18
Hydrological Sciences-J'ournal-des Sciences Hydrologiques, 42(4) August 1997 531 Distributive fairness considerations in sustainable project selection SAM MATHESON Department of Civil and Geological Engineering, Room 342 Engineering Building, The University of Manitoba, Winnipeg, Manitoba, Canada R3T 5V6 BARBARA LENCE Department of Civil Engineering, 2324 Main Mall, University of British Columbia, Vancouver, British Columbia, Canada V6T1Z4 JOSEF FÛRST Institute for Water Management, Hydrology and Hydraulic Engineering, University of Agricultural Sciences, Muthgasse 18, A-1190 Vienna, Austria Abstract This work develops general fairness measures that may be used as criteria for sustainable project selection. Sustainable development, fair allocation objectives and empirical distance-based measures of fairness, and their evaluation are dis- cussed. Generalized fairness measures are developed and extended for both intratemporal and intertemporal fairness comparisons. A preliminary application of the extended distance based fairness measures is then performed for a case study of the selection of an electricity supply project. The case study involves selecting between a dispersed diesel energy supply and centralized energy supply with land line energy distribution. Due to data limitations, the perceived fairness is measured in terms of the annual energy costs per megawatt-hour that result from implementing each alternative. The applied fairness measures indicate that intratemporal fairness, in terms of the distribution of user unit costs, may be increased by choosing the land line alternative and that there is no significant difference among alternatives with respect to intertemporal fairness. These results provide limited insight into the energy supply problem, however, and it is suggested that further analyses should be conducted when information regarding the environmental impacts and reliability of power supply for each of the alternatives becomes available. Considérations d'allocation équitable dans la sélection de projets durables Résumé Ce travail définit des quantificateurs généraux d'équité pouvant être utilisés comme critères de sélection de projets durables. Le développement durable, les objectifs d'allocation équitable et des mesures empiriques d'équité fondées sur une notion de distance ainsi que leur estimation y sont discutées. Des mesures d'équité généralisée sont définies et étendues en vue de comparaisons d'équités tant synchroniques que diachroniques. Une application préliminaire de ces mesures d'équité généralisée a été réalisée dans une étude de cas concernant un projet d'alimentation en électricité. L'étude de cas consiste à choisir entre une production décentralisée utilisant des générateurs diesel et une production centralisée impliquant un réseau de distribution. En raison du peu de données disponibles, la perception de l'équité a été mesurée par les coûts d'énergie annuels par mégawatt-heure produit de chacun des termes de l'alternative. Les mesures d'équité utilisées indiquent que la meilleure équité synchronique, exprimée en termes de coût unitaire pour l'utilisateur, est obtenue en choisissant la solution centralisée et qu'il n'existe pas de différence significative entre les deux solutions en ce qui concerne l'équité diachronique. Ces résultats ne fournissent toutefois qu'une vision limitée du problème d'alimentation en énergie et des analyses plus approfondies devront être réalisées lorsque des données concernant les impacts environnementaux et la fiabilité des deux termes de l'alternative seront disponibles. Open for discussion until I February 1998

Distributive fairness considerations in sustainable project selection

Embed Size (px)

Citation preview

Hydrological Sciences-J'ournal-des Sciences Hydrologiques, 42(4) August 1997 531

Distributive fairness considerations in sustainable project selection

SAM MATHESON Department of Civil and Geological Engineering, Room 342 Engineering Building, The University of Manitoba, Winnipeg, Manitoba, Canada R3T 5V6

BARBARA LENCE Department of Civil Engineering, 2324 Main Mall, University of British Columbia, Vancouver, British Columbia, Canada V6T1Z4

JOSEF FÛRST Institute for Water Management, Hydrology and Hydraulic Engineering, University of Agricultural Sciences, Muthgasse 18, A-1190 Vienna, Austria

Abstract This work develops general fairness measures that may be used as criteria for sustainable project selection. Sustainable development, fair allocation objectives and empirical distance-based measures of fairness, and their evaluation are dis­cussed. Generalized fairness measures are developed and extended for both intratemporal and intertemporal fairness comparisons. A preliminary application of the extended distance based fairness measures is then performed for a case study of the selection of an electricity supply project. The case study involves selecting between a dispersed diesel energy supply and centralized energy supply with land line energy distribution. Due to data limitations, the perceived fairness is measured in terms of the annual energy costs per megawatt-hour that result from implementing each alternative. The applied fairness measures indicate that intratemporal fairness, in terms of the distribution of user unit costs, may be increased by choosing the land line alternative and that there is no significant difference among alternatives with respect to intertemporal fairness. These results provide limited insight into the energy supply problem, however, and it is suggested that further analyses should be conducted when information regarding the environmental impacts and reliability of power supply for each of the alternatives becomes available.

Considérations d'allocation équitable dans la sélection de projets durables Résumé Ce travail définit des quantificateurs généraux d'équité pouvant être utilisés comme critères de sélection de projets durables. Le développement durable, les objectifs d'allocation équitable et des mesures empiriques d'équité fondées sur une notion de distance ainsi que leur estimation y sont discutées. Des mesures d'équité généralisée sont définies et étendues en vue de comparaisons d'équités tant synchroniques que diachroniques. Une application préliminaire de ces mesures d'équité généralisée a été réalisée dans une étude de cas concernant un projet d'alimentation en électricité. L'étude de cas consiste à choisir entre une production décentralisée utilisant des générateurs diesel et une production centralisée impliquant un réseau de distribution. En raison du peu de données disponibles, la perception de l'équité a été mesurée par les coûts d'énergie annuels par mégawatt-heure produit de chacun des termes de l'alternative. Les mesures d'équité utilisées indiquent que la meilleure équité synchronique, exprimée en termes de coût unitaire pour l'utilisateur, est obtenue en choisissant la solution centralisée et qu'il n'existe pas de différence significative entre les deux solutions en ce qui concerne l'équité diachronique. Ces résultats ne fournissent toutefois qu'une vision limitée du problème d'alimentation en énergie et des analyses plus approfondies devront être réalisées lorsque des données concernant les impacts environnementaux et la fiabilité des deux termes de l'alternative seront disponibles.

Open for discussion until I February 1998

532 Sam Matheson et al.

INTRODUCTION

In the process of project selection and implementation, a best compromise solution for a problem is often achieved by considering conflicting criteria, or objectives, and after the project is implemented it may be modified as appropriate based on initial impacts and additional information as this becomes available (United Nations, 1988). Project selection criteria may consider economic, financial, biophysical and social impacts of a given project alternative. Project-related impacts may be either benefits or costs. Simonovic et al. (1995) recently identified three additional criteria for including sustainability in project selection. These are: intergenerational equity, project and impact reversibility and risk management over time. While common project selection criteria such as economic efficiency are widely applied, applications of project selection based on equity, or fairness, are less common. This paper develops intratemporal and intertemporal fairness criteria for project selection. These criteria are examined in a preliminary application to a case study that evaluates alternative electrical power supply technologies required to meet the 40-year load forecast for seven northern Manitoba communities.

In civil engineering, in general, a project may be seen as an allocation of different impacts that may originate during the construction and operational phases of a project's design life, and may affect the biophysical, social and economic sub­systems of a region. Impacts may persist after the project has been dismantled and may also affect other subsystems in different regions and thus may be seen to act on a local, regional, or global scale. An example in water resources engineering is the construction of a structure which controls both spatial and temporal quantities of surface water in order to harness the biophysical system's potential energy. The spatial and temporal manipulation of surface water from its natural state may distribute social, economic and biophysical impacts within the region that may be seen as unfair by affected or interested groups of people.

Examples of project selection and evaluation, based in part on a fair distribution of project related impacts, in water resources engineering and management science include Brill et al. (1976), Cohen (1978), Sampath (1991) and McAllister (1976). Brill et al. (1976) examine both efficiency and equity aspects of waste discharge water quality management programmes for the Delaware Estuary. They define equity as the equality of removal efficiencies among dischargers and use three different distance-based fairness measures: absolute deviation from the mean waste treatment level, the range between the maximum and minimum waste treatment levels and the maximum of the waste treatment levels. Cohen (1978) discusses a multiobjective river basin development plan for the Rio Colorado River in Argentina in which a regional allocation objective function is formulated in addition to an efficiency objective function. The regional objective function is to minimize the mean absolute deviation of water withdrawals among four provinces in a region. Cohen (1978) also mentions that, for this case study, the decision makers did not agree with a perfect equality allocation objective nor did they reveal their preference for an alternative fair allocation objective. Sampath (1991) employs Theil's Entropy Coefficient (Theil, 1967) to examine fairness in the distribution of access to irrigation water between

Distributive fairness considerations in sustainable project selection 533

agricultural groups in India. McAllister (1976) presents a theoretical framework to evaluate fairness and efficiency for both delivered and nondelivered urban public services to examine the implications of service size and service spacing alternatives. He defines fairness as the degree of equality of service and operationalizes it by comparing the standard deviations of the distances between service centres and demand points.

Authors such as the Brundtland Commission (WCED, 1987), Young (1992), Jacobs (1993) and Weiss (1995) state that sustainability requires consideration of the fairness of impacts both on the present generation and on future generations. Therefore, serious investigation of sustainable project selection should examine issues related to the distributional fairness of project-related impacts both within and between generations as perceived by groups of individuals. The intratemporal and intertemporal distributional fairness criteria presented in this paper may be used in conjunction with other criteria in a multiobjective framework to determine the relative desirability of competing project alternatives. Although not considered in this work, other project selection criteria that may be considered in such analyses include economic efficiency, effectiveness, risk and reversibility. In the next section, basic concepts of sustainability are discussed. This section focuses on elements of sustainable development that are relevant to equity. In the third section, fair allocation objectives, empirical distance-based fairness measures, and the previous application of these measures to case studies are presented in order to provide a foundation for the development of intratemporal and intertemporal distributive fairness measures. This section also describes how fair allocation principles may be related to empirical fairness measures found in the literature. Impact assessment and other approaches for measuring fairness are beyond the scope of this work. The fourth section presents measures of distributive fairness that account for the major issues outlined in sections two and three. Following this, a preliminary application of the distributive fairness measures to a case study involving the choice between two different power supply technologies for a number of northern Manitoba communities is discussed. The alternatives for this problem are dispersed diesel generation and hydropower generation with land line distribution, which must meet a 40-year load forecast for the communities. For each alternative, this work considers the forecasted annual average cost per megawatt-hour, in 1993 Canadian dollars, to consumers within the communities. The limited data for this case study restrict the analysis to an investigation of only one issue related to the fair allocation of project impacts. Indeed, further comparisons of other biophysical, sociocultural and economic impacts that result from each alternative would be required for this problem before a sustainable alternative may be selected. However, the measures presented here may be applied should further meaningful impact forecasts become available. Finally, a discussion of the application and the conclusions are given.

SUSTAINABILITY AND DISTRIBUTIVE FAIRNESS

While equity, or distributive fairness, has long been discussed by management scientists regarding the provision of public services (Savas, 1978), interest in this

534 Sam Matheson et al.

topic has increased in the last 10 years due to the introduction of the concept of sustainable development. According to the Brundtland Commission (WCED, 1987), a sustainable development meets the needs of the present generation while still allowing future generations to meet their own needs. The Brundtland Commission's definition of sustainable development stresses the consideration of the needs of the present generation and of future generations, and has prompted others (Young, 1992; Nachtnebel et al., 1994; Beltratti, 1995; Munasinghe & Shearer, 1995; Simonovic et al., 1995) to promote equity as one of a number of objectives required for sustainability. At this point, two issues need to be addressed if intratemporal and intertemporal distributive fairness criteria are to be defined.

First, the notion of fairness within and between generations implied by the Brundtland Commission's definition of sustainable development is vague and needs to be examined in greater detail. In the context of project selection, for example, several questions related to fairness within and between generations are: (a) how long should equity be a concern?; (b) among whom should equity be a concern?; (c) what is a generation?; and (d) what impacts should be evaluated in terms of equity? Matheson (1997) proposes that equity should be a concern in project selection for as long as project-related impacts affect groups of similarly situated individuals. Similarly situated individuals are individuals that are affected by similar project related impact magnitudes, which may be considered as a group, in order to simplify the analysis. Marsh & Schilling (1994) suggest that individuals may be aggregated into groups on several bases such as: spatial, demographic, physical and temporal. An example of segregating a population into groups on a spatial basis may be to divide a population into countries, provinces, or municipalities. A population may be divided into groups demographically according to, e.g. age, gender, ethnic back­ground, occupation, social class, or income. A population could be segregated into groups on a physical basis according to different drainage basins or different bioregions. Finally, a population may be segregated on a temporal basis, perhaps in terms of generations as was mentioned by the Brundtland Commission (WCED, 1987). While the exact grouping basis is case specific, it should also be noted that a grouping strategy may be a combination of more than one of the above bases.

In addressing temporal considerations while evaluating the sustainability of a project, the affected groups over time must be compared. McKerlie (1989) discusses temporal aspects in fairness evaluations. In comparing impacts on two people, he considers whole life, simultaneous segments of lives and corresponding segments of lives comparisons. The whole life approach compares the total impact acting on each person's life. This approach may not reflect differences that occur during the same time period (e.g. mid-life) of the different lives. The simultaneous segments approach compares the impacts acting on the individuals in some mutual time period in both lives, i.e. impacts that both parties experience at the same point in time (e.g. 1991-1995). The corresponding segments approach compares the impacts acting on each life in the same stage of the respective lives, i.e. impacts that both parties experience at the same time period in their lives (e.g. in mid-life). Regarding the question of generations, Matheson (1997) suggests the use of time steps might be a more flexible approach because project-related impacts are project specific and may

Distributive fairness considerations in sustainable project selection 535

not have a duration that exceeds a generation or may exhibit high variability within a generation. Comparisons based on a generational time step may be too coarse to account for this variability and in such cases, a smaller time step (e.g. 1-5 years) may be more appropriate than a generational time step.

The second issue related to the formulation of intratemporal and intertemporal distributive fairness criteria is that, defining fairness as merely equity may be avoiding a more important question, that is: what is fairness in a distributive context? This question must be addressed if a measurement approach for sustainable project selection is to be formulated. This question has been studied at least since the writings of Aristotle (Thomson, 1985), but the answer remains elusive. The next section addresses the question of how fairness may be measured empirically so as to provide the groundwork for the formulation of fairness measures for sustainable project selection. Here, a project alternative is considered to result in an allocation of impacts that are distributed among different groups of similarly situated people.

FAIR ALLOCATION OBJECTIVES, DESIRABLE CHARACTERISTICS AND EMPIRICAL MEASURES OF DISTRIBUTIVE FAIRNESS

Blalock (1991) proposes that, when considering the reactions of groups to any allocation process, the analyst should consider their perceptions and interpretations concerning the fairness of both procedures and the outcome of the allocation. Procedural aspects are also important, in terms of public participation and awareness, as they help to make distributive outcomes more socially acceptable. In general, groups evaluate the fairness of their outcome by a social comparison process between the perceived project-related impacts and the impacts which these groups feel are fair.

Aristotle (Thomson, 1985), Deutsch (1975), Arthur & Shaw (1978), Blalock (1991) and Almond (1995) identify simple fair allocation objectives that groups may view as representing a fair allocation of project-related impacts: perfect proportionality, perfect equality and satisfaction of need. Young (1994) provides an excellent overview of these and other approaches to fair allocation. The first allocation objective, known as perfect proportionality, focuses on differences among groups and requires that each group should receive goods in proportion to what that group deserves. Achieving this objective may be problematic if the amount each group deserves cannot be assessed or if what is being distributed cannot be divided among groups. An equal distribution may be seen as fair when there is no reason to differentiate among groups. However, situations may arise where groups are different and one group may need more or less of what is being distributed than another depending on if what is being distributed is perceived to be a benefit or a cost, respectively. For this reason, the distribution of impacts according to the amount each group needs or can tolerate is often proposed as an alternative to perfect proportionality and perfect equality. Although that which constitutes a group's need is not well defined, Almond (1995) states that this problem may be overcome if basic needs were determined using statistics such as infant mortality or life expectancy.

536 Sam Matheson et al.

The relative importance of each fair allocation objective is discussed by Deutsch (1975) and Blalock (1991). They identify underlying factors that affect preferences for these objectives, for example, the time and information available for the allocator to make a decision.

While a number of desirable characteristics and principles have been proposed to evaluate fair allocation measures, a consensus in the literature does not exist regarding which characteristics and principles are required for evaluating measures of the various fair allocation objectives. Young (1994) states that all acceptable measures of fairness should satisfy the principles of impartiality and consistency because these principles are fundamental to fairness considerations. Impartiality requires that a fairness measure be applied to that which is being distributed and not based on some other ordering or ranking, such as a group attribute that is not considered to affect the distribution. Consistency requires that the fairness measure be applied across groups in the same manner for all groups. Marsh & Schilling (1994) mention three principles that are most commonly found in the literature on the evaluation of fairness measures. These are the principles of impartiality, transfers and scale invariance. The principle of transfers, proposed by Pigou (1912), requires that when a unit amount of positive (negative) impact is transferred from a better-off group to a worse-off group, an acceptable equality-based measure should reflect an improvement in fairness. The principle of scale invariance requires that only relative differences in impacts should matter not absolute differences. Thus, for two impact distributions, one distribution being a multiple of the other, a scale invariant fairness measure would calculate the same deviation from perfect fairness for both distributions.

In addition to impartiality and consistency, measures based on a perfect pro­portionality fair allocation objective are required to satisfy the fundamental principle (Harris, 1983) and thus exhibit an ordinal relation between a group's impact and a group's reference distribution. Measures based on a perfect equality are also required to satisfy the principle of transfers and the principle of scale invariance (Matheson, 1997). As no distance-based measure of need, nor any literature relating to satisfac­tion of need are available, for need there are no additional requirements suggested in the literature. Matheson (1997) suggests that, for need, satisfaction of the principles of impartiality and consistency are required. Other principles and desirable charac­teristics may be required to evaluate a need-based distance measure when this concept is further developed.

Marsh & Schilling (1994) review 20 empirical distance-based fairness measures that have been applied and propose a generalized framework for their classification. A distance-based fairness measure is a weighted sum of the distances between the ideal values for a distribution of impacts among groups of individuals and the actual values of these impacts. The authors' generalized framework for classification and future evaluation of distance-based fairness measures consists of examining three factors: the reference distribution, scaling and distance exponent of the measure. The authors describe a reference distribution as being a specified or desired effect level for each group. Possible types of reference distributions are peer, mean, or attribute-based distributions. Peer and mean reference distributions are based on perfect

Distributive fairness considerations in sustainable project selection 537

equality as the objective for fair allocation where peer reference distributions refer to comparisons among all peers and mean reference distributions refer to comparisons with the mean of the impacts for all peers. For the purposes of this paper, measures of this type are referred to as based on a perfect equality fair allocation objective.

Marsh & Schilling (1994) also describe an attribute-based reference distribution as being specific to each group and being based on, for example, the level of social need, desire, demand, social merit, or population. An attribute-based reference dis­tribution may be referred to as being based on the fair allocation objectives of perfect proportionality or satisfaction of needs. In this manner, Marsh & Schilling's framework is analogous to that of the fair allocation objectives discussed above. Thus, from this point on, measures of this type are referred to as measures based on proportionality or need.

Scaling is described as being commonly used when group sizes differ to account for large differences in the size of the distances measured. If scaling is performed, it is typically based on a normalization of distances or as a weighting based on the different attributes of the groups. Commonly used distance exponents are either one, two, or infinity. As the magnitude of the distance exponent increases from one to infinity, a greater weight is placed on deviations from the reference distribution. Thus, distance-based fairness measures with different distance exponents may indi­cate different levels of fairness when applied to the same impact distribution.

The review by Marsh & Schilling (1994) reveals that when the 20 common distance-based fairness measures found in the literature are applied to the same data set, different measures give different results. Matheson (1997) evaluates the 20 measures reviewed by Marsh & Schilling (1994), based on how well each measure satisfies the desirable principles associated with each fair allocation objective. He shows that potentially appropriate proportionality-based measures are Walsters Formula and the Equal Excess Formula because these measures satisfy the impar­tiality, consistency and fundamental principles. The Relative Mean Deviation Measure and the Gini Coefficient (Gini, 1912) are potentially appropriate equality-based measures because these measures satisfy the impartiality, consistency, transfers and scale invariance principles. Six potential need based measures are proposed that are variations on common proportionality-based measures. In the following section, the Equal Excess Formula (Heiner et al, 1981), Gini Coefficient (Gini, 1912) and a proposed need-based fairness measure (Matheson, 1997) are discussed and expanded for intratemporal and intertemporal fairness considerations.

The distance-based fairness measures discussed by Marsh & Schilling (1994) are either proportionality-based or equality-based fairness measures. This distinction is not made explicit by Marsh & Schilling (1994) and need-based fairness measures are not given in the literature. As the literature indicates that fairness may be a combina­tion of one or more of the fair allocation objectives, a generalized fairness measure based on a combination of proportionality, equality and need fair allocation objec­tives may be a more useful approach than an approach based on either perfect proportionality or satisfaction of need fair allocation objectives. Furthermore, allowing the present generation to meet its needs while allowing future generations to meet their own needs indicates that a need-based fair allocation objective may be an

538 Sam Matheson el al.

important consideration if one applies the Brundtland Commission's definition of sustainable development. Generalized measures for intratemporal and intertemporal fairness are developed in this paper for use as criteria for sustainability. Such measures may be applied for comparisons between groups located at a given time and between groups located in different times. As described above, individuals and groups tend to view fairness from more than one perspective. Such differences in value systems may be even more pronounced between groups in different time periods. Therefore, in this work it is proposed that generalized fairness measures represent a composite or weighted sum of objectives that may describe a fair allocation. The fairness measures presented here, then, are composite measures that describe how much a project's impacts deviate from being distributed in proportion to each group's contribution, being distributed equally and being distributed such that each group's needs are satisfied.

OPERATIONAL DEFINITIONS OF DISTRIBUTIVE FAIRNESS

Consider a situation in which there is a total of / groups where group i is affected by some impact with a magnitude E(i) that results from some action. In this situation, define A(i) as the impact magnitude that group / feels it deserves and Z(i) as the impact magnitude required to meet the needs of group i. Measures identified by Matheson (1997) that satisfy the desirable principles of impartiality, consistency, transfers and scale invariance, and the fundamental principle are given in equations (l)-(3). These equations are based on the fair allocation objectives of perfect proportionality, perfect equality and satisfaction of needs respectively. They are:

Bt=t\E(i)-A(i)\ (1)

/ / \m-EU)\ 2I2Ë (2)

B3 = tjE(i)-Z(i)\ (3)

where 5, = deviations from a proportional impact allocation; B2 = deviations from an equal impact allocation; B3 = deviations from an impact allocation that exactly satisfies the needs of all groups; / = number of groups being considered; i and j = group indices; E(i) = impact magnitude acting on group i; A(i) = impact magnitude that group i deserves, or group fs contribution towards receiving E(i); Z(i) = amount of E(i) that group i requires to satisfy its needs; and E = average impact magnitude that acts upon the / groups being considered.

Equation (1) is known as the Equal Excess Formula and represents the magnitude of fairness resulting from impacts which deviate from an allocation that is proportional to each group's contribution. Equation (2), known as the Gini Index or

Distributive fairness considerations in sustainable project selection 539

Gini Coefficient, is commonly used by economists to measure deviations from perfect equality, and is presented in the form given in Marsh & Schilling (1994). The two group indices, i and j , are required for equation (2) as the numerator of this equation is the sum of all possible pairwise comparisons of impact magnitudes between groups. The magnitude of the Gini Coefficient represents the decrease in fairness resulting from impacts that deviate from being allocated equally among all / groups. Equation (3) is presented in this work as one way to measure the fairness in an allocation of impacts that deviate from meeting each group's need for that impact. It should be noted that other methods of operationalizing a need fair allocation objective, such as a binary expression of whether or not a need is met, may be possible. Equations (l)-(3) have values that may increase in magnitude from zero, where zero corresponds to complete fairness as defined by each fair allocation objective.

If equations (l)-(3) are to be used in the context of sustainable project selection, the equations must be expanded to address four major concerns. First, these measures are not currently formulated in a way that accounts for the dimensions required for both intratemporal and intertemporal comparisons. Second, projects are likely to distribute more than one impact among groups and a measure of fairness should account for this in some way. Third, equations (l)-(3) represent three different aspects of fairness evaluations and a fairness measure should incorporate all of these objectives because, for example, not all people evaluate fairness by proportionality alone and may employ one or more fair allocation objectives. This may gain more significance when considering the extended temporal horizons associated with intertemporal fairness evaluations. Finally, as evaluations of fairness are case specific, a measure should reflect the variability in the emphasis placed on the different fair allocation objectives by the groups who are affected by a project.

Consider a situation in which there are X project alternatives, each distributing G different impact types to / groups over T time steps. Thus, each project alternative may be thought of as having a G x / x T impact matrix. While it may be possible for each project impact matrix to have different magnitudes for G, I and T, this work assumes for simplicity, that each project alternative impact matrix is the same size. Therefore, the matrix that represents the impact magnitude accruing to group /, of impact type g (i.e. either a benefit or cost), during time step t, resulting from project alternative x may be written as E(i, g, t, x). For each group, /', its contribution toward receiving a certain impact type, g, or the impact it deserves, may vary with time step t and is written as A(i, g, t). Additionally, a group's need for a particular impact may also vary with time and is written as Z(i, g, t). Equations (l)-(3) may be rewritten to correspond to this generalized problem. Equations (4)-(6) represent the three fair allocation objectives expanded for the intratemporal case with any number of impacts and equations (7)-(9) represent the three fair allocation objectives expanded for the intertemporal case, with any number of time steps.

Gl (=i«=iL w„E\E(i, g, t, x)-A(i, g, t)\ (4)

540 Sam Matheson et al.

1 T G

Cri /=lg=l

w,,ÉÉ £(i, g, t, x)-E(j, g, t, x)\

2rE..

1 T G tiy\X) — 2. ZJ

5,'(x) = — I Cri g=l

GT i=i g=lL * '=i

1 £

i^(jc) = — £ G 7 £

G-/ g=i

w ʣ(i,g,f,x)-Z(z',g,0

/ r , i'

7' '/• I

HT,\E(i,g,s,x)-E(i,g,t,x)\

;=1 2TZE„

B'(x) GI g=i

w E S £(/, g, ?, x) -Z(z, g, 0

(5)

(6)

(7)

(8)

(9)

where Bx{x) = average intratemporal fairness measure which is the weighted sum of deviations from a proportional impact allocation for all groups; B2(x) = average intratemporal fairness measure which is the weighted sum of deviations from an equal impact allocation for all groups; B3(x) = average intratemporal fairness measure which is the weighted sum of deviations from an allocation that meets the needs of all groups; B'x{x) = average intertemporal fairness measure which is the weighted sum of deviations from a proportional impact allocation for all groups; B'2{x) = average intertemporal fairness measure which is the weighted sum of deviations from an equal impact allocation for all groups; B'3{x) = average inter­temporal fairness measure which is the weighted sum of deviations from an allocation that meets the needs of all groups; G, I, T, X = number of different impact types, number of groups, number of time steps and number of project alternatives, respectively; i, g, t,x = indices for group, impact type, time step and alternative, respectively, such that l < g < G , 1 <I <I, \<t<T and 1 < x < X; j and s = group and time indexes, respectively, that are required for pairwise com­parisons such that 0 <j < I and 1 < s < T; wg = weights on impact type g, such that:

G

E w = 1 ; E(i, g, t, x) = magnitude of impact type g acting on group i during time «=i g

step t that results from project alternative x; Egu = average impact defined over all

-: • combination of impact type, time step and alternative such that groups for a given a r JV r

Em = — YjE(i,g,t,x) ; Eigx = average impact defined over all time steps for a given /«=.

1 i combination of group, impact type and alternative such that E- =^zHE{i,g,t,x);

A(i, g, t) = magnitude of group i's contribution towards receiving impact type

Distributive fairness considerations in sustainable project selection 541

during time t; and Z(i, g, t) = magnitude of impact type g that meets group f's needs during time step t.

Equations (4)-(6) are based on the three fair allocation objectives applied to the impacts distributed among groups during a given time step which are then averaged over all time steps. Equations (7)-(9) are based on the three fair allocation objectives applied to the impacts distributed over all time steps for a given group which are then averaged over all groups. These distance-based measures may be further combined into overall measures of intratemporal and intertemporal fairness by a weighted average approach. This may be accomplished by using a normalized Cartesian-based distance metric shown below:

a(x) = tql v=l

B,{x)

B.,-B,

2

(10)

\\i(x) = tql V = I

Bl(x) BU-B:

(ii)

where a(x) = magnitude of average intratemporal fairness for a given project alternative x, such that 0 < a(x) < 1 ; y/(x) = magnitude of average intertemporal fairness for a given project alternative x, such that 0 < y/{x) < 1 ; v = index for different fair allocation objectives; qr = weight attached to fair allocation objective

3

v such that: Y,qv = 1 ; Bv(x) m\âB'v{x) = magnitudes of the three intratemporal and v=1

three intertemporal fair allocation approaches; B.(x) andi?»(x) = minimum values of Br(x) and B',(x) across all v given a project alternative x; and Btt(x) mdBL(x) = maximum values of Br(x) &nàB'v{x) across all v given a project alternative x.

Thus, distance-based measures that reflect proportionality, equality and need fair allocation objectives are expanded to account for comparisons both within and between time steps for more than one type of impact. The expanded fairness measures are formulated in this work as averages of fairness comparisons across time steps, groups and different impact types. In equations (4)-(ll) weights wg and qv are used to represent the relative importance of fairness considerations among different project-related impacts and the relative importance placed on the three different fair allocation objectives, respectively. These weights represent the affected group's preferences and may be obtained, for example, by a social survey. According to Cohen (1978), the use of predetermined weights is a simple way of incorporating preferences into the decision making process but assumes that weights remain constant regardless of the objective functions' magnitudes and the willingness to trade-off one objective for another is independent of the magnitudes of the objective function values. This issue requires further investigation, however, it should be noted that the overall fairness measures presented in equations (10) and (11) may represent the social objective function of the groups and their views of intratemporal and intertemporal fairness, respectively.

542 Sam Matheson et al.

Uncertainty in fairness measurement may arise from employing an inappropriate fairness measure. For example, a fairness measure that does not accurately describe the perceived fairness of the groups managed may introduce uncertainty into the analysis. Uncertainty may also arise from unknown values for the weights that reflect the relative importance of each type of impact or fair allocation objective. Another type of uncertainty may arise due to prediction errors, e.g. errors in the predictions of the impact estimates over time. Impacts may increase, decrease, remain constant, or be some combination of these trajectories over time and the number of impacts that affect each group may also change over time. The uncertainty introduced by considering an increased temporal horizon may be so great that a fairness analysis may be untenable. Clearly, additional investigation of uncertainty reduction for the distributive fairness measures is needed.

AN APPLICATION OF THE DISTRIBUTIVE FAIRNESS MEASURES FOR SELECTING A POWER SUPPLY ALTERNATIVE IN NORTHERN MANITOBA, CANADA

The generalized fairness measures presented in equations (10) and (11) are applied in this section for a preliminary analysis of a case study that involves selecting among two different power supply technologies for seven remote communities located in northern Canada's Boreal forest approximately 560 km northeast of Winnipeg, Manitoba. The power supply alternatives being considered to satisfy the 40-year load forecast are dispersed diesel generation and land line connection to Manitoba Hydro's central power grid. The remote seven communities of Oxford House, Gods Lake, Gods River, Red Sucker Lake, Garden Hill, Wasagamack and St Theresa Point currently have electricity supplied by diesel generators in each community. Three additional nearby communities, Nelson House, Cross Lake and Split Lake, are supplied by a land line connection to Manitoba Hydro's central system. These three latter nonremote communities are used as reference communities for the fairness comparisons in this case study, because they are in the vicinity of, and have similar demographic characteristics to, the seven remote communities listed above. The dispersed diesel generation alternative for this region would require the addition of diesel generating units to each community over time in order to meet the forecasted peak loads. The other alternative, a transmission line originating at the Kelsey Generating Station as shown in Fig. 1, would connect Oxford House, Gods Lake, Gods River, Red Sucker Lake, Wasagamack, Garden Hill and St Theresa Point to the provincial power grid. While Manitoba Hydro (1993) considered other power generation methods such as small-scale hydro, central biomass and dispersed biomass, these alternatives were found to be uneconomical compared to the construction of a transmission line connection to the central power grid.

Within each community there are residential consumers and nonresidential consumers who pay for the energy consumed. Nonresidential consumers are composed of commercial and government facilities. As the diesel alternative creates higher energy costs than the land line alternative, each alternative may be seen to

Distributive fairness considerations in sustainable project selection 543

Fig. 1 Vicinity map of the seven remote communities (Manitoba Hydro, 1993).

result in different economic impacts among consumers. Other impacts are likely, such as particulate emissions associated with the diesel alternative, but are not considered here due to a lack of monitoring information for these facilities. Matheson (1997) estimates annual average energy costs from 1997 to 2037 (in 1993 Canadian dollars) for both residential and nonresidential customers in the seven communities that may result from the implementation of a project alternative and the three communities that are included for the sake of comparison. These estimates are based on historical demand data, population projections and historical rate records over a 25 year period.

Due to space limitations, the estimates of annual average energy costs are listed in Tables 1 and 2 for residential consumers and nonresidential consumers, respectively, for the years 1997, 2017 and 2037 only (in 1993 Canadian dollars). As Table 1 shows, both residential and nonresidential consumers in the nonremote communities of Cross Lake, Nelson House and Split Lake are not affected by the project alternatives since these communities are only included for comparison with the seven potentially impacted communities. Therefore, the annual average energy cost per megawatt-hour for these communities in any year remains constant over all alternatives. Also shown in Table 1, for a given community in a given year, is a slight decrease in annual average energy costs for residential consumers in the seven remote communities when the land line alternative is implemented in 1997. This may be a result of an annual average increase in demand per residential meter after a community switches to a land line power supply since Manitoba Hydro's land line

544 Sam Matheson et al.

Table 1 Average energy costs (1993 Can$ per megawatt-hour) for a residential consumer.

Community

Cross Lake Nelson House Split Lake Oxford House Gods Lake Gods River Red Sucker Lake Garden Hill St Theresa Point Wasagamack

Dispersed diesel: 1997

52.21 52.55 51.97 84.22 83.92 73.39 74.64 83.08 81.31 82.50

2017

51.42 51.43 50.98 73.43 73.83 61.95 63.21 74.52 71.50 72.58

2037

50.86 50.87 50.48 67.62 68.18 57.65 58.68 69.33 66.17 67.11

Land line: 1997

52.52 52.55 51.97 59.19 59.13 57.23 57.49 58.97 58.69 58.90

2017

51.42 51.43 50.98 52.07 52.06 51.77 51.81 52.04 52.00 52.03

2037

50.86 50.87 50.48 51.06 51.05 50.88 50.91 51.04 51.02 51.03

Table 2 Average energy costs (1993 Can$ per megawatt-hour) for a nonresidential consumer.

Community

Cross Lake Nelson House Split Lake Oxford House Gods Lake Gods River Red Sucker Lake Garden Hill St Theresa Point Wasagamack

Dispersed 1997

45.69 45.29 45.42

431.73 434.13 436.13 434.35 433.99 432.68 428.61

diesel: 2017

43.82 43.49 42.97

430.49 433.36 433.34 432.32 432.48 431.19 427.86

2037

43.04 42.78 42.20

429.81 432.84 431.99 431.28 431.43 430.39 427.48

Land line: 1997

45.69 45.29 45.42 55.39 61.13 66.32 61.93 60.88 57.58 48.81

2017

43.82 43.49 42.96 45.31 45.63 45.81 41.58 45.61 45.46 44.48

2037

43.04 42.78 42.20 43.59 43.69 43.75 38.89 43.69 43.64 43.28

residential rate structure reflects decreasing marginal costs to the consumer. A decrease in annual average energy costs for nonresidential consumers under the land line alternative is shown in Table 2 for a given remote community in a given year. This occurs due to the decrease in remote nonresidential energy rates to the current level of nonremote nonresidential energy rates.

The overall fairness measures presented in equations (10) and (11) are applied to this case study for an annual time step in order to examine the temporal variability of the average energy cost per megawatt-hour accruing to different energy consumers in different communities. When considering intra-temporal fairness as defined in equations (4)-(6), a given energy consumer type evaluates fairness based on the amount that customers of the same type pay in the other nine communities during a given year. When considering intertemporal fairness as defined in equations (7)-(9), an energy consumer in a community evaluates fairness based on comparing average annual energy costs that act on that consumer, in a given community during a given year, with the average annual energy costs for the same type of consumer in the same community over the remaining 39 years.

Distributive fairness considerations in sustainable project selection 545

Of the proportionality, equality and need based fair allocation objectives considered in this work, the equality approach is judged to be the most applicable for both intratemporal fairness and intertemporal fairness comparisons of user unit costs in this case study because the quantification of each group's contribution to the project and need for electricity require further examination. Allocation objectives of perfect proportionality and satisfaction of need may be important, however, when considering other impact distributions for this case study. The approach taken here is to apply the generalized distributive fairness measures in equations (10) and (11) to the residential consumers and the nonresidential consumers. Equations (5) and (8) are calculated by setting the number of groups, /, equal to 10; the number of time steps, T, equal to 40; and the number of different impact types, G, equal to 1. Here, the E(i, g, t, x) are the annual average energy cost per megawatt-hour data contained in Tables 1 and 2. Equations (10) and (11) are then applied by using the weight of unity for q2, since this work only focuses on equality and the residential and nonresidential data shown in Tables 1 and 2.

The results of this application are shown in Table 3. The intratemporal fairness measure, a(x), indicates that fairness, in terms of average annual energy costs per megawatt-hour, may increase for both the residential and nonresidential consumers if the land line alternative is selected because the average annual cost per megawatt-hour will be more equal to that experienced by the nonremote communities if the land line alternative is selected. The intertemporal fairness measure, yAx), does not indicate a difference among project alternatives as the average annual energy cost per megawatt-hour, for a given consumer group, remains fairly constant over time. The greatest intratemporal fairness increase associated with switching from dispersed diesel supply to a land line supply would be experienced by nonresidential consumers who are worse off under the existing dispersed diesel energy supply. This can be seen in Table 3 by comparing the fairness measure magnitudes of 0.26 and 0.02 for the dispersed diesel and land line alternatives, respectively. Of course, other types of impacts, such as impacts to health and safety may show a greater difference between alternatives for fairness considerations. The analysis of average annual energy costs per megawatt-hour is only one of a number of different impacts associated with each project alternative and thus represents a partial perspective of the perceived fairness present in this system. As data on these and other impacts such as environmental changes and reliability of power supply become available, for each alternative, further comparisons should be made before the analyst could advise the decision maker.

Table 3 Intratemporal and intertemporal measure magnitudes for each alternative.

a(x) if(x) Intratemporal Fairness Measure Intertemporal Fairness Measure

Alternative Residential Nonresidential Residential Nonresidential Dispersed diesel 0.08 0.26 0.01 0.01 Land line 0.01 0.02 0.01 0.01

546 Sam Matheson et al.

SUMMARY AND CONCLUSIONS

While intragenerational and intergenerational equity, or intratemporal and intertemporal fairness, respectively, are often advocated as being fundamental to the concept of sustainable development, these issues are rarely discussed in detail or applied to real engineering case studies. Sustainable development, fair allocation objectives, empirical distance-based fairness measures and their evaluation and the application of these measures to real problems, have been discussed here. This work develops generalized intratemporal and intertemporal distributive fairness measures that may be used as criteria for sustainable project selection and are based on the fair allocation objectives of perfect proportionality, perfect equality and satisfaction of needs. These generalized fairness measures incorporate the three common percep­tions of fair allocation by employing a weighted average of the extended proportionality, equality and need based fairness measures.

The generalized fairness measures are a first attempt at an approach for measuring distributive fairness in different cases and over long time horizons. It should be mentioned that the generalized fairness measures are mathematical abstrac­tions of a social objective function and should be considered empirical measures of fairness. They may serve as a starting point from which more refined fairness measures for sustainable project selection may be developed. Additionally, the formulation of these measures may serve to guide data collection efforts for the purposes of fairness evaluation for sustainable project selection.

The generalized fairness measures presented in this work may be used as criteria for sustainable project selection and may also be useful for other applications in civil engineering. For example, the generalized fairness measures might be modified and used to develop reservoir operating strategies that distribute reservoir related impacts in some fair manner. Efforts to explore this application are currently underway. At this time, the main limitations in applying this approach are seen to be the estimation of the weights required by the fairness measures and impact distributions that result from each project alternative. Impact assessment and impact valuation, particularly for impacts over long time horizons are daunting tasks and remain to be addressed.

The measures developed here are applied to a preliminary case study of the selection of either a dispersed diesel generation or land line based connection to a central power grid in order to meet energy demands in a region over a 40-year horizon. The fairness measures are applied to annual average energy costs per megawatt-hour accruing to residential and nonresidential energy consumers in seven communities as a result of implementing either project alternative. The distributive fairness measures formulated for this case study are based on a perfect equality fair allocation objective. The measures indicate that intratemporal fairness, regarding the distribution of average annual energy costs per megawatt-hour, particularly for the nonresidential energy consumers, may be increased by choosing the land line alter­native. However, this case study only illustrates the application of the generalized fairness measures to one project related impact distribution and should not be con­sidered a comprehensive fairness evaluation for this system. Clearly, additional project related impacts are possible and as data become available, a further évalua-

Distributive fairness considerations in sustainable project selection 547

tion must be conducted before one could present the complete analysis to a decision maker.

While fairness considerations have been discussed for at least two millennia by philosophers (see Aristotle in Thomson, 1985), fairness considerations in the context of civil engineering projects are seldom formally incorporated in actual project selection, and much work remains to be conducted in order to rethink an old concept in the context of real engineering problems. Research efforts should be directed at investigating the relevance of perfect proportionality and satisfaction of need allocation objectives in project selection and at identifying the factors that influence the relative importance placed on different fair allocation objectives. Additionally, much work remains in defining and measuring needs. Moreover, efforts toward reducing associated measurement, weighting and impact prediction uncertainties need to be undertaken and ultimately the various approaches for measuring fairness need to be validated. Civil engineering projects, particularly the provision of public services, may affect a large number of people, thus fairness considerations may be important for effective selection and implementation of such projects.

Acknowledgements The authors thank D. Burn, S. Simonovic and E. Onyebuchi for their insightful comments during many discussions of this work. Additionally, the authors greatly appreciate the excellent comments made by the independent reviewers of this article. The authors are grateful for the support provided by Manitoba Hydro under award no. G105; the University of Manitoba Research Development Fund; the Natural Science and Engineering Research Council of Canada under award no. 0GP041643; and the International Institute for Applied Systems Analysis, Laxenburg, Austria, under the Institute Scholars Program. The views expressed in this work are solely of the authors and Manitoba Hydro does not endorse this work in any way.

REFERENCES

Almond, B. (1995) Rights and justice in the environment debate. In: Just Environments: Inter generational, International and Interspecies Issues ed. by D. Cooper & J. Palmer. Routledge, New York.

Arthur, J. & Shaw, W.(1978) Justice and Economic Distribution. Prentice Hall, New Jersey. Beltratti, A. (1995) Intergenerational equity and environmental preservation. In: Energy, Environment and Economic

Growth, 42.95. University of Torino, Italy. Blalock, H. (1991) Understanding Social Inequality: Modeling Allocation Processes. Sage Publications, London. Brill, E., Liebman, J. & ReVelle, C. (1976) Equity measures for exploring water quality management alternatives, Wat.

Resour.Res. 12(3), 845-851. Cohen, J. (1978) Multiobjective Programming and Planning. Academic Press, California. Deutsch, M. (1975) Equity, equality and need: What determines which value will be used as the basis of distributive

justice./. Social Issues 31(3), 137-149. Gini, C. (1912) Variabitita e Mutabilita. (In Italian) Bologna, Italy. Harris, R. (1983) Pinning down the equity formula. In: Equity Theory ed. by D. Messick & K. Cook, 207-241. Prager,

New York. Heiner, K., Wallace, W. & Young, K. (1981) A resource allocation and evaluation model for providing services to the

mentally retarded. Manage. Sci. 27(7), 769-784. Jacobs, M. (1993) The Green Economy. University of British Columbia Press, Vancouver. McAllister, D. (1976) Equity and efficiency in public facility location. Geogr. Analysis 8, 47-63.

548 Sam Matheson et al.

McKerlie, D. (1989) Equality and time. Ethics 99, 475-491. Manitoba Hydro Corporation (1993) Review of North Central Options. A Generation Planning Technical Memorandum

no. 93/7. Marsh, M. & Schilling, D. (1994) Equity measurement in facility location analysis: A review and framework. Europ. J.

Operational Res. 74, 1-17. Matheson, S. (1997) Distributive Fairness Criteria for Sustainable Project Selection. MSc Thesis, Department of Civil

and Geological Engineering, The University of Manitoba, Winnipeg. Munasinghe, M. & Shearer, W. (1995) Defining and Measuring Sustainability: The Biogeophysical Foundations. The

United Nations and The World Bank, Washington. Nachtnebel, H., Eder, G. & Bogardi, I. (1994) Evaluation of criteria in hydropower utilization in the context of

sustainable development. In: Water Resources Planning in a Changing World (Proc. Int. UNESCO Symp., Karlsruhe, Germany), 13-24.

Pigou, A. (1912) Wealth and Welfare. Macmillan, London. Sampath, R. (1991) A Rawlsian evaluation of irrigation distribution in India. Wat. Resour. Bull. 27, 745-751. Savas, E. (1978) On equity in providing public services. Manage. Sci. 24(8), 800-808. Simonovic, S., Lence, B. & Burn, D. (1995) Sustainability criteria: An application to the hydropower industry. In:

Integrated Water Resources Planning for the 21st Century (Proc. ASCE 22nd Annual Conf. of Water Resources Planning and Management) ed. by M. Domenica, 173-176.

Theil, H. (1967) Economics and Information Theory. North Holland Publishing Company, Amsterdam. Thomson, J. (trans.) (1985) The Ethics of Aristotle: Nicomachean Ethics. Penguin Books, Harmondsworth, Middlesex,

UK. United Nations (1988) Assessment of Multiple Objective Water Resources Projects: Approaches for Developing

Countries. United Nations Press, New York. WCED (World Commission on Environment and Development) (1987) Our Common Future. Oxford University Press,

UK. Weiss, E. (1995) Intergenerational fairness for fresh water resources. Environ. Policy and Law 25(4/5), 231-236. Young, H. (1994) Equity In Theory and Practice. Princeton University Press, New Jersey. Young, M. (1992) Sustainable Investment and Resource Use: Equity, Environmental Integrity, and Economic Efficiency.

Man and the Biosphere Series, vol. 9. The Parthenon Publishing Group Inc., New Jersey.