report on teacher, as well as student, segmentation across MT-YRE tracks, in a context where track assignment preferences are a contributing factor, and more extensively investigate the relationship between student achievement and segregation within MT-YRE schools.
A STUDY OF MT-YRE SCHOOLS IN ONE LARGE CALIFORNIA DISTRICT
Our data allow us to examine MT-YRE academic and social segregation as it has developed in one large California school district. Extensive and detailed demographic and achievement data on 12,174 traditional- and MT-YRE-calendar elementary school students in grades 2 through 6, including professional background information about their teachers, were compiled for statistical analysis (see Appendix A for details). At the time the data were collected (after the close of the 199798 school year), eight (30 percent) of the districts elementary schools operated on an MT-YRE calendar, enrolling roughly 37 percent of the districts elementary students.34 This was, in part, to comply with the requirements of the Year-Round School Grant Program.35 In a personal communication, one district superintendent noted that the fiscal incentives offered by this state grant program were compelling.36 By adopting MT-YRE, the district received higher priority for state school building funds (and MT-YRE grant funding), which made MT-YRE a more attractive option for responding to enrollment growth than double sessions, leasing or purchasing relocatable classrooms, or seeking a school construction initiative on the local ballot.
The number of elementary students assigned to MT-YRE schools in this district rose sharply in 1996, from 28 to 37 percent, to accommodate first-year implementation of Californias class size reduction (CSR) initiative: Two additional elementary schools adopted MT-YRE calendars. The average total school enrollment across all elementary schools in the district in 199798 was 735; across MT-YRE schools, average enrollment was 913.37 A descriptive statistical profile of the districts 12,0001 elementary school students (grades 26), including track-to-track differences for the 4,0001 students in MT-YRE schools, is presented in Appendix Table A1. Information from that table is described in the following sections.
The elementary school student population is ethnically diverse. There is a plurality of white students (43.7 percent), followed closely by Hispanics (41.5 percent). A much smaller proportion of the enrollment is black (9.7 percent), with the remaining 5.1 percent, largely but not exclusively Asian, classified as other. The poverty (National School Lunch Program [NSLP] or free/reduced price lunch qualification) rate is 50.5 percent. English is the predominant home language (75.3 percent), followed by Spanish (21.8 percent), with the remainder classified as other. The proportion of the students classified as limited English proficient (LEP) is 17.7 percent. Another 6.8 percent are classified as fluent English proficient (FEP), with the remainder being English only. There are a bit more Hispanic and other, LEP, and Spanish- and other-home-language students in MT-YRE than in traditional-calendar schools, but somewhat fewer poor students.
Gender and grade are fairly evenly distributed. The second- and third-grade samples are slightly larger than those in the higher grades. There are two types of special education identifiers: gifted and talented (GATE) and special education. The GATE-identified proportion of the sample is 9.8 percent. The special-education-identified students are divided into two subgroups: resource specialist program (RSP) for low-achieving students (3.2 percent of the sample) and designated instructional services (DIS) students with other handicapping conditions (2.7 percent). About one student in six (17.1 percent) was new to the district in 199798. The proportion of boys is higher in MT-YRE schools, compared to traditional-calendar schools, but mobility and the proportion of GATE students is lower.
The average achievement levels in mathematics and reading on the spring 1998 statewide administration of the Stanford Achievement Test, Ninth Edition (SAT-9), were recorded in the normal-curve-equivalent (NCE) metric.38 The NCE scale permits the simultaneous comparison of students across grade levels on a common metric, namely, performance relative to a nationally representative sample of students taking the same tests. Additionally, the NCE scale corresponds to national percentile rank scores at 1, 50, and 99, which helps to give some intuitive sense of how well a student or group of students is performing on a given test level.39 For example, the district-wide average achievement levels for mathematics and reading (45.48 NCE points and 44.07 NCE points, respectively) for this California school districts elementary school students are a little below the normed national mean of 50. Mean achievement in mathematics and reading is about 1 NCE point lower in MT-YRE schools than in traditional-calendar schools.
Nearly 20 percent of the students in the district have teachers on probationary contracts, while 67 percent have tenured teachers, with the remainder having the typically underqualified Other contracts. About one-sixth of the students have a teacher who holds a bachelors degree, while about four of six have a teacher with a bachelors degree plus 30 hours, and the remaining one in six has a teacher who holds a masters or higher degree. More than 90 percent of the students have fully credentialed teachers, but slightly more than 11 percent have teachers who hold some type of alternative credential. Across all students, teachers average 7.3 years of teaching experience. Because of the presence in this district of a substantial number of very highly experienced teachers, this mean experience value is misleading, however. A better estimate of average teaching experience would be the median experience level, which is 3 years of experience. Teachers in MT-YRE schools, on average, have less experience, are less likely to have full credentials, are more likely to have alternative credentials, and are less likely to have postbaccalaureate degrees, though more likely to have tenure, than those in traditional-calendar schools. With this overview of the districts elementary schools in mind, we turn to the examination of our central research questions.
INVESTIGATING ACADEMIC AND SOCIAL SEGREGATION IN MT-YRE SCHOOLS
Our data analysis documents seven key findings related to intertrack differences in the MT-YRE schools. Rather than separate a description of what was learned from explanations of how the data were analyzed, we describe the basis for each finding along with presentation of the finding itself (see Appendix B for additional details regarding the statistical methods employed).
INTERTRACK ACHIEVEMENT GAPS ARE LARGE
Based on the differences between track-level means for mathematics and reading achievement, MT-YRE attendance tracks are academically segregated to such an extent that children in the lowest-achieving track (B) are academically about 1.5 years behind their peers in the highest-achieving track (C).40 C-Tracks mean reading score is 50.78, fully 15.70 points above that for B-Track. C-Track also outperforms A- and D-Tracks by 7.23 and 6.21 NCEs, respectively.41 The mathematics story is similar. The A- and D-Track difference of 1.02 NCE points is not significant, but both tracks are significantly above B-Track. C-Track has the very highest math achievement at 52.91 NCEs, 16.30 points above B-Track, 10.68 above A-Track, and 8.17 above D-Track.
Another way to observe these dramatic track-to-track differences is to examine the full distribution of achievement at the track level rather than just the average achievement. This can be done by plotting shift functions, so named because they reveal how much one achievement distribution is shifted above or below another across the entire measured range.42 In the present case, the achievement distribution of the traditional-calendar schools is used as the reference function (comparing its NCE scores at each 5th percentile or vigesile of the score distribution, with the scores at each 5th percentile of the four MT-YRE track score distributions). The achievement differences (positive or negative) between the traditional-calendar group and the achievement levels of the four MT-YRE tracks determine their respective shift function valuesthese values are shown in Figure 1 for both mathematics and reading achievement.
As illustrated in Figure 1, track-to-track achievement differences are quite striking, particularly at the center and high-performance end for mathematics achievement and at the low end and entire upper half for reading achievement. Either by examination of the mean achievement levels for each track, as shown in Appendix Table A1, or by the shift functions in Figure 1, it is possible to see that the within-MT-YRE-school academic segregation by track has three strata: (a) C-Track consistently has the highest performance across the entire achievement distribution in both mathematics and reading; (b) in the middle, A- and D-Tracks have nearly identical achievement distributions that are similar to, though slightly lower than, the achievement distribution in the traditional-calendar schools; and (c) B-Track consistently has the lowest performance across the entire achievement distribution.
INTERTRACK STUDENT DEMOGRAPHIC SEGREGATION IS ALSO SUBSTANTIAL
In addition to substantial academic segregation, the MT-YRE tracking system exhibits very substantial demographic segregation. Children in the lowest-achieving (B) track are almost 2.5 times as likely to be poor as those in the highest track (C). They are more than 5.5 times as likely to be from a non-English-speaking home and almost twice as likely to be members of a nonwhite ethnic group (see Table 2). In addition to the three student differences just noted, there is a large gap in the proportion of students identified for GATE between the B- and C-Tracks, as well as a notable difference in the student mobility rate. Program differentiation is almost certainly an important contributor to this segregation.
Though some of the demographic stratification (i.e., white vs. nonwhite and English vs. non-English) singularly distinguishes B-Track from all of the other tracks by about the same amount, more typically the differences reproduce the three strata found in the achievement segregation described above. B-Track is lowest, C-Track is highest, and A- and D-Tracks have similar intermediate values. C-Track is most sharply distinguished from all other tracks when it comes to GATE student enrollment and year-to-year student mobility. On these two variables A-Track is significantly more disadvantaged than D-Track . In the case of student mobility, A-Track has a rate even higher than B-Track. We should also note that home-language differences involve more than an English vs. non-English linguistic separation. The proportion of students from homes where other languages (predominantly Asian languages) are spoken is highest for A-Track, which also has the highest proportion of students with other ethnicity (see Appendix Table A1). Clearly, the demographic segmentation across MT-YRE attendance tracks observed here is more complex than the initially obvious polar separation of B- and C-Tracks with A- and D-Tracks occupying indistinguishable middle positions. Nevertheless, the demographic segregation found among the four MT-YRE tracks is remarkably similar to the academic segmentation discussed above. As documented in Achievement Differences Are Closely Linked to Demographic Segregation, these demographic differences arising from student enrollment account for a very substantial part of the intertrack achievement differences.
TRACKS EXHIBIT A CORRELATED SEGREGATION AMONG TEACHERS
Intertrack segregation is not limited to student achievement and demographics. In MT-YRE schools, students are sharply differentiated in their access to experienced and credentialed teachers. On average, students in the track with the lowest-achieving students (B-Track) have teachers with four fewer years of teaching experience and are almost four times more likely to have teachers with alternative credentials than the far more fully resourced C-Track students (see Table 3). The C-Track also has the highest percentage of students whose teachers have tenure, a full credential, and postbaccalaureate degrees, with the B-Track lowest on these measures of teacher qualification as well (actually, D-Track has a slightly lower percentage of students with teachers holding a full credential). As with demographic segregation, A- and D-Tracks have some similarities (median teacher experience and teacher education level) and some differences (teacher credential and contract status), such that the middle two tracks are distinguishable, and not always in the middle relative to the B- and C-Tracks.
ACHIEVEMENT DIFFERENCES ARE CLOSELY LINKED TO DEMOGRAPHIC SEGREGATION
Student demographic segregation accounts for a very substantial amount of the intertrack achievement differences observed in this school district. That is, when a linear regression is used to predict mean student achievement by track using the demographic characteristics of each student, much of the variation across MT-YRE attendance tracks is accounted for. Column I in Table 4 (Uncontrolled) reports the MT-YRE track means as they are found in the school testing data. Column II (Student factors) shows how well intertrack differences in student achievement are explained by demographic differences among the student groups enrolled in each track.
As shown in the row of Table 4 labeled Proportional reduction of variance, using all of the demographic-segregation data reported in Appendix Table A1, the effects of student differences are removed from the estimated track means, thereby reducing the variance in these means by an impressive 89.8 percent in mathematics and an even more potent 93.6 percent in reading. Thus, it is safe to conclude that just about nine-tenths of the intertrack achievement differences in reading and mathematics result from the fact that the tracks are serving demographically distinct groups of students.
ACHIEVEMENT DIFFERENCES ALSO ALIGN WITH TEACHER DIFFERENCES
Teacher segregation also contributes to the emergence of intertrack achievement differences. As seen in Table 4, column III (Teacher factors), when the teacher experience, education, contract, and credential variables are substituted for the student demographic variables in a linear-regression model, a more modest, but nevertheless highly reliable, proportion of the intertrack achievement differences is accounted for. This statistical analysis procedure answers the question How well are the intertrack differences in student achievement predicted by differences among the teachers to which the students are assigned? Though the relationship is not nearly as strong as for student demographics, 18.4 percent of the variance in track-level mean mathematics achievement is accounted for by the collection of teacher variables, and 12.1 percent of the variance in track-level mean reading achievement is accounted for by the same teacher factors.
IN COMBINATION, TEACHER AND STUDENT SEGREGATION ACCOUNT FOR ABOUT 95 PERCENT OF ALL ACHIEVEMENT DIFFERENCES
To answer the question How well are the intertrack differences in student achievement explained by a combination of student and teacher differences? a fourth regression model includes all variables for both groups. The results of this regression model are shown in column IV of Table 4, labeled Teachers and students. Taken together, the stratification of both students and teachers accounts for 94.2 percent of the intertrack achievement differences in mathematics and 95.9 percent of the intertrack differences in reading achievement.
Another way to see the dramatic impact of student and teacher stratification on intertrack achievement differences is to look at the magnitude of the difference between the highest-achieving track (C-Track) and the lowest-achieving track (B-Track). Without consideration of the potential impact of student and teacher segregation, C-Track has a mean mathematics achievement score that is 16.31 NCE points greater than that of B-Track (this is the equivalent of about 1.5 years of normal achievement growth). When both student and teacher factors are included in the linear-regression model, however, the remaining difference between these two tracks is only 3.33 NCE points (the equivalent of only about three months of ordinary achievement growth). For reading, the C-Track mean begins at 15.70 NCE points above the B-Track mean. This difference is reduced to 3.20 NCE points after differences among students and teachers assigned to the four different MT-YRE attendance tracks are accounted for.
ACHIEVEMENT DIFFERENCES BECOME LARGER WITH EXTENDED EXPOSURE TO MT-YRE
Our last finding is the result of exploring how the patterns of student, teacher, and achievement segregation might have been reinforced by MT-YRE school attendance tracking. To conduct this exploration, we turn to one additional variable: the number of years each student has been enrolled in an MT-YRE school. Though its interpretation is fairly subtle, the hypothesis to be tested is straightforward. Put simply, by testing whether students with longer MT-YRE exposure have achievement test scores that contribute more than those of their less exposed peers to intertrack segregation (after controlling for student and teacher demographics, of course), it is possible to determine whether the intertrack segregation is a dynamic and cumulative process, rather than a one-time effect created by initial student and teacher track assignments. Though it would have been more convincing to use multiyear learning trajectory data and track-to-track migration patterns for this analysis, we have data from a single year and thus can make only a post hoc inference regarding the dynamics of MT-YRE participation effects.43
The test of interest is performed by conducting a Track by Exposure analysis of covariance (using the student and teacher demographic variables as covariates). If students with longer exposure to MT-YRE play a dominant role in creating intertrack achievement differences, their contributions will show up as a significant Track by Exposure interaction effect, indicating that continued exposure changes the nature of intertrack differences. Once we find this significant interaction effect, examination of the mean scores for each Track by Exposure group will reveal that continued exposure reinforces rather than ameliorates track differences.
Before looking at the statistical output, we should note that any differences found in this way could be the result of either or both of two quite different causes: (a) track-to-track enrollment mobility might exacerbate student body segmentation by having higher-achieving students congregate in the high-achieving track (and low-achieving students move to congregate in the low-achieving track) or (b) educational programs on the different tracks might be differentially effective, raising (or lowering) the relative achievement of students with continued exposure. If track by exposure achievement differences are the result of mobility, family choices would be responsible for segregating students; if they are the result of educational-program differences, then unequal opportunities to learn are producing segregated achievement groups. Without longitudinal achievement data we cannot distinguish these two explanations. The results reported here establish only that existing track-to-track differences in student achievement are linked to student longevity in MT-YRE schools for a particular track. That is, this statistical test for a significant interaction between years in an MT-YRE school and the attendance track on which the student is presently enrolled establishes that achievement differences across MT-YRE tracks are compounded over time.
Table 5 reports the results of the Track by Exposure analysis of covariance. The table shows the relationship between intertrack achievement differences and the number of years a student has participated in an MT-YRE school. Track-to-track differences can be read down the columns, year-to-year differences across the rows. The first thing to note about this analysis is that there is no systematic relationship between achievement and the number of years a student has attended an MT-YRE school. The initially significant intertrack achievement differences not only remain, they generally grow larger as students have more exposure.
The important finding here is that the Exposure by Track interaction is significant. The magnitude of intertrack achievement differences changes as students attend MT-YRE schools for longer periods, with the result that in general, longevity produces increasing differentiation among the track scores. Thus, it is appropriate to conclude that achievement changes over time, as children continue for longer periods in MT-YRE schools, and that the magnitude and direction of the changes depend significantly on which of the four YRE tracks they are enrolled in. The overall character of this significant interaction effect can be seen most easily in the bar graph plot of the marginal means presented in Figure 2. Note that the initially higher means for A-, C- and D-Track students can be seen along the left side of the graph. Among students in their second MT-YRE year, A-Track drops below B-Track in both mathematics and reading, though the other tracks continue to outperform the B-Track students. By the third year, B-Track has again become the lowest-performing group, while C-Track greatly extends its margin of superiority in mathematics.
Two key points are underscored by these bar graphs. First, the C-Track has a noteworthy advantage. First-year C-Track students are somewhat ahead of their peers, while students with three years of MT-YRE experience on the C-Track have a substantially larger lead over their peers in other tracks in both mathematics and reading achievement. Additionally, across all four tracks, the longer students are in enrolled in MT-YRE schools, the greater the divergence among their current MT-YRE track means. Thus, we can safely conclude that the dynamics of MT-YRE tracking are such that initial differences created largely by teacher allocation and student demographic segmentation become exacerbated as children remain in these settings.
It is not clear whether these profound intertrack differences should be attributed to instructional-program differences or to migration of students and teachers in ways that concentrate resources and opportunities in the C-Track. While this issue needs to be studied with better data than we now have, we suspect that initial track differences become exacerbated primarily by the dynamics of student and teacher intertrack mobility.44 Nonetheless, since extended exposure to the MT-YRE tracking system is associated with greater intertrack achievement and demographic differences, it must be the case that either (a) families and teachers recognize track-to-track differences and work to relocate themselves in ways that increasingly segregate track membership or (b) the unbalanced resources available to the different tracks significantly affect childrens learning opportunities. As the curriculum-tracking literature has amply demonstrated, the kind of demographic and academic segregation found in these multitrack schools is almost certain to have a cumulative and continuing negative effect on the long-term educational success of some of the schools most vulnerable students.
SUMMARY OF THE FINDINGS
Multitrack YRE is associated with substantial social and academic segregation of both students and teachers.45 First, we note that taken as a whole, MT-YRE schools differ from traditional-calendar schools. MT-YRE schools have somewhat lower achievement, a bit more challenging student populations, and slightly less adequate teaching resources than traditional- calendar schools.46 These differences, though not profound, were observed to be statistically significant in the present case.
Second, and more importantly, there is a very substantial segregation of students and teachers among the four attendance tracks within MT-YRE schoolsdifferences not well studied in previous research. Data reviewed here show that MT-YRE school attendance tracks differ sharply in student composition and academic achievement. Segmentation in this year-round school population is initially substantial and, over the three years, appears to expand intertrack achievement differences. The C-Track, with its vacation schedule most like that of the traditional calendar and most popular with parents and students who actively choose tracks, is the highest-achieving track and solidifies its advantage for students with extended enrollment. Over time, the D-Track, with academic performance in the midrange among the attendance tracks, loses some of its initial advantage. The B-Track, which is least like the traditional school in both population and attendance schedule (and typically houses bilingual-education programs) starts out behind and gets further behind as student enrollment continues. Ninety-five percent of the intertrack differences in 199798 are accounted for by demographic and programmatic segregation of students in combination with unequal access to highly qualified teachers.47
HOW DOES SUCH ACADEMIC DISPARITY ARISE?
Data from this study demonstrate that MT-YRE calendar tracking tends to take on the very features of curriculum tracking that have been the focus of so much recent analysis and criticism. When students attend classrooms tracked by calendar, they wind up in groups also characterized by segmented demographics and program services, with lower-performing students more likely to be in classrooms with less fully qualified or less experienced teachers. Children are not typically assigned to tracks in response to their performance, but through the exercise of preferences (or constraints thereon), leading to differentiated learning opportunities as a consequence of MT-YRE track selection.
The demographic segmentation of student and teacher groups appears to be sufficiently powerful that we do not need to look to differences in instructional practice in order to account for intertrack achievement differences. Track groups are as differentiated by social status as by school services.48 This is not to say that instructional practices may not differ radically across tracks, but that student and teacher segregation accounts for track-to-track differences in achievement about as well as any other explanation that might be offered. An active and powerful sorting system is operating within the MT-YRE schools of this California district.
One important consequence to highlight is this: In cases like the one studied here, where districts have desegregation policies (or are under court order to desegregate), we are likely to see significant social resegregation at the site level. To use Bourdieus language, the most culturally privileged groups appear to be finding their way into tracks capable of reinforcing their advantage. In all likelihood, they do so by pyramiding their collective social capital to join preferred tracks and facilitate the accumulation of educational advantage.49
Additional research is needed, however. It is not clear whether intertrack achievement differences should be viewed as entirely the consequence of the sociopolitical process of student and teacher assignment or as involving significant educational factors as well. It is possible that initial assignment differences create inequalities in educational effectiveness that snowball into substantial achievement differences.50 It is equally likely, however, that initial differences are compounded by parent and teacher awareness of track differentials that lead them to exercise their choice options in ways that further exacerbate the initial segmentation. While the data available for this study cannot distinguish between these possibilities, data monitoring intertrack movement among students and teachers would show whether the large achievement differences found here are created by student migration rather than instructional effectiveness differences. We plan just such a study in the near future.
DIFFERENT BY DESIGN
In sum, our study is consistent with earlier case studies finding that the modest differences in educational opportunity initially created by the establishment of multitrack year-round calendars work to produce very substantial differences in the distribution of students, teachers, and programs among the different attendance tracks. Selection of tracks by families and teachers and the accompanying alignment of programs and services in response to these choices account for nearly all of the large academic-achievement disparities observed among the four MT-YRE attendance tracks. Beginning with the earliest elementary school years, enrollment in particular attendance tracks becomes the gateway for access to high-achieving classmates, experienced and qualified teachers, and enriched curricular opportunities. Before children in kindergarten have a chance to blossom, before the schools provide the opportunity for children to learn to read, write, or calculate, they are segregated and tracked within their neighborhood MT-YRE schools. Enrollment opportunities are distinct administrative designs that structure both choice opportunities and resource allocationsand the consequences are substantial.
Family and staff choice play the dominant role in this process. These choices, when exercised in the MT-YRE environment, appear to have roughly the same effect that they have in the housing market: segregating advantaged and disadvantaged groups and creating a system that separates strong, high-performance schools (or attendance tracks) from weak and low-performing ones. If one primary purpose for establishing a free, mass, compulsory public education systemsupported by the taxing authority of the state to provide resources and the police power of the state to compel participationis the creation of more equitable life chances for all children,51 MT-YRE programs like those found in our sample have to be viewed as a threat to that goal. In recent years, education policy has been expanding choice on the grounds that it will induce competition for excellence among the public schools; we see nothing in the data reviewed here to support this proposition. Instead we see the competitive process being used to differentiate and concentrate educational quality without raising overall achievement in any measurable way.
DETAILED DESCRIPTION OF DATA
Student achievement data for this study are drawn from Californias state-mandated achievement test administered in the spring of 1998 to students in grades 2 through 6 (Stanford Achievement Test, Ninth Edition, Form T). The reading comprehension and mathematics total battery NCE scores are used throughout this analysis.52 For each student, the data set also includes gender, ethnicity, home language, grade, NSLP participation, English-language proficiency, identification for special education or gifted education services, and interdistrict mobility between annual test administrations. These variables are well known to be associated with differences in student academic achievement.53 The NSLP variable serves as a poverty indicator.54 The interdistrict-mobility variable identifies new or transient students. Student English-language proficiency is coded as limited English proficient, fluent English proficient, or English only. Special education services are coded as not identified for special education services, identified for the resource specialist program (RSP), or identified for designated instruction services (DIS).55 For the purpose of analysis, and reflecting the student population in the district, home language is coded as English, Spanish, or Other. Similarly, student ethnicity is coded as white, Hispanic, black, or Other.
The students school and classroom assignment data make it possible to identify attendance track and teacher. The four MT-YRE tracks are labeled A through D. The year-round schools cycle on a fiscal calendar (July through June). The tracks are off in reverse alphabetical order when the school year begins in July. D-Track has the first summer vacation month in July, C-Track in August, B-Track in September, and A-Track in October (A-Tracks third vacation month comes in June each year). Thus, C-Track is closest to the traditional schedule, and B-Track is least like the traditional schedule, with many families perceiving D-Track, which has the traditional summer vacation month of July, as more like the traditional schedule than A-Track.
Student and teacher track assignments in MT-YRE schools were obtained for three consecutive schools years: 199596 through 199798. Unfortunately, fully comparable student achievement data across all three years were not available. As such, the MT-YRE attendance trajectories of students could be determined, but not their achievement trajectories in both mathematics and reading. However, for the purpose of comparing the relative achievement ranking of each MT-YRE attendance track, mean mathematics and reading achievement levels were calculated (statistics not reported here).
Teacher data from the California Basic Education Data System (CBEDS) Professional Assignment Information File (PAIF) were linked to the student-level data file through the school, grade, and teacher name fields reported in both files. The variables taken from the CBEDS PAIF are (a) total years of teaching experience, (b) number of years of teaching experience within the district, (c) education level, (d) credential status, and (e) contract status. Education level is coded here as a bachelors degree (BA), bachelors degree with 30 or more semester hours of advanced postsecondary education (130), or at least a masters degree (MA or higher). Two dichotomous credential status variables are used: the teacher has a full credential or not, and the teacher holds an alternative credential or not.56 In addition, the teachers contractual status in the district is coded in three categories: Tenured (beginning with the third full contract year using a Preliminary or Clear credential), Probationary (two years or less experience or when using a temporary credential while eligible for regular contract status), and Other (a very small group with typically little or no experience and not qualified for a probationary or tenured contract).
About 10 percent of the sample is excluded as a result of unavailability of either data from the student records or CBEDS teacher data. After eliminating cases with missing data, the total sample size dropped to 12,174 students. Teaching experience in the district is highly correlated with total years of teaching experience. Thus, the years-of-teaching-in-the-district variable was redundant and dropped from further analyses.57 As previously noted, a track-by-track breakdown, along with totals for the MT-YRE schools, traditional-calendar schools, and the sample as a whole, for all of the variables in this study are shown in Appendix Table A1.58
DETAILED DISCUSSION OF STATISTICAL METHODS
The test used for whether or not any particular factor significantly accounts for the variance (Z2) in student mathematics or reading achievement is the analysis of variance (ANOVA) F-test. Multiple pairwise comparisons of MT-YRE attendance track group means (or proportions for categorical variables, e.g., race/ethnicity, home language, special education services, teacher education level) are tested using Bonferroni-adjusted significance levels.
As shown in Wilcox (see note 42), shift functions are calculated by first establishing the score distribution for a reference group (here, all students attending traditional-calendar schools). The score at each decile (the value at each tenth percentile of the achievement distribution) of the reference distribution is then subtracted from the score at each decile of the treatment groups (here, the four groups are the students attending each of the four MT-YRE attendance tracks), leaving the residuals differences for each MT-YRE track to be plotted relative to the reference group (the values at the endpoints of the distributions, i.e., 0 and 100 percent, are not included in the plots); the reference group decile scores are also subtracted from the reference group deciles, thus setting the values for the reference group to zero across the full range of the distribution. However, rather than using deciles to construct our shift functions, we use vigesiles (the value at each fifth percentile of the achievement distribution, i.e., at the 5th, 10th, 15th, . . . , 95th percentiles). By doing so, we obtain greater continuity and resolution, which is justified since our large sample sizes for each track permit reliable estimates for each 5th percentile.
Multivariate estimates of marginal mean differences among the MT-YRE tracks are computed using linear-regression coefficients. The mean achievement for students assigned to each of the four tracks is estimated using the unstandardized regression coefficients estimated for dummy-coded variables for tracks A, B, and C, with D as the reference group. However, because B-Track has the lowest mean achievement level, which makes it the best reference group for discussion purposes, the value of its regression coefficient is subtracted from the value of each of the other three (A, C, and D). This procedure sets the value of the B-Track coefficient at zero and causes the D-Track coefficient to have a nonzero value.
Statistical controls are used to adjust achievement scores for differences in student demographics, program assignments, and teacher qualifications across tracks. Models are run separately for total mathematics and for reading comprehension achievement subtest NCE scores. Prior to controlling for the effects of both student and teacher characteristics, analyses are undertaken to determine how well track mean achievement can be predicted by student characteristics and by teacher qualifications separately.
To ascertain whether MT-YRE tracking has a dynamic and continuing impact on family choices and student assignments, we test for a significant interaction between MT-YRE track assignment and the number of years a child attended MT-YRE classes. In other words, in addition to all of the student and teacher variables included above, a variable for the number of years a child attended MT-YRE classes and the interaction terms between this variable and the track to which the student was assigned are included in the linear-regression models previously specified.
This work was supported by the California Educational Research Cooperative, Graduate School of Education, University of California, Riverside. The authors would like to acknowledge Robert Hanneman, Steven Brint, Jon Lorence, Jane Hannaway, Barbara Schneider, Amy Stuart Wells, and the Teachers College Record editor, as well as anonymous reviewers, for constructive comments on previous drafts. Full responsibility for the content rests solely with the authors. An earlier version of this article, submitted under the title Organizational Segregation of Student Achievement within Elementary Schools: The Influence of Multi-track Year- Round Schools, was presented at the 94th Annual Meeting of the American Sociological Association, submitted under the title Organizational Segregation of Student Achievement within Elementary Schools: The Influence of Multi-track Year-Round Schools.
ROSS E. MITCHELL is a research scientist in the Gallaudet Research Institute at Gallaudet University. His research interests include education policy analysis and evaluation, sociology of education, demography, and deafness. His recent publications include The Political Economy of Education Policy: The Case of Class Size Reduction and Demographic and Achievement Characteristics of Deaf and Hard of Hearing Students.
DOUGLAS E. MITCHELL is Professor of Education in the Graduate School of Education, University of California, Riverside. His research interests include education policy formation and implementation, organization and control of school systems, labor relations and teacher incentives, citizen influence, and school politics. His forthcoming book is titled The New Foundations of Educational Administration, Policy and Politics: Science and Sensationalism (Erlbaum).