Student-Level School Resource Measures

Robert Berne and Leanna Stiefel

Robert F. Wagner Graduate School of Public Service
New York University

The purpose of this paper is to discuss the types of student-level resource measures that would be preferred if student-level resource data are collected on a regular basis.

Introduction

Americans are obsessed with their public schools. They are especially focused on productivity- what student outcomes do their tax dollars buy? How can better student performance be obtained with the same or lower spending? If more money is to be spent, where will it do the most good? In response (or anticipation) to the public, analysts are studying the question of which resources and types of organizations are more productive than others, but a major problem with these studies is the unavailability of appropriate resource data. Either the data are collected for another purpose resulting in an incorrect unit of analysis, or the data lack sufficient detail. Data are most commonly available at the district level on a per pupil basis. However, these data are averaged over a great variety of students and programs and, as a result, they provide very imprecise measures of resources consumed by any particular student in a district. In addition, student-level data are often incomplete. There may be measures of average total expenditures or average experience of teachers, but seldom is there a relatively complete account of all the relevant resources received by an individual student.

The purpose of this paper is to discuss the types of student-level resource measures that would be preferred if student-level resource data are collected on a regular basis. In section one, the paper explores the types of questions that could be answered with student resource data. Section two reviews selective literature to show how analysts have approached answers to such questions with data now available. Section three discusses cost accounting concepts that are useful for thinking about how to collect appropriate student-level resource data. Section four recommends alternatives for NCES to consider. These alternatives take into account other related NCES data collection efforts, such as the National Education Longitudinal Study of 1988 (NELS:88) , National Assessment of Educational Progress (NAEP), and Common Core of Data (CCD). Section five concludes the paper.

How Would Student Resource Measures Be Used: Questions Student Resource Data Could Answer

There are three types of frequently asked questions whose answers require student-level resource data. The three questions concern production functions, equity of resource distribution, and intent of resource distribution. All three are asked by policy makers, the public, and analysts. This section describes the nature of the questions, and the next section reviews selective literature where they have been addressed.

1. Production function questions, with both the student and school as the unit of analysis: What is the relationship between outputs and inputs, especially school inputs? Several more specific questions require production function knowledge:

   • Resource effectiveness questions: Examples: Do additional resources for children lead to additional outcomes? What kinds of resources and resource use lead to the largest outcome gains? These questions can be at a general level, or for a specific program (professional development or mixed grade classes), or for a specific type of student (at-risk or bilingual).

   • Cost-effectiveness questions: What is the cost effectiveness of one program versus another, for example, reading recovery versus cross-age tutoring?

2. Equity questions: Examples: Within a district, for students in different schools, or for different students in the same school, what are the resources directly attributable to them and are they equitable? What resources are they consuming? Both attribution and consumption questions are important. A resource may be attributed (that is assigned or allocated to a student) but the student may or may not actually consume (receive) all of that re- source. The consumption question is even more difficult to answer than the attribution question. A key issue is whether student differences and resource differences can be matched to obtain valid and reliable vertical equity and equal opportunity measures.

3. Resource intent questions: For students who have been identified as entitled to specific resources (e.g., handicapped, bilingual, or compensatory education), do they receive additional resources? Does this match the intent of the financing system? This question can be addressed within schools and districts and across students, schools and districts.

These three types of questions all require some kinds of student-level resource measures, that is, resource measures that capture the variation across different students. All of these questions have been addressed in previous studies, not always with the appropriate data. New data collection efforts should attempt to address the concerns raised by each of the questions.

Selective Literature Review of Studies that Use Student-Level Resource Allocation Measures

The purpose of this section is to identify the kinds of resource data that have been used in previous production, equity, and intent studies, and to evaluate whether those data have been adequate for the research. A few well-known studies have been selected in each area and, therefore, this section is illustrative rather than comprehensive. We wish to both identify the need for student-level data and show how analysts have dealt with that need in the relatively recent past.

Production Function Studies

Table 1 summarizes the types of resource variables identified in the studies we reviewed. The categories of variables in Table 1 are taken from a recent study by Richard King and Bettye MacPhail-Wilcox (King and MacPhail-Wilcox 1994). The rest of this section gives a brief description of each article and our conclusions about resource variables based on the article.

One of the most "famous" recent production function studies is Eric Hanushek's 1986 review of 147 estimated equations in the Journal of Economic Literature (Hanushek 1986). From the point of view of types of resource data used, Hanushek found that most studies included three school inputs that are the largest determinants of instructional expenditures- teacher experience, teacher education, and class size. In addition, approximately 55 percent of the equations he reviewed included either teacher salary or expenditures per pupil. Because of additional evidence that some characteristics of teachers are significant determinants of changes in student achievement, Hanushek discusses studies that include teacher specific dummy variables as a way of getting at a "total teacher effect." For the purposes of this paper, it is important to note that most of the studies reviewed use data that are available in administrative records for districts or schools. Presumably more detailed data at the student level would be beneficial because they would allow researchers to explore whether individual components of school resources affect outcomes. At a minimum, some detail on teachers should be made available.

King and MacPhail-Wilcox (1994) recently reviewed the results of a large number of production function studies. They describe a wide variety of school inputs that researchers have used. Many of these are resource variables, or closely related to such variables. The variables are summarized in Table 1, with the identifiers K and M. For teacher characteristics, King and MacPhail-Wilcox note that years of experience, training, verbal achievement, and salary are the most often analyzed variables and that "teacher experience and teacher abilities bear a stronger, and more consistent, relationship with pupil performance on achievement tests than do other characteristics." (p. 53) For policy and administrative arrangements, the authors note that class size, pupil-teacher ratio, and ability grouping have been analyzed frequently. For classroom-based research, the authors write:

Researchers have argued for many years that studies would be improved if individual children and classrooms were the unit of observation rather than the school or district,... if resources were identified as those available to a specific child rather than by average resources in a school or district, and if processes were to include the quality and intensity of student-teacher interactions and time on task. (p. 59)

From this study, we conclude that researchers from different disciplines have emphasized different kinds of resource variables. Purely financial resource variables seem to have been taken from available records; more detailed classroom data are obtained by special, mostly one-shot, studies. Once again, it would be valuable for one source to combine as many of these different kinds of variables as possible and to combine them in an on-going way.

In an article that concludes "that when targeted and managed wisely, increased funding can improve the quality of public education," Ronald Ferguson (1991, 488) uses district-level data from Texas to estimate production function relationships. The resource variables he uses are summarized in Table 1, and are indicated by F. Regarding data availability, Ferguson begins his article:

Allocating resources efficiently and equitably in public primary and secondary schools has been an elusive goal. Among the primary reasons is the surprising scarcity of data appropriate for establishing the relative importance of various schooling inputs. As a result, recent research to discover how increasing spending might affect the quality of schooling and how improving the quality of schooling might affect how much children learn has reanalyzed old data or has relied on data sets that are very limited in size and scope. (p. 463)

Ferguson makes a case for his district-level data in Texas, but also notes the need for disaggregated data in some cases. We conclude that even when researchers are careful to put together a comprehensive and unique data set, they cannot always obtain resource variables at the correct unit of analysis.

In an innovative study, Byron Brown and Daniel Saks use detailed time allocations to students and subjects they study and their relationship to learning in reading and mathematics (Brown and Saks 1987). The authors explore technology and teacher values within the classroom. The resource data they use are summarized under Classroom-Based Research in Table 1, and are indicated by B and S. The study addresses the issues in this paper because it is one of the few that collects details at the student level and because it finds small but significant positive effects on learning when more time is spent on a subject. Thus, there may indeed be a need to collect such detailed data more systematically and over time, so that others can explore their effect.

Brown and Saks make the following (now familiar) comments on the type of data available for production function studies:

But even the best of the previous studies have suffered from one or more serious defects. First, the production functions have related output ... not to the inputs of teaching actually received by the student but to some average input available to the class or school. Because the allocation of inputs within schools and classrooms is itself highly variable, such data can provide highly misleading estimates of a school's underlying production technology... Second, previous work has been severely limited in its ability to deal with the consequence of heterogeneity of teachers and students.... Third, because the teacher is producing multiple outputs in the classroom... It becomes fundamental... to discover what the teacher is trying to accomplish in that classroom, to discover just what the teacher's objectives or preferences are. (p. 320)

Both through what they declare and what they find through the use of their data, Brown and Saks make a strong case for collecting data at a very micro level.

Finally, a recent article by David Monk examines the history (to the present) of production function studies in the context of the national interest in productive schools (Monk 1992). He notes that studies of process, often at the classroom level, seem to be less noticed than the more common production function studies that use more aggregated data. He also notes a recent trend back toward more aggregate data in published articles about production functions. Finally, based on his assessment of the history of production function research, he advocates a classroom approach, which would involve collection of student and classroom process data.

The pattern of inconsistent and largely insignificant results reported in this article points in a promising direction for future productivity research in education, and this direction involves raising the classroom to a higher level of importance in the conduct of productivity research. Thus, I am calling for a more disaggregated approach than has been characteristic of recent attempts to estimate production functions. I am also raising a concern over placing too much emphasis on school-level analyses, something that I believe has happened as a by-product of early effective schools studies. And I am arguing that more can be done with the economically oriented process studies that I reviewed earlier. My goal here is to motivate a classroom-oriented line of inquiry into education production that is deductively driven and that complements the already developed school-oriented studies. (p.320)

In conclusion, these authors, and the authors whose studies they reviewed, seem to agree either explicitly or implicitly on the following points. First, studies often use administrative data because it is available rather than because it is right. Second, data that are disaggregated to the student level are better than data that are averaged from the district level. Third, a wide range of resources might be relevant to the determination of gains in performance. The number depends both on the exact nature of the production relationship that is hypothesized and on empirical findings on what variables are significant.

Equity Studies

There is a long history of school finance equity studies, most of which have used district-level data. More recently, resource equity across schools is being examined, for example in court cases, such as the one in Los Angeles, or in school-based reforms, such as those in Chicago and Kentucky. We reviewed two studies that have used school-level data to look at variations across schools and ask if the data have been adequate to the task.

Berne and Stiefel (1994) analyzed the distribution of New York City budgets and expenditures per pupil by school and sub-district (community school district). They used published budget data and administrative data on expenditures that separated streams of funding into general education, special education, and reimbursable funding (e.g., Chapter I, state compensatory education etc.). They related the spending per pupil to the percent of pupils in poverty (by school or subdistrict) to assess one aspect of vertical equity.

Several problems with the data emerge. First, pupil counts for special education do not match funding streams. Second, although reimbursable funding can be attributed to schools, there is no way to know if the funding is spent on the children for whom it is targeted. Third, much of the district spending is not allocated at the school (or subdistrict) level. For example, fringe benefits, transportation, school lunches, and utilities are not allocated. A study such as this could be done with much more accuracy if there were student-level resource measures that were defined to be inclusive and to differentiate between kinds of programs and students. The data would be useful if it were gathered at the school level or, if it were a sample of individual student-level data that was representative at the school level. As part of this study of New York City school budgets, a model of disaggregated resource variables needed to be developed. Whether or not NCES collects such data for school districts, the existence of a model way to represent resource variables would help school districts produce their own data and analysis.

Lawrence Picus, as part of a larger project, studied school-level allocations using existing government data (the NCES Schools and Staffing Survey, 1987-88 and the U.S. Census Bureau's 1987 Census of Governments) (Picus 1993). Picus' goal was to look at patterns; he was not explicitly looking at equity. We place the study in the equity category because some of the patterns can be reframed to reveal answers about distributions that are generally thought about in terms of equity (distributions with respect to demographics or geography). The school is the unit of analysis in the Picus study and substantial effort was required to construct the school-level data. The merged data set contains considerable resource information, but as with the Berne and Stiefel study, more could be learned if there were a standard way of reporting school-level resource data. While Picus is not explicitly interested in student-level data, it would improve his analysis, because the school-level data in the study are averages over different kinds (and proportions) of students receiving different kinds of resources (e.g., some schools receive Chapter I funding and some do not).

We conclude from the school-based studies reviewed here that a well-defined set of student resource variables would improve equity studies at the school level including studies that use administrative data, particularly if those variables are capable of serving as models for other data sets. They would be useful if collected by NCES for equity studies only if they were representative of schools in a specific unit such as a district or state, or inclusive of all schools in those units.

Intent Studies

The four studies we review in this section have used unique cost collection methods for answering questions about how resources flow to programs or schools. Only one study (Chambers et al. 1993) is designed to explicitly answer an intent question. The other studies are classified as intent because they could be reframed to answer such questions.

The resource cost methodology (RCM) utilized by Jay Chambers and Thomas Parrish and based on an ingredients approach developed by Henry Levin (1983) is employed to study the use of Chapter I funding (Chambers et al. 1993) and the cost of alternative programs for students with limited English proficiency (Parrish 1994). The method uses a bottom-up approach that begins with the program and client of interest and assigns costs to those programs. RCM is data intensive and generally expensive to employ. In the two studies reviewed here the authors have used what they call a "purposive" rather than a random sample of districts or schools, in part because of the need to collect extensive data rather than using existing administrative records. Clearly this collection effort increases the cost of a study.

The RCM does provide one way to get accurate cost data by program or client. As described by Parrish (1994):

Essentially, the resource cost model system used in this study involves three steps: 1) disaggregating and listing the relevant set of service delivery systems or models required for any educational program, 2) determining the specific resources utilized in each delivery system, and 3) attaching prices to each of these resources to determine specific program costs. Overall and per pupil costs are determined on the basis of these programmatic standards and the number of pupils enrolled in each program. (p. 260)

If NCES were to develop a model way of collecting resource costs by students that included types of programs and types of funding for each student, researchers could approximate the RCM method for the students in the sample. However, the sample of students would need to be representative of the programs or funding sources (or chosen randomly with respect to these) in order to make generalizations. The review of this method is useful primarily because it emphasizes the collection of detailed ingredients of resources by program and illustrates the need to go further for some purposes than an aggregate average per pupil expenditure number.

Bruce Cooper and colleagues have developed a School-Site Allocation Model (SSAM) that is a cost accounting framework for obtaining costs by function at the school level (Cooper et al. 1994). They have developed and tested the model in 10 school districts across the United States. The model is capable of generating costs by type of school (elementary, middle, high school) by five functions for the school site and for the central office. The five functions are administration, facilities and operations, staff support and development, pupil support, and instruction. The article does not give estimates of the cost of collecting these data, but does note that school districts are required to "use the same definitions of cost items and place them into the correct functional categories- meaning that clarity, constancy, and accuracy are essential." (p. 71) Thus, like the Chamber and Parrish RCM, the data are most probably better than when obtained from normal administrative records, but the cost is high. And again, if NCES develops a prototype for student resource data collection, it may be easier for schools and districts to follow.

Yet another recent effort to find costs at the school level is the analysis by Coopers & Lybrand on behalf of the Mayor of New York City (Coopers & Lybrand 1994). Coopers & Lybrand label their model the School District Budget Model (SDBM). It crosswalks existing New York City Board of Education budget data into three classifications: functions, school type, and program category. The functions are: instruction-schools, instructional support-schools, operations-schools, operations-central and districts, pass-throughs, and debt service. The school types are: pre-k, elementary, middle school, high school, and non-school. The program categories are: special education, bilingual/ESL, other categorical, and regular education. In contrast with the RCM and SSAM, the SDBM collects no new data. Rather it finds ways to recategorize existing data. This is an important distinction, because it probably makes the SDBM less expensive per unit of data, but also less accurate. If the original data provide no clear indication of costs generated at the school versus the district or central level, then some assumptions need to be made to place the costs in those areas. The other two models must also make assumptions, but they are made earlier in the data collection process when categories of costs are created and schools or programs are instructed how to put each type of budget item into a category.

All three cost models reviewed here are efforts to obtain data by school program, or function data that are more disaggregated than those available from normal administrative records. All three are expensive efforts and there is some possibility that a comprehensive school resource definition by NCES could begin to lessen the expense by establishing a more well-accepted model.

We should also note that the three systems struggle with similar issues regarding allocation of overhead costs, central office allocations to schools or programs, and the appropriate level of disaggregation. In part their categories are driven by the purposes of the models. Cooper et al. are in part interested in determining what they call the "productivity" of various kinds of expenditures and, in particular, hypothesize that additional expenditures at the classroom level will be productive. As Coopers & Lybrand state, "Development of the SDBM evolved from the belief that the primary mission of schools is the direct instruction and support of students in the classroom. All other functions exist to support this basic mission." (p. 2). Chambers et al. and Parrish use their RCM to trace costs to programs or to find out how much supplemental spending there is on special programs. There is certainly no one right way to collect disaggregated data; methods depend on the purpose of the data collection.

Useful Cost Concepts

The cost accounting literature provides methods for measuring costs that are useful in decision making. In the context of this paper, cost accounting is valuable because it emphasizes the need to conceptualize the use of the cost data before the data are collected. This section reviews some of the key distinctions among types of costs for use in our assessment of alternative ways for NCES to construct student resource measures.

The distinctions described are:

• differences between departmental and product costing;

• the ways that product full costs can be subdivided (direct and indirect; variable and fixed);

• the ways that components of product full costs can be allocated to students (process versus job-order costing and methods in between);

• real resource versus dollar costs;

• the relevance of allocating components to students.

Departmental costing finds the costs of administrative units, such as agencies, departments, responsibility centers, etc. A primary purpose of departmental costing is to help managers administer units efficiently. In other words, decisions using departmental cost data are centered on how to perform the department's activities with an efficient use of resources. Product costing, on the other hand, finds the costs of producing various kinds of products. Some primary purposes of product costing are to appropriately price a product, to determine reimbursement levels, and to help decide whether a product should be produced in-house, obtained by outside contract, or not produced at all.

In our study, the concepts of product costing are relevant because the questions to be answered with the resource data are centered, for the most part, on the student (product) and not on the administration of the districts or schools (units) that "produce the education" for the students. We are interested in whether changing the way resources are allocated to students will change outcomes for the students, whether resource allocations to students are equitable, and whether students are receiving the resources that are intended for them. All three of these questions require ways to analyze resources linked to students rather than ways to assess the effective use of resources by particular organizational units. Naturally, the two questions of management effectiveness and student costs are related; but the distinction is useful when deciding how to design a resource measurement system.

Products are assigned their full costs if, after adding together the costs of all products in the organization, the organization's total costs are determined. Full costs can be subdivided in numerous ways, but two important distinctions are direct versus indirect and variable versus fixed. These are two different and not completely overlapping distinctions. That is, direct costs are not always variable and indirect are not always fixed.

Direct costs are those that can be assigned uniquely to one product. They are clearly incurred as a result of that product's production. Indirect costs, on the other hand, are incurred as a result of production of many products; they cannot be assigned, except by a rule, to just one product. They are often called overhead costs. When students are the product, instructional supplies such as textbooks, pencils, and writing pads are clearly direct costs while the time of administrators such as the superintendent's time is clearly an indirect cost.

Variable costs change as more units of the product are produced. For students, these are costs that increase when more students are added to a school. The amount of food for lunch is an example of a variable cost. Variable costs may occur in steps (semi-fixed), for example, when class size reaches the point that an extra aide or teacher is needed. In such cases, groups of additional students result in changes in costs. Fixed costs stay constant as more units of a product are produced. For students, the school's physical plant is a fixed cost (until capacity is reached). The cost of the principal is a fixed cost. The distinction between fixed and variable costs depends on the time frame and on the circumstances. Physical plant becomes a variable cost at the point when capacity is met or when the plant needs replacement. The cost of the principal becomes variable when more students require the hiring of assistants to help manage the school. Some costs have both variable and fixed components (semi-variable costs). For example, transportation costs for students involve a fixed component (the buses and drivers) and a variable component (the gasoline required to travel to pick up additional students). Again, the fixed costs become variable when the bus is full or needs replacing.

There are two systems for assigning direct costs to products. In reality some combination of the two systems is generally used, but when conceptualizing a costing system, the distinction between the two systems is useful. Job-order costing determines the costs of each individual unit of a product. A common example is the determination of the cost of hand-made, custom ordered furniture, where the specific material, and amount and kind of labor would be determined for each piece of furniture. Process costing determines the costs of groups of identical units and then divides by the number of units to obtain an average cost. A common example of process costing is the determination of the cost of a box of a certain kind of breakfast cereal. The time needed to track costs to each box would be great and since there is unlikely to be much difference between boxes, it is unnecessary for decision making. Job-order costing provides more accurate information; however, it is more difficult and expensive to collect. Many actual cost accounting systems are hybrids of job order and process costing.

Assigning resource costs to types of students could be done either way, although some process costing would probably be involved even when job-order system is dominant. For example, we could take individual students classified by type of school, type of special funding if any, type of academic program, etc. and then use a sample to determine as exactly as possible how much teacher time, other personnel time, and instructional supplies are used for the student in a period of time. However, there might be some direct costs, for example physical education teacher time, for which an assessment of time spent with each individual student would be unnecessary. These costs might be assigned to whole groups (classes or schools) of children and then divided by the number of children, producing process costing in this dimension. At the other end of the spectrum, process costing could be used at various levels of aggregation. For example, costs could be assigned by classroom (or grade, school, district, academic program, or type of funding) and then divided by the number of students.

Real resource costs are stated in terms of a resource's natural units. Teacher resources would be determined as positions or time per student; computers as numbers of computers or amount of time per student. The advantage of real resource costs is that comparisons across time or parts of the country can be made without worrying about the different purchasing price of dollars. For example, if beginning teachers of equivalent quality are paid differently in southern versus northeastern states, costs stated as positions or time will more accurately reflect what students receive. On the other hand, dollar costs have advantages. Dollars are a common measure; they allow us to aggregate different kinds of resources into one number. In addition, they can be adjusted for differences in purchasing power and if they are, they can reflect differences in quality in ways that counting positions or minutes cannot. For example, a more expensive teacher might also be a more effective one- and dollars can help indicate this.

Recommendations for NCES Student-Level Resource Measures

This section builds on the analysis of the uses of resource measures, the selective literature review, and the basic components of cost accounting systems to formulate recommendations for NCES as it considers whether to invest in the collection of student-level resource data. In addition to the review of the policy literature in section two, we also have reviewed various NCES activities that relate to the development of a student-level resource measure. First, Chapter 5, "Cost Accounting for Educational Programs," of Financial Accounting for Local and State School Systems, 1990 , presents an application of cost accounting to LEAs, and discusses uses, designs, and applications. Second, NCES has a successful history of student-level data collections, most notably the National Longitudinal Study of the High School Class of 1972 (NLS-72) , High School and Beyond (HS&B), and the National Education Longitudinal Study of 1988 (NELS:88). Both activities have implications for our recommendations.

Our recommendations should be viewed in a cost-benefit framework. While mindful that NCES has limited resources, a full assessment of alternative courses of action is clearly beyond the scope of this paper. The conclusion brings together our recommendations, with some consideration of costs, although it is a partial view from the NCES perspective.

The initial question that frames the recommendations on the development of a student-level resource measure is the purpose to which it will be used. This is particularly important here because the design of cost systems can vary according to their purpose. We identified three broad purposes in section two: effectiveness of resource use, equity, and intent. All three purposes could be served by a student-level resource measure given the questions facing our education system, current state of knowledge, availability of data, and potential benefits from a student-level resource measure. However, the question of effective resource use should be placed at a slightly higher priority than the other two purposes. It may be that an effort to develop a student-level resource measure can serve all three purposes simultaneously, but we suspect that choices will need to be made and need to take into account the higher priority purpose.

If the student-level resource measure will be used to assess the effectiveness of resource use, then the measurement system will need to link resources with student outcomes. That is, not only is it necessary to measure resources, usually in terms of dollar costs, it is necessary to know the different outcomes that are achieved by students so that the outcomes and costs can be linked. Note that NCES already has experience with complex data sets with the student as the unit of analysis, where outcome measures are included (e.g., NELS:88).

Based on this highest priority purpose, the effectiveness of resource use, NCES should move ahead to develop a framework for a student-level resource measure, regardless of whether the data are actually collected. This development should take place before a decision is made to actually collect the data to permit substantial input into its definition and design. Given the likely amount of data that will need to be collected and the expected cost of such an effort, there should be broad input into the development of this framework. Note that even if the actual data collection effort by NCES does not move ahead, the development of a framework could benefit other efforts at the state or local levels to collect data on student-level resources.

A student-level resource measure will need to build upon the existing financial accounting system. The financial accounting system produces the basic elements that will be categorized and combined into a cost accounting system. Thus, the student-level resource measure will need to take into account the basis of accounting (usually modified accrual), which means that there are not precise distinctions between operating and capital items, and the fund structure, which sometimes creates accounting-based divisions of expenditures that are not appropriate for a student-level resource measure.

The framework for a student-level resource measure will need to consider the following elements. First, will the measure be based on organizational units (departmental costing) or students (product costing)? Given the primary purposes for gathering the cost information, a student-based system (product costing) will be required. While an organizational model can provide information on certain divisions of resources, such as instruction versus non-instruction, without more specific links to outcomes at the student level, questions of resource effectiveness cannot be addressed at an appropriate level of disaggregation.

Second, will the student-based, student-level resource measure distinguish among the following concepts?

• Full versus partial costing. To determine resource use, it is desirable to obtain measures of full costs, not a partial measure that may exclude some "overhead" or other costs. The need for full costs is important because when these costs are broken down, the list will give analysts choices of resource definitions identical across students. If full costs are not gathered, then there is a risk that different types of overhead costs will be omitted by different schools or that the specific list an analyst would find valuable is missing. A full cost definition will need to be applied consistently in all settings. In several areas, (for example, teacher pensions, which may be partially or wholly state- financed), the state and local accounting systems may make inclusion- of full costs difficult.

• Direct versus indirect activities. We believe that this is an essential distinction to assess resource effectiveness, and to translate the results of the analysis into practical recommendations. Indirect costs must often be allocated by formula to students and thus will not always be useful to analysts. Also, with this distinction and the further delineation of types of direct and indirect costs, the effectiveness of different resources and in different combinations can be assessed. Any cost accounting system will be better (reliably and validly) able to specify direct costs compared to indirect costs. If the distinction is used, the data should also contribute to the debates over whether resources spent in the classroom, or on instruction, or for specific types of instruction (such as professional development), are more productive than other patterns of resource use. Note that the conceptual distinction may be useful but may be affected by the nature of the accounting system. For example, fringe benefits (health coverage, pensions, etc.) would be ideally defined as direct, but they may be accounted for in a manner (for example, at the state or district level) such that they are precluded in a direct costing method.

• Different educational programs. This would include, for example, regular instruction at elementary, middle and high schools, special education at different levels, vocational education, etc., as well as support programs such as school administration, district administration, facilities maintenance, transportation, etc. This distinction is important from both educational (how learning is organized) and organizational (how tasks are structured for management) points of view. All activities should be categorized in one (or more) programs.

• Process versus job order costing. Recall that process costing is used when the resource use does not vary across the units using the resources and the units themselves do not vary. For most educational issues, both the resource use varies (differential teacher time for different pupils) and the units vary (students are different from one another). Thus, to the degree possible, a student-level resource measure should employ a job order costing system to reflect the variation in resource use across different pupils.

• Alternative funding sources (e.g., local, state, and federal). From a policy perspective, this is an important- variable because the revenue distribution systems vary by level of government, but from a cost accounting point of view this may be one of the more difficult aspects of the resources to capture. At the teaching and learning level, resources from all sources are mixed, and in fact resources themselves cannot be examined to determine the funding source in the same way as they can for other distinctions we examine (program, direct versus indirect, etc.). To some degree if we learn about resource effectiveness, it is a somewhat lower priority to learn the source of funding, but given the strong influence of the intergovernmental system, it would be desirable to include this distinction.

• Methods for allocating indirect costs. Indirect costs, by their nature require cost allocation to obtain full costs. Key decisions in the indirect costing system beyond which costs are indirect include the basis for the allocation (students, square feet, teachers, etc.), and the method for the allocation (one step, two step, etc.). The basis and method of allocation should be made very clear; for many purposes analysts may want to omit the indirect costs if the formula is mechanistic.

• Capital versus operating resources. One of the shortcomings of LEA accounting systems, and all govern- mental accounting systems in general, is the imprecise treatment of capital costs. Ideally, some annual contribution of capital (for example the school building or the school buses) should be included in a student-level resource measure. Unfortunately, because accounting concepts such as depreciation are not used uniformly throughout school district accounting systems, capital items will probably require special treatment. While it is possible to think of using proxies for depreciation, such as annual debt service (principal and interest payments), one would need to proceed very carefully. It is quite possible capital is financed quite differently across districts in terms of the percentage that is debt-financed versus the percentage that is purchased with current funds.

Also, some states may help finance debt without accounting for the debt service flow through the district's financial records. Any cost accounting systems must be very careful to treat capital uniformly and to make sure capital purchases do not not cause or mask variations in other resources.

• Dollars versus real resources. In section three dollars and measures of real resources, for example, number of teachers, guidance counselors, etc. were compared. We concluded that due to the different aspects of resources that they each capture, it would be desirable to measure both dollars and real resources. It is likely that the inclusion of real resources along with dollars will increase the cost of the student-level resource measurement system, and while real resources probably have a slightly lower priority than dollar measures, we recommend that they be considered for inclusion.

• Variable versus fixed costs. Once the relationship between costs and outcomes are determined, from policy and management perspectives, it is useful to know whether the costs are variable or fixed. However, because the distinction between fixed and variable costs is dependent on the time frame used to assess costs, we deem this distinction to be a second order priority. Also, although direct (indirect) costs are not always variable (fixed), they often are, and analysts may be able to approximate variable (fixed) costs by using selected items of direct (indirect) costs.

Conclusions

NCES should move ahead with the development of a framework for a student-level resource measure. Whether or not NCES actually collects the data needed to determine student-level resource measures will depend on an agency-level cost benefit analysis. We believe that the case is strong enough for the necessary development to proceed and make the following recommendations:

1. Student-level resource measures should be designed primarily to answer questions of resource effectiveness and secondarily to answer questions of equity and intent. When conflicts arise, NCES should choose the resource effectiveness goal because questions of productivity are the most pressing for the public, policy makers, and researchers.

2. Development of a system to measure student-level resources should build on other NCES efforts such as financial accounting frameworks and longitudinal data bases.

3. Resource measures should be linked to student outcomes.

4. Measures should be student-based (product as opposed to departmental) costing ones and should include full not partial costs.

5. The measures should include direct and indirect costs, with detailed breakdowns in each category.

6. Measures that include costs of educational programs are a high priority.

7. There should be heavy reliance on job order costing where appropriate.

8. If possible, there should be - tions of alternative funding sources (federal, state, and local).

9. NCES should include the methods used to allocate indirect costs in the description of the resource measures- sures.

10. The measures should distinguish between operating and capital costs.

11. If possible, NCES should include measures of real resources as well as the dollar value of resources.

12. The measurement of fixed versus variable costs is a lower priority.

A student-level resource measure can be adopted to one of NCES's existing efforts, for example NELS:88. One of the crucial issues is whether the costing methodology needs to be carried out for every student in the district, because of the nature of cost allocations, or whether accurate costing methods can be developed based on a sample similar to the one in NELS:88. We believe that a sampling method is possible.

Finally, we want to emphasize the importance of developing a good set of student-level resource measures to accompany NCES data bases such as NELS:88. It is crucial that we begin to make progress on the question of effective use of resources in education. The NELS:88 data base is certainly one of the best existing data bases for researching that question, but without an appropriate set of resource numbers these data will go unused for this purpose.

References

Production Function Studies

Brown, B.W., and D.H. Saks. 1987. "The Microeconomics of the Allocation of Teachers' Time and Student Learning." Economics of Education Review 6: 319-332.

Ferguson, R.F. Summer 1991. "Paying for Public Education: New Evidence on How and Why Money Matters." Harvard Journal on Legislation 28: 465-498.

Hanushek, E.A. 1986. "The Economics of Schooling: Production and Efficiency in Public Schools." Journal of Economic Literature 24: 1141-1177.

King, R.A., and B. MacPhail-Wilcox. Summer 1994. "Unraveling the Production Equation: The Continuing Quest for Resources that Make a Difference." Journal of Education Finance 20: 47-65.

Monk, D.H. Winter 1992. "Education Productivity Research: An Update and Assessment of Its Role in Education Finance Reform." Educational Evaluation and Policy Analysis 14: 307-332.

Equity Studies

Berne, R., and L. Stiefel. Winter 1994. "Measuring Equity at the School Level." Educational Evaluation and Policy Analysis 16: 405-421.

Picus, L.O. March 1993. "The Allocation and Use of Education Resources: School Level Evidence from the Schools and Staffing Survey." Center for Research in Education Finance, Working Paper 37.

Resource Intent Studies

Chambers, J., T. Parrish, M. Goertz, C. Marder, and C. Padilla. January 26, 1993. "Chapter 1 Resources: Supplementing an Equal Base?" Final Report: final draft, submitted to U.S. Department of Education.

Cooper, B., and Associates. "Making Money Matter in Education: A Micro-financial Model for Determining School-Level Allocations, Efficiency, and Productivity." Journal of Education Finance 20: 66-87.

Coopers & Lybrand L.L.P. October 4, 1994. "Resource Allocations in the New York City Public Schools, draft." on behalf of Mayor Rudolph W. Guiliani and Herman Badillo, Esq.

Levin, H. 1983. Cost-Effectiveness: A Primer. Volume 4. Sage Publications, Beverly Hills, CA.

Parrish, T.B. Winter 1994. "A Cost Analysis of Alternative Instructional Models for Limited English Proficient Students in California." Journal of Education Finance 19: 256-278.

Other References

Finkler, Steven A. 1994. Cost Accounting for Health Care Organizations: Concepts and Applications. Gaithersburg, MD: Aspen Publishers, Inc.

Fowler, William J., Jr. 1990. Financial Accounting for Local and State School Systems, 1990.U.S. Department of Education. Office of Educational Research and Improvement. Washington, DC: National Center for Education Statistics.

U.S. Department of Education. Office of Educational Research and Improvement. National Center for Education Statistics. March, 1990. National Education Longitudinal Study of 1988. User's Manual. Washington, DC: NCES 90-482.

[Previous Paper]

[Next Paper]