Selected Papers in School Finance 1994
Henry M. Levin and Cyrus E. Driver
National Center for the Accelerated Schools Project
Stanford University
About the Authors
Henry M. Levin is the David Jacks Professor of Higher Education and Economics at Stanford University. He is also Director of the Center for Educational Research at Stanford (CERAS) and was the founding Director of the Institute for Research on Educational Finance and Governance (IFG). Levin received the Ph.D. in Economics from Rutgers University in 1966. Prior to his arrival at Stanford in 1968, he was a Research Economist at the Brookings Institution.
Levin has been a Fellow at the Center for Advanced Studies in the Behavioral Sciences, a Fullbright Professor at the University of Barcelona, and a Distinguished Visiting Professor at the University of Beijing. He is a past-president of the Evaluation Research Society and a winner of its Myrdal Prize. He served as a member and President of the Palo Alto School Board. He is the immediate past-editor (1991-94) of the Review of Educational Research. In 1991, the New York Times named Levin as one of nine "national standard-bearers" in educational innovation. In 1992, Levin received the Charles A. Dana Award for Pioneering Achievement in Education.
Levin is a specialist in the economics of education and human resources. His work is focused specifically on cost-effectiveness, educational finance, educational and workplace productivity, and investment strategies for educationally at-risk students. He has published 13 books and about 200 articles in scholarly journals.
Much of his recent work focuses on the establishment of accelerated schools to bring all students into the educational mainstream and make them academically able. Evaluation results show remarkable success at very low cost. In 1994-95, there were over 800 elementary and middle schools in 39 states pursuing the accelerated schools program.
--------------------------------------------------------------------------------------------------------------------------------
Cyrus E. Driver is a Ph.D. candidate in the education school at Stanford University, and a research assistant with the National Center for the Accelerated Schools Project at Stanford. His research interests focus on the economics of education with specific work in applying economic theories of contracts to school district employment relations. Prior to coming to Stanford, he worked for 10 years as a community organizer in Chicago. From 1985 to 1991 he worked for the Education Research and Advocacy Organization and Designs For Change, organizing parents, teachers, and others in support of the Chicago school reform effort during the late 1980s and early 1990s.
Estimating the Costs of an
Educational Voucher System1
Henry M. Levin and Cyrus E. Driver
National Center for the Accelerated Schools Project
Stanford University
More than three decades ago, Milton Friedman proposed a radically different way of financing education (Friedman 1955, 1962). Parents would receive a government voucher that could be used for tuition at any approved school. Such schools would compete for students by offering programs that were designed to meet their needs and minimal standards for curriculum and any other requirements in order to be eligible to redeem the vouchers. Friedman argued that this voucher approach would create a more efficient schooling system by providing a wide range of choices to meet parental and student needs and by using vouchers as an incentive for schools to compete for students. Friedman also maintained that the use of vouchers would speed technological progress in education by building incentives for schools to gain a competitive edge in the marketplace.
The Friedman proposal was initially viewed as an academic novelty, unknown beyond university classes in public finance. The fact that Friedman's proposal was published at a time when states were using voucher-based schemes to avoid the U.S. Supreme Court's mandate for desegregation tended to undermine his argument that opportunities for African-American and poor students would be improved. However, the early education failures of the Johnson Administration's War on Poverty, as well as the failure to end school segregation in the North, stimulated a search for radical strategies for school reform in the late 1960s. Christopher Jencks (1966) proposed that vouchers be used to vastly improve education in urban ghettos, and Sizer (1967, 1969) called for an educational voucher plan for the poor. In a more skeptical light, Levin (1968) called for experiments with vouchers to ascertain their consequences. By 1969, the U.S. Office of Economic Opportunity (OEO) had initiated preparations for a voucher experiment, resulting in a plan to provide larger vouchers for poor children (Center for the Study of Public Policy 1970).
The OEO was unable to find a school district willing to accept the experiment in a state that would provide enabling legislation for private schools to receive public funds. The result was a compromise in which the voucher experiment was reduced to a competition in San Jose, California, among public schools that had been reconfigured. Teachers at each participating school established three or four "minischools" at their school site, each with a unique education philosophy. Parents could choose any school or minischool within the Alum Rock School District. Although the experiment provided useful insights into public choice of schools, it was not adequate to test the impact of educational vouchers (Weiler 1974).
By the late 1970s the main proponents of vouchers were pushing for state legislation or state constitutional initiatives that would establish state voucher systems (Coons and Sugarman 1978). To some degree, their momentum was sidetracked by efforts to legislate an alternative to vouchers--tuition tax credits--that could reduce the tax liability of parents, by some portion, for tuition paid at private schools (James and Levin 1983). The failure to enact tuition tax credits and the publication of an important book advocating vouchers by Chubb and Moe (1990) restored this momentum. Chubb and Moe evaluated a large data set on high schools and found that students in schools with greater autonomy tended to have higher test scores. They concluded that the best way to increase school autonomy was a system of publicly financed scholarships (vouchers) that could be used in a competitive marketplace of public and private schools.
Although the statistical methods, conclusions, and interpretation of results by Chubb and Moe (1990) were challenged by critics (Witte 1992), the wide publicity given to their claims and recommendations generated a renewed interest in vouchers. In the early 1990s, constitutional initiatives to adopt the voucher approach were proposed and defeated in California, Colorado, and Oregon, but a fervor for public school reform and the growing effectiveness of political coalitions advocating vouchers augurs well for a continued push for vouchers among the states.
Although conflicting claims were made about the effects of educational markets generally, and educational vouchers specifically (Levin 1991a,b; West 1991a,b), there were few empirical data to validate either view (Witte 1992). Only in Milwaukee has a voucher plan been adopted for students from low-income families. Evaluation showed that, after 3 years, voucher recipients in private schools were performing no better than were similar students in public schools (Witte, Bailey, and Thorn 1993).
Regardless of its educational impact, a shift from the prevalent system of state financing and governance of education to one based on educational vouchers would require a profound transformation of institutions that support the schooling system. In California, for example, a shift to a voucher system would require state authorities to keep records and administer vouchers to nearly 6 million youngsters rather than deal with about 1,000 local school districts. To assure adequate access to alternatives, it is probable that information centers would be needed to enable parents to make informed choices and that an expanded system of publicly funded transportation would also be needed. In addition, a system of adjudication would need to be provided for parents wanting a partial refund of vouchers in order to change schools during the academic year. Finally, a state system of monitoring and assessment would be needed to establish voucher eligibility for both students and schools.
Although there has been much debate about the overall impact of educational vouchers on individual schools, considerably less attention has been devoted to the changes required for, and the cost implications of, a system to support the educational voucher framework.
The purpose of this report is to begin estimating the cost of the supportive framework for an educational voucher system. Creating accurate estimates requires the identification of categories for which a voucher system would entail additional or expanded state services or oversight and calculation of their costs. For reasons given in the next section, this effort can only be exploratory, not exhaustive. Therefore, it is important to stress that the purpose of this paper is to provoke discussion and to allow further refinement. In the following sections, we discuss issues that must be addressed in order to estimate the cost of a voucher system, with illustrative cost estimates.
The cost of replacing a system of schooling with a voucher system cannot be estimated without accurate specification of the particular voucher plan being considered and the system it would replace, the setting in which it would be applied, assumptions about the behavior of schools and famillies under the voucher approach, and the method of estimating costs. In this section, we review the importance of each of these concerns.
The Particular Plan
Although the educational voucher system is often referenced generically as though it were a single, unified approach to financing education, the term actually covers a wide variety of arrangements with varying potential consequences for cost. Proposed voucher systems vary from market approaches with little or no government intervention, other than the funding of vouchers, to highly regulated educational marketplaces with elaborate provisions for disseminating related information, transportation, school admission policies, and participation requirements. This variability suggests that they should be viewed not as either/or proposals, but as policy designs having different effects on cost--and on families and schools (Hoenack 1994).
Educational voucher plans differ in their regulations, information dissemination requirements, and systems of finance (Levin 1991a). The original Friedman proposal had few participation requirements for schools. This proposal made no provision for information dissemination and stipulated that parents would receive a flat amount for each child, which they could supplement. In contrast, the OEO proposal called for various admission restrictions, specific information from schools, an extensive system of related information that would be made available to parents, and vouchers tailored to students' needs, including larger voucher amounts for poor families. Parents would not be allowed to add to the allotted voucher. The Friedman proposal would not require elaborate systems of information dissemination, transportation, school monitoring, or evaluation of students. In contrast, the OEO plan would necessitate rather extensive outlays in all of these areas.
To provide even an approximate picture of costs for a supportive state framework for vouchers, it is necessary to know the specific details of the voucher plan. It is also crucial to know precisely what type of system the voucher plan would replace. Some states have very large numbers of local school districts, while others have relatively few. Some states have elaborate systems for monitoring and regulating schools, while others do not. In most states, oversight of private schools is less than might be required under a voucher plan. In other states, such as Minnesota, or local jurisdictions, such as 16 school districts in Massachusetts, parents have a wide variety of public education choices for their children. In those instances, at least some provision is made for disseminating school information to parents, although it is rarely extensive. Virtually all states provide some transportation for students who are distant from their schools r have severe handicaps. Because the cost of these services must be deducted from a particular voucher arrangement to receive the net cost for a voucher framework, it is necessary to know the characteristics of the public system it would replace.
The Setting
The setting in which vouchers would be used will affect the system's cost. Some states have large numbers of students in private schools. In densely urban areas, transportation costs would be lower because (1) the market would offer many transportation choices and (2) because public transportation is widely available. The provision of information on alternatives might also benefit from economies of scale in these urban areas. In contrast, the cost of transportation might be considerably higher in rural areas because of the longer distances between schools. Monitoring schools, providing voucher-related information, and evaluating students for a voucher that meets their individual needs might also be more costly. Even in urban areas, population density and numbers of school-age children vary. For these reasons, it is necessary to know the specific setting in which a voucher plan would be implemented in order to estimate the cost of such a plan.
Behavioral Responses
In addition to the specifics of the voucher plan and the existing plan used, as well as the setting in which a voucher plan would be applied, the behavioral responses of households to a voucher plan must be determined. To the degree that a voucher approach would establish a new set of incentives and increase options for parents, one can expect that it would stimulate changes in both attendance patterns and school choice. As an example, Hoenack (1994) developed an econometric simulation to predict how voucher amounts of different sizes would affect the shift from public to parochial schools in Minnesota. In general, the costs of the framework for a voucher system would depend, in part, on the direction of such behavioral responses to new opportunities. What proportion of students would change schools (Lankford and Wyckoff 1992)? What proportion would require transportation to their new school? Would attendance be concentrated in particular neighborhoods and schools, or would the distribution be more random? How many schools would arise in response to the voucher system requirements of monitoring and approval? Would the residential mobility of households be affected by school choice and availability? What proportion of families would avail themselves of different types of voucher-related information services? The answers to such questions would be determined by how families responded to the opportunities and incentives inherent in a given voucher system, and how these responses would affect the cost of supporting that system.
Costing Method
Finally, the method of estimating the costs associated with a voucher approach will be one of the central determinants of whether implementing such a system is feasible. The method that should be used is straightforward. With the information stipulated above, it is possible to project the activities needed to provide a structure that meets government specifications for a voucher system. Based on parental choices, the number of children to be transported, as well as transportation patterns from particular neighborhoods to particular schools, may be calculated. The type of organization that would provide voucher-related information, the type of information this organization would collect, and how it would obtain and disseminate that information should be factored into the cost estimate, as should the types of monitoring activities required of schools, student evaluations required to allot vouchers, and the recordkeeping system required both for students and schools.
Each of these functions can be expressed in specific activities and services, and the resources or ingredients required to produce them can be identified. It is then possible to ascertain their prices and to estimate the cost of specific activities, as well as the cost of the overall voucher framework. This method has been used in education costing and meets standard methodological criteria (Hartman 1981, James 1983).
It is important to note that many of the costs estimated using the costing method discussed above might not be borne by the educational sector or government sources, but by households. For example, if provisions were not made for transportation, parents would have to make their own arrangements. Whether or not they were to use public transportation with its private costs and public subsidies, additional costs would be imposed. The same is true for information dissemination. If publicly funded information on school options were meager, some parents might choose to seek information independently from schools or private counselors. Indeed, the overall cost of these services might actually be greater. Therefore, it should not be assumed that restrictions of specific services offered by a voucher plan would limit its cost if the responsibility for obtaining those services were simply shifted to households. Both public and private costs must be taken into account when estimating the cost of supporting a voucher plan.
We assert that the cost of the infrastructure we consider necessary to support an educational voucher system would exceed the costs associated with the existing public school system. We believe that the information dissemination, transportation, and centralization of records required under a voucher system would entail added costs beyond what these services require in a more traditional schooling approach. To focus attention on and initiate discussion about such costs, we provide a ballpark cost approximation for each item under a given set of assumptions.
Although Lieberman (1993) argues correctly that published data are inadequate for comparing existing public and private education costs, his assertion that the cost of public schools exceeds the cost of a market system is based on assumption rather than on careful analysis and measurement. In an effort to clarify what the costs are, he lists many categories, with little attempt to define them carefully and show how they would differ between public and market systems. For example, in his catch-all approach, he lists the electoral process, operation of school boards, and education legislation as costs of public schools that would be eliminated under a market system (pp. 136-137). However, Encarnation (1983) has demonstrated that, far from eliminating legislation, government support of private entities historically leads to increased government intervention and regulatory oversight to protect the public interest. Further, school board elections and governance are hardly the dead weight loss asserted by Lieberman. Tyack and Hansot (1981) argue that public schools represent one of the few foci for democratic discourse, not only about education, but also about society itself. They assert that
"...public schools are everywhere close at hand and open to all children. They generate valuable debates over matters of immediate concern and offer a potential for community purpose that is unparalleled in our society."
None of these arguments resolves the issues surrounding the relative cost of different systems, but they illustrate why these issues cannot be settled with polemics or tendentious checklists.
However, it is important to stress that even if costs are higher for a market system, this is not a prima facie criticism of vouchrs. The real issue is whether any benefits of a voucher approach over the existing system would be justified by the additional cost or whether the additional cost for a supportive framework is offset by savings at the individual school level. This issue cannot be ascertained from the present analysis, but should be viewed as an open question.
In this section, we provide illustrative estimates of the costs of shifting to a voucher system. These estimates are merely illustrative because we lack specifics on a particular voucher plan, what type of system it would replace, where it would be applied, and behavioral responses to the new arrangements. Without these ingredients, it is not possible to provide even an approximate cost for a state or smaller entity, nor a definitive comparison of current costs with those incurred through a voucher plan. However, it is possible to estimate costs for hypothetical situations that would be consistent with a shift to a voucher plan in order to provide the beginning of a dialogue on costs and to illustrate methods of obtaining costs. We address these costs in five areas: accommodating additional students, recordkeeping and school monitoring, transportation, information dissemination, and adjudication of disputes.
In each of these areas, we begin by discussing the issue and why it is central to a voucher plan. Next, we demonstrate how to measure costs for a particular area and provide illustrative costs, showing the degree to which some of these costs would be offset by cost reductions in the existing system. Finally, we suggest the magnitude of potential cost differences for the function.
Accommodating Additional Students
Under a voucher system, students who attend nonpublic schools would be eligible to receive publicly-funded vouchers. This means that, even in the absence of a shift from public to private schools or the provision of services to create an efficient system of choice, there would be an additional cost to the public sector, and a windfall gain for most families with children already in private schools. The amount of this public cost depends upon three factors: the number of students in nonpublic schools eligible to receive vouchers, the amount of the voucher, and the cost of attending those schools. The number of eligible students would depend on the regulations in voucher plans and the willingness of nonpublic schools to participate in the voucher system. With minimal regulations, it is likely that all or most schools would participate and be eligible. However, restrictions on admission policies, tuition charges, curriculum requirements, testing, etc., would limit the number of nonpublic schools willing to participate. For purposes of estimation, we will assume that under the least restrictive arrangement, 100 percent of existing students in nonpublic schools would be eligible to receive vouchers. We will assume a 75 percent rate of participation for estimating costs under the more regulated plans.
The amount of the voucher is obviously crucial in determining the additional public cost of accommodating students in nonpublic schools. The larger the voucher, the greater the cost. For purposes of estimation, we will assume that the maximum voucher amount would be equal in size to the average per-pupil expenditure in U.S. public schools in 1990-1991 (See Table 1).
The final criterion in estimating the cost of a voucher system is the amount of private school charges that would be eligible for voucher reimbursement. Although many private schools charge less than the average per-pupil public school expenditures, they hold fund-raising activities or require fees from parents. In addition, private schools often receive contributions in kind, such as donated or subsidized facilities and voluntary labor or staff who are willing to accept below-market wages because of a school's dire financial straits. However, we would expect schools eligible for vouchers to charge the full amount of the voucher. The additional school income would be used to improve staffing and salaries (Chambers 1987) and provide better facilities and services--that is, to raise the quality of the school. This tendency would be virtually certain in a market situation where schools and parents would have a strong incentive to use the full voucher.
Table 1. Total and per-pupil costs of K-12 public education: School year, 1990-91
Cost per pupil Total cost in average daily attendance -------------------------------------------------------------------------- $228.9 billion $5,872 --------------------------------------------------------------------------
NOTE: Average daily attendance is the aggregate attendance of a school during a reporting period (normally a school year) divided by the number of days school is in session during this period as defined in the Digest of Education Statistics, 1993.
SOURCE: U.S. Department of Education, National Center for Education Statistics, Digest of Education Statistics, 1993.
However, voucher costs might nevertheless be lower for nonpublic schools than for public schools for three reasons. The first reason is that a voucher plan might provide a somewhat lower allocation for nonpublic schools because they need not meet the same public accountability requirements as public schools.
A second reason is that few private schools provide either special education services for the disabled or vocational services--two of the most expensive public school offerings. By federal law, public schools must provide education for all students in the least restrictive environment, and ths cost of special education is nearly two-and-one-half times that for a "regular" student (Chaikind, Danielson, and Brauen 1993). The cost of vocational education is two to five times that of the academic curriculum at the high school level (Hu and Stromsdorfer 1979).
The third reason is that nonpublic schools have a higher proportion than do public schools of children at the elementary level where per-student costs are considerably lower than they are at the secondary level. Therefore, for purposes of estimation, we will assume that the minimum cost of voucher students in nonpublic schools would be 80 percent of the average per-pupil expenditure in public schools in 1990-1991. It should be noted that this lower cost per student is due to the fact that private schools enroll students with lower educational needs, it does not represent lower costs for students with equivalent educational needs. Thus, we are assuming no net savings for equivalent services when a student shifts from one sector to the other.
Table 1 provides total and per-pupil public school costs for the Nation
2 for the 1990-1991 school year from the Digest of
Education Statistics, 1993.
As seen in Table 1, the overall cost of nearly $230 billion breaks down to nearly $5,900 per student in average daily attendance. The Digest of Education Statistics, 1993 (U.S. Department of Education 1993) also reveals that slightly more than 10 percent of all students attended private school that year--about 4.7 million of the total 46 million students enrolled in school. 3 Had all of the schools attended by the students in our model been eligible for vouchers, additional public costs for education would have been about $27.4 billion higher for school year 1990-1991. Had only 75 percent of these students attended schools participating in a voucher plan, additional costs would have been about $20.6 billion. Had the amount of the student's voucher been only 80 percent of the average per-pupil cost in the public sector in school year 1990-91, the cost would have been about $22 billion for all students, or about $16.5 billion at an eligibility rate of 75 percent.
Table 2. Potential public education cost increases using vouchers equal to 100 percent, and to 80 percent of public school costs, under a voucher plan according to proportion of private school students participating (in 1990 dollars)
----------------------------------------------------------------------------
Percent of
private school Cost using vouchers Cost using vouchers
students equal to 100% equal to 80% of
participating of public school costs public school costs
----------------------------------------------------------------------------
100 % $27.4 billion $22.0 billion
75 % 20.6 billion 16.5 billion
----------------------------------------------------------------------------
SOURCE: U.S. Department of Education, National Center for Education Statistics, Digest of Education Statistics, 1993.
Table 2 summarizes potential increases in public costs. Most of these costs would represent a shift from the private to the public sector (from present private school families to taxpayers), although some of these costs represent an overall increase in the social cost of education to the degree that overall expenditures on education would rise under a voucher plan.
Recordkeeping Systems
Voucher plans would require an elaborate system of recordkeeping by the state for three reasons: to assure that all children who are required to attend school are enrolled, to determine the appropriate voucher amount for each child based on educational need (such as grade level, disability, socioeconomic status, and language), and to assess the eligibility of schools to redeem vouchers. More specifically, each child required by law to be in school and those who continue their education through high school would need to be monitored with respect to enrollment and voucher provision. Such monitoring should be active rather than passive because of student mobility through changes in school selection and family mobility, as well as the normal shifts from one school level to another.
At the same time, children would be eligible for different educational services with appropriate differences in voucher amounts. For example, secondary schools have a higher cost than elementary schools, and services for disabled, educationally disadvantaged, and language minority students are more costly. Accordingly, students would have to be evaluated in terms of services required and voucher size. Finally, only schools meeting the regulations determined for participation in the voucher plan would be eligible to redeem their vouchers. Schools would then have to be evaluated, certified, and monitored for eligibility.
Currently, there is some cost to local educational agencies for recordkeeping and student evaluation, and some cost to the state for maintaining records on school programs. With voucher plans it is likely that these functions will be more extensive and complicated, given the need for more active monitoring, assessment and determinations. In addition, issuing vouchers, enforcing compulsory attendance laws, and regulating and monitoring schools would logically rise to a state agency rather than relying on local school districts to assure uniformity and consistency. Under the existing system, local school districts carry out many of these functions, including, complying with compulsory attendance laws, maintaining student records, and evaluating educational services. Because these functions can be integrated into a single agency that is close to the families and neighborhoods served, costs are likely to be lower than when they are handled by a centralized agency that is independent of the schools. Both the extensive nature and centralization of these functions will likely increase costs under voucher plans.
Accurate costs can only be determined once the specifics of a voucher plan are known. Without a specific voucher plan, we must instead look to analogous systems to get an estimate of potential costs. In this case, we examined the federal Social Security Administration (SSA), which maintains an ongoing record of eligibility of all 38 million SSA beneficiaries (U.S. Social Security Administration 1987). This system is partially comparable to a voucher plan, because SSA goes through essentially the same steps of any centralized office that determines initial and ongoing eligibility. Once initial benefits are determined--with higher processing costs for the more complicated disability claims--SSA then continues to distribute uninterrupted, equal monthly payments to beneficiaries. At different points in time, though, benefits may be adjusted to reflect changes in the situation of the individual or his or her family. This situation is analogous to a change in voucher amounts in a plan that allows vouchers of variable amounts.
We have chosen the SSA analogy because it describes a situation in which a relatively simple eligibility determination can be made, the retirement insurance case, and one in which eligibility requires a more complex determination (the disability insurance case.) This analogy parallels two voucher plan situations: one in which eligibility criteria are fairly simple, and one in which the eligibility criteria are more complex as a result of more variables. In the simple case of retirement claims, social security law requires determination of benefit payments based on a variety of formulae applied solely to earnings records of beneficiaries. This process is similar to a straightforward voucher plan in which the size of the voucher would depend on the school level in which the child is enrolled. The average annual processing costs for these claims in 1986 was $43 (in 1990 dollars) (U.S. Social Security Administration 1987).
In a more variable voucher plan where such factors as low-income status of families or special education needs provide for larger vouchers, the analogy of SSA retirement claims processing costs may actually be too low. A more complex voucher system would have ongoing processing costs similar to those of SSA disability claims processing, but probably less than the 1986 average of $372, which included costly periodic medical evaluations (in 1990 dollars) (U.S. Social Security Administration 1987). Therefore, a likely amount for a voucher system may lie somewhere between the processing costs of ongoing retirement and disability claims, perhaps something like the amount shown in Table 3, which is the average of these two processing costs. The cost of $78 per claim is based on the heavy predominance of retirement claims over disability claims. However, to the degree that the overall budget of the SSA also includes routine maintenance of accounts prior to retirement or disability, some of that cost is attributable to other functions. Thus, a value somewhat below $78 per student per year would be comparable. Table 3 provides "ballpark" per pupil annual costs based on the SSA analogy. Here in Table 3, the simple (low cost) case assumes vouchers are all the same amount and that all schools and students participate in a given area; the complex case (high cost) assumes that voucher amounts vary.
Table 3. Possible high and low average annual costs for a recordkeeping system in a voucher plan (in 1990 dollars)
---------------------------------------------------------------------------- Low Cost High Cost ---------------------------------------------------------------------------- Less than $43 per student $78 per student ----------------------------------------------------------------------------
NOTE: November 1993 U.S. Department of Labor Consumer Price Index (CPI) Detailed Report's "CPI-U" used to adjust the SSA amounts to 1990 dollars.
SOURCE: Social Security Administration, Executive Handbook of Selected Data, May 1987.
Table 3 provides the approximate costs of a recordkeeping system, but they are not necessarily the net additional costs of recordkeeping a voucher system. To determine the net additional cost, we need to add the costs of the application process for new schools to certification and monitoring for existing schools for their eligibility to participate in a voucher system. We would then have to deduct from this total the costs expended on all of these functions (recordkeeping, evaluating services, and determining eligibility and monitoring schools) under the present system to determine whether these costs are actually greater under a given voucher system.
We have argued that the costs of the recordkeeping system are likely to be lower than they would be under a voucher plan because of the relative ease of gathering and monitoring information on a local level, with the integration of pupil information and the evaluation of educational needs provided by local educational agencies. Further, monitoring schools is far more routine than it would need to be in a dynamic marketplace of births, deaths, and changes in schools. Moreover, there are likely to be far more schools to monitor under a voucher system--perhaps twice as many--if the size of private schools is an indicator of what will happen to school size under this type of plan (Chambers 1981). For example, in 1987-88, over half of private schools with 4th grades and almost half of schools with 12th grades had less than 150 students enrolled, in contrast to 11 percent and 14 percent, respectively, of public schools at the same grade levels. In contrast, only about 8 percent of private schools with a 12th grade had 750 or more students enrolled, while almost 36 percent of such public schools were that large (U.S. Department of Education 1992).
The evidence seems to be overwhelming that the savings that would occur in reduced administration of the present system would be more than offset by the more extensive system required for regulating and monitoring schools and the investigative costs of assuring compliance with compulsory attendance laws as well as the centralization of that system. In that case, the higher cost of $78 per student per year might serve as a first approximation of the overall cost of recordkeeping and monitoring, and the lower cost of $43 per student from the SSA illustration might represent a first approximation for the net cost.
Transportation Costs
Any voucher plan that provides for meaningful school choice must include transportation. Although home schools and distance learning schools that use computers and interactive television might be eligible for vouchers if they met certain requirements, the vast majority of students eligible under a voucher plan would likely be found in school settings away from home. Because schooling must be produced and experienced in such settings, accessibility to a range of school sites would require adequate transportation.
Transportation costs would be a function of several factors, including the number of students being transported, the number of schools served and their locations, the geographic distributions of students and the schools they choose, and the costs of different modes of transportation. Other factors include, population density (rural students might have to travel farther to school and would be less likely to have access to adequate public transportation), climate, and labor costs. Transportation costs under a voucher system would probably be higher than under the present system, for two reasons: freedom to choose should lead to more students attending schools outside their immediate neighborhoods, and certain routes are likely to be lower in rider density and regularity.
We assume that a voucher plan would lead to a net increase in new schools because existing nonpublic schools are considerably smaller than their public counterparts. This larger number of schools would have both advantages and disadvantages for transportation. Increasing numbers of school choices would mean that students would have more school options closer to their homes, particularly in urban areas. But the current economies of scale--using larger buses with regular routes--would likely give way to the need for smaller vehicles with costly and irregular routing. Labor costs for drivers are similar for large and small vehicles, as are many maintenance and insurance costs. Assuming a further shift in student enrollment from nearby (perhaps formerly public) schools to other schools, both the number of students and the number of schools to which students need transportation would increase. Also, administrative costs would likely increase because more students and schools would be served. This would be particularly true if private schools continued to be considerably smaller than public schools, entailing a far larger number of schools--even beyond existing private schools--that would be added to the transportation system.
What do public school transportation costs look like? A 1987 study (Deloitte, Haskins, and Sells 1987) of data collected from 34 California districts that provided transportation service directly or subcontracted with private companies in school year 1985-86 provides ranges of per pupil transportation costs (see Table 4; costs given in 1990 dollars).
Table 4. High and low per-pupil and per-mile costs (in 1990 dollars) of public school transportation in California: School year 1985-86
----------------------------------------------------------------------------- Per-pupil cost Per-mile cost Low cost $218 $0.72 High cost 3,782 3.90 -----------------------------------------------------------------------------
SOURCE: Deloitte, Haskins, & Sells, School Transportation Study, 1987.
Table 4 shows that there is enormous variation in student transportation costs, even among those with relatively systematic and regular routes transporting both regular and special education children. In the 1986-87 school year, the national per-pupil average transportation cost was $292. By 1990-91, it had risen to $394 in 1990-91 dollars (U.S. Department of Education 1993). These costs reflect benefits of economies of scale in districts that can establish regular routes and patterns of transportation among a fixed number of schools and with school assignment policies that can minimize the number of students bused and the distances they must travel. Under a voucher plan (within reasonable transportation boundaries), the choice of schools would be decided by parents and students, and they might have less of an incentive to economize on transportation costs if they were paid by the government.
The recent St. Louis school desegregation plan is somewhat analogous. Students from St. Louis were permitted to choose to attend schools in the suburbs, and those in the suburbs were permitted to choose to attend schools in St. Louis. The Missouri Department of Elementary and Secondary Education prepared calculations to show the total cost of transporting nearly 14,000 children among St. Louis and these communities; this cost was about $25 million, with a per-pupil cost of $1,800 in the 1992-1993 school year. 4 The reason for this seemingly high cost is the type of transportation provided: a combination of 1,100 traditional large school buses and 15 passenger van-style buses were used by the contracted bus companies, along with 80 to 90 taxis, on round trips ranging from only a few miles to about 80 miles. This combination might well be used under a voucher system.
School buses are not the only means of transporting children; in most regions, particularly urban and suburban areas, public transit is available at a seemingly low cost. In the San Francisco Bay area, for example, the Bay Area Rapid Transit (BART) system's average cost per mile for consumers in 1988-89 was about 11.6 cents per mile (in 1990 dollars) (Bay Area Rapid Transit District 1989).5 However, this cost excludes the enormous subsidies that BART receives in such forms as local and state tax revenue and federal grants. Taking into account these other sources of revenues and their concomitant taxpayer expenditures, the average cost per mile increases from 11.6 to 49.5 cents. This higher figure represents a more accurate assessment of the costs of this public transit system, even if they are only partially borne by its users. Of course, the marginal cost of transporting an additional student is small, but for an entire voucher system, the student demands would be more massive than those at the margin, justifying the use of an average cost estimate.
Table 5. The cost of a 10-mile round trip on the Bay Area Rapid Transit System over a 180-day school year: 1988-89 (in 1990 dollars)
-------------------------------------------------------------------------
Fare paid $209
Actual cost 892
-------------------------------------------------------------------------
SOURCE: Bay Area Rapid Transit District, 1989.
Table 5 shows the annual cost of a 10-mile round trip (5 miles each way) on BART in 1990 dollars. It becomes apparent that public transit has a low cost for patrons, though inclusion of subsidies substantially raises these costs. Nonetheless, public transit appears to be less costly than the more traditional bus service described above, assuming that public transit can serve school commuting routes within a reasonable commuting time. Of course, it is unlikely that parents would permit their elementary school children to use public transit without older children or adults acting as chaperones.
Depending on the number of new schools that would arise under a voucher system, its transportation systems might look more like those of St. Louis than do most systems we see today that only use large buses (except for those providing services for the disabled). In many large cities, private schools provide transportation at extra cost using minibuses or vans. These large buses likely would be replaced by more vans used to transport a fraction of the children who have chosen a given school.
An efficient transportation system primarily using vans currently serves the San Francisco Bay area's three airport facilities. Several companies offer very competitive door-to-airport service. Interviews with company staff suggested that a similar customized arrangement would work under a voucher plan if necessary. These interviews also elicited the per-pupil cost estimates under such a system given in Table 6.
Table 6. Projected per-pupil 10-, 20-, and 40-mile daily and yearly transportation costs with "customized" service under a voucher system
----------------------------------------------------------------------------
10-mile 20-mile 40-mile
cost cost cost
Daily cost $7 $10 $12
Yearly cost 1,260 1,800 2,160
----------------------------------------------------------------------------
SOURCE: Interviews with staff of the San Francisco Bay Porter Express and Express Shuttle bus companies.
What would be the magnitude of these cost differences? As noted above, the 1990-91 national per-pupil average transportation cost was $394; we estimated that the average cost under a voucher system would probably be well over $1,000. Even the most efficient customized transportation system would likely cost between $1,000 and $2,000 per student per year, with $1,500 being a reasonable figure, depending on the plan, setting, and school choice. The validity of this range is supported not only by the St. Louis busing data, but also by the $2,000 per student report for the Milwaukee voucher experiment. 6
Not only would the average cost of transportation rise considerably under a voucher system, but the number of students requiring transportation also would rise, with important cost consequences. Nearly 60 percent of the students who attended public schools were bused in 1990-1991 (U.S. Department of Education 1993). Had this number been 80 percent of both public and private school students, an additional 11 million students would have been bused at an additional cost of about $4 billion dollars, assuming the above-mentioned per-pupil cost of $394. However, had the cost per pupil been about $800 per year per student because of less economical routes, less expensive modes of transportation, and smaller schools, this higher cost per student, calculated to include the additional bused students, would have amounted to an additional $17 billion a year--about half of it attributable to the additional students and half to the higher cost. This figure is considerably lower than those derived from costs cited for St. Louis and other cities, or the projected costs of the use of vans.
Information Dissemination Costs
The competitive efficiency of educational market systems depends greatly on consumer knowledge of alternatives (Levin 1991a). Families of different racial, socioeconomic, and linguistic backgrounds would require equal access to information about the available schools and their own choices of schools. At a minimum, families would need to know that they have choices, the appropriateness of particular choices for their children, and the probable consequences of such choices. They would require details on such matters as a school's philosophy, curriculum, personnel, facilities, and test scores; the success rate of a school's placements after graduation; the nature and rate of complaints; and its student turnover rate. Schools would be required to provide such information to a central data system, and to establish a capacity for school visits and interviews by both parents and school staff.
Costs would be entailed for schools, parents, and government: for schools the costs include, collecting the appropriate data and making them available to parents and government agencies and providing for school visits and interviews; for parents, time and transportation needed to gather the data at both a central information agency and school sites as they narrow their choices; and for government, maintaining and updating a data base on schools and providing an efficient system for disseminating that information.
How can we project these information costs? One way to begin is to consider the steps involved in sharing that information. The first step is to specify, collect, and store the information. The second step is the actual sharing and dissemination of the information. In the first step, costs would increase according to the depth and quality of the information compiled, as well as the frequency of updating it. Consider, for example, cost differences between a one-time public notification that a voucher plan exists and a requirement that every school develop and publish an annual report in several different languages. In the second step, costs would depend on how much information needs to be shared, as well as how it is shared. Printed materials circulated through libraries, for example, should be less costly than information centers with specially trained staff. Nonetheless, a balance should be struck between the cost of sharing more information and its potential to improve decisionmaking by families.
It is important to stress that information costs could be borne to a great extent by schools and families in addition to government. The cost of information development could be borne by schools and might not be new or larger for private schools than at present unless the information were to be disseminated across a larger audience. However, such costs would be extended to a larger number of schools, because under a voucher plan, all schools would be private, or existing public schools would also need to provide information to be competitive. On the other hand, the lion's share of the costs of obtaining information could be borne by parents as they inquire about schools through both formal and informal means. In the Milwaukee voucher program, for example, parents frequently learned about the program through informal communication, a reasonable approach when a maximum of only 12 schools were involved (Witte et al. 1993).
Government might bear a large share of information costs where equal access is an important consideration. Government would need to play a large role in helping parents with meager resources to obtain and evaluate information. It should be noted that such systems of choice have been characterized by much lower levels of familiarity with, and poorer use of, information by nonwhite and low-income parents than by white and middle-income parents (Bridge 1978; Archbald 1988).
Massachusetts provides an example of how information can be conveyed to families who have increased school choice (Glenn, McLaughlin, and Salganik 1993). Sixteen cities in Massachusetts recently began to allow parents to choose the public school their child attends. Parents obtain information about schools through a variety of means, including school visits, conversations with friends and neighbors, and brochures distributed by schools. Thus, both families and schools contribute to the costs of sharing information. We have not seen data on the extent to which these costs have resulted from the school choice policies of these districts, but one new way of sharing information has emerged--Parent Information Centers (PICs)--funded with state desegregation monies. PICs have served to supplement information from other sources and have been particularly helpful to families with limited English-language skills. At these centers, parents can get written information and learn from trained counselors. Such centers might well be needed in a voucher plan situation in which parents would be faced with a myriad of options and would require unbiased, well-informed assistance in selecting a school.
What are the costs of these PICs? The 16 Massachusetts cities have a combined enrollment of about 200,000 students, and the PICs serve the families of about one- third of these students each year (65,000) as they choose a school or move from one school level to another. The total annual costs of the PICs are about $2.5 million, so the per-pupil cost is about $38 per year for this fairly modest information dissemination approach.7 A more extensive approach, which would ensure substantial services in terms of extensive data and counselors to interpret that data, would be more costly. Further, the Massachusetts data do not include costs to parents or schools, only to government.
Costs of Adjudication
If there is no flexibility in a voucher plan--that is, if all families were to receive a voucher of a uniform size for each of their school-age children--then the only cause for legal dispute might be a child's real age. In such cases, there would be few disputes, and we believe costs to resolve them would be minimal. As we have noted, though, several factors might complicate a voucher plan; these include the eligibility of certain classes of students and schools and the variability in voucher size. We suggest that as a voucher system becomes increasingly complex and requires multiple judgments regarding eligibility for a given voucher amount, disputes surrounding these judgments will increase, and the cost of resolving them will rise. A potential for dispute can be seen when parents wish to move their child from one school to another because of a residential move, a job change, or the child is not doing well, and the parent believes that the school misrepresented its ability to meet his or her child's needs. In the latter case, a school that has included the child's voucher in its budget for the year would likely be reluctant to share hat voucher amount with a new school. A system that provides voucher amounts according to family income would risk disputes about relative voucher amounts, whereas a system that provides a set amount would not.
A recent California proposition (Proposition 174) provided for a uniform voucher amount for all students (State of California 1993). In November 1993, this proposition was voted down and an initiative that would provide larger voucher amounts for children from low-income families and children entitled to special education services recently has been circulated. This example shows how adding the single complicating factor of variability might give rise to disputes, and it is hard to imagine that some families of marginally low income, but not so-deemed by law, would not dispute their status under a voucher plan. However, even with uniform vouchers, a change of schools during the school year would require either a prorated refund (to allow the child to register at a different school) or a system for adjudicating the conditions under which the student could change schools.
Adjudication costs would be a function of the number of cases in which the need for adjudication arises and the cost of creating and maintaining the adjudication system. The former would be determined by the complexity of the voucher system (partially by the frequency of school changes), and the latter would depend on the choice of dispute resolution systems. For example, a system that limits resolution to the decision of one arbiter would be less costly than one in which a panel of arbiters offers decisions that could be appealed.
What might these costs look like? To provide some indication, we turned to Salzer's study detailing the costs of special education mediation conferences and state due process hearings in California (Salzer 1987). Public Law (P.L.) 94-142 provides for due process hearings for parents of special education children who disagree with school district actions that affect their children. Although P.L. 94-142 and many state companion laws are highly complicated and require many contestable actions, Salzer's study of mediation conferences and state due process hearings begins to provide a range of dispute resolution costs that could apply under a voucher system.
Salzer describes three alternative routes for dispute resolution in California. The first route--mediation conferences--involves parents and school district representatives in a one-day session with a mediator. If this fails, one or both parties can request a state due process hearing (the second route), which must be held within 45 days after the mediation conference. The third route is direct recourse to a state due process hearing, eliminating the intermediate mediation step. Salzer focused on both parent and school district costs with regard to the three routes. Table 7 shows the costs that Salzer found in 1981 (in 1990 dollars).
A third set of costs, which Salzer discusses briefly, are the salaries of mediators and due-process hearing officers. These costs, although not insignificant, are quite small compared with the costs shown in Table 7. For example, the cost of a mediator for a one-day hearing (the first route) was $190 (in 1990 dollars).
Salzer notes several reasons why district costs were so much higher than parent costs. The most important reason was the cost of legal representation: districts tend to hire more expensive attorneys than do parents. In addition, Salzer notes what she calls "indirect costs," which were the costs of either parent or district staff time. Her valuation of the cost of district staff time was considerably greater than that of parent time.
Table 7. Average costs of special education dispute resolution in California by dispute resolution method: 1981 (in 1990 dollars)
------------------------------------------------------------------------ Cost Cost Combined Method to parent to school district cost Mediation (route 1) $496 $1,965 $2,461 Mediation and due process (route 2) 2,277 4,642 6,919 Due process (route 3) 3,282 6,089 9,371 ------------------------------------------------------------------------
SOURCE: Evaluation and Cost Estimation of Special Education Mediation Conferences in California-dissertation submitted to Stanford University's School of Education and the Committee on Graduate Studies by K.W. Salzer, 1987.
How relevant is this analogy for disputes that might result from decisions surrounding vouchers? As noted above, the increasing complexity of a voucher system would probably increase the likelihood of disputes, and probably their cost, because parents and/or administrative staff would need more time to assure that they were preparing a case that covered all the bases. In addition, a potential and important source of disputes would involve the mobility of families during the school year. Vouchers could be prorated so that when families move, they receive a voucher that covers school costs for the balance of the year. But this will not solve a problem that schools could face: consider that at the start of a school year, schools invest in resources--not only supplies, but staff and facilities--for that year. Much of this investment will represent fixed costs that will be lost if some students leave the school. Therefore, schools would not want to return partial or prorated vouchers. One method of resolving this problem is to allow schools to charge higher prices at the outset to cover any losses they may incur if families move. The other method is to resolve disputes through a process, perhaps greatly increasing the number of disputes that may occur.
In any case, there will be a minimum cost associated with each dispute resolution, no matter what method is used to resolve the dispute. Even excluding attorney costs, there will be both parent and administrative time, as well as the cost of a mediator or other officer to aid or decree the resolution. To determine a minimum cost, let us assume that parent, administration, and mediator costs are only half of Salzer's first dispute resolution route cost, and further assume that both parties represent themselves. To determine a high cost figure, let us use half of her third resolution route figures, including the cost of legal representation. Table 8 shows these low and high cost estimates for dispute resolution.
Table 8. Projected low and high cost estimates per case for voucher-related dispute resolution under a voucher plan (in 1990 dollars)
--------------------------------------------------------------------------- Low cost: $1,632 High cost: $5,854 ---------------------------------------------------------------------------
SOURCE: Evaluation and Cost Estimation of Special Education Mediation Conferences in California-dissertation submitted to Stanford University's School of Education and the Committee on Graduate Studies by K.W. Salzer, 1987.
Finally, it is important to mention Salzer's finding that fewer than 1 percent of all special education cases entered mediation or due process (p. 38). Therefore, even though our cost of resolution appears quite high, the number of cases requiring this type of resolution probably would be very small.
The purpose of this study was to explore the costs that might be associated with the construction of a publicly-funded voucher approach to education. Our analysis shows there probably will be additional costs in several important categories over and above those arising in the current provision of education. For a variety of reasons, the costs presented should be considered provocative rather than definitive. To provide more precise figures, it would be necessary to know the specifics of the voucher plan, features of the setting in which it would be applied, the behavioral responses of families and schools to the new framework, and the method of determining costs and their distribution among families, schools, and government. In addition, we would need to know what, if any, cost savings would result from replacing the existing public school system. In the absence of these data, it is impossible to provide a reasonably precise picture of the net costs associated with a voucher plan. However, we have identified probable costs and used existing programs that are somewhat analogous to an educational voucher system to illustrate the potential magnitude of the costs, hoping to launch the policy discussions necessary for informed public education decisions.
Archbald, D. 1988. "Magnet schools, voluntary desegregation, and public choice theory: Limits and possibilities in a big city school system." Diss. U of Wisconsin.
Bay Area Rapid Transit District, 1989. (Perp. 34). Annual Report 1988/1989. San Francisco, CA.
Bridge, G. 1978. "Information imperfections: The Achilles' heel of entitlement plans." School Review 86: 504-29.
Bryk, A., & Lee, V. 1992. "Is politics the problem and markets the
answer?" Economics of Education Review 11(4):
439-52.
Center for the Study of Public Policy. 1970. Education vouchers. A report on financing elementary education by grants to parents. Cambridge, MA.
Chaikind, S., Danielson, L. C., & Brauen, M. L. 1993. "What do we
know about the costs of special education? A
selected review." Journal of Special Education 26(4): 344-70.
Chambers, J. G. 1981. "An analysis of school size under a voucher system." Educational Evaluation and Policy Analysis 3: 29-40.
Chambers, J. G. 1987. "Patterns of compensation of public and private school teachers." Economics of Education Review 4: 291-310.
Chubb, J., & Moe, T. 1990. Politics, markets, and America's schools. Washington, DC: The Brookings Institution.
Coons, J., & Sugarman, S. 1978. Education by choice. Berkeley, CA: University of California Press.
Deloitte, Haskins, & Sells. 1987. School transportation cost study
report prepared with School Services of
California for the Office of the Legislative Analyst, State of California,
Sacramento.
Encarnation, D. J. 1983. "Public finance and regulation of nonpublic education: Retrospect and prospect." In Public Dollars for Private Schools. Ed. T. James & H. M. Levin. Philadelphia: Temple University Press, 175-95.
Friedman, M. 1955. "The role of government in education." In Economics and the Public Interest. Ed. R. Solo. 1955. New Brunswick: Rutgers University Press.
Friedman, M. 1962. "The role of government in education." Capitalism and Freedom. Chicago: University of Chicago Press.
Glenn, C. L., McLaughlin, K., & Salganik, L. 1993. Parent information for school choice: The case of Massachusetts. Report No. 19. Boston: The Center on Families, Communities, Schools, and Children's Learning.
Hartman, W. 1981. "Estimating the costs of educating handicapped children: A resource-cost model approach--summary report." Educational Evaluation and Policy Analysis 3(4): 33-48.
Hoenack, S. A. 1994. Alternative designs of an educational voucher system in Minnesota. St. Paul: Choice In Education Foundation.
Hu T., & Stromsdorfer, E. 1979. "Cost-benefit analysis of vocational education." In Handbook of Vocational Education. Ed. T. Abrahamson, C. Tittle & L. Cohen. 1979. Beverly Hills: Sage Publications.
James, T., & Levin, H. M. 1983. Public dollars for private schools: The case of tuition tax credits. Philadelphia: Temple University Press.
Jencks, C. Winter 1966. "Is the public school obsolete?" The Public Interest 2: 18-27.
Lankford, H., & Wyckoff, J. 1992. "Primary and secondary school choice among public and religious alternatives." Economics of Education Review 11(4): 317-39.
Levin, H. M. 1968. "The failure of the public schools and the free market remedy." The Urban Review 2: 32-37.
Levin, H. M. 1983. Cost-effectiveness: A primer. Beverly Hills: Sage Publications.
Levin, H. M. 1991a. "The economics of educational choice." Economics of Education Review 10: 137-58.
Levin, H. M. 1991b. "Views on the economics of educational choice: A reply to West." Economics of Education Review 10: 171-76.
Lieberman, M. 1993. Public education: An autopsy. Cambridge: Harvard University Press.
Salzer, K. W. 1987. Evaluation and cost estimation of special education mediation conferences in California. Diss. Stanford University.
Sizer, T. R. 1967. "Reform and the control of education" (unpublished manuscript, Harvard University Graduate School of Education).
Sizer, T. R. January 1969. "The case for a free market." Saturday Review 93(1): 34-42.
State of California, Secretary of State. 1993. California Ballet Pamphlet: General Election. Sacramento, CA.
Tyack, D. B., & Hansot, E. 1981. "Conflict and consensus in American public education." Daedalus 110: 1-26.
U.S. Department of Education. Office of Educational Research and Improvement. . The Condition of Education. 1992. Ed. N. Alsalam et al. Washington, DC.
U. S. Department of Education. Office of Educational Research and Improvement. . 1993. Digest of Education Statistics, 1993. Washington, DC.
U. S. Social Security Administration. 1987. Annual Report to the Congress, 1987. Washington, DC.
Weiler, D. 1974. A public school voucher demonstration: The first year at alum rock. Santa Monica: Rand.
West, E. G. 1991a. "Public schools and excess burdens." Economics of Education Review 10: 159-70.
West, E. G. 1991b. "Rejoinder." Economics of Education Review 10: 177-78.
Witte, J. F. 1992. "Private school versus public school achievement:
Are there findings that should affect the
educational choice debate?" Economics of Education Review
11(4): 371-94.
Witte, J. F., Bailey, A. B., & Thorn, C. A. 1993. Third year report. Milwaukee Choice Program (unpublished report to the Department of Public Instruction, Milwaukee Public Schools, University of Wisconsin-Madison).
[A Call for Multiple Intrastate Studies] 
[Intrastate Cost Adjustments]
