Alemayehu Ambo** Keyword(s): Multicriteria decision making; New road projects; Attributes; Weights.

Transcription

1 PRIORITIZATION METHODOLOGY FOR PLANNING NEW ROAD PROJECTS IN DEVELOPING COUNTRIES* Alemayehu Ambo** Abstract The objectve of ths paper s to develop an alternate methodology to prortze new road projects n developng countres for future detaled economc and prelmnary engneerng studes. To date, an effectve and effcent methodology has not been employed and the selecton and prortzaton of new road projects has largely reled on assumptons rather than a systematc quantfcaton of road project attrbutes. The proposed methodology s not as complex and sophstcated as those developed for ndustralzed countres. It s rather desgned to effcently and effectvely handle data that are commonly avalable n developng countres. A hypothetcal data set of varables commonly avalable n studyng developng countres for road projects was used to test the proposed methodology. The same data were also used to test fve exstng methodologes that were developed n ndustralzed countres for makng decsons on projects, nvestments, strategc plans, etc and comparatve analyses were conducted on the results. The proposed methodology produced dentcal project rankngs as the two exstng methodologes and closely correlated wth the thrd methodology but was weakly correlated wth the fourth methodology. The ffth methodology was found to be the workable gven the data set. The alternate methodology was proved to be a valuable screenng tool to assst developng countres. Keyword(s): Multcrtera decson makng; New road projects; Attrbutes; Weghts. * The fnal verson of ths artcle was submtted n December ** Montgomery Street., Fredercton, New Brunswck E3B 2Y, Canada, e-mal: angtata@yahoo.com 25

2 Alemayehu Ambo: Prortzaton Methodology.... INTRODUCTION Road transport s vtal n developng countres where the majorty of the people lve n rural areas about 75 percent of the people are engaged n tradtonal agrculture although, value added from agrculture s low compared to other sectors. However, the road densty (km of road per km 2 of area) s stll very low and as a result, lmted access to agrcultural areas has become one of the major nhbtng factors to explot rural resources. Most road projects n developng countres are fnanced by regonal banks lke the Afrcan Development Bank, the Asan Development Bank and the Inter-Amercan Development Bank; by multlateral development banks lke the World Bank and the European Bank for Reconstructon and Development; and multlateral fnancng nsttutons lke the Nordc Development Fund. Securng funds from lendng and grantng agences requre a thorough analyss of project prortzaton as well as detaled apprasal studes. Benefcares are requred to produce well-prepared studes wth defendable results. Normally, road project studes begn at the plannng stages of dentfcaton and prortzaton, before ntensve feasblty studes are carred out. Ths paper focused on the plannng aspect of new road project studes. There are dfferent economc, socal and poltcal factors that are consdered durng the plannng stage of new road projects n developng countres, and one factor could be more mportant than the other regardng the natonal nterest. The economc factors could be mportant to a country as compared to socal or poltcal factors and there could even be dfferent degrees of mportance amongst the economc factors under consderaton. In addton, the purpose of a road can dffer from one locale to the other or from one country to the other dependng on economc and socal resources. In developng countres, new roads may be constructed for one or a combnaton of the followng: transportaton of agrcultural surplus, dstrbuton of food n defct areas, provson of access to people n remote areas, explotaton of forest and lvestock resources, regonal ntegraton, etc. Based on the above, ths paper deals wth the prortzaton of new road projects and marshalng them nto queue for future rgorous economc and engneerng feasblty studes. In ths regard, suffcent effort has not been done to develop a plausble methodology n developng countres, although a sgnfcant amount of new road network s requred. Ths paper s meant to contrbute n ths drecton. It s to be 26

3 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 noted that dfferent models, such as the Hghway Development and Management Tool (HDM-4), have been establshed for use n developng countres at the plannng and programmng stages of exstng road projects. But ths model apples to new road projects, not to the exstng ones, whch has not been gven attenton n developng countres. 2. EXISTING MULTI-CRITERIA DECISION MAKING APPROACHES IN DEVELOPED COUNTRIES There have been a number of research studes undertaken regardng mult-/multple crteron decson makng models (MCDM) based on data and arguments typcally avalable for developed countres. Trantaphyllou (2000) consdered dfferent MCDM methods and found the followng fve to be the most wdely used n developed countres: Weghted sum model (WSM), Weghted product model (WPM), Analytc herarchy process (AHP), Elmnaton and choce translatng realty (ELECTRE), and Technques for Order Preference by Smlarty to Ideal Solutons (TOPSIS). The followng sectons provde an outlne of these approaches. Weghted Sum Model (WSM) In respect of WSM, Trantaphyllou (2000) stated that f there are m alternatves and n crtera, the best alternatve s the one that satsfes the expresson: A * WSM - score = max n a j w j for =,2,3,..., m where: A * WSM-score s the WSM score of the best alternatve, n s the number of decson crtera, a j s the actual value of the -th alternatve n terms of the j-th crteron, and w j s the weghty of mportance of j-th crteron. Trantaphyllou further stated that 'addtve utlty assumpton' governs ths model,.e. the total value of each alternatve s equal to the sum of the products. WSM can be used wthout any dffculty wth sngle-dmensonal cases where all the unts are the 27

4 Alemayehu Ambo: Prortzaton Methodology... same (e.g. dollars, feet, seconds). However, the dffculty emerges when t s appled to mult-dmensonal MCDM problems. In ths regard, Trantaphyllou concluded that the addtve utlty assumpton s volated when dfferent unts are combned. Weghted Product Model (WPM) Regardng the weghted product model (WPM), the followng equaton was establshed by Trantaphyllou: w R (A K /A L ) = ( a / a ) j where: n s the number of crtera, a j s the actual value of the -th alternatve n terms of the j-th crteron, and w j s the weght of mportance of the j-th crteron. n k j l j As stated by Trantaphyllou, WPM s smlar to WSM except that WSM uses addton whle WPM deals wth multplcaton. In WPM, f the term R(A K /A L ) s greater than or equal to one, then A K s more desrable than A L. The best alternatve s the one that s greater than or at least equal to all other alternatves. Trantaphyllou further stated that snce the structure of WPM elmnates any unts, t can be used n sngle- and mult-dmensonal MCDM wth the advantage of usng relatve values nstead of actual ones. Analytc Herarchy Process (AHP) Accordng to Trantaphyllou the analytc herarchy process (AHP) model decomposes a complex MCDM problem nto a system of herarches wth the fnal structure dealng wth m x n matrx where, m s the number of alternatves and n s the number of crtera. In ths case, the matrx s constructed by usng the relatve mportance of the alternatves n terms of each crteron. The best alternatve s: n A AHP score = aj max w for =,2,3,..., m j As stated by Trantaphyllou, the dfference between WSM and AHP s that the former uses actual values whle the latter most rely on relatve values. In addton, the vector 28

5 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 (a j, a 2, a 3,..., a n ) for each s the prncpal egenvector of an n x n recprocal matrx whch s determned by parwse comparsons of the mpact of the m alternatves on the -th crteron. Elmnaton and Choce Translatng Realty (ELECTRE) Trantaphyllou says that the ELECTRE method focuses on "outrankng relatons" wth par-wse comparsons usng separate alternatve crtera. For example, under two alternatves A and A j, the outrankng relatonshp between the two descrbes that even when the th alternatve does not domnate the j th alternatve quanttatvely, the decson-maker may stll take the rsk of regardng a as better than a j. Ths means that one alternatve s domnated by another whch makes t excel n one or more crtera whle remanng equvalent n other crtera. However, Trantaphyllou noted that snce the method yelds a system of bnary outrankng relatons between alternatves where the system s not necessarly complete, the method s sometmes unable to dentfy the most preferred alternatve, only producng a core of leadng alternatves. Regardless of the above shortcomngs, the method has a clear vew of alternatves by elmnatng less favorable ones whch s convenent when there are decson problems that nvolve a few crtera wth a large number of alternatves. Technques for Order Preference by Smlarty to Ideal Solutons (TOPSIS) Accordng to Trantaphyllou, TOPSIS was developed as an alternatve to the ELECTRE method. The basc concept of TOPSIS s that the selected alternatve should have the shortest dstance from the deal soluton and the farthest dstance from the negatve-deal soluton n a geometrcal sense. The method assumes that each crteron has a tendency of monotonously ncreasng or decreasng utlty wth easy defntons of deal and non-deal solutons. Under ths method, a normalzed decson matrx s constructed frst (the same as under ELECTRE method) followed by a weghted normalzed decson matrx. Then, deal and negatve-deal solutons are determned followed by separaton measures. The best alternatve s the one that has the shortest dstance to the deal soluton. Saaty (980) proposed the AHP method n decson makng n order to determne strengths wth whch varous elements n one level nfluence the elements on the next 29

6 Alemayehu Ambo: Prortzaton Methodology... hgher level so that the relatve potency of the mpacts of the elements of the lowest level on the overall objectves can be computed. He descrbed the development of weghts to be assocated wth ndvdual actvtes, based on group quantfed judgments and sad that f there are n actvtes (C, C 2,..., C n ), the quantfed judgments on pars of actvtes (C, C j ) are represented by n-by-n matrx. Saaty further stated that, there are two rules for entres of elements n the matrx: rule one says a j s the recprocal of a j, and rule two says, f C s judged to be of equal mportance to C j, then, both a j and a j wll be equal to. Fnally, the weghts and entres are treated to develop selecton crtera. Saaty proposed crude estmates of vector calculaton as follows: ) Add the elements n each row and normalze them by dvdng each sum by the total of all the sums, so that the results add up to unty. The frst entry of the resultng vector s the prorty for the frst actvty; the second entry for the second actvty, and so on. ) Take the sum of the elements n each column and form the recprocals of these sums. Dvde each recprocal by the sum of the recprocals to normalze them so that thus add to unty. ) Dvde the elements of each column by the sum of that column and the elements n each resultng row and dvde ths sum by the number of elements n the row. Ths s a process of averagng out the normalzed columns. v) Multply the n elements n each row and take the nth root normalze the resultng numbers. Pomerol and Barba-Romero (2000) also dealt wth a number of mult-crtera decson makng models. They stressed the absence of a recognzed rule for qualtatve crtera and thus, proposed normalzaton procedures. They stated that f evaluaton of m alternatves are gven by a j for a gven crteron j, a normalzaton procedure wll transform the vector (a j, a 2j,..., a mj ) nto a normalzed vector ( υ j, υ 2j,..., υ mj ) usng one of the followng four procedures: Procedure : υ = a max a procedure 2: υ a mn a = max a mn a where : 0<v 0 v 30

7 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 Procedure 3: a a Procedure 4: ( a a 2 2 ) / where : 0<v < 0<v < Pomerol and Barba-Romero further suggested that the nterpretaton of Procedure s made by consderng the percentage of maxmum a; Procedure 2 as percentage of range (max a - mn a ); Procedure 3 as percentage of the total sum, and procedure 4 as the th component of unt vector. After normalzaton, weghts (w 's) are calculated wth the sum of. Fnally, by applyng the weghted sum model, evaluatons are made such that R(a ) = w ja, (=,2,..., m) and the chosen alternatve a j wll be j the one whch gets the hghest value of R(a ). If there s a te, ether one can be taken. 3. MCDM APPLICATION IN THE PLANNING OF NEW ROAD PROJECTS IN DEVELOPING COUNTRIES Applcaton of MCDM n Developng Countres MCDM models can ncorporate both qualtatve and quanttatve data. As stated above, the three general factors that are employed n the selecton of development projects are: Alternatves whch focus on road project; crtera whch are attrbutes of road projects such as resources potental; and factor weghts. The frst two are quanttatve n nature whle the thrd one s qualtatve. Generally, the decson makng process nvolves fve stages: road project dentfcaton, establshment of attrbutes, development of weghts, analyses and decson makng. Road Project Identfcaton Ths actvty s usually undertaken by government and publc agences. Agrcultural offces, both publc and prvate rely on road nfrastructure to transport farm nputs to and outputs from development centers. Relef agences requre road nfrastructure to 3

8 Alemayehu Ambo: Prortzaton Methodology... transport emergency food and other commodtes to affected areas. Regonal admnstratve offces requre publc transport to provde access to the people n remote areas and to foster regonal ntegraton. Potental road projects are dentfed by desgnated government offces takng nto account all these actvtes. Establshment of Attrbutes Ths secton deals wth the dentfcaton of relevant crtera whch provde the bass for project evaluaton and the quantfcaton of attrbutes. Common natonal economc, socal and other crtera can be establshed for desgnng all road projects. Although these crtera dffer from country to country, they are more or less generc for most rural road projects. In developng countres, roads may be requred to transport one or a combnaton of the followng: subsstence/cash crop surplus to market centers, relef supples to deprved areas, passengers to ther destnatons, and forest and lvestock surplus to ndustres and consumpton centers. Roads may also be needed for purposes of regonal ntegraton whch could be entrely poltcal. Once approprate crtera are dentfed, pertnent government agences can undertake the quantfcaton work. Agrcultural nsttutons can estmate cash and subsstence crop as well as lvestock and forest reserve potentals; relef agences and nongovernment organzatons can calculate the prevalent and long-term nutrent requrements n food defcent areas; statstcal offces can estmate the number of people n project areas; and mappng nsttutons can estmate how far dentfed project areas are located from the nearest urban centers. The next step s to establsh weghng factors for each crteron. Ths s dffcult due to the subjectve nature of the task. Judgments could be napproprate due to bas and msunderstandng of the facts. Ideally, the natonal nterest must be the ultmate objectve. Weghts should be gven for each attrbute/crteron after a careful comparson wth others. In general, all attrbuters are not consdered to have equal values. For example, a country can be n need of foregn currency and thus exportable agrcultural produce such as coffee and cocoa may be valued more than other agrcultural produce such as wheat, maze, etc. On the other hand, the concern of a country may be the poltcal strengthenng of the naton; thus, regonal ntegraton may be gven a hgher prorty. Thus, weghts can be assgned to each attrbute based on the partcular need of a developng country or regon. 32

9 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 However, reasonable estmates can be developed f knowledgeable and unbased ndvduals are nvolved n the task. As stated by Souder (980), a Delph method can be employed snce the opnons of several experts can sometmes be 'averaged' to cancel out the errors/bases of ndvdual predctons. Smlarly, Nutt (984) noted that a Delph survey systematcally solcts and collates judgments to form a group consstng of dfferent organzatons. Accordngly, partcpants can be selected from government organzatons, unverstes and prvate agences. Experts from the offces of fnance, central plannng, statstcs, transport, agrculture, ndustry, trade and relef agences can represent governments; economsts, statstcans and researchers can represent hgher academc nsttutons; and traders, transport operators, consumers and farmers can represent prvate stakeholders. Accordng to Nutt, a seres of questonnares can be used. The ntal questonnare asks general questons wth subsequent questonnares bult on the responses to the precedng questonnare. Offces of central plannng and road authorty can be mandated to manage the study. Weghts can be absolute or normalzed values. The analyses can be done by road authortes/mnstres of transport, systems analysts or consultants and the study can be montored by a steerng commttee. Once the study s completed, decsons can be taken jontly by the offces of road agences/mnstres of transport, central plannng agences and hgher level polcy makers. Proposed Alternate Methodology (AM) An alternate methodology (AM) s developed consderng the requrements of developng countres wth a focus on the resource types that are commonly utlzed n those regons. The methodology focuses on attrbutes of new road project wth the allocaton of weghts to prortze and streamlne the projects for future detaled economc and engneerng studes. The approach developed through ths paper s stated below: Assumptons: (a) Let P denote a set of dentfed new road projects, (b) P s composed of a fnte number of n alternatves, thus, P=(x, x 2,..., x n ) (c) Let A denote a set of attrbutes (d) A s composed of a fnte number of m attrbutes. Thus, B represents the 33

10 Alemayehu Ambo: Prortzaton Methodology... attrbutes of the subject roads wth m number of attrbutes. Thus, B=(B, B 2,..., B m ). Next, weghts are allocated. For example: w for attrbute B, w 2 for attrbute B 2 and so on untl w m for B m. Thus, normalzed weghts would be: w 2 + w w m = () (Note: The values of w' s can be whole numbers and the sum can be greater than one provded that the ratos between them reman the same). Then, weghts are dstrbuted among road projects based on the magntude of attrbutes. The road projects and attrbutes are llustrated n the matrx below: Attrbutes Projects Sum of Attrbutes x x x n x n j x 2 x x 2n x n x j x j x jn x m x m x mn 2 j n x j = n x m j For each cell, a score s calculated based on the followng formula: V j m x j x jn w j (2) 34

11 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 where: =,2,3,..., n j =,2,3,..., m w = weghts Subsequently, all scores for each road project are summed up as follows: m P = v j = v + v v m m P 2 = v2 j = v 2 + v v m P n = m v j = v n + v 2n v mn On the bass of ths calculaton, the project wth the hghest aggregated score wll be selected to be the frst for detaled feasblty study. Ths methodology s ntutve and can be utlzed wthout dffculty by experts n developng countres. The methodologes formulated for use n ndustralzed countres are more complex and mathematcally sophstcated, requrng traned experts, currently scarce n developng countres. In addton, most of the exstng MCDM models cannot be appled n developng countres wthout modfcaton. Analyses and Comparson of Results Data whch were postulated to descrbe the attrbutes of varous road projects are presented n Table. The attrbutes are the major crtera that are normally consdered durng road project evaluatons n developng countres. MCDM analyses were carred out usng the proposed alternate methodology (AM). The results generated by AM were compared wth those calculated usng the WSM, WPNM, AHP, Saaty and TOPSIS methods. The ELECTRE method was not used snce TOPSIS s smply an mproved verson of t. In addton, as mentoned by Trantaphyllou, the ELECTRE, method yelds a system of bnary outrankng relatons between the alternatves where the system s not necessarly complete and sometmes t s unable to dentfy the most preferred alternatve only producng a core 35

12 Alemayehu Ambo: Prortzaton Methodology... of leadng alternatves. Thus, the EL/ECTRE method has not been used for the purposes of comparson. Fve new road projects were consdered wth seven economc attrbutes. The weghts are random values based on assumed mportance of the attrbutes. The values n brackets (n Table ) are randomzed fgures whose sum equals. Analyses were done usng the AM method consderng both randomzed and non-randomzed weghts. Analyses were carred out usng the followng sx methods: Alternate Methodology (AM), Weghted Sum Model (WSM), Weghted Product Model (WPM), Analytc Herarchy Process (AHP), Saaty and Technque for Order Preference by Smlarty to Ideal Soluton (TOPSIS). The proposals by Saaty descrbed above as (a), (b) and (d) do not work usng ths data. Ths s because (a) and (b) nvolve addton of attrbutes whch s mpossble n ths case, snce for example, crop surplus and populaton cannot be added together. Smlarly, (d) s not workable snce t nvolves the multplcaton of attrbutes whch result n values of zeros because of the "0" values provded n the data. The results ncorporatng values and rankngs are shown n Table 2. As shown n Table 2, AM (non-normalzed) and AHP produce dentcal results, WSM and TOPSIS produce dfferent results from the former two. It was not possble to conduct analyses usng WPM because WPM utlzes ratos of attrbutes. Ratos cannot be produced wth "0" values of denomnators. WSM produced dfferent values from AM and AHP because the degree of magntudes of attrbutes were not consdered. Instead, multplcatons of weghts and attrbutes are consdered whle AM and AHP apply weghts after respectve attrbutes have been dvded by the correspondng sums. The relatve weghts of the attrbutes wthn each project alternatve need to be calculated and should be used as a mechansm of prortzaton. Ths s so because the relatve magntudes of attrbutes wthn projects can be measure and ncorporated. In the case of TOPSIS, the dstances for deal and negatve-deal solutons were measured from the maxmum and mnmum values, respectvely, based on a normalzed decson matrx. The model, however, consders beneft and cost/loss crtera as fcttous alternatve to measure solutons whch do not usually appear n attrbutes of road projects. Rankng under AM, AHP and Saaty (v) are smlar. 36

13 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 Spearman rank correlatons were calculated usng Anderson et al. (98) to measure the degree of closeness between the rankngs. The correlaton between AM/AHP/Saaty (v) and WSM was 0.2 whch s weak, the correlaton between AM/AHP/Saaty (v) and TOPSIS was 0.65 whch s reasonably strong whle the correlaton between WSM and TOPSIS was whch s negatvely correlated ndcatng that the rankng are opposte to the rankng of AM/AHP/Saaty (v), under the gven data. Table : Hypothetcal Projects and Correspondng Attrbutes Attrbutes Subsstence Crop Surplus (ton) Subsstence Crop Defct (qts.) Projects A A2 A3 A4 A Populaton Cash Crop Surplus (qts) Forest Reserve (m 3 ) Lvestock (no. of cattle) Regonal Integraton (dstance from nearest urban centre n km) Weght (w) 4 (0.6) 3 (0.2) 4.5 (0.7) 5.0 (0.9) 3.5 (0.3) 3.5 (0.3) 2.5 (0.0) Table 2: Score and Ranks of Hypothetcal Road Projects Models Project AM WSM WPM AHP Saaty (c) TOPSIS Value Value R R Value R Value R Value R Value R ('000) A 0.2* NW NW A NW NW A NW NW A NW NW A NW NW * Based on Normalzed weghts + Based on Non-Normalzed weghts, NW= Not Workable, R=Rank (Note: The results under Saaty (c) are the same wth results under AHP/Non-Normalzed weghts under AM dvded by the number of attrbutes whch s 7 n ths case.)] 37

14 Alemayehu Ambo: Prortzaton Methodology CONCLUSIONS In developng countres, approprate prortzaton methodologes for new road projects are not yet developed. The methodology proposed n ths paper helps to streamlne potental road projects for future detaled economc and prelmnary engneerng feasblty studes n developng countres. Usng typcal varables mportant to road projects n developng countres, a seres of analyses were conducted utlzng the proposed alternate methodology (AM) and contrasted wth the exstng models developed specfcally for developed countres. The exstng models utlzed n the analyses are: Weghted Sum Method (WSM), Weghted Product Method (WPM), Analytc Herarchy Process (AHP), Saaty () and Technque for Order Preference by Smlarty to Ideal Soluton (TOPSIS). The WPM model was found to be the workable snce t ncludes "0"s as denomnators n ts calculatons. The results under WSM were dfferent from AM snce comparsons amongst the attrbutes wthn each road alternatve were not conducted as the model proposes. The results under TOPSIS were also dfferent because ths method requres data descrbng beneft and cost/loss crtera whch are not common n road project attrbutes n developng countres. The results under AM (non-normalzed) and AHP were dentcal. Regardng rankngs, AM, AHP and Saaty () produced smlar results. AM, AHP and Saaty () are plausble for use n developng countres although the last two are more sophstcated requrng greater effort. Saaty () s the same as AM (non-normalzed) and AHP except that t proposes values to be dvded by the number of attrbutes. AM produced normalzed as well as non-normalzed values. It s smple to use and does not requre sophstcated computer utlzaton as proposed by Saaty and Trantaphyllou. AM can easly handle data normally used n developng countres wth no complcaton. 38

15 Ethopan Journal of Economcs, Volume IX No., Aprl 2000 REFERENCES Anderson, D.R., Sweeny D.J. and Wllams T.A Introducton to Statstcs: An Applcaton Approach, West Publshng Company, New York. Cook, W.D. and Kuhn, T.E. eds Plannng Processes n Developng Countres: Technques and Achevements. North-Holland Publshng Company, Amsterdam. Dalkey, N. et al The Delph Method, IV: Effect of Percentle Feedback and Feed-n of Relatve Facts. Rank Corporaton, Calforna. Ftch, K. et al The Rand/UCLA Approprateness Method User's Manual, Prepared for Drectorate General XII, European Commsson, Calforna. Nutt, P.C Plannng Methods for Health and Related Organzatons. A Wley Medcal Publcaton. John Wley & Sons, New York. Pomerol, Jean-Charles and Barba-Romero, S Multcrteron Decson n Management, Prncples and Practce, Kluwer Academc Publshers, Boston/Dordrecht/London. Retveld, P. 98. Multple Decson Methods and Regonal Plannng, North-Holland Publshng Company, Amsterdam. Saaty, T.L The Analytc Herarchy Process, McGraw-Hll Internatonal Book Company, New York. Souder, W.E Management Decson Methods for Managers of Engneerng and Research, Van Nostrand Renhold Company, New York. Trantaphyllou, E Mult-Crtera Decson Makng Methods: A Comparatve Study, eds. P.M. Pardalos and D.Hearn. Kluwer-Academc Publshers, Boston. World Bank World Development Indcators Database, July Washngton, D.C. Zeleny, M Multple Crtera Decson Makng, McGraw-hll Book Company, New York ed. Multple Crtera Decson Makng Kyoto 975 Lecture Notes n Economcs and Mathematcal Systems, New York. 39