Abstract: Inter- and intra-district public school choice, vouchers, tuition tax credits and other forms of school choice have been advocated for decades, in large part on grounds that the market forces engendered will improve public education. There are many studies of school choice policies and programs and a large theoretical literature on school choice, but thus far no studies have used a large national sample and common metric to perform a multi-level, multi-district analysis of relationships between school choice policy and student achievement. This study links a national sample of NAEP student achievement data, with district level information on magnet-based school choice policy, and with demographic data from the U.S. Census. Using three-level hierarchical linear modeling we find substantial effects of school and district demographic variables on student achievement, but, after adjusting for multi-level demographic characteristics, we find only small differences in student achievement in school districts with magnet schools and school choice policies as compared with districts with attendance area based student assignment, and no magnet schools. NAEP achievement scores are marginally lower in the sample of students within districts reporting magnet schools and associated school choice policies.
School choice in a variety of forms has expanded nationally and internationally, in large part on the belief that market forces can improve public education (Peterson 2001; Teelken, 1999; Van Zanten, 1996; Whitty, Power, & Halpin, 1998). In the United States, magnet schools were the first school choice policy widely implemented in public education (Blank, Dentler, Baltzell & Chabotar, 1983; Levine & Havighurst, 1977). Researchers have studied magnet schools to learn whether this approach to school choice affects achievement, but in almost all cases studies have focused on just a single or a few districts. This study merges existing national data sets to perform a multi-level, multi-district analysis. It charts new ground by linking student achievement data from the National Assessment of Education Progress (NAEP) with district level policy information on school choice and demographic data from the U.S. Census.
Of the many different forms of school choice in existence (or proposed), magnet schools implement a regulated form of school choice--neither private schools nor charter schools are involved, and parents' choices may be restricted in the interests of racial desegregation of schools. Nonetheless, the presence of magnet schools in a district still produces a great deal of parental choice, raising questions about whether school districts with magnet-based school choice might have higher achievement than districts lacking school choice. This is the question we examine.
Theory and Research on Magnet Schools, School Choice, and Student Achievement
The Growth of Magnet Schools and School Choice
While charter schools and voucher programs have been getting more attention recently, magnet schools remain one of the most common approaches in the U.S. (CER, 1998; Steel & Levine, 1994). The "magnet school boom" (Warren, 1978) of the late 70s and continuing into the 80s was the precursor of the widespread adoption of public school choice policies we see today. Since 1976 when only 14 districts had magnet schools (Blank, Dentler, Baltzell & Chabotar, 1983), the number of districts with magnet schools has grown rapidly along with enrollment in magnet schools (Blank, 1989). By 1990 estimates were that almost half of urban districts had magnet schools (Steel & Levine, 1994).
When a district implements a magnet-based school choice policy, it designates a portion of the schools in the district as magnet schools. Magnet schools are open to enrollment by choice on a district-wide basis, subject to racial balance guidelines (sometimes regions within districts are used if the district is geographically large). Magnet schools typically have special features, programs, or instructional methods that distinguish them from "traditional" schools. Magnet schools may offer distinctive instructional approaches such as Montessori or Waldorf; curricular emphases such as art, math, science, technology, or foreign language immersion; or disciplinary climates involving above average regimentation and required uniforms. (1) In addition, magnet schools sometimes offer special features like smaller class sizes, instructional aides, or newer facilities, features made possible by federal funds from the Magnet Schools Assistance Program that has supported magnet schools' racial integration function since the mid-70s. Districts with magnet schools often extend choices to cover other schools in the district as well. Parents may choose and enroll a child in any school as long as the choice does not increase racial desegregation in the district. A child in a school where his/her race is in the majority can transfer to any school where his/her race is in the minority.
Theory and Research on Magnet-based School Choice
A large academic and practitioner-oriented literature supported the growth of magnet schools and public school choice. Magnet-based school choice was supported because it helped voluntary racial integration (e.g., Greeley, 1987; Levine & Havighurst, 1977; Rossell, 1990), expanded educational alternatives for parents, and stimulated competition in public schools (Barr, 1982; Blank, 1989; Clinchy & Cody, 1978; Finn, 1985; Finn, 1987; Kolderie, 1985; Nathan, 1989; Manley-Casimir, 1982; Raywid, 1985; Toch, Linnon, & Cooper, 1991, May 27).
Researchers and policymakers have long been interested in achievement effects associated with magnet schools (e.g., Archbald, 1995; Blank, 1989; Blank & Archbald, 1992; Gamoran, 1996; Goldhaber, 1999; Larson, Witte, Staib, & Powell, 1990). Two inter-related arguments have supported the growth of magnet schools, both suggesting achievement benefits. One argument is that the expansion of educational alternatives improves the fit between a school district's educational program and its clientele's preferences. Michaelson (1981) has referred to this as "allocative efficiency." It reflects the assumption that not all parents want the same thing, and so rather than force all parents to accept the same kind of school program, school systems should have differentiation among schools reflecting different educational philosophies, pedagogies, and curriculum specializations. In theory, magnet schools do this. While they may not be for everyone, they make it possible for a school district to accommodate more parents with varied preferences than would otherwise be possible with a "one size fits all" philosophy. This viewpoint is prominent in the literature supporting magnet schools (e.g., Fliegel, 1993; Clewell & Joy, 1993; Nathan, 1989; Raywid, 1985).
The second argument reflects free market principles. Advocates of magnet schools have argued that they foster healthy competition in school systems. Schools that have long been assured a designated attendance area clientele, are in a more competitive position when a school system implements magnet-based school choice policies. Children from the neighborhood are no longer guaranteed to a school. When the market is opened up, as the theory goes, schools will try harder to satisfy parents. Observers disagree whether "trying harder" necessarily results in improving curriculum, instruction, and school discipline in ways that raise academic achievement. Still, these two inter-related arguments have been very influential in expanding the adoption of public school choice, and still propel the growth of charter schools and voucher programs.
Ideally, one would want to compare achievement in a sample of districts with magnet-based school choice over time to a matched sample of districts without these policies, but this has not been done for a variety of reasons, not least of which are the methodological difficulties such a study would pose. One research method compares magnet students' achievement to the achievement of other students in the district not in magnet schools. These comparisons typically find magnet school students have higher test scores. For instance, Archbald (1995) compared achievement scores of elementary magnet students, neighborhood school students, and students who had "choiced out" of a neighborhood school, but not to a magnet. The results showed a statistically significant effect of magnet school enrollment, although the effect was fairly small (about 3 NCE points on average). Larson et al. (1990) compared high school magnet students to a sample of nonmagnet students on mathematics and science tests. Since it was known from unadjusted cross-sectional comparisons that the magnet students had higher test scores (on average) and that several of the magnet programs were selective in their admissions, (3) the comparison group was randomly sampled from college-prep courses in the nonmagnet high schools. Their analyses found modestly, statistically significantly, higher mean test scores in the magnet programs, even when adjusting for prior eighth grade achievement.
An important methodological limitation of the above magnet-nonmagnet comparisons is selection bias. Ideally, students should be randomly assigned to magnet and nonmagnet schools, but, of course, students are in their schools based on their own and their parents' decisions, and so the two groups are possibly different on unmeasured covariates of the characteristic that distinguishes the choosers from the nonchoosers. If these covariates are associated with student achievement, then this confounds the magnet-nonmagnet comparisons. Studies of differences between choosing and nonchoosing parents in school choice systems find that parents who take advantage of school choice opportunities often differ in ways that would predict higher achievement among their children (Archbald, 1996; Levine...