Ask the Scholar

Document scope · 1 page
doc
Scholar
Ask about this object, its catalog metadata, its source description, or the page inventory. For page-specific OCR and visual context, open one of the page chats.

Scholar Source Context

Document identity
localId
351086345
label
"A Benefit-Cost Approach to the Prioritization of Rehabilitation Research" [1980]
core
doc
dtoType
document
pageCount
1
Source metadata
Source extras
naId
351086345
levelOfDescription
fileUnit
recordType
description
ocrSource
nara-archive
Single page context
seq
1
pageIndex
0
type
document
mediaId
5fd8e595df97dd81
ocrText
Originally Processed With FOIA(s): FOIA Number: S S FOIA MARKER This is not a textual record. This is used as an administrative marker by the George Bush Presidential Library Staff. Record Group/Collection: Donated Historical Materials Collection/Office of Origin: Frieden, Lex, Collection Series: Printed Materials Subseries: Papers/Books OA/ID Number: 52110 Folder ID Number: 52110-009 Folder Title: "A Benefit-Cost Approach to the Prioritization of Rehabilitation Research" [1980] Stack: Row: Section: Shelf: Position: A BENEFIT-COST APPROACH TO THE PRIORITIZATION OF REHABILITATION RESEARCH Report Grant HEW 12-P-59036/6-03 The Institute for Rehabilitation and Research Baylor College of Medicine September 25, 1980 Houston, Texas A BENEFIT-COST APPROACH TO THE PRIORITIZATION OF REHABILITATION RESEARCH Report Grant HEW 12-P-59036/6-03 Prepared by: David Cardús, M.D., Project Director Marcus J. Fuhrer, Ph.D. Robert M. Thrall, Ph.D. With the Assistance of: John H. Boynton, M.S. David S. Bunch, M.S. Sarah Taylor, M.P.H. September 25, 1980 Houston, Texas CONTENTS Page PREFACE - EXECUTIVE SUMMARY 1 PART I - BACKGROUND AND THEORY I.1 A PROCESS FOR ALLOCATING REHABILITATION RESEARCH FUNDS ..... 8 I.1.1 Formulation and Prioritization of Rehabilitation Issues/Problems I.1.2 Delineation and Analysis of Researchable Components I.1.3 Development of a Research Agenda I.1.4 Formulation of Project Concepts I.1.5 Prioritization of Project Concepts I.1.6 Choice of Project Formulation I.1.7 Prioritization of Competing Proposals and Selection for Funding I.1.8 Role of Budget Constraints I.2 THE BENEFIT-COST MODEL 16 I.3 BENEFITS AND COSTS OF REHABILITATION RESEARCH 18 I.3.1 Identification of Benefits I.3.2 Identification of Costs I.4 OPERATIONALIZING THE TERMS OF THE BENEFIT-COST MODEL 23 I.4.1 Assessment of Benefits I.4.2 Target Population I.4.3 Probability of Success I.4.4 Probability of Utilization I.4.5 Introducing Costs into the Benefit-Cost Ratio I.5 IMPLEMENTING THE MODEL 30 I.5.1 The Monetary Submodel I.5.2 The Non-Monetary Submodel I.5.3 Combining Monetary and Non-Monetary Benefits I.5.4 Using the Benefit-Cost Ratio to Prioritize Projects I.6 COMPUTER PROGRAM FOR MODEL APPLICATIONS 44 PART II - PROCEDURES II.1 ESTIMATION OF TARGET POPULATION SIZE 45 II.2 ESTIMATION OF MONETARY BENEFITS FOR INDIVIDUALS 50 II.3 SCALING OF NON-MONETARY BENEFIT DIMENSIONS 53 II.3.1 Procedure for Quantifying Individual Non-Monetary Benefits II.3.2 Procedure for Quantifying Non-Individual Non-Monetary Benefits II.4 ESTIMATION OF THE PROBABILITY OF SUCCESS 58 II.5 ESTIMATION OF THE PROBABILITY OF UTILIZATION 60 II.6 WEIGHTING BENEFIT DIMENSIONS 63 ii CONTENTS (Continued) PART II - PROCEDURES (Cont.) Page II.7 ARPCOM USER'S GUIDE 66 II.7.1 Accessing ARPCOM II.7.2 ARPCOM Instructions II.7.3 Using ARPCOM II.7.4 Data Inputting II.7.5 Conclusions GLOSSARY 76 REFERENCE LIST 77 APPENDIX 1 - THE ROLE OF THE BENEFIT-COST RATIO IN THE SELECTION OF ALTERNATIVE COURSES OF ACTION 78 1.1 The Expected net Benefit of a Research Proposal 1.2 Benefit-Cost Analysis which Considers Constraints 1.3 The Role of the Denominator in the Benefit-Cost Ratio 1.4 Illustrative Numerical Examples APPENDIX 2 - BENEFIT CLUSTERING 87 iii PREFACE EXECUTIVE SUMMARY : It is virtually inevitable that the National Institute for Handi- capped Research (NIHR) will be confronted with far more worthwhile pro- posals for rehabilitation research than can be funded. This situation will underscore the need for an approach to prioritizing proposed re- search that is reasonable, systematic, timely, relatively economical and open to the scrutiny of the rehabilitation community, the Admini- stration, and the Congress. This executive summary serves to introduce the scope of this report and describes briefly an approach to the benefit-cost analysis of proposed rehabilitation research, which is the primary thrust of the main body of the report. Particular emphasis is directed to the evaluation of research project concepts which, at a minimum, are understood as speci- fying: (1) a designated rehabilitation issue or problem, (2) problem- related gaps in knowledge, (3) an overall research strategy, (4) applicable research methodologies, and (5) estimates of benefits and costs. It is expected that research project concepts of this kind are likely to loom large in the formulation and implementation of NIHR's long-range plan for rehabilitation research. In 1971, a team of multi-disciplinary professionals was formed and funded by the Social and Rehabilitation Service to conduct a study entitled "Analytic Aids for Research Proposal Selection (AARPS)." The intent of this group, which includes specialists in rehabilitation research and operations research, was to construct a mathematical evaluation model that 1 would take into account both the monetary and the non-monetary benefits of rehabilitation research. This goal made it necessary to develop means of combining monetary and non-monetary benefits in the mathematical analysis, a goal that has been successfully achieved in the AARPS benefit-cost model. This executive summary serves therefore as an introduction to the model and to its specific administrative applications. A computer program has been proposed and partially developed to perform the routine calculations required by the model so as to facilitate its use. This program is more fully described in a later section of this report. The AARPS Benefit-Cost Model Following a lengthy theoretical analysis and through several math- ematical developments, the following expression for benefits which might be expected from a particular research project was constructed: The probability of success, Ps, is defined as the likelihood that the objec- tives of the project will be achieved, and this parameter varies between zero and one. The probability of utilization, PU, is the likelihood that one or more individual benefits will be utilized by those in the target popula- tion (N), either collectively or individually. The parameter N represents the size of the target population, which includes all of those individuals (or defineable groups, institutions, or other single units) which are seen to benefit directly from the project. The individual expected net benefit, B₁, is a measure of those advantages or conditional improvements which are expected to accrue to the individuals or units within the specified target population, less the costs of achieving those benefits. The indirect benefit, Bs, provides an accounting of those benefits and costs that depend upon project success but are not proportional to the size of the target popu- 2 lation. Similarly, the indirect benefit, BF, accounts for those benefits and costs which occur irrespective of project success or any target group. Rehabilitation professionals were surveyed systematically to deter- mine the possible benefits to be used in the model. The result was a list of 243 separate benefit factors which was subsequently reduced in number to 46 and then to 18 in an attempt to achieve a manageable number for the benefit-cost model. These 18 benefit factors were ultimately trans- formed into seven benefit dimensions, which are as follows: Monetary Benefits All benefits that can be represented in monetary units which accrue to either an individual, a group of individuals, one or more institutions, or society as a whole. Enhanced Quality of Services Those benefits exemplified by improved access to services, improved individualization of services, improved coordination among services or improved continuity of services. Improved Individual Client Outcomes Those benefits exemplified by minimization of functional limitations and personal disability, the encouragement of greater individual social participation, or improved vocational status and material well-being. Improved Administrative Bases for Service Provision Those benefits exemplified by improved management of information systems yielding timely and relevant administrative decision support, identifying operational constraints on operations and implementation, establishing more explicit procedures for program prioritization, and more effective sequencing of program development. Improved Policy Bases for Rehabilitation Those benefits exemplified by improved legislative impact and coordination of government entities; the development and communication of policies, plans, and procedures; and the facilitation of societal change. Indirect Benefits, Given Project Success Those benefits exemplified by expanded knowledge bases, the identification of new areas of research, and the spinoffs of technology and/or procedures. Indirect Benefits, Regardless of Project Success Those benefits exemplified by an enhanced public awareness of an issue or problem or a sustained effort focussed on a research problem. 3 To perform a benefit-cost analysis, all research project costs must be accounted for in using the above benefit equation. Two quantities can be defined to show how costs enter the evaluation process. The first quantity is the "expected net benefit," B, which results when all costs necessary to realize a particular benefit have been subtracted from the value of that benefit. All of the benefit terms (B₁, Bs, and BF) in the benefit expression are expected net benefits. An example of this quantity would be an environmental control device that might ultimately save the quadriplegic consumer $1000 but costs $200 to purchase and maintain. The expected net benefit would be $800. The second quantity is the "benefit-cost ratio," B/CR, and is the value of the total expected net benefits, B, divided by the first-year cost of research, CR. The cost of research is included, of course, in the total cost necessary to realize the total expected benefits (i.e. C*). Costs, therefore, can be seen to figure into the benefit-cost analysis as follows: total expected net benefits, B, are equal to total expected benefits, B*, less total expected benefit costs, C*, or simply B = B* - C*; and the benefit-cost ratio, B/CR, becomes the total expected net benefits divided by the first-year cost of research. Features of the Benefit-Cost Model Several unique features of the AARPS model are available when it is used for administrative planning and evaluation. The most significant is the ability to combine both monetary and non-monetary benefits. In the term Ps, full consideration is taken of the likelihood of failure to which research is vulnerable. For such assessments, peer-review is deemed to be critical. Both the direct and indirect benefits considered by the model are viewed as being contingent upon the estimate for probability of success. 4 When several different target populations are expected as beneficiaries of a project, it is also possible that the benefits differ from one population to another. With the separate term (N) in the model, the capability exists for considering these population differences, i.e., each separate benefit can be applied differentially to each specific population. In the same context, not all members of a target population will use, or take advantage of, each benefit to the same extent. Provision for the probabil- ity of utilization, PU, accounts for this estimated variation in usage and is the remaining parameter in the individual benefit product (PUNB₁). Historically, most approaches to prioritizing proposed research have only considered benefits accruing to persons to whom the research was directed. It was apparent to the AARPS group, however, that many research project concepts may benefit others besides the obvious beneficiaries, so the Bs term was added to the model and made contingent upon project success. The final net benefit, BF, was included in the model to account for benefits and costs that are independent of project success. An example is the cost of the research itself, since this is a "negative benefit," or benefit cost, which must be paid regardless of the success of the project. For an administratively more meaningful analysis of benefits and costs, certain "weighting" and "scaling" factors can be applied to one or more terms of the model. These apply not only to estimates for the probabilities and target populations, but also importantly to the combination of monetary and non-monetary benefits. The importance of the weighting factors is that the values and perspectives of decision-makers can be taken into account. These, and other mathematical concepts related to using the model, are ex- plained in much greater detail in the remainder of this document. 5 Administrative Aids to Using the Model A parallel activity has been the generation of a basic computer program which has the potential of being developed into a more sophisticated computer program that employs all possible features of the benefit-cost model. The present computer program ARPCOM, together with an available user's guide, can be used to compute the total expected net benefits according to the seven benefit dimensions. In other words, ARPCOM provides for the combination of monetary and non-monetary benefits. In addition, research costs are intro- duced and corresponding benefit-cost ratios are calculated. The framework for a more elaborate computer program, ARPSIN, is only envisioned at this writing, but if developed later by the AARPS group, it will become a supplement to this report. In conducting this project, the AARPS group has stressed the importance of first achieving a fully developed, idealized conceptual model and then considering the kinds of adaptations necessary for it to be introduced into administrative practice. This approach contrasts with developmental efforts that embrace practical expedience at the onset. The virtue of the AARPS group's approach is that, as adaptations are made to meet practical constraints, their significance, both conceptually and operationally, can be better assessed. Some of these adaptations are discussed in the body of this report; others have been envisioned by the AARPS group and could be refined after future interactions with the NIHR staff. Conclusions As an aid to the management task of prioritizing research project concepts and competing research project proposals, the AARPS group has developed a mathematical benefit-cost model that features multidimensional benefit parameters and has the capability for considering both monetary and non-monetary 6 benefits. The model provides for the consideratation of probabilities of project success and benefit-utilization, target populations and subpopulations for each benefit, and for direct and indirect benefits. The model also allows for benefits which may be realized even if the project is unsuccessful. Properly applied, the model can be a powerful tool for the administrator who must consider a multitude of worthwhile research projects, all competing for limited available funds. 7 PART I BACKGROUND AND THEORY I.1 A PROCESS FOR ALLOCATING REHABILITATION RESEARCH FUNDS It is generally recognized that a benefit-cost (B-C) analysis is a useful adjunct to decision making. A prerequisite to the application of benefit-cost analysis is the defin- ition of decisions to be made. The B-C model developed. by the AARPS group was intended to support the rehabilitation research management process and in particular to provide workable concepts and practical procedures for evaluating proposed rehabilitation research. Implicit in this work is an overall approach to research management which has grown out of a close working relationship with the RSA research management staff. The AARPS group believes that the proposed benefit-cost model and the management concepts to which it relates apply equally well to the National Institute of Handicapped Research (NIHR). In the following sections, the prioritization of rehabilitation research is considered from two key perspectives: (1) The array of management tasks that are inherent in the process of allocating research funds, regardless of the administrative mechanisms used, (2) The approach to research management which has been implicit in the AARPS work to date and which, we believe, bears upon NIHR's approach to the allocation of research funds. The component tasks of the research allocation process are discussed separately in the following sections and are duplicated in their sequential order in figure 1. A glossary of certain non-standard terms used in this report is presented at the conclusion of Part II. 8 FORMULATE PRIORITIZE ANALYZE/DELINEATE REHABILITATION REHABILITATION RESEARCHABLE ISSUES/PROBLEMS ISSUES/PROBLEMS COMPONENTS DEVELOP A FORMULATE PRIORITIZE CHOOSE RESEARCH RESEARCH RESEARCH PROJECT AGENDA PROJECT CONCEPTS PROJECT CONCEPTS FORMULATIONS BUDGET CONSTRAINTS PRIORITIZE SELECT ALLOCATE/ COMPETING OPTIMAL AUTHORIZE RESEARCH PROJECT(S) RESEARCH PROPOSALS FOR FUNDING FUNDS FIGURE 1. - PROCESS FOR ALLOCATING REHABILITATION RESEARCH FUNDS 9 I.1.1 Formulation and Prioritization of Rehabilitation Issues/Problems By legislative mandate, the research to be supported by the NIHR is to be strongly mission oriented. According to its own statement of functions, that mission is to " improve the lives of people of all ages with physical and mental handicaps, especially the severely disabled." The legislation is mute, however, regarding the full range of specific issues and problems that might be resolved by research in order to improve the lives of handicapped people and to support the rehabilitation service system. The long-range plan for rehabilitation research doubtlessly will provide the principal basis for issue identification. Hopefully, it will cover the domain of rehabilitation issues in a relatively balanced and comprehensive manner, indicating what are the problems and unmet needs. No less important will be the effort to specify the relative importance and urgency of the problems that are surfaced. I.1.2 Delineation and Analysis of Researchable Components Some of the concerns and issues that are identified will not be solvable by the provision of additional knowledge but rather by the redeployment of funds, new policy decisions, or the intensification of program efforts. Consequently, each high priority issue or problem must be submitted to competent analysis to identify components reflecting knowledge gaps that may be filled by systematic investigation. Some issues may be characterized by one predominant gap; others may consist of several different knowledge gaps. When performed systematically, this kind of assessment frequently takes the form of a state-of-the-art review specifying what needs to be known, what is known, and what is likely to be knowable given current knowledge and the existing investigative capabilities. As a number of such reviews becomes available, they can be 10 studied together to identify research possibilities common to two or more issues. Such possibilities for research are more attractive candi- dates for development than those pertaining to single issues. I.1.3 Development of a Research Agenda If performed adequately, the state-of-the-art review provides the critical groundwork for constructing a research agenda to acquire the needed knowledge. The agenda provides some notion of the order in which sub-problems must be addressed, i.e., which problems presuppose the solutions of others. An attempt is made to specify the alternative outcomes of component studies and to consider their implications for the design of succeeding studies. At each juncture an effort is made to specify the kinds of investigative resources that are required. The preparation of research agendas and their critical review for soundness and promise by duly qualified experts were not visible parts of the RSA approach to rehabilitation research management. Efforts by NIHR to establish a capability for research agenda planning should include an attempt to become familiar with similar efforts in other federal agencies (NCI, NET and NIMH in particular), and means should be explored of coordinating efforts of NIHR staff members and rehabilitation researchers throughout the country in agenda construction. I.1.4 Formulation of Project Concepts Having broadly identified how investigation is to proceed to fill the knowledge gap associated with a high priority issue or concern, it is possible to plan a specific research effort that implements the agenda. The resulting project concept may take a variety of forms, but at a minimum, it is likely to identify the rehabilitation issue or 11 problem to which it pertains, highlight an issue-related gap in knowledge, articulate an overall research strategy, characterize applicable research meth- odologies, describe how the knowledge to be gained can be disseminated and its use encouraged, and supply a cost estimate for conducting the research. I.1.5 Prioritization of Project Concepts The aggregate costs of the total array of project concepts will inevitably surpass the funds available to NIHR. The difficult decision must be faced of selecting a subset of project concepts that have two characteristics: (1) they are particularly meritorius, and (2) their aggregate costs conform to the total funds available. The AARPS benefit- cost model has been developed to facilitate this selection process because of the key role it plays in the overall management strategy. It should be noted, however, that only modest revisions of the operational procedures are required to assess investigator-initiated research proposals as well. I.1.6 Choice of Project Formulation Project concepts which have been chosen for funding must be developed further in the form of specific proposals for research. The proposal may be prepared by the agency staff in the form of a contract for which bidders will be sought nationally or in the form of an agency- solicited request for proposals addressed to a particular project concept. Some attention also must be directed to investigator-initiated proposals because of the prominent role these play in the funding allocation strategy of other agencies, including NIH and NSF. Were NIHR fully committed to the management strategy that has been outlined, investigator-initiated 12 proposals might still be condoned (but effectively discouraged) by requiring that these achieve priority for funding solely in terms of the extent of their relationship with high priority project concepts. Alternatively, a portion of the available funds for research and demonstration support can be laid aside specifically for investigator-initialted proposals and a prioritization process can be developed to choose those to be funded. In the present context, it was noted before that the AARPS benefit- cost model can, with only a moderate procedural revision, be made applicable to evaluating investigator-initiated proposals versus project concepts. The reason for this applicability is that investigator-initiated proposals can be viewed as containing two principal components: (1) the direct equivalent of a project concept (i.e., identification of a rehabilitation issue, specification of an issue-related knowledge gap, characterization of an over- all research strategy, etc.), and (2) the formulation of a specific research proposal aimed at addressing the project concept. The evaluation of such project proposals, as well as contract proposals and responses to agency- solicited proposals, is discussed in the following sections. I.1.7 Prioritization of Competing Proposals and Selection for Funding The AARPS model for evaluating competing project concepts has been adapted to assessing competing project proposals. In the related pro- cedures, a prominent role is assigned to peer review. The points discussed previously are essential components of a process for allocating rehabilitation research funds. Some of the linkages in this process can immediately be perceived as sequential lines of action. Others can be conceived as feedback loops which impart more accurate direction toward the desired goals of the whole process, as shown in figure 1. 13 I.1.8 Role of Budget Constraints The allocation of research support by NIHR must be managed with sensitivity to opportunities for research or demonstration that are delineated explicitly in the authorizing legislation (PL 95-602). That legislation authorizes, for example, research bearing on the rehabilitative needs of handicapped children, individuals aged 60 or over, rural populations, native americans, etc. Though line-item appropriations have not been associated with each problem area, the agency may well wish to assure that some effort is supported in most, if not all, such areas. One step that could be taken to foster the desired effort is to encourage the development of one or more well-formulated project concepts in each legislatively specified area. This development could be aided by an appropriate choice of staff members and external consultants who would be responsible for developing the concepts. However, if the benefit-cost model alone is used for funding decisions, it cannot be assumed that the project concepts in each and every area would compete successfully against the entire array of proposals. This situation requires the use of the benefit-cost model be coupled with a management strategy in which both maximum and minimum funding limits are assigned to each legislatively designated area. The project concepts for each would be formulated with cognizance of these limits, and at the same time provision could be made for their modification if it became clear that the required research costs must exceed the upper funding limit. In connection with this approach, the benefit-cost model might have two uses, one for the case in which two or more competing project concepts had been developed in the same designated area, thus necessitating prioritization; the other use would be to evaluate the project concepts in each designated area against the entire array of 14 proposals to estimate the degree to which the pre-assignment of funding levels was at the expense of achieving maximum benefit-cost outcomes. The following sections are particularly addressed to the development of project concepts and to the application of the model to prioritization. The development of the model and the operational procedures for evaluating project concepts has been the focus of the AARPS group's effort. See GLOSSARY, which follows Part II, for definitions of key words used in this report. 15 I.2 THE BENEFIT-COST MODEL The benefit-cost (B-C) model developed by the AARPS group to priori- tize rehabilitation research was derived from a theoretical analysis. The general formulation for the benefit portion of the model is: [1] where B = total benefits, Ps and PU = probabilities of success and of utilization, respectively; N = size of the target population; B₁ = benefits directly accruing to individuals or units; Bs = indirect benefits of successful research; BF = indirect benefits of funding. Equation [1 represents both the one-dimensional and multi-dimensional cases for each of the parameters shown. In the multi-dimensional case, each of the benefit terms (B₁, Bs, and BF) is a vector. The specific connotation of these terms as vectors, as well as a full explanation concerning the multi-dimensional aspects of the model, can be found in sec. I.3, Benefits and Costs of Rehabilitation Research. The complete form of the model includes, of course, the costs. There are various ways in which the costs may be included. For our purposes, the benefit-cost ratio we will use is: R = B/CR [2] where, for application to research prioritization, CR is the cost of research. Equation [2]is formally correct in the one-dimensional case only. In the multi-dimensional case, B is replaced by V (see section I.5.3). The application of benefit-cost analysis requires a careful consideration of the concepts of benefits and costs. When the benefit-cost ratio is used, it is important that there be a clear understanding of what goes to the numerator and what to the denominator. A benefit is generally defined as a specific advantage that one or more persons may realize as a result of some action. The action involved in this particular case is the performance of rehabilitation research. The benefits 16 gained by the utilization of the results of rehabilitation research may be increased personal earning power, reduction of expenses for medical care or less dependency on others. Such benefits can be measured in monetary terms. Benefits can also be expressed in terms of improved "quality of life," a greater sense of "well-being," or a more positive "role in society." These less tangible benefits are difficult to express in monetary terms but, nonetheless, exist and are taken into account in the benefit-cost model that the AARPS group has developed. As such, the model is an important advance over conventional benefit-cost models which ignore non-monetary benefits. 17 I.3 BENEFITS AND COSTS OF REHABILITATION RESEARCH In the AARPS benefit-cost model, three types of benefits are posited: (1) those serving individuals or institutions directly if the research is successful and utilized (B₁); (2) those serving individuals, groups and institutions or society-at-large, but indirectly and only if the research is successful (Bs); and (3) those serving any person or group of persons regardless of research success (BF). I.3.1 Identification of Benefits When the AARPS study began, an attempt was made to develop an exhaustive list of potential benefits that may be derived from rehabili- tation research. To accomplish this search, the administrative staff of the appropriate Federal agency (SRS) was asked to list all benefits they would consider important. After eliminating obviously redundant items, a list of 243 potential benefits was generated. Although the survey was certainly effective in regard to achieving comprehensiveness, the list was far too lengthy for direct use in the model. It was imperative, therefore, to reduce the potential benefit list. This reduction was accomplished by conducting a second stage of data collection, which was submitted to a cluster analysis. The benefit list was thus reduced to 23 personal and 23 non-personal benefits. Such a list makes it evident that each of the three types of benefits (B₁, Bs, and BF) is a multi-dimensionalterm in the benefit-cost model. The practicality of any benefit-cost analysis decreases substantially with an increase in the number of dimensions being considered. With a potential of 23 benefit factors for each type of benefit in the equation, the task of rating a group of projects could become prohibitive. The AARPS group sought, therefore, to reduce further the list of potential benefits into a 18 more manageable number of dimensions. A subsequent two-stage reduction analysis (Fuhrer, Cardus, and Rossi, 1979) resulted in the final list of benefit dimensions shown in table I. This benefit clustering process, together with a comprehensive chart of all benefit factors, is presented in Appendix 2. The reader is reminded that each benefit dimension in table I includes both benefits and costs (which may be considered negative benefits). As was described in sec. I.2, the benefit portion of the AARPS model is: B = Pₛ(P₁NB₁+Bₛ) + + + BF [1] This equation represents both the one-dimensional and the multi-dimensional case for each of the benefit terms, with the multi-dimensional condition meaning that all of the B terms are vectors. This concept is more readily understood when the vectors are broken down into their constituent components, each component collectively represented by variable subscripts: [3] In this expression, Bi is the expected net benefit for the i-th dimension. If all of the Bi's are expressed in the same units, they can be added together to get a single number (a value V) for the total expected net benefit. In the AARPS model, each of the seven benefit dimensions (table I) has: different units, and a slightly modified notation can be used to avoid confusion: [4] Thus, aseven-dimensionalvector, X, is generated by the model and is then used to produce another vector, X(M), in which all seven components are expressed in the same units (normally dollars). The procedure is described in section I.5.3. Equation [4]indicates there are twenty-four terms in the model: However, it is clear from an examination of table I that some of the benefit 19 dimensions are defined in a manner that would make several of the twenty- four terms have a zero value. For example, Improved Individual Client Outcomes (benefit dimension 3) would have a numerical value for B1,3, but would have zero values for Bs,3 and BF,3, by definition. In fact, Monetary Benefits (benefit dimension 1) is the only dimension that can have contributions from all three benefit categories (B₁, Bs, and BF). Each of the six remaining benefit dimensions has a contribution from only one out of the three. This orthogonality of benefits is an important aspect of the model and results from the benefit clustering process described in the previous section and in Appendix 2. The table below illustrates this important aspect and clarifies the correspondence between the vectors BI, Bs, and BF and the seven benefit dimensions (table I). The nature of the table below and of table I are described in Appendix 2. Benefit Dimension 1 2 3 4 5 6 7 Model BI BI,1 B₁,₂ BI,3, BI,4 B₁,5 - - Benefit Bs Bs,1 - - - - BS,6 - Term BF BF,1 - - - - - BF,7 20 TABLE I LIST OF BENEFIT DIMENSIONS AND BRIEF DESCRIPTIONS 1. Monetary Benefits- All benefits that can be represented in monetary units which accrue to either an individual, a group of individuals, one or more insti- tutions, or to society as a whole. 2. Enhanced Quality of Services- Those benefits exemplified by improved access to services, improved individualization of services, improved coordination among services or improved continuity of services. 3. Improved Individual Client Outcomes- Those benefits exemplified by minimization of functional limitations and personal disability, the encouragement of greater individual social participation, or from improved vocational status and material well-being. 4. Improved Administrative Bases for Service Provision- Those benefits exemplified by improved management of information systems yielding timely and relevant administrative decision support, identifying operational constraints on operations and implementation, establishing more explicit procedures for program prioritization, and more effective sequencing of program development. 5. Improved Policy Bases for Rehabilitation- Those benefits exemplified by improved legislative impact and coordi- nation of government entities, the development and communication of policies, plans and procedures, and the facilitation of societal change. 6. Indirect Benefits, Given Project Success- Those benefits exemplified by expanded knowledge bases, the identifi- cation of new areas of research, and the spinoffs of technology and/or procedures. 7. Indirect Benefits, Regardless of Project Success- Those benefits exemplified by an enhanced public awareness of an issue or problem, or a sustained effort focussed on a research problem. 21 I.3.2 Identification of Costs In order to achieve a maximal total expected net benefit for a given research budget, the benefit-cost ratio can be employed. To achieve this goal, the traditional benefit-cost ratio of total expected benefits to total expected costs (both with future terms discounted to present values) is not suitable. Instead, we use the more correct ratio of expected net benefits (total expected benefits minus total expected costs) to current-year research costs. Details of this modified ratio are given in section I.4.5 and Appendix 1. The essence of the distinction between the two is that what is significant to the decision-maker is the amount of the scarce item, i.e., current research appropriations, that is required for a given project or set of projects. 22 I.4 OPERATIONALIZING THE TERMS OF THE BENEFIT-COST MODEL The terms of the model include three types of benefits (B₁, Bs and BF), the target population size (N), two probabilities (Ps and PU) and the costs. The following discussion contains recommended guidelines for operational- izing each term of the model. I.4.1 Assessment of Benefits As was described in sec. I.3 all non-monetary individual benefits and costs (B₁,i for i=2,3,4, & 5) are subsumed in benefit dimensions 2, 3, 4, and 5. Non-monetary indirect benefits and costs, given project success (Bₛ,6), are subsumed in benefit dimension 6. Non-monetary in- direct benefits and costs, regardless of project success (BF,7), are subsumed in benefit dimension 7. All monetary benefits and costs (BI,1, Bs,1, and BF,1) are subsumed in benefit dimension 1. Of the seven benefit dimensions described, it is often easiest to estimate monetary benefits (see procedure in sec. II.2) since they can be totalled in identical units (dollars). When benefits are non-monetary, the measurement process becomes less conventional. To quantify a non-monetary benefit, some kind of utility scale must be devised which represents value, in a defined framework or from a professional viewpoint, of that benefit to those who would realize a gain from it (an individual client, a population of clients, an institution, etc.). Although this judgement of value is not as straight- forward as for monetary benefits, there are nonetheless several satisfactory procedures for estimating such values. These procedures are discussed in greater detail in the following paragraphs. Rehabilitation professionals can rate project concepts with respect to each benefit dimension, based on their knowledge and experience. Several subjective scaling methods have been developed which 23 could be used in assigning values to the non-monetary benefit dimensions. These methods include paired comparisons, ratio estimation (constant sum method), rank ordering (order of merit), equivalence, rating scales and the method of magnitude estimation. In selecting the method to be recommended, the following criteria have been applied: (1) the method should be readily comprehensible to raters who may not be particularly skilled in psychometric measurement; (2) written instructions should be sufficient to explain the method to the judges; (3) the entire judgement process should be able to be carried out by any one rater in a reasonable amount of time (e.g., a few hours or so); and (4) the method should be capable of producing a ratio scale. Using these criteria, the method of magnitude estimation was selected as the most appropriate for the purpose of scaling non-monetary benefits. Details about the procedure are given in section II.3. I.4.2 Target Population Each project concept, either explicitly or implicitly, specifies a set of individuals who are the intended potential beneficiaries. This set of persons is called the target population of the project concept. The World Health Organization (WHO) has proposed a system of class- ification on three levels that is very useful in describing target populations. The first level is that of impairment: "a permanent or transitory psycho- logical, physiological or anatomical loss and/or abnormality." Impairments frequently relate to clinical diagnoses, e.g amputation, paralysis due to polio, hypertension, etc. The second level consists of functional limitations: "the partial or total inability to perform (underlines added) those activities necessary for motor, sensory, or mental functions within the range and manner 24 of which a human being is normally capable, such as walking, counting, taking an interest in and making contact with surroundings." It is noted further that such limitations may last varying lengths of time, be permanent or rever- sible, progressive or regressive, but "should be quantifiable wherever possible." The third level consists of disabilities, described as follows: "Disability in which functional limitation and/or impairment is a causative factor, is defined as an existing difficulty in performing one or more activities, which in accordance with the subject's age, sex and normative social role, are generally accepted as essential, basic components of daily living, such as self-care, social relations, and economic activity. (under- lines added)." In terms of the WHO distinctions, target populations are frequently specified in terms of multiple criteria, i.e., in terms of several impair- ments or of a combination of impairments, functional limitations and dis- abilities. Thus, the target group for a new orthotic device to improve a stroke patient's gait might be defined in terms of a number of functional limitations, e.g. weakness of leg muscles or range of motion of the ankle and knee joint, as well as a number of functional abilities, e.g., mental ability to understand instructions required for use of the device and ade- quate visual/spatial perception. Seldom will survey data be available for the target populations of proposed rehabilitation research. The reason is that the available data are based upon very broad characterizations of populations, e.g., an inability to work or the occurence of injury to the spinal cord. Such general descriptors do not accord satisfactorily with the restrictive manner in which target populations of specified research proposals need to be defined. 25 A number of approaches are available for estimating a target popu- lation if the pertinent survey data are unavailable. One approach is to obtain global estimates from a number of experts familiar with the popula- tion in question. A Delphi-type process could be used to refine such judgements. Alternatively, the AARPS group has described an explicit, analytic process for generating target-population size estimates. As described in greater detail in section II.1, this process requires that an estimate be made of a parent population, i.e., one containing the entire target population as a proper subset or subpopulation. The relationship between estimates of parent and target populations is explained next. Estimation of target population size: Estimates of target populations can be generated if two requirements are met. First, national prevalence estimates concerning the pertinent parent populations should be available. Second, authoritative estimates based upon local experience must be available con- cerning the size of specific target groups relative to (e.g., a percentage of) the appropriate parent populations. With availability of these two estimates, it is possible to estimate the national prevalence of the target population. Consider, for example, a proposal regarding an innovative medication to eliminate urinary tract infections in spinal cord injured individuals. The two prerequisites would then be: (1) an estimate of the nationwide prevalence of spinal cord injury, and (2) an authoritative estimate based on local program experience of the proportion of such persons who have, or who are at risk of having, such infections. Variance of estimation for target groups can be assessed by obtaining information from multiple sources (see section II.1). 26 Implementation of this approach to target population estimation would place considerable emphasis upon NIHR becoming a repository for prevalence data bearing upon the various kinds of disabilities and upon functional limitations and impairments associated with disabilities. In addition, substantial responsibility would be borne by research proposers to seek out the best expertise available for estimating the size of relevant target populations. I.4.3 Probability of Success In evaluating project concepts, some means are necessary for esti- mating the extent to which each objective will be met. The estimated prob- ability of success of some project concepts may be quite low. It may be the case, however, that these are precisely the ones which yield the most substantial potential net benefits. This observation underlines the importance of benefits and probability of success being treated as independent values that are assessed separately. With this provision, high-risk projects can score higher than low-risk ones because the potential benefits are larger or because a larger target population is affected. Estimation of probability of success: It is critical that the judgements of the probability of success be rendered by individuals with expertise and demon- strated competency in the problem area being evaluated. Because the focus is on project concepts, however, the applicable criteria must differ from those that ordinarily pertain to the peer review of proposed research projects. For example, the peer review of project proposals often involves close scrutiny of the experimental design, data collection methods, and the investigator's experience in conducting such research. Specific information of this kind will 27 ordinarily be unavailable for project concepts. Instead, the investigative methodology will be characterized only in very general terms, e.g., the state of the art regarding supporting knowledge or technology, and the availability of experienced investigators and supportive facilities needed to address the problem. Operationally, two or more qualified judges can be asked to apply such criteria to each objective of a specific project concept. Each evaluator then would be asked to provide a single combined judgement for that project concept on a rating scale ( sec. II.4 has recommended procedure). I.4.4 Probability of Utilization As previously defined, the probability of utilization appearing in the AARPS benefit-cost model is the likelihood that a given benefit will accrue to the target population or subpopulation of either individuals or organi- zations. An important feature of the model is that the probability of utili- zation can be assessed in a benefit-cost analysis for each specific benefit and for each target population or subpopulation. Estimation of Probability of Utilization: Estimates of probability of util- ization may be obtained in a manner similar to that described for the prob- ability of success. However, the professionals who are called upon to pro- vide these judgements might well be of a different expertise. Whereas judgements provided by administrators at the institutional level could be instructive, a survey of practitioners (e.g., medical rehabilitation pro- fessionals or vocational counselors) or representative members of the target population might be more meaningful. A rating scale similar to that proposed 28 for estimating probability of success may be employed, with judgements being guided by criteria that include: relevance to agency goals, ease of under- standing, ease of implementation, accessibility to the outcomes of the research, the relative advantage of those outcomes compared to what is currently available, the compatibility of the research outcomes with existing values and norms, etc. (See section II.5 for a recommended procedure). I.4.5 Introducing Costs into the Benefit-Cost Ratio Costs are introduced into the benefit-cost model as follows: a. One calculates, and designates by B*, the total expected value of all benefits (see Assessment of Benefits, sec . I.4.1). b. One calculates, and designates by C*, the total expected value of all costs. c. One calculates, and designates by B, the difference of expected benefits minus expected costs. The project is considered, in principle, worthwhile if B is positive. d. One determines, and designates by CR, the current research funds requested for conducting the project. e. One calculates the benefit-cost ratio, R, using: R = B/CR = (B* - C*)/CR [5] bearing in mind that CR is part of C*, as discussed in Appendix 1. This ratio is the one used to rank the various projects according to merit and is therefore used by the administrator to make funding decisions in the presence of specified constraints, a process explained more fully in Appendix 1. 29 I.5 IMPLEMENTING THE MODEL The mathematical structure of the benefit-cost model can be better described by considering separately its principal parts. These are: the monetary submodel, the non-monetary submodel, the combination of monetary and non-monetary benefit dimensions and the derivations and use of the benefit-cost ratio. I.5.1 The Monetary Submodel The AARPS model requires that each project be evaluated from the standpoint of each of the seven benefit dimensions listed. in table I. (section I.3.1). Then, each project is represented by a set of values, constituting a vector X, whose components are x1 Monetary benefits x₂ Enhanced quality of services x3 Improved individual client outcomes x4 Improved service administration x5 Improved service policy bases x6 Indirect benefits, given project success X7 Indirect benefits, regardless of project success In this section, the procedure for obtaining x1 is explained. In section I.5.2, the procedure for dealing with the remaining benefit dimen- sions is discussed in detail. According to the small table in section I.3.1, B₁, Bs, and BF all con- tribute to the new parameter x1 (benefit-in this case monetary-expressed as a vector): x1 = PSPUNBI,1 + PsBs,1 + BF,1 [6] The term containing BI,1 represents the monetary gain realized by those individuals or units which have been designated as beneficiaries (target population) of the project. For this monetary gain to be realized, the 30 research must be successful and the results must be utilized. In general, the target population, N, must be partitioned into sub- populations T1, T₂, ,Tₖ overeach of which the expected net benefit and the probability of utilization per individual are approximately constant, so that the PSPUNBI,1 term in [6 Jis replaced by a sum of similar terms, one for each of the k subpopulations. The extent of the population refinement should be determined as a compromise between the precision desired and the costs of the analysis. As already mentioned, the benefits discussed above are all expected net benefits; that is, they have been aggregated over a number of years, the costs have been subtracted and the discount rates have been factored in. The calculations of individual monetary benefits can now be illustrated through equation [6]and its generalization with two appropriate examples. Example 1 Suppose that for the possible action of developing an improved wheel- chair, the estimated target population is 1000 and is composed of the follow- ing subpopulations differentiated by four levels of functional limitations: SI = those able to walk with some aid (cane, walker, etc.) S2 = those partially dependent on a wheelchair (e.g. during times of fatigue, pain, stress, etc.) S3 = those who are totally dependent on a wheelchair for mobility S4 = those who are immobile (i.e. bedridden). The monetary benefits of this action are attributed to restoration of earning power and more effective self-care. Some direct and indirect costs to be considered are: purchase and maintenance of the equipment, job training expense, and other employment-related expenses, such as new transportation costs. 31 To assess earnings over a lifetime, the target population should be further refined into age groups. Table II lists the number of persons in each functional limitation-age group (chosen in decades from 10 to 60). TABLE II DISTRIBUTION OF 1000 INDIVIDUALS IN FUNCTIONAL LIMITATION-AGE GROUPS(N&) Age Range, yr 10-20 21-30 31-40 41-50 51-60 SI 65 50 41 56 60 S2 31 45 52 98 126 S3 21 37 53 40 66 S4 5 18 29 44 64 To illustrate the estimation of the monetary portion of B₁, a sample beneficiary aged 35 from group S3 can be taken. His or her benefit would be a sum of terms, each of which is a product of three terms: (1) Bj, expected net benefits to age j, (2) Qj, probability of survival at age j, and (3) a discount factor (1+r)⁻ⁱ⁺³⁵. The sum runs from age 35 to age 80, with Qj = probability of survival* to age j =(0.98) Also assume that: Job training cost $1000 Ej Expected earnings $10,000/yr Savings in self-care $1000/yr Cj Maintenance fee $100/yr Other employment expenses $400/yr Price of wheelchair $2500 *This formula is for illustration only, since it underestimates survival to intermediate ages and overestimates survival at the larger ones. 32 Then Bj = Ej - Cj, where: Ej 10,000 + 1000 35 <j <65 1000 j >65 2500 + 1000 + 100 + 400 j = 35 Cj 100 + 400 35 <j < 65 100 j > 65 Adding the terms together, one gets BI = $90,000 as an average figure for the S3 and 31-to-40 age group. Another sample is taken for age 25 from S3 and the 21-to-30 age group, and so forth. The 20 values for individual monetary benefits are summarized in table III. TABLE III CALCULATED MONETARY BENEFITS FOR EACH FUNCTIONAL LIMITATION-AGE GROUP B₁, α , Monetary Benefits: in $1000 Age Group, yr 10 - 20 21 - 30 31 40 41 - 50 51 60 S1 125 130 120 100 90 S2 110 115 105 90 80 S3 98 100 90 80 70 S4 9 10 8 5 2 Table IV presents hypothetical probabilities of utilization for each subpopulation. Actual numbers could be based on estimates such as the per- centage of people who would actually buy these improved wheelchairs, how much they depend on this improvement for the purpose of employment, for self-care, etc. 33 TABLE IV WHEEL-CHAIR PROBABILITY OF UTILIZATION BY EACH FUNCTIONAL LIMITATION-AGE GROUP PU,α Values Age Group, yr 10-20 21-30 31-40 41-50 51-60 S₁ 0.01 0.01 0.01 0.005 0.005 S₂ 0.02 0.25 0.25 0.200 0.200 S₃ 0.40 0.45 0.45 0.400 0.300 S4 0.02 0.02 0.01 0.005 0.001 Suppose that Ps = 0.3. Substituting this value for Ps and using values for Nα,B₁ₓ and from Tables II, III and IV in the expression PsPU,α BI,1, , followed by adding the 20 products, yields an approximate value of PSPUBI,1 = $4,400,000. This value is the first component (x1) of the vector X for this project. The twenty population-group breakdown is unworkable in practice since it requires twenty assessments of target population size, probabilities of utilization and individual benefits. The work is extremely tedious, if not impossible, when a similar partitioning technique is applied to a large number of project concepts. Once again, some compromise between precision and practicality is called for. One modification under consideration is to reduce the number of subpopulations and use a single average number for individual benefi ts for all subsets sharing the same functional limitation. Having obtained a dollar value for the first item in equation [6], the remaining two terms must now be considered. The term Bs,1 contains monetary benefits and costs that accrue indirectly and only if the project is successful. Such benefits and costs are independent of target popu- lation size. 34 One example of the cost included in Bs,1 is the development costs associated with implementing the new wheelchair. It is assumed that the government will pay for such development in order to facilitate the availability of the expected new product. A manufacturer must be hired to do the necessary engineering and testing, to build a prototype and pro- duction-model wheelchair, and to perform the manufacturing design and tooling for full-scale production. Such activity could easily approach a cost of $100,000. In addition, a training program for wheelchair users, complete with a curriculum and any support materials, must be developed (with an estimated cost of $15,000). The final term, BF, contains the benefits and costs of funding the project regardless of project success. An obvious cost included in this term is the research cost in funding the project over a 3-year period. Assuming a yearly budget of $300,000 and a discount rate of 10 percent: BF,1 = - $300,000 - (300,000)/(1 + 0.1) - (300,000)/(1 + 0.1)² = - $820,000. Using all previous information for the three terms in eq. [6] the estimated net monetary benefits for the wheelchair are obtained: x1 = $4,400,000 - 0.3($115,000) - $820,000 = $3,550,000. Example 2 A project is proposed to develop a rehabilitation management system which would use new computer techniques. If applied to a state agency, the new system would eliminate 30 percent of the paperwork and ten percent of the administrative staff. This reduction in operational workload would have a significant budgetary impact over a ten-year span, but would also require a sizeable initial investment. 35 In the numerical representation of table V, the fifty state agencies (the target population for the project) are partitioned into subclasses. For political or financial reasons, the project is not applicable to 17 of the states. Values for Co (initial cost setup), Ej (annual savings); and Cj (annual maintenance costs) are listed for each subclass. The approach for developing the system has already been established as very reliable; therefore, Ps can be assumed to be one and, for convenience, the discount rate can be assumed to be zero. Table V contains the calculations for the individual monetary benefits. The total expected net benefits associated with the B₁,₁ term is $25.26 million. It is assumed that Bs,1 = 0 and that BF,1 = -CR, the cost of research, which is estimated at $260,000. Then X1 for this project woud be: X1= N 7 PSPU,aNαBI,1,a + = $25 million TABLE V Monetary Parameters ($ millions) Subclass Co Ej Cj = =10(Ej-Cj)-Co {α} 1 - - - - 0 17 - 2 0.17 0.15 0.05 0.83 0.3 10 2.49 3 0.17 0.15 0.05 0.83 0.7 6 3.49 4 0.27 0.20 0.05 1.23 0.5 4 2.46 5 0.35 0.25 0.05 1.65 0.5 7 5.78 6 0.35 0.25 0.05 1.65 0.8 4 5.28 7 0.50 0.42 0.05 3.20 0.9 2 5.76 Total 25.26 36 1.5.2 The Non-Monetary Submodel An approach to constructing utility scales for estimating non- monetary benefits is now described. Along with the analytic description a hypothetical example to illustrate the procedure is given. Suppose there are three competing project concepts (PCj) to be prioritized. PC1 is the action of developing an improved wheelchair (example 1, section I.5.1) PC2 is the development of a new computer system to be used in three state agencies (example 2, section I.5.1) PC3 a project that fulfills a congressional mandate. Here, the population breakdown is 20 populations for PC1, 3 populations for PC2, and one population for PC3. The size of each subgroup may be denoted by NJ, where the superscript j identifies the project and the subscript identifies the subpopulation. Then, for each benefit dimension (i=2,3,4, and 5): X is a sum of products one such product for each subpopulation), and Now, a set of judges (assume four) is asked to assess all of the non-monetary benefits for each subpopulation by the method proposed in section II.3. For example, in table VI judge 1 assesses the achievability of x₂, or "enhanced service quality," of the three projects (refer to table II for the population breakdown of PC1). 37 TABLE VI RATINGS GIVEN BY JUDGE 1 ON THE BENEFIT DIMENSION "ENHANCED SERVICE QUALITY" FOR THREE PROJECT CONCEPTS 1 1 1 2 2 1 1 1 2 2 PC1 2 2 2 3 3 3 3 3 4 4 PC2 55 55 55 PC3 70 The normalized values for judge 1 are given in table VII. Each entry in table VI is divided by 278 (the sum of all entries) to get the corresponding entry in table VII. TABLE VII NORMALIZATION OF RATINGS GIVEN BY JUDGE 1 (Table VI) 0.0036 0.0036 0.0036 0.0072 0.0072 0.0036 0.0036 0.0036 0.0072 0.0072 PC1 0.0072 0.0072 0.0072 0.0108 0.0108 0.0108 0.0108 0.0108 0.0144 0.0144 PC2 0.19784 0.19784 0.19784 PC3 0.2518 The purpose of the normalization is to give equal total weight to each judge while preserving his/her relative values. 38 I.5.3 Combining Monetary and Non-Monetary Benefits Once monetary and non-monetary benefits have been estimated and each project is expressed by a vector X, a procedure must be used to merge monetary and non-monetary benefits so that all benefit dimensions can be converted to the same unit (dollars). The process is as follows: Step 1. The j-th competing research project is represented by a vector X whose seven components are ratings on the respective benefit dimensions. Step 2. Each benefit dimension is assigned a weight which is an estima- tion of its relative importance as perceived by the responsible decision-maker. These weights define a vector 1, , as explained in section II.6. Step 3. Another vector Y is selected to represent all competing projects. The components of vector Y are the maximal values for each of the seven benefit dimensions of the project vectors xj. Step 4. The representative vector Y and the vector expressing the relative importance of the benefits ()) are compared and proportionality coefficients (mᵢ) are obtained to allow the conversion of all non-monetary benefit dimensions into dollar equivalents. An illustrative example is presented next: Step 1. Suppose we have five project concepts and suppose the five X vectors rating them are: 39 1 X x² x3 X4 x5 x1 4,400,000 1,000,000,000 26,550,000 850,000 1,100,000 X₂ 8,000 2,400 65,000 12,800 3,500 x₃ 53,200 25,000 3,333 41,350 9,900 x4 14,000 10,880 78,000 353'00 106,000 X5 828 1,100 3,000 9,500 1,500 x6 8 1 32 90 45 x7 11 0 18 55 88 2. Each benefit dimension is given a weight of relative importance (vector ). Let us suppose 1 0.50 X2 0.15 }3 0.20 A = 14 = 0.05 J 0.05 ¹₆ 0.03 ⁷₇ 0.02 Step 3. We select a vector Y whose i-th component yᵢ (for each i) is the largest of the i-th components of the vectors xj, with Y given by: 1,000,000,000 65,000 53,200 Y = 106,000 9,500 90 88 40 Step 4. We determine multipliers m1, m₂ m₇ SO that the products m1y1, m2y2,....,m7y7 are proportional to the relative importance weights Since the first dimension already has a well-defined unit of measurement (the dollar) it is taken as basic, and by setting m1 = 1, we can interpret each mᵢ as the induced dollar value of one unit in dimension i. (the formula for mi is Next, each vector X is transformed into a vector X³(M) defined by: [m1x1j XJ(M) = m7x7j For example, for project 1: 4,400,000 39,923,200 400,000,000 X¹(M) = 13,207,600 8,715,528 5,333,330 5,000,000 The components of this vector are the dollar values contributed by the seven benefit dimensions. Thus, the sum, v¹, of the components of x1 (M) is: v1 = 4.7358 X 10⁸ 41 and is the computed dollar value of project 1. The same procedure is followed to obtain total dollar values (vi) for the remaining projects. I.5.4 Using the Benefit-Cost Ratio to Prioritize Projects The problem of choosing projects among PC¹, PCⁿ for grant awards is obviously under the constraint of limited funds. Suppose the research cost for PCi is CRJ and the total amount of funds available is S. With the calculated data, vi, and the benefit-cost ratios Rj = VJ/CRi, we can formu- late the process of selection as an integer programming problem. max v¹z¹⁺ subject to CR¹ z¹⁺ +. CRⁿ Zn, S where zj = 1, if PCi is to be funded, or zj = 0, if PCi is to be rejected. One intuitive solution (see Appendix 1) is to arrange the PC! in order according to their benefit-cost ratios beginning with the largest RJ and then implement projects in order until the total resource S is allocated. Suppose, for example, that S = $1,750,000 and that the values vj, CR and RJ for the previous five project concepts are those listed: PC¹ PC² PC³ PC⁴ PC⁵ vj 4.7x10⁸ 1.22x10⁹ 4.863x10⁸ 5.89x10⁸ 2.755x10⁸ CRJ 3x10⁵ 1.05x10⁶ 8x10⁵ 1.05x10⁵ 6.5x10⁵ Rj 157.9 97.7 60.8 393 42.7 Thus, R⁴ > R¹> R². > R3 > R⁵, and CR 4 + CR¹ + = 1,500,000 < S CR 4 + CR 1 + CR² + CR³ = 2,300,000 > S Therefore, the final decision is to accept projects 1, 2, and 4. 42 There is an apparent defect in this "order and choose" rule. Suppose, for instance, S = $1,300,000. Then CR⁴ + CR¹ + CR² > S and hence only two projects, PC⁴ and PC¹ , should be funded, However, the implementation of PC⁴, PC¹, , and PC³ is within the range of total funds and implies more benefit than the previous selection of only two projects. This problem may be solved in either of two ways: (a) apply a computer algorithm to solve the so-called "knapsack" programming problem, or (b) allow the funds S or the research funding CRJ to be somewhat flexible SO that the project listed on the margin of total budget is also included in the funded set. 43 I.6 COMPUTER PROGRAM FOR MODEL APPLICATIONS Sections I.4 and I.5 describe the rationale and operationalization of the AARPS benefit-cost model and its special application to the prioritization of research concepts. By necessity, the model must have a certain degree of complexity. Since this complexity could be a deterrent to its utilization, a computer program can be used to facilitate its application. The computer program is being generated in two stages. In the first stage, a computer subroutine (ARPCOM) was developed which accepts as input data the benefit assessments (X vectors), the research cost for each project concept, and the relative-importance weights for the benefit dimensions. ARPCOM performs all of the necessary mathe- matical calculations and places in rank order the resulting benefit-cost ratios. This subroutine is capable of doing for large sets the same thing that was done manually for the small set of data shown in the illustration in section I.5.3. Instructions for use of ARPCOM are given in section II.7. For the future, a much more elaborate program (ARPSIN) is envisioned that will use as input data the estimations of probability of utilization, probability of success, target population, and benefits. The ARPSIN output will be the benefit vectors which are needed as input data for ARPCOM. ARPSIN development, pending funding, will be completed once experience in collecting and processing input data has been acquired. 44 PART II PROCEDURES The procedures included in this part are the instructions and methods intended for use by judges in attaining numbers for the terms of the AARPS model (i.e., Bₗ, Bs, BF, N, Ps, and PU) and the weights for benefit dimensions (¹ₛ) and to rate the judges, if desired. Judges are asked to give ratings only, which are used as input to a special computer program (developed in two stages, as discussed in sect. II.7). II.1 ESTIMATION OF TARGET POPULATION SIZE Figure 2 displays the successive steps in a process that begins with attention to the project concept's or research proposal's stated objectives. These objectives imply the target population of the research and therefore provide the basis for constructing this specification. As such, a target population is simply the total group that potentially falls within the focus of the project. Included are all individuals to whom the findings might apply. In addition to being consistent with the project's objectives, the target population must also be designated in satisfactorily operational language. For example, defining the target population as "handi- capped people" is hardly adequate unless the proposer can specify the opera- tions by which "handicapped" persons are to be identified. Having delineated the target population, the question becomes whether or not the national prevalence of the population and of its significant subpopulations have been established (Step 2, figure 2). For the overwhelming proportion of research proposals, satisfactory prevalence data will not be available. The reason for this void is that available survey data deal with broadly defined populations, such as "per- 45 STEP 1. STEP 2. STEP 3. ADOPT, DELINEATE SPECIFY DETERMINE WHETHER YES INDICATING PROJECT TARGET SATISFACTORY TARGET UNCERTAINTIES OBJECTIVES GROUP(S) POPULATION PREVALENCE OF ESTIMATING ESTIMATES ARE AVAILABLE NO STEP 4. STEP 5. STEP 6. IDENTIFY CRITICAL IDENTIFY ENUMERATE TARGET PARENT POPULATION ACCESSIBLE GROUP MEMBERS AND & ITS PREVALENCE POPULATION EXPRESS AS PROPORTION OF ACCESSIBLE POPULATION STEP 7. OBTAIN ESTIMATED TARGET POPULATION PREVALENCE BY MULTIPLYING OUTCOMES OF STEPS 4 & 6, ADJUSTING FOR INFERRED BIASES AND INDICATING UNCERTAINTIES OF ESTIMATING FIGURE 2. - SEQUENCE FOR OBTAINING TARGET POPULATION PREVALENCE VALUES 46 sons unable to maintain employment" or "spinal-cord injured individuals." On the other hand, the target populations of specific research proposals are of necessity defined in relatively restrictive terms. If an estimate of a target population needs to be generated anew, a critical parent population (CPP) must be designated (Step 4). The notion of a parent population is critical to understanding the CPP. A parent popluation is one that contains the entire target population as a proper subset, i.e., as one of several subpopulations. In principle, a given target population may have a large number of different parent populations, some hierarchically organized and others completely independent of one another. A target population of mobility-impaired rheumatoid arthritics has as parent populations rheumatoid arthritics generally (both with and without mobility impairments) as well as arthritics generally (those with both the rheumatoid and non-rheumatoid forms of the disease). Yet another parent population is "mobility-impaired persons," including those with impair- ments due to rheumatoid arthritis as well as to other impairments. The CPP is a parent population with two additional characteristics: (1) reasonable estimates of its national prevalence are available, and (2) it approximates an accessible population that can be studied to determine the proportion of individuals fulfilling the restrictive definition of the target population. The interplay of these two essential characteristics of the CPP can be illustrated by considering an example of proposed research with a target population consisting of persons with spinal cord injury who are considered to be candidates for receiving training in independent living skills. Assume, further, that the target population is defined specifically as 47 consisting of individuals who are wheelchair users, are from 16 to 50 years of age, have completed a comprehensive inpatient rehabilitation program and have expressed an interest in such training. The CPP for this kind of hypothetical target population might well be "persons with traumatic spinal cord injury," since a credible prevalence estimate of this population has been provided by the National Center for Health Statistics. Ideally, estimation of the target population's prevalence would proceed by drawing a statistically valid, random sample from this CPP. It is hardly likely that resources would be authorized to mount a properly designed national study of this kind. It might be realistic, however, to acquire data from a sample drawn from an accessible population of "persons with traumatic spinal cord injury" (Step 5, figure 2). In some instances, the accessible population will be presented by clinical records that can be studied retrospectively. The data may repre- sent the experience of a specific program or possibly the aggregate experience of several similar programs. For the traumatic spinal cord injury population, for example, aggregate data are being compiled by the National Spinal Cord Injury Data Research Center. The available record file may be exhaustively surveyed or randomly sampled with the purpose of determining the proportion of individuals conforming to the definition of the target population (Step 6). Some target populations are defined by attributes that are not routinely reported in clinical records. In such instances, the data must be acquired directly from a random sample of subjects drawn from an accessible population. This procedure might be necessary in the example being considered, since an expression of interest in receiving training in independent living skills is not likely to be documented routinely in clinical records. The feasibility of acquiring the needed data will depend substantially upon the 48 requirements for detecting or measuring the attributes in question. For the present example, a relatively economical telephone survey of appropriately selected respondents would probably suffice. The number of target population members in the sample may be expressed as a proportion of the accessible population, and that proportion may be applied to the prevalence estimate for the CPP to obtain the needed target population size estimates (Step 7, figure 2). It is important that the estimate be adjusted upward or downward depending upon judgements of the degree to which the accessible population represents the CPP. Degrees of CPP representation may arise, for example, because of differences in operational definitions. In the example being considered, traumatic spinal cord injury may have been identified in the CPP by means of health status interviews, but specified in the accessible population by means of a formal neurological examination. It is also useful to compare descriptive statistics bearing upon the CPP and accessible population. Of particular concern are differ- ences between the populations on variables that may interact with those critical to defining the target population. For the hypothetical project being considered, interest in receiving the independent living skills training may be correlated with the level of average personal income. Thus, evidence that the accessible population and CPP differ markedly in average income might be used as a basis for modifying the estimate based upon data from the accessible population. 49 II.2 ESTIMATION OF MONETARY BENEFITS FOR INDIVIDUALS In section I.5, it was noted that equation [3]applies to each subclass (subpopulation) of the parent population. We are now concerned with evaluation of the monetary component of the expected net benefit, BI,1' for an "average" individual member of the subclass. Noble (1977) sum- marizes benefits and costs of rehabilitation in table VIII (table 6 of that reference). Not all of Noble's listed items will apply to any given project. A well formulated project concept will indicate which benefits and costs are considered to be relevant to it and should also give estimates in dollars of their magnitude for the average individual. Recall that one criterion for subclass selection is approximate constancy of these dollar amounts from individual to individual within the subclass. Our definition of B₁,₁ includes use of an approved discount rate to bring all future benefits and costs back to present values. Thus, selection of a discount rate to be used in all benefit and cost analyses is a critical step in the whole evaluation process. Setting this rate is an important management prerogative. and responsibility. As the discount rate increases, an individual's age becomes less important. For example, for a 10 percent discount rate, values twenty years hence are divided by 7, and for 15 percent, by 16. With such rates, it becomes reasonable to use some average age in the calculation and avoid refinement of the target population into age groups as was done in the illus- trative example in section I.5.1 (see tables I, II, and III). Suppose that the target subpopulation includes a large current population as well as expected annual increments. Then, one discounts values for these incremental groups from their date of entry into the target population back to the present time. If the current target popu- 50 lation is N, if the annual increments all have approximately the same size (denoted by n), if the discount rate is r, and if the discounted expected net benefit for the average individual in the current target population is BI,1, then the total net benefit for the overall target subpopulation (present and future) is: (N +n/r)B1,1 [7] and for this subpopulation, the contribution to the economic component X 1 will be PsPu(N + n/r)B₁,1 [8] Use of equation [7] permits a substantial reduction in the extent of refine- ment that will be needed for the entire target population. The interactive computer program , ARPSIN, is planned to relieve the evaluator of the burden of the numerical calculations inherent in our benefit- cost model. The monetary component clearly involves more of both arithmetic and value judgements than are needed in the non-monetary components. Similar considerations apply to the other monetary terms, Bs,1 and BF,1 51 TABLE VIII BENEFITS AND COSTS OF REHABILITATION BY ANALYTIC PERSPECTIVES * Analytic Perspective Benefits & Costs Indiv./Families Employers/Pvt. Sector Gov't Society A. Benefit Increases 1. Earnings X a. .Net of taxes X b.Taxes X 2. Homemaker Services X X 3. Unpaid Work X X 4. Life Satisfaction X X 5. Family Member Earnings a.Net of taxes X b.Taxes X 6. Decreased Nursing, Med- ical & Custodial Costs (a + b + c = 1) aX bX cX X 7. Lower Turnover in the Labor Markets X X B. Costs 1. Case Service Expendi- tures (a + b + c = 1) aX bX cX X 2. Administrative & Over- head Costs (a + b = 1) aX bX X 3. Income Loss & Foregone Earnings in the Program a. Net of taxes X b. Taxes X 4. Research, Training, and Facility Costs (a + b = 1) aX bX X *From J. H. Noble, Jr., 1977 (p. 352). 52 II.3 SCALING OF NON-MONETARY BENEFIT DIMENSIONS The forms in this section provide raters with magnitude estimation procedures for rating a project concept according to each of the seven benefit dimensions given in table I. With slight adaptation, the first type of form is appropriate for rating each of the individual non-monetary benefit dimensions (numbers 2, 3, 4 and 5). The second type of form is designed for rating dimensions 6 and 7, which are non-individual and usually non-monetary in character. To rate the individual benefit dimensions, cognizance must be taken of a project's target populations, i.e., all individuals to whom the findings potentially apply. Since not all target population members may benefit to the same extent, provision is made for dividing target populations into subclasses that are relatively homogeneous in degree of expected benefit. The subclasses are termed Project-Target Populations (PTP's). The PTP pertaining to each project concept will be explicitly designated for raters. It should be noted that dimensions 6 and 7 are rated in terms of the overall project, not each of the PTP's. 53 II.3.1 Procedure for Quantifying Individual Non-Monetary Benefits (2 to 5) DIRECTIONS: The projects are to be evaluated in a specific sequence, so please do not rearrange them. You are being asked to judge the ex- tent to which a typical member of a Project-Target Population stands to benefit in terms of the benefit dimension specified below, assuming the project is successful and the results used. Identify PTP's in col.(2) 1. Identify yourself in the space marked "judge." 2. To proceed, rank order the Project-Target Populations (PTP's) associated with the first project. Assign the number "1" to the PTP that stands to benefit the most. Rank the remaining PTP's accord- ing to the extent to which you feel they stand to benefit. Write the rank you assign each PTP in column (3) below marked RANK. Go on to the next project, again assigning a rank of "1" to its top- most PTP, and ranking the remaining PTP's as you feel appropriate. 3. After you have ranked the PTP's within all of the projects, it is necessary to assign the PTP's points on a ratio scale. To do this, go back to the first project and give its topmost PTP (i.e., the PTP ranked "1") a rating of 100. Write the rating you assign in column (4) below marked RATING (100). Assign a rating of zero to all PTP's of this project to which this benefit does not seem to apply. Then take a second ranked PTP (assuming it has not been assigned a zero) and decide how much a typical individual would benefit relative to an individual in the PTP that has been assigned the score of 100. For example, if the benefit would be half as much, assign a rating of 50; if two-thirds as much, assign a rating of 66, etc. Do this for all PTP's associated with the project and go on to the next project. 4. Having done this rating for all projects, it is now necessary to compare the top-rated PTP across project concepts, i.e., those that received a rating of 100. First, find among all project concepts the top-rated PTP which you consider would benefit more than any other top-rated one. Assign this PTP a rating of 1000. Write the rating 54 you assign in column (5) below. Now search for the next most valued PTP that has been assigned a 100, and assign it a new rating between zero and 1000 that is proportional to the expected benefit of the PTP rated as 1000. As before, if you feel the benefits will be half or three-fourths as much (or any other proportion), assign the appropriate numerical value less than 1000 that reflects this judgement (e.g. 500 or 750). Do this rating for all the 100-rated PTP's. The scaling process is now complete (the computer program ARPSIN will establish the ratings of the remaining non-top-rated PTP's using the ratings you have provided once this program becomes available). EXAMPLE: Benefit Dimension No. 2 Enhanced Service Quality-Improved access to services, improved individualization of services, better coordination of ser- vices, improved continuity of services Judge PROJECT NO. PTP NO. RANK RATING (100) RATING (1000) (1) (2) (3) (4) (5) 55 II.3.2 Procedure for Quantifying Non-Individual, Non-Monetary Benefits (6&7) DIRECTIONS: 1. Identify yourself in the space marked "judge." 2. Rank all projects according to the extent to which each would provide benefits of the kind described by the factor specified below, assuming each project was completed successfully and the results were used. Assign the number "1" to the highest ranked project. Rank the remaining projects according to the extent to which you feel they would provide the benefit dimension described below. Write the ranks you assign in the column below marked RANK. 3. Next, it is necessary to assign the projects to a ratio scale. Assign the project ranked 1 a score of 1000. Assign a rating of zero to all projects to which this benefit does not seem to apply. Then take the second ranked project (assuming it has not been assigned a zero) and decide how much it would provide the benefit described below relative to the project that has been assigned to the rating of 1000. For example, if you feel it would yield only three quarters of the benefit provided by the first, then assign it a rating of 750. If you feel, instead, that it would provide only half the benefits of the first, assign a rating of 500. Write the ratings you assign in the column below marked RATING (1000) for all projects considered in the rating exercise. 56 EXAMPLE: Benefit Dimension No. 6 Indirect Benefits Given Project Success - Expanded knowledge bases, identification of new areas of research, spinoffs of technology/procedures Judge PROJECT NO. RANK RATING(1000) PROJECT NO. RANK RATING(1000) 1. 13. 2. 14. 3. 15. 4. 16. 5. 17. 6. 18. 7. 19. 8. 20. 9. 21. 10. 22. 11. 23. 12. 24. 57 II.4 ESTIMATION OF THE PROBABILITY OF SUCCESS The following forms represent an initial effort to provide judges with a procedure for rating probability of success for project concepts on an absolute scale. The probability of success should be judged by individuals with research experience and demonstrated competency in the problem area being evaluated. The following criteria should be used in rating : the probability of success. a. The state of the art regarding supporting knowledge b. The state of the art regarding technical knowledge c. Availability of qualified manpower and resources. The format suggested for quantifying the probability of success is presented on the following page. 58 Estimation of the Probability of Success DIRECTIONS: You are asked to rate the probability of success of the project concept based on the information provided by the project proposer. First, identify yourself and the project concept you are rating in the spaces provided below. Now, using the scale below, please indicate your assessment of the likelihood of successful completion of the project concept. Certain 10 Highly Probable 9 Quite Probable 8 Judge Somewhat Probable 7 Project Concept Slightly Probable 6 Equally Probable 5 Slightly Improbable 4 Somewhat Improbable 3 Quite Improbable 2 Highly Improbable 1 Impossible 0 59 II.5 ESTIMATION OF THE PROBABILITY OF UTILIZATION The accompanying form is an initial effort to provide judges with a procedure for rating the probability of utilization on an absolute scale. PU should be judged by administrators or staff within the agency and by potential users of the information generated by research (e.g., patients, practitioners, researchers, engineers, administrators, policy makers, and legislators). The following criteria were developed to be used in determining PU. These criteria are variably applicable depending upon the concreteness and specificity with which the outcomes of the research can be envisioned. Relevance: Importance of proposed research in relation to agency goals. Payoff: In relation to users in terms of improved efficiency/effectiveness. Ease of understanding: Degree to which an innovation is easily under- standable by prospective users. Ease of implementation: Degree to which an innovation is easily implementable by prospective users. Compatibility: With existing values, norms, past experience, and pro- cedures and facilities available to users. Accessibility: Degree to which an innovation is available to users. Observability: Opportunity for users to see a convincing and practical demonstration of the usefulness of the innovation. Trialability: Ease with which it can be tried out on a small scale before major or irreversible commitments are made. Credibility: Arising from evident validity of the findings and degree to which the researcher has won the trust of potential users Relative advantage: Over existing practices that will more than offset the effort of changing to something new. Utilization plan: Provision made to identify users, needs and to dis- seminate and promote utilization of research results. 60 To rate the probability of utilization, cognizance must be taken of a Project-Target Population (PTP), i.e., all individuals to whom the findings potentially apply as described in project benefits (section II.3). Since not all target population members may utilize the findings to the same extent, provision is made for dividing target populations into subclasses that are relatively homogeneous in degree of utilization. The subclasses are termed PTP's. The PTP pertaining to each project concept will be explicitly designated for the judges. The format suggested for rating the probability of utili- zation is presented on the following page, 61 Quantifying the Probability of Utilization DIRECTIONS: You are asked to rate the probability that the findings generated by the project concept will be utilized by each of the PTP's within each project concept. First, identify yourself and specify which project concept and PTP you are rating in the spaces below. Now, using the scale below, indicate your assessment of the likelihood that the specified target population will utilize project concept findings. Certain 10 Highly Probable 9 Judge Quite Probable 8 Project Concept No Somewhat Probable 7 Project-Target Population Slightly Probable 6 Equally Probable 5 Slightly Improbable 4 Somewhat Improbable 3 Quite Improbable 2 Highly Improbable 1 Impossible 0 62 II.6 WEIGHTING BENEFIT DIMENSIONS In choosing the most appropriate procedure to use in weighting benefit dimensions, it was decided that the method should allow for both numerical and graphical conceptualizations of weighting factors. It was believed that, although some raters may feel more comfortable using numbers, others might be more comfortable using graphs. A distinct advan- tage of using both methods is that discrepant judgements yielded by the two can be called to the judge's attention, thus affording the chance to resolve the differences. In the following example, weights are generated both numerically and graphically. The estimation part of the procedure is included in steps 1 to 4. The rater could then finish the procedure using a simple pocket calculator for computational steps 5 to 8. Alternatively, the rater could conclude the weighting analysis (steps 5 to 8) using the ARPCOM computer program (see section II.7), particularly as the number of weighting judgements increases. The suggested procedure for weighting benefit dimensions is presented on the following two pages (which may be removed or copied, as with other formats in this book, for evaluative use). 63 Weighting Benefit Dimensions DIRECTIONS: 1. Identify yourself as "judge." 2. Assign a number between 0 and 100 to each of the seven benefit dimensions below, indicating how much weight should be given to each benefit dimension compared to how much weight would be given to the other benefit dimensions. The actual value chosen does not matter. Only the value relative to the other numbers counts. For example, if you feel all benefits should be weighted equally, assign the same number (such as 1) to each. Or, if you feel that benefits of one kind are worth as much as all of the others combined, assign a 6 to that benefit and a 1 to each of the other six. Judge Benefit Dimension (see table I, p. 21) Weight 1. Monetary Benefits 2. Enhanced Quality of Services 3. Improved Individual Client Outcomes 4. Improved Administrative Bases for Service Provision 5. Improved Policy Bases for Rehabilitation 6. Indirect Benefits, Given Project Success 7. Indirect Benefits, Regardless of Project Success 3. Without looking at the numbers you have assigned, graph what you feel would be appropriate relative weights for these benefits. Do not estimate numbers, simply graph your judgments: EXAMPLE: Weight 1 2 3 4 5 6 7 Benefit Dimension So as to minimize distraction from your previous estimations, it is suggested you rate the benefit dimensions in a different order of numbers from those you assigned just above the graph (for example, 4-3-5-2-6-7-1). Although you may choose to rate these values in a different order, the numerical assignment of the benefit dimensions (i.e., numbers 1 to 7) must remain consistent. Now turn to the two scales on the next page (step 4). 64 GRAPH A GRAPH B 15 1.0 10 0.5 5 0 01234567 01234567 4. Now write in numbers on the scale at the left on Graph A. assume that the bottom of the scale is zero and note that the top of the scale. is fifteen (with intermediate points 5 & 10). Determine the number value for each column (bar) in your graph by using the scale at the left. Write the determined value above each column (bar). 5. Add the numbers which you have written above each column. Then divide each number by the total score and plot the results on Graph B above. Do this for all seven benefit dimensions. 6. Now, referring to the preceeding page, add together the scores assigned to the seven benefit dimensions. Next, divide each number by the total score and plot the results, again on Graph B. It is suggested that, if you used a pencil to plot in step 5, use a pen for step six, or vice versa. 7. If the two graphs on Graph B do not agree, decide what each value should be. Repeat any of the above steps if it's helpful. 8. Write your final weights for each benefit dimension in the appro- priate space below. These values should add up to 1.0 exactly. 1 2 3 4 5 6 7 The arithmetic for steps 5 to 8 is done by computer as part of ARPCOM. 65 II.7 ARPCOM USER'S GUIDE ARPCOM is an interactive computer program that calculates bene- fit-cost ratios for project concepts. The program obtains data from the user, performs calculations using the AARPS benefit-cost model, sorts projects in order of decreasing benefit-cost ratio, and prints the results. One purpose of ARPCOM is to demonstrate the decision-making utility of the AARPS mathematical model. The project-concept benefit assess- ments used as data in this program are generated by another, more-complex pro- gram (ARPSIN). ARPCOM illustrates how the decision-maker may generate administratively useful numbers, i.e., benefit-cost ratios, once the raw benefit data are available. II.7.1 Accessing ARPCOM ARPCOM is stored in the memory of the General Electric Mark III Foreground computer system under the ID of the AARPS Research Group. The computer is accessed by dialing the local GE computer network tele- phone number, and then coupling the phone to a terminal via a modem. The following modes are used: Modem - Full duplex Terminal - Half duplex Parity - Odd Baud - 1200 for CRT, 300 for regular terminal. II.7.2 ARPCOM Instructions Within five seconds of coupling the telephone to the modem-type "HHH," push the "return" button. The computer responds by printing the characters "U#=." Type in the user number, password, and project identification (separated by commas); then push return. The computer 66 prints "USED 2.32 UNITS" (the number may vary). The computer is now waiting for a command. To run ARPCOM, type "RUN ARPCOM" and push return. See Table IX, a sample ARPCOM RUN, at the end of this section. To escape from ARPCOM in the middle of a run, push the "break" button. This command aborts the program and returns control to the computer, which prints "READY" and waits for instructions. To run ARPCOM again, type "RUN ARPCOM." To sign off, type "BYE." II.7.3 Using ARPCOM ARPCOM is an interactive computer program. This term means that there is a constant "dialogue" between the computer and the program user, via a computer terminal. When the computer wants data or some kind of response from the user, it will prompt the user with a question mark (?). For example: ENTER THE NUMBER OF PROJECTS, NOT TO EXCEED FIFTY. ? The user types in the appropriate number, and then pushes RETURN: ENTER THE NUMBER OF PROJECTS, NOT TO EXCEED FIFTY. ? 12 R Note: In this user's guide, all user responses are underlined to differentiate them from characters printed by the computer. The symbol R means that the user is to push the "RETURN" button. In actual use, no underlining or R's will appear. When entering more than one item of data, all values are typed in consecutively on one or more lines, separated by commas: ENTER THE SEVEN BENEFIT ASSESSMENTS, AND THE ESTIMATED COST FOR EACH PROJECT. PROJECT 1 ?4400000, 839, 1140, 23, 5000, 90, 200, 2500000 R 67 In this case, the computer is looking for eight numbers. If too few numbers are entered, the computer will keep prompting until all data are entered: PROJECT 1 ?4400000, 839, 1140, 23, 5000 R ?90,200 R ?2500000 R If too many data are entered, the computer will use the first eight numbers and ignore the remainder. It should be noted that, since commas are used to separate data, they cannot be used in the conventional sense of signifying thousands. For example, 4.4 million is written "4400000" instead of "4,400,000." One important feature of ARPCOM is that it allows the user to correct his mistakes before proceeding to another part of the program, e.g.: IF ALL DATA ARE CORRECT, PUSH RETURN. IF NOT, ENTER THE NUMBER OF A PROJECT CONTAINING INCORRECT DATA ? 12 R RE-ENTER DATA FOR PROJECT 12 ? Once the above principles are understood, using ARPCOM becomes straightforward. The computer asks the user for the data it needs (see Data Inputting, next) and gives the user opportunities to correct any mistakes made in previous steps. The computer then calculates and prints the final results. For an example, see sample of ARPCOM RUN (table IX). The user may then change the Lambda data (next section), and rerun the 68 calculation step. This procedure enables the decision-maker to assess how sensitive the final results are to changes in Lambda (see sample ARPCOM RUN). II.7.4 Data Inputting ARPCOM uses the following input data: 1. Number of projects, which must not exceed fifty. 2. Project concept assessments, up to and including eight entries per project. The first seven entries are benefit assessments and the eighth is the project cost. The seven benefit dimensions are listed in table I and must be entered in the same corresponding order as in the table. In the future, the entire data set will be generated by ARPSIN for access by ARPCOM. 3. Lambda data, the understanding of which is crucial to using ARPCOM. The Lambda values are normalized weights indicating the relative impor- tance of the seven benefit dimensions. Thus, a "Lambda" is a normalized set of seven numbers. Here is an example: Benefit Dimension: 1234567 [100 50 60 30 15 25 5 = Benefit Weights This set of numbers was generated by the rater who was asked to indicate the relative importance of the seven benefit dimensions. In this example, and those that follow, the graphical technique for estimation des- cribed in section II.6 was not used. This rater believed that the economic benefits (dimension 1) were the most important and therefore arbitrarily assigned B1 a value of 100. B₂ was about one-half as important as B1, so the rater assigned B₂ a value of 50, continuing until each had a weighted value. While this procedure is a convenient method for assigning weights, these numbers cannot be used as they are. They must be "normalized," i.e., adjusted so that their sum totals to one. The result of normalizing the 69 above benefit dimension weights is referred to as "Lambda": [0.35 0.17 0.21 0.11 0.05 0.09 0.02]= Lambda Lambda is generated by adding up all seven numbers to get a total T and then dividing each of the seven numbers by that total (T). NOTE: The ARPCOM program performs all normalizations for the user, thus allowing input of benefit weights in the form initially shown. ARPCOM also allows the user to generate a composite Lambda. A com- posite Lambda is a value obtained by averaging a set of two or more indiv- idual Lambdas. This procedure allows the decision-maker to solicit bene- fit weights from a group of "experts" and then put them together to get a final, composite value representing all raters. The decision-maker may even wish to assign "weights" to each of the experts, depending on their perceived-expertise, as in the example below: Rater Benefit Dimension Weights Rater's Weights 1 [ 100 50 60 30 15 25 5] 100 2 [1000 300 500 200 200 100 100] 80 3 [ 5 10 2 2 3 1 5] 10 Lambdas 1 [ 0.35 0.17 0.21 0.11 0.05 0.09 0.02] 0.53 2 [ 0.40 0.12 0.20 0.08 0.08 0.04 0.08] 0.42 3 [ 0.18 0.36 0.07 0.07 0.10 0.04 0.18] 0.05 The composite Lambda = 1(0.53) + 2(0.42) + 3(0.05) = [0.36 0.16 0.20 0.09 0.07 0.07 0.05] 70 The previous example illustrates some important points. First, it is obvious why normalization is important in this instance. If the benefit dimension weights are not normalized prior to the final calcu- lation, Rater 2's weights would have exerted a disproportionate effect on the final results. Second, the example shows the importance of the decision-maker's role. Raters 1 and 2 believe that B₁ is more important than B₂, whereas Rater 3 thinks that B₂ is more important than B₁, yet the latter was assigned a lower weight by the decision-maker than the first two. The final result reflects this intention: the opinions of Raters 1 and 2 completely dominate those of Rater 3. ARPCOM gives the decision-maker both Lambda options: the rater may enter a Lambda of choice or may generate a composite Lambda from a set of individual Lambdas, along with the respective weighting values. This latter process is illustrated in the sample ARPCOM RUN at the end of this section (table IX). II.7.5 Conclusions ARPCOM is not a practical program in and of itself but is designed for demonstration purposes. In the future, project concept data for ARPCOM will be generated by another, more-complex computer program, ARPSIN (yet to be developed). In actual day-to-day use, entering all project data, as is required by ARPCOM, would be cumbersome indeed. A future version of ARPCOM will include an option to access a data file created by ARPSIN, yet the interactive portion of ARPCOM, which allows entering dif- ferent Lambdas, would remain essentially the same. 71 RUN ARPCOM ARPCOM 09:08EDT 07/01/80 WELCOME TO ARPCOM THIS INTERACTIVE COMPUTER PROGRAM CALCULATES BENEFIT - COST RATIOS FOR PROJECT CONCEPTS. PLEASE SEE THE ARPCOM USER GUIDE FOR COMPLETE OPERATING INSTRUCTIONS. ENTER THE NUMBER OF PROJECTS TO BE PROCESSED, NOT TO EXCEED FIFTY. 75 ENTER THE BENEFIT AND COST DATA USING THE FOLLOWING FORMAT: PROJECT x7x1,x2.x3,x4,X5.X6.x7.PROJECT COST (PUSH RETURN) WHERE X1 = MONETARY BENEFITS X2 = ENHANCED QUALITY OF SERVICES X3 = IMPROVED INDIVIDUAL CLIENT OUTCOMES X4 = IMPROVED ADMIN. BASES FOR SERVICE PROVISION X5 = IMPROVED POLICY BASES FOR REHABILITATION X6 = INDIRECT BENEFITS, GIVEN PROJECT SUCCESS X7 = INDIRECT BENEFITS, REGARDLESS OF PROJECT SUCCESS PLEASE LIMIT EACH ENTRY TO EIGHT DIGITS. PROJECT 17100000,22500,1000.500,250,250,300x350000 PROJECT 27500000,3000,1500,600,1000,300,200,475000 PROJECT 37267000,500,500.167.167.233,267.500000 PROJECT 47255000,2500,2750.5000,500250,0y600000 PROJECT 57325000,250.500,100.2000,45.500x9999 PROJECT DATA P# X1 X2 X3 X4 X5 X6 X7 PROJ COST 1 100000. 22500. 1000. 500. 250. 250. 300. 350000. 2 500000. 3000. 1500. 600. 1000. 300. 200. 475000. 3 267000. 500. 500. 167. 167. 233. 267. 500000. 4 255000. 2500. 2750. 5000. 500. 250. 0. 600000. 01 325000. 250. 500. 100. 2000. 45. 500. 9999. FIGURE 3. SAMPLE ARPCOM COMPUTER PROGRAM RUN 72 RUN ARPCOM ARPCOM 09:08EDT 07/01/80 WELCOME TO ARPCOM THIS INTERACTIVE COMPUTER PROGRAM CALCULATES BENEFIT - COST RATIOS FOR PROJECT CONCEPTS. PLEASE SEE THE ARPCOM USER GUIDE FOR COMPLETE OPERATING INSTRUCTIONS. ENTER THE NUMBER OF PROJECTS TO BE PROCESSED, NOT TO EXCEED FIFTY. 75 ENTER THE BENEFIT AND COST DATA USING THE FOLLOWING FORMAT: PROJECT x7x1.x2.x3,x4.x5.X6.X7.PROJECT COST (PUSH RETURN) WHERE X1 = MONETARY BENEFITS X2 = ENHANCED QUALITY OF SERVICES X3 = IMPROVED INDIVIDUAL CLIENT OUTCOMES X4 = IMPROVED ADMIN. BASES FOR SERVICE PROVISION X5 = IMPROVED POLICY BASES FOR REHABILITATION X6 = INDIRECT BENEFITS, GIVEN PROJECT SUCCESS X7 = INDIRECT BENEFITS, REGARDLESS OF PROJECT SUCCESS PLEASE LIMIT EACH ENTRY TO EIGHT DIGITS. PROJECT 17100000,22500,1000,500,250,250,300,350000 PROJECT 27500000,3000,1500.600,1000,300.200:475000 PROJECT 37267000,500,500,167.167:233.267:500000 PROJECT 47255000,2500,2750.5000,500,250.0,600000 PROJECT 57325000,250.500,100,2000,45x500,9999 PROJECT DATA P# X1 X2 X3 X4 X5 X6 X7 PROJ COST 1 100000. 22500. 1000. 500. 250. 250. 300. 350000. 2 500000. 3000. 1500. 600. 1000. 300. 200. 475000. 3 267000. 500. 500. 167. 167. 233. 267. 500000. 4 255000. 2500. 2750. 5000. 500. 250. 0. 600000. 01 325000. 250. 500. 100. 2000. 45. 500. 9999. FIGURE 3. - SAMPLE ARPCOM COMPUTER PROGRAM RUN 72 IF ALL DATA ARE CORRECT, PUSH RETURN. IF NOT, ENTER THE NUMBER OF A PROJECT CONTAINING INCORRECT DATA. 75 RE-ENTER DATA FOR PROJECT .73250,500,250,1000,2000.45.500.300000 IF ALL DATA ARE CORRECT, PUSH RETURN. IF NOT, ENTER THE NUMBER OF A PROJECT CONTAINING INCORRECT DATA.? PROJECT DATA P# X1 X2 X3 X4 X5 X6 X7 PROJ COST 1 100000. 22500. 1000. 500. 250. 250. 300. 350000. 2 500000. 3000. 1500. 600. 1000. 300. 200. 475000. 3 267000. 500. 500. 167. 167. 233. 267. 500000. 4 255000. 2500. 2750. 5000. 500. 250. 0. 600000. 5 325000. 500. 250. 1000. 2000. 45. 500. 300000. IF ALL DATA ARE CORRECT, PUSH RETURN. IF NOT, ENTER THE NUMBER OF A PROJECT CONTAINING INCORRECT DATA. ? IF LAMBDA IS COMPOSITE, ENTER THE NUMBER OF COMPONENTS, NOT TO EXCEED EIGHT. IF LAMBDA IS NOT COMPOSITE, ENTER 1. ?1 ENTER THE SEVEN BENEFIT WEIGHTS.7100,40.50.30.20.10.10 1.000E+02 4.000E+01 5.000E+01 3.000E+01 2.000E+01 1.000E+01 1.000E+01 IF THE BENEFIT WEIGHTS ARE CORRECT, PUSH RETURN. TO RE-ENTER THE WEIGHTS, ENTER 1.7 LAMBDA: 0.385 0.154 0.192 0.115 0.077 0.038 0.038 PROJECT BENEFIT DATA IN DOLLAR EQUIVALENTS 1 1.000E+05 2.000E+05 9.091E+04 1.500E+04 1.250E+04 4.167E+04 3.000E+04 2 5.000E+05 2.667E+04 1.364E+05 1.800E+04 5.000E+04 5.000E+04 2.000E+04 3 2.670E+05 4.444E+03 4.545E+04 5.010E+03 3.350E+03 3.883E+04 2.670E+04 4 2.550E+05 2.222E+04 2.500E+05 1.500E+05 2.500E+04 4.167E+04 0. 5 3.250E+05 4.444E+03 2.273E+04 3.000E+04 1.000E+05 7.500E+03 5.000E+04 PROJECTS IN ORDER OF DECREASING BENEFIT-COST RATIO PROJECT TOTAL BEN. COST B-C RATIO 5 5.3967E+05 3.0000E+05 1.799 2 8.0103E+05 4.7500E+05 1.686 1 4.9008E+05 3.5000E+05 1.400 4 7.4389E+05 6.0000E+05 1.240 3 3.9579E+05 5.0000E+05 0.792 FIGURE 3A. - SAMPLE ARPCOM RUN (Continued) 73 TO RE-RUN ARPCOM WITH NEW LAMBDA DATA, ENTER 1. IF NOT, PUSH RETURN. 71 IF LAMBDA IS COMPOSITE, ENTER THE NUMBER OF COMPONENTS, NOT TO EXCEED EIGHT. IF LAMBDA IS NOT COMPOSITE, ENTER 1. 72 ENTER THE SEVEN BENEFIT WEIGHTS FROM RATER 1.7100,50,3020101 ???? ?11111,111,1111 1.000E+02 5.000E+01 3.020E+06 7.770E+02 1.111E+04 1.110E+02 1.111E+03 ENTER THE SEVEN BENEFIT WEIGHTS FROM RATER 2.730,30,100,80,10,10,10 3.000E+01 3.000E+01 1.000E+02 3.000E+01 1.000E+01 1.000E+01 1.000E+01 IF ALL DATA ARE CORRECT, PUSH RETURN. IF NOT, ENTER THE NUMBER OF THE INCORRECT COMPONENT.?1 ENTER THE SEVEN BENEFIT WEIGHTS FROM RATER 1.7100,40,50,30,20,10,10 1.000E+02 4.000E+01 5.000E+01 3.000E+01 2.000E+01 1.000E+01 1.000E+01 IF ALL DATA ARE CORRECT. PUSH RETURN. IF NOT, ENTER THE NUMBER OF THE INCORRECT COMPONENT.? LAMBDA 1: 0.385 0.154 0.192 0.115 0.077 0.038 0.038 LAMBDA 2: 0.111 0.111 0.370 0.296 0.037 0.037 0.037 ENTER THE RATER WEIGHTS.750,50 5.000E+01 5.000E+01 IF THE WEIGHTS ARE CORRECT, PUSH RETURN. IF NOT, ENTER 1. ? NORMALIZED RATER WEIGHTS: 0.500 0.500 COMPOSITE LAMBDA: 0.248 0.132 0.281 0.206 0.057 0.038 0.038 PROJECT BENEFIT DATA IN DOLLAR EQUIVALENTS 1 1.000E+05 2.672E+05 2.064E+05 4.152E+04 1.437E+04 6.346E+04 4.569E+04 2 5.000E+05 3.563E+04 3.096E+05 4.983E+04 5.747E+04 7.615E+04 3.046E+04 3 2.670E+05 5.939E+03 1.032E+05 1.387E+04 9.598E+03 5.914E+04 4.066E+04 4 2.550E+05 2.969E+04 5.675E+05 4.152E+05 2.874E+04 6.346E+04 0. 5 3.250E+05 5.939E+03 5.159E+04 8.305E+04 1.149E+05 1.142E+04 7.615E+04 FIGURE 3B. - SAMPLE ARPCOM RUN (Continued) 74 PROJECTS IN ORDER OF DECREASING BENEFIT-COST RATIO PROJECT TOTAL BEN. COST B-C RATIO 4 1.3596E+06 6.0000E+05 2.266 2 1.0591E+06 4.7500E+05 2.230 5 6.6809E+05 3.0000E+05 2.227 1 7.3865E+05 3.5000E+05 2.110 3 4.9940E+05 5.0000E+05 0.999 TO RE-RUN ARPCOM WITH NEW LAMBDA DATA, ENTER 1. IF NOT, PUSH RETURN. ? TO RE-RUN ARPCOM, ENTER 1. IF NOT, PUSH RETURN.? PROGRAM STOP AT 1950 USED 18.67 UNITS BYE 00036.72 CRU 0000.43 TCH 0007.29 KC OFF AT 09:27EDT 07/01/80 FIGURE 3C. - SAMPLE ARPCOM RUN (Concluded) 75 GLOSSARY Benefits - Direct or indirect advantages accruing to a defined set of persons which include discounted monetary benefits and costs (negative benefits) Benefit Cluster - A galaxy of benefit factors, each related in a defined manner to a rather general system of benefit-oriented values Benefit-Cost Analysis - An analytic technique for evaluating entities with respect to their potential benefits and costs Benefit-Cost Ratio - Net expected benefits divided by the sum of critical costs Benefit Dimension - A general grouping of several related benefits or benefit factors which then allows the research evaluator or administrator to select specific benefits within the grouping for emphasis in a research evaluation Benefit Factor - A class of benefits from research which can be further broken down into specific and realizable benefits to individuals or groups Critical Cost - The limiting portion of total costs which come from the decision-maker's resources, such as a budget line-item for research Discount Rate - The percentage reduction in future monetary benefits and costs to make them compatible with current values Direct Benefits - Advantages from a project which accrue directly to persons Expected Net Benefits - Expected benefits, expressed numerically, with all discounted costs subtracted from discounted benefits on a probable distribution Indirect Benefits - Advantages from a project which accrue indirectly to persons Monetary Benefits - Advantages from a project which can be expressed in money units Multidimensional Parameter - A variable in an analysis which can be measured or considered on more than one value scale and therefore of one or more dimensions Non-Monetary Benefits - Advantages which accrue to either persons or groups but which are measured or estimated on a non-monetary value scale Parent Population - A group of individuals large enough to include all of one or more smaller target populations or subpopulations Prioritization - The process of ranking issues or projects according to relative merit Probability of Success - The likelihood that a project will fulfill its stated objectives Probability of Utilization - The likelihood that project benefits will be used as intended Project Concept - A planning proposal which identifies research issues, highlights research gaps, articulates research strategy, addresses information dis- semination, and embraces cost estimations where research funding is limited Ratio Scale - A scale where the zero point is prescribed and where only ratios are meaningful Research Project - An area of investigation that if funded may provide information or technology which apparently fills a void in the state-of-the-art Subpopulation - A subset of persons in a larger population (target population) defined by identifiable common characteristics Target Population - The total array of persons about which knowledge is to be developed by a research project or demonstration program Weighting Factor - A relative multiplying factor which indicates a chosen assignment of relative importance for one parameter or value with respect to others in the set. Usually, all weighting factors are made to sum to one. 76 REFERENCE LIST Allen, K.H. First Findings of the 1972 Survey of the Disabled: General Characteristics. Office of Research Statistics, Soc. Sec. Adm., 1977. Cardus, D. & Thrall, R.M. An Overview: Health and Planning of Health Care Systems. Preventive Med. 6:134-142, 1977. Cardús, D., Hammons, D.B. & Thrall, R.M. Multiple Objective Benefit-Cost Modeling for Decision Makers. Decision Information, Academic Press, pp. 73-93, 1977. Cardus, D. & Thrall, R.M. The Concept of Positive Health & Planning of Health Care Systems. "Health System Modeling and the Information System for the Coordination of Research in Oncology." Proceedings of the HASA Biomedical Conf., Dec. 8-12, 1975. D. D. Venedicton, Ed., pp. 211-233, 1977. Cardús, D., Fuhrer, M.J. & Thrall, R.M. Quality of Life in Benefit-Cost Analysis of Rehab- ilitation Research. Arch. Physical Med. & Rehab. In press. Dudek, R.A. Human Rehabilitation Techniques: A Technology Assessment. Texas Tech. Univ., pp. 1-56, 1975. Fuhrer, M.J., Cardus, D., and Rossi, D. Judgements of the Potential Benefits of Rehabil- itation Research. Arch. Phys. Med. & Rehab. 60:239-246, 1979. Fuhrer, M.J., Cardús, D., & Thrall, R.M. Estimating Target Population Size for Proposed Rehabilitation Research and Demonstration Projects. Arch. Phys. Med. & Rehab. In press. Nagi, S.Z. An Epidemiology of Disability Among Adults in the U.S. Milbank Memorial Fund Quarterly, 54:439-467, 1976. National Center for Health Statistics: Impairments Due to Injury, U.S., 1971; Vital and Health Statistics Series 10, No. 87; DHEW Publ. HRA 74-1514, Dec. 1973. National Center for Health Statistics: Limitations of Activity and Mobility due to Chronic Condition, U.S., 1972; Vital and Health Statistics Series 10, No. 96: DHEW Publ. HRA 75-1523, Nov. 1974. National Spinal Cord Injury Data Res. Center. Model Systems SCI Digest, Good Samaritan Hospital, Phoenix, Ariz., Spring, 1979. Noble, John H. Rehabilitating the Severely Disabled: The Foreign Experience. J. Health, Politics, Policy & Law. 14:221-249, 1979. Thrall, R.M. & Cardús, D. Benefit-Cost and Cost Effectiveness Analysis in Rehabilitation Research Programs. Methods & Information in Med. 13:147- , 1974. Thrall, R.M. Benefit-Cost Estimation, Alternative Requirements, Advantages and Disadvan- tages. Computer Applications in Health Care Delivery. Symposia Specialists, Miami, Fla., pp. 27-35, 1976. Thrall, R.M. & Cardús, D. Benefit-Cost Modeling in the Presence of Multiple Decision Criteria. "Health System Modeling & the Information System for the Coord. of Research in Oncology." Proceedings of the IIASA Biomedical Conf., Dec. 8-12, 1975. D. D. Venedicton, Ed., pp. 225-237, 1977. Thrall, R.M., Cardús, D., & Fuhrer, M.J. Multicriterion Decision Analysis. Am. Assoc. for Advancement of Science. In publication. Urban Institute: Report of the Comprehensive Service Needs Study. Wash., June 23, 1975. 77 APPENDIX 1 THE ROLE OF THE BENEFIT-COST RATIO IN THE SELECTION OF ALTERNATIVE COURSES OF ACTION The concepts "benefit-cost" and "cost/effectiveness" have led to considerable controversy in areas to which either or both have been applied, and the health care area is no exception. In the sense that a decision-maker weighs expected advantages and disadvantages before under- taking a course of action, benefit-cost assessments are almost universally accepted, as is the consideration of the most economical (cost/effective) way to implement a decision once it is made. The controversy relates to the acceptability of methods for measuring benefits, costs, and, in some appli- cations, "effectiveness." 1.1 THE EXPECTED NET BENEFIT OF A RESEARCH PROPOSAL The following discussion is concerned with: (1) the formal definition of the expected net benefits (ENB) of a research proposal, (2) the uses of benefit-cost ratios in the presence of budget constraints, and (3) the dis- tinctive features of the AARPS benefit-cost model concerning the selection of costs which appear in the denominator of the benefit-cost ratio. To set the stage for benefit-cost comparisons, we first consider a single research project P. If the research is successful and the results implemented, there is an associated train of benefits and costs sometimes running far into the future. Some of the costs are operational in nature and can be directly associated with one or more of the benefits; such costs can be interpreted as "negative" benefits. Other costs relate to funding of the initial research itself. 78 If a discount rate of r has been selected, then the total benefit B*(P) associated with P will have the form: B*(P) = B₁*(P) +(1/1 +r)B₂*(P)+ + [1.1] where Bt*(P) is the total dollar value of benefits expected in year t and N is the total time span of the project (in years). Sometimes, "r" is called the social discount rate. Similarly, the total cost, C*(P), is: C*(P) = C₁*(P) + (1/1 +t)C₂*(P)+ + .+[1/(1 + r)N⁻¹]CN*(P) [1.2] The difference B(P) = B*(P) - C*(P) [1.3] is the expected net benefit (ENB) of the research project P. It is difficult to justify undertaking a research project P unless its ENB is positive. If we recognize that the measurement of benefits and costs may not be precise, we might set some safety threshold T and require B(P) > T, instead of merely B(P) >0. We may be faced with a set of several alternatives (e.g., multiple research project choices) Pᵣ ,Pₙ and be asked to select that or those which are best. If we renumber the set in order of decreasing ENB SO that B(P₁) ≥ B(P₂) ≥ ≥B(Pₙ) [14] then, except for a tie, the choice of P1 is clearly optimal. More generally, if we could undertake some or all of the alterna- tives, it seems reasonable to reject any for which the ENB is negative (or below a selected threshold T) and in the absence of constraints (other than operational costs) to consider favorably the acceptance and implementation of all that remain. 79 1.2 BENEFIT-COST ANALYSIS WHICH CONSIDERS CONSTRAINTS In the presence of additional constraints, the situation is different. Suppose that the only constraint is a limitation on the number, S, of alternatives to be accepted. If the top "s" alternatives are all accep- table, we would maximize the total ENB by selecting P₁,..,P₃. Otherwise, we would select P₁,...,P r' where r is chosen so that Pr is acceptable but Pr + 1 is not. A more common and important type of constraint is provided by a limi- tation of one or more resources needed for implementation of the proposed courses of action. For example, if P₁,...,Pₙ are research anddemonstra- tion projects, then the total funds available to the decision-maker for implementing such projects may be a binding constraint. Suppose that to implement Pi, the amount of scarce resource needed can be measured in cost terms as CR(Pi), which is part of the total cost C*(Pᵢ). The constraint imposed by the scarce resources takes the form: [1.5] where C(S) measures the total amount of the scarce resource which is avail- able and where, for each i (i=1,...,n) : Zi=1 = if alternative Pi is implemented, and zi=0 = if alternative Pi is rejected. [1.6] The decision problem is then to choose the Zi so as to maximize the sum of the ENB's. B(P₁)z₁+B(P₂)z₂ +...+B(Pn)Zn [1.7] subject to equations [1.5]and [1.6]. Strictly speaking, this conceptualization is the integer programming exercise known as the "knapsack" problem* If the bound C(S) is slightly flexible, the solution is to (1) arrange the Pi in order according to the benefit cost-ratios: R(Pᵢ) = B(Pᵢ)/CR(P₁) [1.8] "Danzig, D.B. Linear Programming and Extensions. Princeton Univ.Press,pp.517-520,196 80 beginning with the largest ratio, and (2) implement projects in order until the total resource C(S) has been allocated. In more detail, suppose that the alternatives have been numbered so that: R(P₁) R(P₂)≥ ≥R(Pₙ) [1.9] Let "k" be the largest integer for which CR(P₁) + + CR(Pₖ) C(S), [110] thenset z₁ = ... = Zh, ²h+1 = ... = Zn = 0, where either h = k or is k + 1 (i.e., small overruns are acceptable). This method is known as "selection by the benefit-cost ratio." In the situation where C(S) is inflexible (e.g., costs incurred in emergency life-saving equipment), an exact solution of the maximization problem can be found by using more sophisticated mathematical algorithms (e.g., dynamic programming), but the naive selection of h = k gives a feasible solution, which under most circumstances might be acceptable. If there are other constraints, in addition to the budget limita- tion, selection via the benefit-cost ratio may need to be replaced by more comprehensive mathematical programming algorithms. However, if the addi- tional constraints involve some lower bounds to funding of special classes of research projects, the ratio can still be used with the following changes: a. Arrange the projects in each special class in descending order of their benefit-cost ratios b. Select from the top of the list just enough projects to meet the lower bound(s) of the additional constraint(s) c. Return to the full set of projects and delete those already funded d. Select from the top of those remaining until C(S) is exhausted. 81 1.3 THE ROLE OF THE DENOMINATOR IN THE BENEFIT-COST RATIO There is a number of alternative formulations of the benefit-cost ratio, and these are far from equivalent. The major source of variation is in the selection of the cost which appears in the denominator. For a research project P we would write C*(P) = CO(P) + CR(P) [1.11] Here, CR(P) denotes the "cost of research" and is defined as the amount of the current research budget C(S) that is required to fund P. The term CO(P) refers to "other costs." This term includes set-up costs, opera- tional costs and also downstream research funding for P. The traditional benefit-cost ratio discussed in most economics texts is the quotient B*(P)/C*(P) [1.12] of total expected benefits divided by total expected costs. This ratio exceeds one if, and only if, B(P) (the ENB) is positive. The equation: B(P)/C*(P) = [B*(P) - C*(P)/C*(P) = [B*(P)/C*(P)] - 1 [1.13] shows that, for comparing projects, the ratio, ENB/total cost, is equiv- alent to the traditional benefit-cost ratio. The ratio we have used before: B(Pᵢ)/CR(Pᵢ) [1.8 ] has a much smaller denominator and hence a greater value. We now provide some numerical examples to illustrate why[1.8]is considered superior to [1.12]or [1.13] Clearly, if two projects P₁ and P₂ differ only in some inessential details, they should be assigned the same (or nearly the same) value measurement. To illustrate this concept in more detail, suppose that some proposed project P₁ has total expected benefits B*(P₁) and costs C*(P₁). Let P₂ differ from P₁ only in that it includes (a) borrowing D dollars for one year at interest rate "r" equal to the social discount rate (see [1.1] above) and (b) investing those D dollars with a guaranteed return rate, also equal 82 to r. We may describe (a) and (b) as "wash items." Clearly, projects P₁ and P₂ are of identical merit. Now, suppose that B*(P1) = $1,000,000; CR(P1) = C*(P₁) = $200,000; and C = $10,000,000. Then, since the interest rate is equal to the social discount rate, the total expected benefit for P₂ is B*(P₂) = $11,000,000 and its cost is C*(P₂) = $10,200,000. The ratios then are B*(P₁)/C*(P₁) = 5 and B*(P₂)/C*(P₂) = 1.078. Clearly, in this case, the ratio of all benefits to all costs is not an appropriate measure for comparing the merits of the two projects. By contrast, for the ENB's we have B(P₂) = $800,000 and CR(P1) = CR(P₂) = $200,000 so that R(P1) = R(P₂) = 4, which agrees with our intuitive assessment that P₁ and P₂ have equal merit. However, there is more to the story. Consider a third project P₃ with B*(P₃) = $1,300,000 and CR(P₃) = C*(P₃) = $500,000. Now, B(P1) = B(P₃) = $800,000, so that if one used ENB as the sole criterion of merit, projects P₁ and P₃ would rate equally. Yet, to achieve this common ENB, project P₃ uses over twice as much of the research budget as does P₁. This fact is reflected in the benefit-cost ratios, since R(P1) = 4, whereas R(P₃) = 1.6. In this case, the traditional ratios 5 and 2.6 would also reflect the fact that, per dollar of research funding, P1 is more productive than is P3. Our use of the ratio ENB is proper if the overall objective is to use the total available current research funds to maximize the total ENB. The traditional benefit-cost ratio does not do this. 83 1.4 ILLUSTRATIVE NUMERICAL EXAMPLES Table 1.1 contains data for examples which further illustrate various features of the benefit-cost ratio. The first three cases have already been discussed. Projects P₄ and P5 differ only in the distribution of costs between CO and CR. The larger ratio R for P5 is responsive to its lower research budget. Let P₆ be a research project and P₇ a project modification based on the same research but involving additional downstream benefits and costs. For example, the addtional feature could be individual purchase and use of a newly designed wheelchair where the total added benefit ($600,000) was 1.5 times the added cost ($400,000). The tradi- tional ratio B*/C* penalizes P₇ and obscures its higher ENB per research dollar. Projects P₈, P9, and P₁₀ all have the same research costs but differ in the other costs and the resulting benefits. Project P₉ comes out on top with either benefit-cost ratio but they interchange the posi- tions of P₈ and P₁₀. Projects P₁₂ toP 15 are all variants of P₁₁; their rank under B/CR is P₁₃, P₁₅,P₁₁,P₁₄,P₁₂ and under B*/C* is P₁₃P₁₄₁,P₁₁,P₁₂,P₁₅ the low position of P 15 under B*/C* once again illustrates the fact that the traditional ratio not only penalizes projects with wash items but also those where additional downstream expenditures can result in desire- able downstream benefits. The last two projects, P₁₆ and P₁₇, in table 1.1 illustrate a situation in which neither ratio is completely satsifactory. 84 P₁₆ is a project which lasts only one year, whereas P₁₇ is the same project spread over two years (in this example we assume that the discount rate is zero, with half of the research funding in each year). The deferred half of the research costs are accounted for as "other costs." The result is that the one-year project is heavily penal- ized under the ratio B/CR. On the other hand, the ratio B*/C* gives no credit for deferring some of the costs until the following year. It is not difficult to modify the definition of CR so that it represents the present value of all research costs for a project. One would then use total (dis- counted) research costs for determining the ratios for new research starts and first-year costs to determine the cut-off point. If the funding process consists of first providing for continuing projects and then taking for C(S) what remains from the total research appropriation, then the use of current-year research costs for CR permits maximization of the total ENB for projects funded in the current year. However, if the funding process is done under a longer range view which has as its objective to maximize ENB over a period of years, then the in- clusion of total research costs in CR may be preferred. The choice among the definitions of CR is a policy decision which should not be made by the modeler. If, in fact, almost all projects have three-year durations, then it makes no essential difference which alternative is followed. We could use current year costs for CR and take care of a handful of one- year projects by giving their ratios an enhancing multiplier. It is important to recognize that the benefit-cost ratio principle is flexible enough to model whichever variant of research cost accounting turns out to be most appropriate, and if, as time progresses, changes are needed, they can be readily incorporated. 85 TABLE 1.1 BENEFIT-COST DATA FOR NUMERICAL EXAMPLES Project B* * B CR B/CR B*/C* P₁ 1,000,000 200,000 800,000 200,000 4 5 P₂ 11,000,000 10,200,000 800,000 200,000 4 1.078 P₃ 1,300,000 500,000 800,000 500,000 1.6 2.6 P₄ 100,000 50,000 50,000 30,000 1.67 2 P5 100,000 50,000 50,000 20,000 2.50 2 P6 600,000 200,000 400,000 100,000 4 3 P₇ 1,200,000 600,000 600,000 100,000 6 .2 P₈ 400,000 200,000 200,000 100,000 2 2 P₉ 600,000 250,000 350,000 100,000 3.5 2.4 P₁₀ 600,000 350,000 250,000 100,000 2.5 1.71 2,000,000 800,000 1,200,000 480,000 2.5 2.5 P₁₁ 2.100,000 850,000 1,250,000 530,000 2.36 2.47 P₁₂ 2,250,000 850,000 1,400,000 530,000 2.64 2.65 P₁₃ 2,150,000 850,000 1,300,000 530,000 2.45 2.53 P₁₄ 2,350,000 1,000,000 1,350,000 530,000 2.55 2.35 P₁₅ 500,000 200,000 300,000 100,000 3 2.5 P₁₆ 500,000 200,000 300,000 50,000 6 2.5 P₁₇ 86 APPENDIX 2 BENEFIT CLUSTERING As stated in the body of this report, an original comprehensive list of benefits numbering 243 separate items was obtained from administrators of the Social and Rehabilitation Services (SRS) agency in 1973. This list was obtained at the beginning of a Delphi-like study in which 50 individuals were asked to participate. Of those 50, 42 (84 %) responded to the initial query. The result of the preliminary benefit query produced 119 personal benefits and 124 non-personal benefits, as shown in figures 2.1 and 2.2. Clearly, comprehensive as they were, the two lists contained an unwieldy number of items to incorporate into a benefit-cost analysis di- rectly, so a data-reduction process was undertaken in a later stage of the Delphi query. In what was the beginning of a "clustering" process, the participants identified similarities among the items within each list. The similarity judgements were then submitted to Johnson's hierarchical cluster analysis to identify groups of similar benefits within each list. The process starts with considering each element as a cluster and then merges those elements which are closest according to the predefined metric. The next closest elements or clusters are then identified and merged to form new clusters. The process can be stopped at any stage and continued until all elements or clusters have been merged into a single cluster containing all of the original elements. In the initial Delphi-type study, the analysis reduced the original lists (119 and 124) to a total of 46 clusters, 23 each for personal and non-personal benefits. These two sets of 23 "benefit factors" are also shown in figures 2.1 and 2.2. In a later Delphi-type study, participants were, first asked to judge each of the 46 benefit factors to determine if any should be 87 deleted as superfluous or not as meaningful as others on the lists. Only two items were removed as being of less importance (Child Welfare and Ameliorating Societal Disturbance) in preparation for the final clustering (which yielded nine clusters for each category, personal and non-personal). These 18 clusters represent the far- right columns in figures 2.1 and 2.2. They were considered as "second order" by the AARPS team, whereas the "first-order" benefits were the total of 46 benefit factors described earlier. A final query consisted of ordering the 18 second-order benefit clusters according to relative importance. Table 2.1 shows the final order produced from this query and also shows the mean ratings and standard deviations produced by 96 participants during this phase. Because the eventual list of 18 benefit clusters was still impractical in size to be used with facility in a benefit-cost analysis using the AARPS model, the final clustering results were reassessed using an alternative clustering criterion, one yielding fewer, more generic clusters. Figure 2.3 shows how the 18 clusters were finally reduced to five benefit categories, all of which applied to either of the three benefit terms in the model (B₁, Bs, and BF). Table 2.II indicates which of these final five categories applies to which of the three benefit terms. It was recognized that all of the 18 clusters mentioned previously could have monetary benefits under certain circumstances, so lines of connection are drawn in figure 2.3 between the items in the list of 18 clusters and the monetary benefit dimension (no. 1, shown second at the far right). Since the BF and Bs terms are unique in the model, two benefit dimensions were created to address these terms separately, thereby reducing the final list of what could be considered 15 (five for each of the three benefit terms in the model) to a total of seven. This total of seven make up the contents of table I in the body of this report (page 21), and they are the seven benefit dimensions ultimately intended by the AARPS team to be used with its benefit-cost model (eq. [1]). Thus, the completion of a rather elaborate 88 clustering process over a period of several years resulted in reducing a list of 243 very specific potential rehabilitation-research benefits to a more practical number of seven benefit dimensions to facilitate addressing each of the multi-dimensional benefit terms in the AARPS benefit-cost analysis model. 89 TABLE 2.1 IMPORTANCE RATINGS FOR SECOND-ORDER BENEFITS * Consensus Second-Order Benefit Factors Mean Standard Order Rating Deviation 1. Enhancing quality and accessibility of services 86.80 +18.87 2. Enhancing individual coping skills 79.22 19.45 3. Minimizing functional limitations and personal disability 78.15 21.36 4. Improving personal vocational status and material well-being 76.33 21.62 5. Enhancing effectiveness of service providers 76.21 20.77 6. Improving program development and evaluation 71.98 19.83 7. Expanding knowledge base 69.49 22.19 8. Encouraging individual's social participation 66.65 23.65 9. Improving program performance and performance measures on berefit-cost criteria 65.51 26.41 10. Fostering consumer involvement 65.54 23.94 11. Facilitating societal change 59.42 26.84 12. Improving legislative impact and coordination of government entities 57.49 27.35 13. Improving physical environment 57.44 26.91 14. Developing and communicating policies, plans and procedures 56.73 25.20 15. Facilitating administrative flexibility and improvement 54.77 26.71 16. Containing personal costs and need for services 52.83 23.01 17. Promoting generalizability of services 50.26 26.39 18. Containing institutionalization 49.93 26.38 Mean value of all scores 65.26 22.25 *From Fuhrer, Cardus, & Rossi (1979). 90 TABLE 2.II.- BENEFIT DIMENSION APPLICABILITY TO TERMS OF MODEL Benefit-Model Term Benefit Dimension B1 Bs BF 1. Enhancing Service Quality* X - - 2. Monetary Benefits to Individual* X X X 3. Improving Individual Client Outcomes X - - 4. Improving Administration Bases for Service Provision X - - 5. Improving Public Policy Bases for Rehab- ilitation X - - 6. Indirect Benefits, Given Project Success - X - 7. Indirect Benefits, Regardless of Success - I X *NOTE: The benefit dimension, monetary benefits to individuals, was later made more general to include all three benefit terms and is now referred to as simply Monetary Benefits, applying to any benefits which can be expressed in monetary units. Because this benefit dimension was seen to apply to all three terms of the model, it was interposed with item 1 above in the final list (as presented in table I in the text) so that monetary benefits would be listed first. 91 DECREASED NUMBER OF CHILDREN IN FOSTER CARE MORE EARLY ADOPTION PLACEMENTS FIGURE 2.1-PERSONAL BENEFIT FACTORS FEWER SCHOOL DROP-OUTS CHILD FARE IMPROVED WELFARE OF CHILDREN EARLIER RETURN OF CHILDREN TO OWN HOME MORE REALISTIC EXPECTATION OF SERVICE EFFECTIVENESS PERCEIVED EQUITY OF QUALITY OF SERVICES AVAILABLE IMPROVED USAGE OF PREVENTIVE SERVICES IENT EXPECTATION SKILLS FOR OBTAINING SERVICES OF SERVICES KNOWLEDGE FOR OBTAINING SERVICES IMPROVED UNDERSTANDING OF VALUE OF SERVICES FOSTERING CONSUMER INCREASED AWARENESS OF AVAILABILITY OF SERVICES INVOLVEMENT INCREASED UNDERSTANDING OF RIGHTS/CELIGATIONS UNDERSTANDING RIGHTS AND ORI IGATIONS MORE CHANCES FOR CLIENT PROGRAM PARTICIPATION INCREASED AWARENESS OF OPPORTUNITIES FOR FEEDBACK CONSUMER PARTICIPATION LESS INCONVENIENCE IN ESTABLISHING ELIGIBILITY, IMPROVED ACCESS TO FAMILY COUNSELING LESS PERSONALLY DISRUPTIVE SERVICE BETTER COORDINATION OF SERVICES AVAILABILITY OF NEW SERVICES IMPROVED QUALITY OF SERVICES MORE HUMANE SERVICES QUAL ITY OF SERVICE REFERRAL COUNSELING JVERY BETTER QUALITY PROVIDERS IMPROVED ACCESS TO SERVICES IMPROVED CONTINUITY OF SERVICES IMPROVED DURABILITY OF SERVICE RESULT ENHANCING CUM ITY AND IMPROVED INDIVIDUALIZATION OF SERVICES ACCESSIBILITY SERVICES INCREASED COMMUNITY AMARENESS OF SERVICE PROGRAMS INCREASED ACCESS TO EDUCATION COUNSEL ING. EDUCATION INCREASED ACCESS TO TRAINING AND TRAINING IMPROVED ACCESS TO COUNSELING REHABILITATION OF INDIVIDUALS NOT NOW SUITABLE EXPANDING BENEFITS OF IMPROVED POTENTIAL FOR REHABILITATION ABILITY TO PERCEIVE/ACT ON CUES FOR DYSFUNCTIONS SERVICES PREVENTION OF INSTITUTIONAL PLACEMENT CONTAINMENT OF CONTAINING POSSIBILITY OF DE-INSTITUTIONALIZATION INSTITUTIONAL IZATION INSTITUTIONALIZATION REDUCED NEED FOR PUBLIC SUPPORT DECREASED NEED FOR SERVICE SERVICE CONTAINMENT CONTAINING PERSONAL COSTS DECREASED TIME IN SERVICE PROGRAM AND NEED FOR SERVICES DECREASED COST OF SERVICE CONTAINMENT OF CLAIMS PAID MORE RAPIDLY PERSONAL COST REDUCTION IN PERSONAL COST OF ONGOING TREATMENT FREE OTHER HOUSEHOLD MEMBERS FOR OTHER PURPOSES IMPROVED ABILITY TO CARE FOR CHILDREN IMPROVED PARENTAL FUNCTIONING BETTER FAMILY RELATIONSHIPS FAMILY FUNCTIONING FAMILY BETTERMENT AND STABILITY IMPROVED FAMILY FUNCTIONING BETTER MARITAL RELATIONSHIP PREVENTION OF FAMILY BREAK-UPS GREATER NUMBER OF PERSONAL SOCIAL CONTACTS IDENTIFICATION WITH ADVOCACY GROUP IMPROVED INTERPERSONAL RELATIONS IMPROVED GROUP RELATIONS ENCOURAGING INDIVIDUAL'S MORE SOCIALLY OUTGOING SOCIAL INVOLVEMENT SOCIAL PARTICIPATION INCREASED SOCIAL COMPETENCE INCREASED SOCIAL MOBILITY- MORE COMPENITY PARTICIPATION GREATER VARIETY OF PERSONAL SOCIAL CONTACTS .CCESS TO DESIRED ROLES (MARRIAGE,CHILDREN,ETC) PERSONAL -SOCIAL FULFILLMENT BETTER HOUSING BETTER TRANSPORTATION BETTER PHYSICAL IMPROVING PHYSICAL ENVIRONMENT IMPROVED SATISFACTION WITH LIFE ENVIRONMENT INCREASED SELF-RESPECT FEELING OF WELL-SEING GREATER CONTENTMENT IMPROVED SELF-IMAGE INCREASED HAPPINESS SELF-ACTUALIZATION PERSONAL WELL-BEING SELP-UNDERSTANDING AND IDENTITY INCREASED FREEDOM MORE CONTROL OVER FUTURE COMFIDENCE TO PARTICIPATE MORE ENJOYMENT OF LEISURE SENSE OF NOT BEING ANONYMOUS INCREASED EMOTIONAL STABILITY INCREASED POTENTIAL FOR SELF-DEVELOPMENT CLIENT SELF-PROPELLING CLIENT INITIATIVE ENHANCING INDIVIDUAL ABILITY TO CHANGE ENVIRONMENT COPING SKILLS INCREASED ABILITY TO PLAN EFFECTIVELY ADAPTIVE BEHAVIOR RELIEF FROM STRESSES IMPAIRING SUCCESSFUL FUNCTIONING BETTER INSIGHT INTO IMPROVED BEHAVIOR INSIGHTS AND BEHAVIOR BETTER LINKED IMPROVED COGNITIVE FUNCTIONING COPING BEHAVIOR BETTER INTELLECTUAL SKILLS INCREASED ABILITY FOR PERSONAL PRODUCTIVITY IMPROVED CAPABILITY FOR SOLVING ONGOING PROBLEMS PERSONAL USAGE OF EFFECTIVE REGIMENS FOR SELF-CARE INCREASED ABILITY TO PERFORM HOMEMAKING SERVICES ABILITY FOR DEAF TO USE TELEPHONE INCREASED PHYSICAL MOBILITY PERSONAL FUNCTION OF RESTORATION OF FUNCTION ABILITY TO WORK AGAIN- HANDICAPPED ABILITY TO USE HANDS AGAIN IMPROVED ABILITY FOR SELF-CARE MINIMIZING FUNCTIONAL ABILITY TO COMLETE SCHOOL FOR HANDICAPPED INITATIONS IMPROVED PERSONAL PRACTICES FOR HEALTH MAINTENANCE PERSONAL DISABILITY DECREASED DISABILITY DAYS DECREASED ACCIDENTS CONTAINMENT OF IMPROVED HEALTH- PERSONAL ILLNESS DECREASED DAYS OF ILLNESS DECREASED PAIN AND ANGUISH SAVINGS IN THE BANK (OUT OF DEBT) INCREASED INCOME PERSONAL ECONOMIC SELF-SUPPORT IMPROVEMENT ASSURANCE OF CONTINUATION OF SELF-SUPPORT IMPROVED WORK AND VOCATIONAL SKILLS INCREASED ABILITY TO WORK IMPROVING PERSONAL VOCA- INCREASED JOB MOBILITY INDIVIDUAL VOCATIONAL TOTAL STATUS AND INCREASED JOB SATISFACTION SUFFICIENCY MATERIAL WELL-BEING INCREASED MOTIVATION TO WORK NOREASED EMPLOYMENT OPPORTUNITIES MPROVED QUALITY OF LIFE FOR HANDICAPPED MPROVED STANDARD OF LIVING IMPROVED QUALITY OF LIFE MATERIAL CLIN ITY OF IMPROVED LIVING CONDITIONS LIFE IMPROVED ACHIEVEMENT OF INDEPENDENT LIVING GOALS 92 TRANSFERABILITY TO OTHER TARGET GROUPS PREVENTION OF COMMUNITY DISINTEGRATION FIGURE 2.2-NON-PERSONAL BENEFIT FACTORS IMPROVED STATUS OF POPULATION GROUPS IMPROVED URBAN/RURAL LIFE IMPROVED ENVIRONMENT REDUCED CRIME RATE REDUCED ALIENATION ENHANCE SOCIAL JUSTICE INCREASED SOCIAL MIXING AMELIORATING SOCIETAL REDUCE SOCIAL DISRUPTIONS DISTURBANCES REMOVE BARRIERS TO ADOPTION MORE EFFECTIVE USE OF HUMAN POTENTIAL REDUCE DEPENDENCY OF POPULATION GROUPS LESS ADVERSE SOCIAL PRESSURE ON RECIPIENTS IMPROVED PROGRAM RELEVANCE TO TARGET POPULATIONS TRAIN AND INFORM STAFF RESPONSIBLE FOR PROGRAMS BROADER STAFF UNDERSTANDING BETTER MIDDLE MANAGEMENT SELECTION AND TRAINING IMPROVED STAFF TRAINING INFORMATION FOR MANPOWER NEEDS, DISTRIBUTION, UTILIZATION IMPROVED EFFICACY OF WHAT PRACTITIONERS/PROFESSIONALS DO ENHANCING EFFECTIVENESS OF SERVICE PROVIDERS ESTABLISH REQUIREMENTS FOR UTILIZATION OF NEW PROCEDURES IMPROVED EFFICIENCY OF PRACTITIONERS/PROFESSIONALS IMPROVED PROVIDER'S IMPACT OF PROGRAM ON INSTITUTIONAL AND INDIVIDUAL PROVIDERS EFFECTIVENESS ESTABLISH REQUIREMENTS FOR USE OF NEW DEVICES BY PRACTITIONERS ADVANCING THE FRONTIERS OF SCIENTIFIC KNOWLEDGE STIMULATION OF FURTHER NEEDS FOR RESEARCH IMPROVED RESEARCH CAPABILITY- ADAPTION OF NEW SOCIAL SCIENCE THEORY EXPANDED R&D POTENTIAL METHODOLOGY FOR MEASURING SOCIAL CHANGE SERENDIPITOUS REVELATIONS OF NEW BENEFITS EXPANDED DEMONSTRATION OF VALUE OF RESEARCH CONTRIBUTIONS TO DEVELOPMENT OF SOCIAL SCIENCES,DATA PROCEDURES INFORMATION SUGGESTING CHANGES NEEDED TO REALIZE GOALS INCREASED DATA BASE FOR PROGRAM ANALYSIS EXPANDING KNOWLEDGE BASE IMPROVED INFORMATION INFORMATION FOR PROGRAM MODIFICATION SYSTEMS BETTER MANAGEMENT INFORMATION SYSTEM YIELDING TIMELY AND RELEVANT ADMINISTRATIVE DECISION SUPPORTS UTILIZATION OF COMPUTER TECHNOLOGY INCREASE QUALITY OF STATISTICS IMPROVED DATA QUALITY SUGGEST NEW ORIENTATIONS TO ISSUE ANALYSIS VERIFICATION/DETERMINATION OF BENEFIT-ASSUMPTION VALIDITY BASIS FOR ASSESSING POTENTIAL SOCIETAL IMPACT OF PROGRAMS VALIDATION OF BENEFITS INCREASED COST/EFFECTIVENESS OF PROGRAMS BETTER TECHNIQUES FOR DETERMINING COSTS/BENEFITS OF PROGRAMS COST/EFFECTIVENESS AND BENEFIT/COST INCREASED EFFICIENCY IN PAYMENT OF BILLS IMPROVING PROGRAM PERFORMANCE AND PERFORMANCE MORE EFFICIENT UTILIZATION OF RESOURCES MEASURES ON BENEFIT-COST CRITERIA USING TAX DOLLARS MORE EFFECTIVELY IMPROVED FINANCIAL MANAGEMENT IMPROVED PRODUCTIVITY OF SAVING TAX DOLLARS SPENT TAXES FEWER INSTANCES OF ABUSE AND WASTE REDUCTION IN COSTS PER UNIT OF SERVICE IMPROVED TECHNOLOGY TO MINIMIZE EXPENDITURES/MAXIMIZE SERVICES IMPROVED TECHNIQUES FOR THERAPEUTIC/SERVICE INTERVENTION IMPROVED ORGANIZATION OF SRS SERVICE DELIVERY INNOVATIONS IN SERVICE DELIVERY- SERVICE PROCESS REFINEMENT REDUCED TIME FOR SERVICE DELIVERY PROVIDE FOLLOW-UP SERVICES FOR DEVELOPPENTALLY DISABLED ELIMINATE NON-EFFECTIVE PROGRAMS DETERMINE PROGRAM MALFUNCTION PROGRAM CONTAINMENT EVENTUAL REDUCTION OF PROGRAMS NEEDS BETTER MEASURES TO EVALUATE SERVICE EFFECTIVENESS IMPROVED TECHNICAL EVALUATION OF PROGRAMS BETTER MEASURES OF PROGRAM OUTPUT IMPROVED EVALUATION OF MEASURES OF DIFFERENTIAL CASE DIFFICULTY PROGRAMS IMPROVING PROGRAM DEVELOPMENT & EVALUATION ESTABLISHMENT OF MEASURES OR INDICATORS COMPREHENDING IMPLICATION OF EXISTING/POTENTIAL PROGRAMS IMPROVED UNDERSTANDING OF PROGRAM OPTIONS USEFUL TESTING OF ALTERNATIVE APPROACHES PROGRAM IMPROVEMENT BETTER JUSTIFICATION OF RELATIVE PROGRAMS STRATEGIES ESTABLISH MORE EXPLICITLY PROCEDURES FOR PROGRAM PRIORITIES CLARIFY THE MAGNITUDE AND SOLVABILITY OF SRS PROGRAM ISSUES KNOWLEDGE OPTIMIZING SERVICE PROGRAMS IN DIFFERENT SETTINGS/CONDITIONS UNDERSTANDING HOW TO CONSTRUCT MORE EQUITABLE PROGRAMS MORE EFFECTIVE SEQUENCING OF PROGRAM DEVELOPMENT IDENTIFY APPROPRIATE SCOPES OF PROGRAMS PROGRAM IMPROVEMENT TACTICS MORE FUTURE-ORIENTED PROGRAM DEVELOPMENT RESOURCES REQUIRED FOR EFFECTIVE PROGRAM IMPLEMENTATION BETTER UNDERSTANDING OF MOTIVATION AND GROUP PROCESSES IDENTIFY WAYS OF DEALING WITH SOCIAL PROBLEMS IDENTIFY FACTORS IN SOCIAL CHANGE MEDIATING SOCIAL CHANGE IDENTIFY CAUSES OF SOCIAL PROBLEMS BETTER INTEGRATION OF COMMUNITY RESOURCES FACILITATING SOCIETAL CHANGE BETTER PUBLIC UNDERSTANDING OF PROGRAMS IMPROVED IMAGE OF PROGRAMS INCREASED CONSUMER PARTICIPATION PUBLIC ACCEPTANCE OBTAIN COMMUNITY SUPPORT OF PROGRAMS KNOWLEDGE OF HOW DIFFERENT POPULATIONS REACT TO VARIOUS SERVICES GENERAL IZABILITY OF SERVICES PROMOTING GENERALIZABILITY OF SERVICES TRANSFERABILITY TO OTHER TARGET POPULATIONS AND SUBPOPULATIONS EFFECTIVE COMMUNICATION OF PROGRAM OUTCOMES TO POLICY MAKERS INFORMATION DISSEMINATION EXTENSION OF KNOWLEDGE TO INDIVIDUALS/AGENCIES NOT COVERED IN PROGRAMS FEEDBACK INFORMATION TO ASSIST IN PLANNING IMPROVED LONG-RANGE PLANNING- PLANNING PROCESSES IMPROVED CAPACITY FOR BUILDING, PLANNING, ETC. DEVELOPING & COMMUNICATING POLICIES. PLANS. AND PROCEDURES KNOWLEDGE ABOUT BENEFITS OF ALTERNATIVE POLICY CHANGES INFORMATION ON AND METHODS TO IMPLEMENT POLICY CHANGES ESTABLISH POLICY RECOMMENDATIONS FOR IMPROVING SERVICES POLICY REFINEMENT ESTABLISH POLICY RECOMMENDATIONS FOR IDENTIFYING TARGET GROUPS INCREASED AGENCY OPENNESS TO SELF-EVALUATION INCREASED AGENCY OPENNES TO CHANGE SIMPLIFY ADMINISTRATION BETTER ADMINISTRATION ADMINISTRATIVE FLEXIBILITY CACILITATING AGMINISTRATIVE FLEXIBILITY REDUCE PROCEDURAL COMPLEXITY AND IMPROVEMENT AND IMPROVEMENT MANAGEMENT DECISION-MAKING SUPPORT IMPROVED EXECUTIVE UNDERSTANDING OF PROGRAMS ESTABLISH LEGAL BASIS FOR HAHDICAPPED'S NEEDS(HOUSTNG,RECREATION.ETC) LEGISLATIVE BASIS FOR PROGRAM IMPLEMENTATION ON A NATIONAL SCALE IMPROVED CONGRESSIONAL UNDERSTANDING OF PROGRAMS IMPROVED JUDICIAL UNDERSTANDING OF PROGRAMS NEW LEGISLATION TO ALTER OLD PROGRAMS NEW LEGISLATION TO ALTER NEW PROGRAMS ADVOCACY AND PROTECTIVE LEGISLATION REVISE OBSOLETE LEGISLATION ELIMINATE LEGISLATIVE GAPS/LOOPHOLES EGISLATIVE IMPACT SIMPLIFICATION OF CRITICAL LEGISLATION CONSOLIDATE MAJOR LEGISLATIVE AUTHORITY INFORMATION AND METHODS TO IMPLEMENT LEGISLATION INTERACTIONAL EFFECTS AMONG/BETWEEN LANS/POLICIES RATIONAL INTEGRATION OF POLICY/LEGISLATION WITH RELATED PROGRAMS IMPROVING LEGISLATIVE IMPACT & COORDINATION BETTER MEANS TO STRUCTURE LEGISLATION BASED ON MEASURED NEEDS/BEMEFITS OF COVERNMENT ENTITIES RELATIVE EFFECTIVENESS OF STATE ADMINISTRATIVE ORGANIZATIONS ESTABLISH STANDARDS FOR STATE AGENCIES FEDERAL LINKS WITH PROVIDING GUIDANCE TO OTHER COUNTRIES OTHER GOVERNMENTS BETTER FEDERAL-STATE,STATE-LOCAL,FEDERAL-LOCALETC.RELATICNS DIFFERENTIATE AMONG FEDERAL, FEDERALYSTATE, AND STATE RESPONSIBILITIES 93 FIGURE 2.3-FINAL BENEFIT CLUSTERING FOSTERING CONSUMER INVOLVEMENT ENHANCING QUALITY AND II. ENHANCED QUALITY OF SERVICES ACCESSIBILITY OF SERVICES CONTAINING INSTITUTIONALIZATION I. .MONETARY BENEFITS CONTAINING PERSONAL COSTS AND NEED FOR SERVICES Personal Benefit Factors ENCOURAGING INDIVIDUAL'S SOCIAL PARTICIPATION IMPROVING PHYSICAL ENVIRONMENT ENHANCING INDIVIDUAL COPING SKILLS III. IMPROVED INDIVIDUAL CLIENT OUTCOMES MINIMIZING FUNCTIONAL LIMITATIONS AND PERSONAL DISABILITY I IMPROVING VOCATIONAL STATUS AND MATERIAL WELL-BEING ENHANCING EFFECTIVENESS OF SERVICE. PROVIDERS - EXPANDING KNOWLEDGE BASE I IMPROVING PROGRAM PERFORMANCE AND PERFORMANCE MEASURES ON B-C CRITERIA IV. IMPROVED ADMINISTRATIVE BASES FOR SERVICE PROVISION IMPROVING PROGRAM DEVELOPMENT Non-Personal Benefit Factors AND EVALUATION FACILITATING SOCIETAL CHANGE PROMOTING GENERALIZABILITY OF SERVICES DEVELOPING AND COMMUNICATING POLICIES, PLANS AND PROCEDURES V. IMPROVED POLICY BASES FOR REHABILITATION FACILITATING ADMINISTRATIVE FLEXIBILITY AND IMPROVEMENT IMPROVING LEGISLATIVE IMPACT AND COORDINATION OF GOVERNMENTAL ENTITIES VI. INDIRECT BENEFITS, GIVEN PROJECT SUCCESS VII. INDIRECT BENEFITS, REGARDLESS OF PROJECT SUCCESS 94