Sustanablty 2014, 6, 2658-2684; do:10.3390/su6052658 Artcle OPEN ACCESS sustanablty ISSN 2071-1050 www.mdp.com/journal/sustanablty Low Carbon Suppler Selecton n the Hotel Industry Cha-We Hsu 1, *, Tsa-Ch Kuo 2, Guey-Shn Shyu 1 and P-Shen Chen 3 1 2 3 Department of Travel and Eco-toursm, Tungnan Unversty, New Tape Cty 22202, Tawan; E-Mal: gsshyu@mal.tnu.edu.tw Department of Industral and Systems Engneerng, Chung Yuan Chrstan Unversty, 200 Chung Pe Road, Chung L 32023, Tawan; E-Mal: tckuo@cycu.edu.tw Insttute of Envronmental Engneerng and Management, Natonal Tape Unversty of Technology, Tape Cty 106, Tawan; E-Mal: t102608004@ntut.edu.tw * Author to whom correspondence should be addressed; E-Mal: jcwhsu@mal.tnu.edu.tw; Tel.: +886-286-625-958 (ext. 734); Fax: +886-286-625-957. Receved: 29 January 2014; n revsed form: 18 Aprl 2014 / Accepted: 29 Aprl 2014 / Publshed: 7 May 2014 Abstract: Ths study presents a model for evaluatng the carbon and energy management performance of supplers by usng multple-crtera decson-makng (MCDM). By conductng a lterature revew and gatherng expert opnons, 10 crtera on carbon and energy performance were dentfed to evaluate low carbon supplers usng the Fuzzy Delph Method (FDM). Subsequently, the decson-makng tral and evaluaton laboratory (DEMATEL) method was used to determne the mportance of evaluaton crtera n selectng supplers and the causal relatonshps between them. The DEMATEL-based analytc network process (DANP) and VlseKrterjumska Optmzacja I Kompromsno Resenje (VIKOR) were adopted to evaluate the weghts and performances of supplers and to obtan a soluton under each evaluaton crteron. An llustratve example of a hotel company was presented to demonstrate how to select a low carbon suppler accordng to carbon and energy management. The proposed hybrd model can help frms become effectve n facltatng low carbon supply chans n hotels. Keywords: suppler selecton; carbon management; hotel ndustry; FDM; DANP; VIKOR
Sustanablty 2014, 6 2659 1. Introducton Wth the ncreased conscousness on the ssue of clmate change, the mplementaton of energy conservaton and carbon reducton n the hotel ndustry has become sgnfcant to address global warmng [1 5]. The hotel ndustry, a major sub-sector of the toursm ndustry, consumes a sgnfcant amount of energy, whch equates to the amounts of ndrect greenhouse gas (GHG) emssons assocated wth the energy consumpton of the hotel sector [6,7]. The Tawan Green Productvty Foundaton [8] reports that the top 50 most ntensve energy users n Tawan s hosptalty ndustry, mostly tourst hotels, produced 363,810 tons of carbon emssons n 2008. To acheve the target of low-carbon operatons, hotel companes have adopted ether ISO 50001 (energy management systems) or ISO 14064 (greenhouse gas systems) to ncrease energy effcency and mtgate carbon emsson. These companes nclude the Marrott Washngton DC Hotel, Regal Arport Hotel n Hong Kong, NH Hotels, Mramar Garden Hotels, and Evergreen Hotels n Tawan. The World Busness Councl for Sustanable Development and the World Resources Insttute ndcate that at least 80% of carbon emssons are produced n the total supply chan [9]. Ths fndng s consstent wth that of Sundarakan et al. [10], who emphaszed that carbon emsson across stages n a supply chan consttutes a sgnfcant threat that warrants careful attenton n the desgn phase of the supply chan. In controllng the carbon footprnt across a supply chan, Wttneben and Kyar [11] underlned that GHG emssons from supplers need to be consdered to adequately assess the contrbutons of busnesses to clmate change. The 2010 supply chan report of the Carbon Dsclosure Project states that more than half of ts surveyed members expressed that n the future, they wll cease dong busness wth supplers that do not manage ther carbon emssons [12]. Ths fndng mples that carbon footprnt can affect the optmal choce on sourcng decsons [13,14], operatons decsons n nventory management [15], and product development [16]. Low-carbon suppler management s clearly a crtcal actvty n purchasng management to acheve low-carbon operatons wthn the hotel ndustry. Bonlla-Prego et al. [17] ponted out that tour operators are requred to measure and manage the carbon performance of ther supplers. Teng et al. [5] stated that selectng a suppler that adopts energy conservaton and carbon reducton, workng wth local farmers or vendors to reduce food mles, and purchasng local or seasonal food and products/materals can facltate low-carbon hotel operatons. Accor launched Accor Procurement Charter 21 and ntegrated sustanable development crtera nto all phases of ts suppler relatons, from specfcatons n ts calls for bds to specfc clauses ntegrated nto suppler certfcaton contracts. At the end of 2012, more than 2000 certfed supplers 60% of the total sgned Accor Procurement Charter 21. Accor Hotels requres ther supplers to evaluate the envronmental mpact that ther stes, products, and servces exert on the envronment and to set objectves on the quanttatve reducton of GHG emssons [18]. Reflectng these trends, companes n the hotel ndustry must therefore requre ther supplers to oversee ther GHG emssons and energy management for a long-term collaboratve partnershp n the low-carbon supply chan. Recently, suppler selecton and evaluaton of carbon management has become mportant n makng low carbon purchasng decsons [5,19 24]. Nevertheless, to the best of our knowledge, suppler selecton that specfcally consders carbon or energy management competence n the hotel ndustry s rarely found n prevous lterature. A few studes have attempted to ncorporate carbon management
Sustanablty 2014, 6 2660 nto the process of suppler selecton n specfc manufacturer ndustres [20,25,26]. By ncorporatng the carbon performance nto the suppler selecton process, Hsu et al. [20] proposed a framework that develops a carbon management model wth 13 crtera used to manage supplers n the Tawanese electroncs ndustry. Ther study used the Decson-makng Tral and Evaluaton Laboratory (DEMATEL) approaches to recognze the nfluental crtera of carbon management and mprove the overall carbon performance of supplers. Later, Shaw et al. [24] ncluded the crteron of carbon emsson n suppler selecton to develop a low carbon supply chan n the Indan garment manufacturng. The fuzzy analytc herarchy process was appled before analyzng the weghts of crtera, and the fuzzy mult-objectve lner programmng was used for suppler selecton. Ths formulaton ntegrates carbon emsson nto the objectve functon and takes the carbon emsson cap (Ccap) of sourcng as a constrant whle selectng a suppler. Smlarly, n terms of optmzng green supplers, Peng [26] ntegrated the crteron of energy consumpton nto green suppler selecton n a large manufacturng enterprse. The analytcal herarchy process (AHP) and grey relatonal analyss were used to evaluate green supplers. To construct a green and low carbon suppler evaluaton model, Lee et al. [25] used the fuzzy analytc network process to evaluate varous aspects of supplers. Goal programmng was then appled to allocate the most approprate amount of orders to each of the selected suppler. Cho [27] proposed a two-stage optmal suppler selecton scheme n whch phase one flters the nferor supplers and phase two helps to select the best suppler among the set of non-nferor supplers by mult-stage stochastc dynamc programmng. The mpacts brought by dfferent formats of carbon emsson tax are explored. Suppler selecton and evaluaton s a mult-crtera decson-makng (MCDM) problem [28,29] that provdes an effectve framework for comparng supplers. In the current study, a hybrd MCDM model s proposed to dentfy the evaluaton crtera of carbon performance usng the Fuzzy Delph Method (FDM). By consderng the nterrelatonshp between crtera, the decson-makng tral and evaluaton laboratory (DEMATEL) method s used to recognze cause-effect relatonshps and to construct the cognton map of the evaluaton crtera. The DEMATEL based on an analytc network process (ANP), also called the DANP method, s used to calculate the nfluence weghts of the crtera. Fnally, the VIseKrterjumska Optmzacja I Kompromsno Resenje (VIKOR) wth DANP weghts s used for the evaluaton of the carbon performance of supplers and to determne performance scores and gaps. An llustratve example of a hotel frm n Tawan s used to demonstrate the proposed framework for approprate suppler selecton n terms of carbon management. The remander of ths paper s organzed as follows. Secton 2 revews the lterature on suppler selecton based on carbon performance. Sectons 3 brefly descrbes the FDM method, the DEMATEL method, the DANP nfluental weghts, and the VIKOR technque, whch are used to buld a hybrd MCDM model for selectng a low-carbon suppler. An emprcal case of a hotel company s used to demonstrate the proposed model n Secton 4. We present and dscuss the results of proposed framework n Secton 5. The concluson and suggeston for future research are presented n Secton 6.
Sustanablty 2014, 6 2661 2. Carbon Management Crtera n Suppler Selecton Several useful crtera assocated wth carbon management and ther categores are ponted out n the lterature. Informaton about them was utlzed to construct a framework for competency n carbon management aware suppler selecton n hotel supply chan. Twelve crtera were fnally ncluded. 2.1. Energy Effcency of Products In mplementng energy management systems of ISO 50001-certfed, organzatons wll requre ther supplers to provde energy effcency nformaton on ther products or equpment [30]. Green hotel assocatons and some government webstes provde nformaton on the energy effcency of products, such as prntng paper, tolet/tssue paper, computers, refrgerators, ar condtoners, and employee unforms [5]. Wth the avalablty of energy effcency nformaton on products, hotel operators can purchase hghly effcent products and facltes nstead of those wth hgh-energy consumpton to acheve low carbon operaton. 2.2. Eco-labelng of Products Hotel operators that adopt green purchasng can reduce energy consumpton and smultaneously reduce operatng costs [31]. For example, the Energy Star program has sgnfcantly reduced economc costs and CO 2 emssons assocated wth electrcty consumpton [32]. The products of supplers are qualfed by eco-labels, such as the energy-savng label, green mark, and water-savng label, hotel operators can mplement green purchasng to reduce energy consumpton. 2.3. Carbon Accountng and Inventory Carbon accountng and nventory s an essental step n developng strateges for controllng GHG emssons and evaluatng ts progress n the operatons of a company, n products, and n supply chan, as companes need to know ther current stuaton [33]. Cogan et al. [34] found that more than 60% of the evaluated companes conducted a GHG emssons nventory. 2.4. Energy Reducton of Food Processng In the food ndustry, hgh levels of energy consumpton are necessary for key operatons, such as food preservaton, santaton, processng, and storage [35]. For example, the U.S. food ndustry consumes 7% of the total electrcty used by the manufacturng sector. Therefore, about 15% of the total energy requrements of the food ndustry are from electrcty [36]. To show an example of fcttous slaughter and meat processng, Frtzson and Berntsson [37] performed dfferent energy effcency measures, such as ncreasng the heat exchanger networks and heat pumps, to acheve the target reducton of 5% and 35% of the total CO 2 emssons. Consderng the low carbon supply of food avalable n the hotel ndustry, supplers from food processng supplers must embrace dfferent measures to save energy and reduce carbon emssons.
Sustanablty 2014, 6 2662 2.5. Carbon Governance Over 90% of Carbon Dsclosure Project (CDP) members have tasked ether a board commttee or another executve body wth the overall responsblty of clmate change management to ensure that the strategy s effectvely mplemented [12]. Companes that ntegrate clmate change nto ther board and executve structures, as well as ther publc reportng mechansms, are far more lkely to mantan long-term commtments and the comprehensve approaches necessary to effectvely address clmate change rsks and opportuntes across ther entre busness structure [34]. 2.6. Carbon Polcy The CDP [10] reveals that ts members have ntegrated carbon polces nto ther procurement departments and that majorty of these companes (90%) have a carbon emsson reducton plan n place. Accordngly, companes can facltate carbon management practces by establshng a carbon polcy as a manfestaton of ts poston on carbon emssons dsclosure, carbon reducton targets, and carbon emssons certfcaton, among others. Moreover, by mplementng the energy management systems standard ISO 50001, companes wll be able to mplement an energy polcy [38]. 2.7. Carbon Reducton Targets In terms of the mtgaton of clmate change, Wenhofer and Hoffmann [39] argue that GHG reducton targets reflect a long-term need to decrease emssons. Settng targets to reduce GHG emssons has become the norm n corporate clmate change strateges, whch nclude quanttatve emsson reducton targets for ther Scopes 1 and 2, and occasonally even Scope 3, GHG emssons [34]. A company must set ts carbon reducton target at a suffcently hgh level to enable authentc and measurable progress n addressng clmate change. 2.8. Carbon and Energy Management Systems To mtgate carbon emssons, frms attempt to acqure dfferent certfed standards assocated wth carbon and energy management systems. Recently, most companes have appled varous standards on carbon management, such as ISO 14064-Parts I and II and PAS 2050, to conduct nventores and account for GHG emssons. Energy management s the combnaton of energy effcency actvtes and technques, and the management of related processes that result n lower energy costs and CO 2 emssons [40]. Ates and Durakbasa [38] pont out that the energy management system ISO 50001 s expected to compel ndustral organzatons to examne the systems and processes requred to ncrease ther energy performance, energy effcency, and ntensty. 2.9. Transport Effcency The energy effcency and carbon emssons of transportaton should be consdered to facltate the creaton of a low carbon supply chan wthn the hotel ndustry, as transportaton s requred for the mass delvery of food, consstent wth Teng et al. [5]. Ther study argues that food and beverage operators should be aware of ther carbon footprnt and reduce t, as well as mprove the energy effcency of
Sustanablty 2014, 6 2663 road freght transport. For example, the energy requrement contrbuton of transportng foodstuffs for breakfast s sgnfcant [2]. 2.10. Collaboraton of Supplers Workng wth supplers to green supply chan n hotel sector, Internatonal Toursm Partnershp [41] argued that hotel operators should encourage local busnesses to cut down on transport energy by sourcng locally. Clmate change s not a sngle ssue that can be addressed by only one company or even one sector. Companes need to collaborate wth ther suppler to clmate change of adaptaton and mtgaton. Accordng to Scott and Becken [42], Carla Agurre from VstSweden reported on ther experence to encourage and motvate potental supplers, and show leadershp on sustanablty and clmate change ssues. 2.11. Carbon Reducton and Energy Conservaton Measures To mtgate carbon emssons, most companes no longer concentrate solely on nfluencng polcy debates. Instead, they have begun to pursue varous frm-specfc practcal actons aganst clmate change wthn the framework of a corporate clmate strategy [43]. Companes can take nternal and external measures on ther carbon doxde emssons [33,44,45]. Internal measures are usually defned as actvtes wthn the busness operatons of the company, whereas external measures represent emsson compensaton measures [39]. 2.12. Food Mle Management Food mles are usually explctly lnked to carbon accountng and clmate change [46]. Internatonally, the demand of the toursm sector for food and ts assocated food mles have a sgnfcant mpact on GHG emssons and thus have mplcatons for clmate change [47]. Through the trackng of food mles and assocated sources, Pratt [48] concludes that ecotoursm operatons, such as those wthn the hotel ndustry, have dentfed and mproved ther sustanablty and ecologcal footprnt by mnmzng GHGs. 3. Buldng a Hybrd MCDM Model of Low Carbon Suppler Selecton The methodology of constructng an evaluaton framework for selectng a low carbon suppler n the hotel ndustry for ths study has three phases. The frst phase emphaszes the dentfcaton of crtera to evaluate the carbon management competence of supplers. In ths study, fve managers from hotel frms and three unversty professors were nvted to screen and ft the crtera usng FDM technques. In the second phase, after dentfyng the consstency of crtera, the DANP method was used to examne the nterrelatonshp between and the nfluental weghts among the crtera. Fnally, VIKOR was used to rank the supplers of an llustratve hotel company n terms of carbon management competence.
Sustanablty 2014, 6 2664 3.1. Recognzng the Evaluaton Crtera by FDM Method The Delph Method has been wdely used and recognzed for makng predctons and for decson-makng snce ts ntroducton n 1963 by Dalkey and Helmer at the RAND Corporaton [49]. The Delph Method was conceved as a group technque that ams to obtan the most relable consensus of a group of experts usng a seres of ntensve questonnares wth controlled opnon feedback [50]. Despte ts recognton as a valuable tool, t has some drawbacks. The tool s tme consumng, and convergng results through repettve surveys s costly [51 53]. Further, the problems of ambguty and uncertanty reman n the responses of experts [51,53,54]. To solve these defects, Murray et al. [55], combned the concepts of the tradtonal Delph Method and the fuzzy set to allevate the ambguty of the Delph Method. Kaufmann and Gupta [56] proposed a more complete FDM procedure, n whch the fuzzy set theory s used by askng partcpants to gve a three-pont estmate (.e., pessmstc, moderate, and optmstc values). Trangular fuzzy numbers (TFNs) were then formed, and ther means were computed. Ths study appled pared TFNs to locate three ponts n the extent of mportance (.e., mnmum, medum, and maxmum values) on a scale of 0 to 10 ponts. We and Chang [57] adopted the same concept to calculate and represent these group average values. The pared TFNs were categorzed nto two, namely, the conservatve TFN ( C L, C M, C U ) and the optmstc TFN ( O L, O M, O U ). The ntersecton of the fuzzy opnons of experts mples the convergence of consensus, as shown n Fgure 1. Fnally, the geometrc means of conservatve, moderate, and optmstc values ( C, a, O ) were computed to acqure the consensus values ( G ) of each tem. In vew of the advantages of FDM n evokng expert-group opnon, varous studes [57 59] have embraced FDM n the creaton of performance ndcators or evaluaton crtera. Some essental FDM steps are as follows [57,60]: Fgure 1. TFNs formed n the FDM. Step 1. The questonnares are dstrbuted. An approprate panel group of experts s organzed to express the experts most conservatve (mnmum) and optmstc (maxmum) values for each tem on a scale of 0 to 10. Step 2. The most conservatve (mnmum) and optmstc (maxmum) values from each expert for each tem are gathered, and the geometrc mean of the expert group s opnons s computed. A group
Sustanablty 2014, 6 2665 average s calculated for the pessmstc (optmstc) ndex of sub-crteron, and the abnormal value, whch s outsde the two standard devatons, s elmnated. The rest of the values, namely, the mnmum ( C L ), geometrc mean ( C M ), and the maxmum ( C U ) of the remanng conservatve values; and the mnmum ( O L ), geometrc mean ( O M ), and maxmum ( O U ) of the remanng optmstc values, are calculated. Step 3. The two TFNs as the most conservatve TFN ( C L, CM, CU ) and the most optmstc TFN ( O L, OM, OU ) are determned based on group average values. Step 4. The expert opnons are examned to determne f they are consstent. The consensus sgnfcance value ( G ) for each tem s calculated. (1) If the pared TFNs do not overlap (.e., CU OL ), then a consensus for tem exsts. The consensus sgnfcance value s calculated as follows: G C M s less than the nterval value of and U M, then the consensus sgnfcance value of each tem s calculated as follows: CU OM OLCM G (2) CU CM OM OL If the pared TFNs overlap (.e., C U O L ) and the gray zone nterval value Z CU OL s more than the nterval value of C and O M OU CM, then the expert opnons have dscrepances. Steps 1 to 4 should be repeated untl each tem converges and G s calculated. (2) If the pared TFNs overlap (.e., C U O L ) and the gray zone nterval value Z CU OL C O M O C 3.2. Buldng a Network Relaton Map Usng DEMATEL DEMATEL s a comprehensve tool for buldng and analyzng a structural model that nvolves causal relatonshps between complex factors [61]. Developed by the Scence and Human Affars Program of the Battelle Memoral Insttute of Geneva from 1972 and 1976, DEMATEL has been used to research and solve a group of complcated and ntertwned problems. DEMATEL was developed wth the belef that poneerng scentfc research methods and ther approprate use could mprove the understandng of a specfc problematc cluster of ntertwned problems, thus contrbutng to the dentfcaton of workable solutons usng a herarchcal structure. The methodology, accordng to the concrete characterstcs of objectve affars, can confrm the nterdependence among varables/attrbutes and restrct the relatonshp reflectng ther characterstcs usng an essental system and a development trend [62,63]. The product of the DEMATEL process s a vsual representaton (.e., an ndvdual map of the mnd) that the respondent uses to organze hs/her own actons. The DEMATEL method s ncreasngly beng used to determne the nterrelatonshps between factors through a cause-effect relatonshp dagram, partcularly to determne the crtcal factors of reverse supply chans [64], SaaS adopton [65], arlne safety management systems [66], and performance evaluaton n hotel ndustry [67]. Therefore, DEMATEL modelng fts the problem examned n the present study best and O M 2 (1)
Sustanablty 2014, 6 2666 offers the advantage of provdng a systematc approach to determne the relatonshps of low carbon suppler management n hotel ndustry. The followng steps show the DEMATEL process: Step 1. The average matrx s calculated. Suppose we have H experts n ths study and n factors to consder. Each respondent s asked to ndcate the degree to whch he/she beleves a factor,, affects factor j. Parwse comparsons between any two factors are denoted by x k j and are gven an nteger score of 0 to 4, representng No nfluence (0), Low nfluence (1), Medum nfluence (2), Hgh nfluence (3) and Very hgh nfluence (4) [68]. Fgure 2 shows an example of an nfluence map. Each letter represents a factor n the system. An arrow from c to d shows the effect that c has on d; the strength of ts effect s 4 (very hgh nfluence). DEMATEL can convert the structural relatons between the factors of a system nto an ntellgble map of the system. The scores provded by each respondent provde an n n non-negatve k k 1 2 H answer matrx X = [ x j ], wth k 1,2,..., H. Therefore, X, X,, X are the answer matrces for k k k each of the H experts, wth each element of X [ xj ] nn beng an nteger denoted by x j. The dagonal k k elements of each answer matrx X [ x ] are all set to 0. The n n average matrx A for all expert j opnons can then be computed by averagng the scores of the H experts as follows: nn a 1 H H k j x j k 1 The average matrx A [ a ] s also called the orgnal average matrx. A shows the ntal drect j nn effects a factor has on and receves from other factors. The causal effect between each par of factors n a system can be outlned by drawng an nfluence map, as shown n Fgure 2. Fgure 2. Example of an nfluence map. (3)
Sustanablty 2014, 6 2667 Step 2. Calculate the drect nfluence matrx. The normalzed ntal drect-relaton matrx D s obtaned by normalzng the average matrx A n the followng method: n n Let S mn max a j, maxa j 1n 1 jn j1 1 (4) Thus, D Α (5) s As the sum of each row j of matrx A represents the drect effects of factor on others, represents the one wth the hghest drect nfluence. Lkewse, as the sum of each column of matrx A represents the drect effects receved by factor, max 1 jn n a j 1 represents the one most nfluenced by other factors. The postve scalar s s equal to the larger of the two extreme sums. Matrx D s obtaned by dvdng each element of A by the scalar. Note that each element of matrx D s between 0 and 1. Step 3. Compute the total relaton matrx. Indrect effects between factors are measured by powers of D. Contnuous decrease n the ndrect 2 3 effects of factors, ncludng the powers of matrx D, namely, D, D,..., D, guarantees convergent solutons to the matrx nverson smlar to an absorbng Markov chan matrx. Note that m 2 3 m 1 lm D [0] and lm( I DD D... D ) ( I D ), where 0 s the n n null matrx and m nn m I s the n n dentty matrx. The total relaton matrx T s an n n matrx and s defned as follows: sum T = [t j ] := 1 D D( I D) k As lm D [0] where D [ d ], 0 1 k d j j nn j nn or one row sum d j equals 1. 1, j = 1, 2,, n (6) d j, and d j d j j. At least one column 0, 1 We also defne r and c as n 1 vectors representng the sum of the rows and the sum of the columns of the total relaton matrx T as follows: where superscrpt denotes transposton. r [r ] n1 = n t j j1 n1 d j max 1n n j1 a j (7) n c [ c j ] 1 n = t j 1 1n (8) Let r be the sum of the -th row n matrx T. Therefore, r shows the total effects, both drect and ndrect, of the -th factor on other factors. Let c j denote the sum of the j-th column n matrx T.
Sustanablty 2014, 6 2668 The value c j shows the total effects, both drect and ndrect, receved by factor j from other factors. Therefore, the sum ( r c ) gves an ndex (.e., the poston) representng the total effects both gven and receved by the -th factor. In other words, ( r c ) shows the degree of mportance that the -th factor plays n the system (.e., total sum of effects gven and receved). Moreover, the dfference ( r c, also called the relaton) shows the net effect; the -th factor contrbutes to the system. When ( r c ) s postve, the -th factor s a net causer; when ( r c ) s negatve, the -th factor s a net recever [69,70]. Step 4. Set the Threshold Value and Obtan the Cognton Map. To obtan the cognton map from the factors, a threshold value p should be establshed to extrcate neglgble effects from the total nfluence of matrx T [71]. Only some crtera, whose effect n matrx T s greater than the threshold value, should be chosen and shown n a network relatonshp map (NRM) for nfluence [70]. 3.3. Combnng DEMATEL and ANP to Calculate the Evaluaton Weghts by NRM ANP s the general form of AHP, whch s used n MCDM to address restrctons on herarchcal structures [72]. However, the survey questonnare of ANP s too dffcult for ntervewees to accomplsh [67,73]. Moreover, the tradtonal ANP assumpton, that s, each cluster s of equal weght n obtanng a weghted supermatrx, s not reasonable [74 76]. To mprove ths shortcomng, we used a novel combnaton of DEMATEL and ANP technque called DANP to determne the nfluental weghts of the crtera based on the NRM of DEMATEL. Recently, DANP has been wdely appled n dfferent areas of toursm polcy [77], best vendor selecton [75], performance evaluaton for hot sprng hotels [67], and web stes of natonal parks [78]. The DANP process has the followng steps: Step 1. Establshng an unweghted super matrx. The total-nfluenced matrx s obtaned from DEMATEL. Each column s summed up for normalzaton. The total-nfluenced matrx T c t j s obtaned by the crtera, and T nxn D D t j s mxm obtaned by the dmensons (clusters) from T c. Next, the supermatrx T c s normalzed for the ANP weghts of the dmensons (clusters) usng the nfluence matrx T D. (9)
Sustanablty 2014, 6 2669 After normalzng the total-nfluence matrx T c through the dmensons (clusters), a new matrx T c s obtaned, as shown n Equaton (8). (10) The normalzaton T s explaned and that of the other 11 c ann T c s the same as above. 11 11 t j, 1,2,...,m 1 (11) d c m 1 j1 (12) Let the total-nfluence matrx match and fall nto the nterdependence clusters. The result s the unweghted supermatrx, whch s based on the transposton of the normalzed nfluence matrx T c by the dmensons (clusters), that s, W =(T c ) '. (13) If the matrx W 11 s blank or 0 as shown as Equaton (14), then the matrx between the clusters or nn the crtera s ndependent and has no nterdependent. The other W value are as above.
Sustanablty 2014, 6 2670 (14) Step 2. Obtanng the weghted supermatrx Each column s added for normalzaton. T D 11 1 j 1n j 1 n nj n1 The total-nfluence matrx T D s normalzed, and a new matrx T D s obtaned, where T D t 11 D / d 1 1 t j D / d 1 t 1n D / d 1 t 1 D / d j / d t n D / d t n1 D / d n t nj D / d n t nn D / d n 11 nn 1 j 1n j 1 n nj n1 nn (15) j t j D = / d. Let the normalzed total-nfluence matrx T D complete the unweghted supermatrx to obtan the weghted supermatrx. W T D W Step 3. Lmtng the weghted supermatrx. 11 W 11 1 W 1 j n1 W 1n 1 t j D W 1 t j D W j t nj D W n t 1n D W n1 t n D W nj t nn D W nn The weghted supermatrx s lmted by rasng t to a suffcently large power k untl the supermatrx converges and becomes a long-term stable supermatrx to obtan the global prorty vectors (called the DANP weghts), such as lm h (W ) h. (16) (17)
Sustanablty 2014, 6 2671 3.4. Rankng the Alternatves Usng the VIKOR Method The compromse rankng method (known as VIKOR) was ntroduced as an applcable technque to mplement n MCDM [79]. It s based on the concept of the postve- and negatve-deal soluton to evaluate the standard of dfferent projects competng wth the MCDM model [80]. The postve-deal soluton represents the alternatve wth the hghest value, whereas the negatve-deal represents that wth the lowest value. Smlar to some MCDM methods, such as TOPSIS, VIKOR reles on an aggregatng functon that represents closeness to the deal. In contrast to TOPSIS, however, VIKOR ntroduces a rankng ndex based on the partcular measure of closeness to the deal soluton; ths method uses lnear normalzaton to elmnate unts of crteron functons [80]. VIKOR ranks and selects from a set of alternatves, determnes compromse solutons for a problem wth conflctng crtera, and asssts decson makers n generatng the fnal decson [81]. Varous studes regarded VIKOR as a sutable technque to evaluate each alternatve for each crteron functon [80,82]. The compromse rankng algorthm VIKOR has the followng steps [81 83]: Step 1. Determne the best and the worst values. The best value s and (19), respectvely. * f j and the worst s f * j f j. These two values can be computed by Equatons (18) max f j, 1, 2,... m (18) f j mn f j, 1, 2,... m (19) where, * f j s the postve-deal soluton and f j s the negatve-deal soluton for the jth crteron. Step 2. Calculate the dstance. In ths step, the dstance from each alternatve to the postve deal soluton s computed. S n w j j1 / f * j f j (20) f j * f j / f j j1, 2,...n (21) Q max w j f * j f * j f j where w j represents the weghts of the crtera from DANP, S ndcates the mean of group utlty and represents the dstance of the th alternatve achevement to the postve deal soluton; and Q represents the maxmal regret of each alternatve. Step 3. Calculate the values R by the relaton [80]. R S v S S S * * * 1 * - v Q Q Q Q * * where S mn S, S mn S, S * max S, Q mn Q, Q max Q. Equaton (22) can be rewrtten as R vs 1 vq, when S * 0 and Q * 0 (.e., all crtera acheve the deal level) and S 1 and Q 1 (.e., the worst stuaton). In the equaton, v s (22)
Sustanablty 2014, 6 2672 ntroduced as the weght for the strategy of maxmum group utlty, and 1-v s the weght of the ndvdual regret. In Equaton (22), when v = 1, t ndcates the decson-makng process that can use the strategy of maxmum group utlty. Conversely, when v = 0, t ndcates the decson-makng process that can use the strategy of mnmum ndvdual regret. In general, v = 0.5 wll be used f the decson process nvolves both maxmum group utlty and ndvdual regret [82,83]. The compromse soluton s determned by VIKOR, and t can be accepted by the decson makers based on a maxmum group utlty of the majorty and a mnmum of the ndvdual regret of the opponent. 4. A Hotel Company as an Example In ths secton, an example demonstrates the proposed model for suppler selecton n terms of carbon management competence. The M hotel, an ordnary tourst hotel wth rooms prced accordngly, had ts grand openng n 2006. It provdes exceptonal, hgh-qualty facltes and servces at reasonable prces to satsfy the demand for accommodatons, food and beverages, and lesure servces of local and foregn toursts. The M hotel also advocates three envronmental vsons, that s, Envronment, Energy Conservaton, Carbon Reducton, ncludng sustanable development processes. To facltate low carbon hotel operatons, the M hotel launched varous measures of energy conservaton and carbon reducton, such as food mle management, electrcty montor systems, and energy effcency mprovement. In 2011, M hotel also acqured the certfcaton of Energy Management Systems-ISO 50001 to mtgate carbon emsson and to manage energy effectvely. However, n achevng low carbon operatons, M hotel encountered crtcal challenges n determnng approprate supplers for long-term collaboratve partnershp n the low carbon supply chan. At least 80% of carbon emssons are produced n the total supply chan. The ISO 50001 requres supplers to provde energy-effcent products. Thus, the M hotel used the proposed framework to select low carbon supplers. In ths study, fve supplers (S1, S2, S3, S4, and S5) of the hotel company n the case study were demonstrated to assess the carbon performance of the 10 crtera dentfed. Three managers n the case company conductng the assessment were responsble n the felds of suppler management, procurement management, and energy management. Managers used a fve-pont scale (.e., 0 bad, 1 low, 2 moderate, 3 good, and 4 excellent performance) to evaluate the supplers. After that, the authors evaluated these merchants usng the hybrd MCDM model that combnes DANP wth VIKOR. 4.1. Identfyng the Consstency of the Evaluaton Crtera Consderng the stuaton of carbon management of supplers n the Tawanese hotel ndustry, a draft of the evaluaton framework should be confrmed frst by experts. Eght experts were nvted n the FDM process to express ther opnons on dentfyng the consstency of evaluaton crtera for the selecton of low carbon supplers. Consderng the practce experence n the feld of carbon management n the hotel ndustry, the study dentfed fve managers from hotel frms, who were responsble for the mplementaton of green procurement and energy management, and three unversty professors whose research were related to carbon and energy management n the hotel ndustry. The 12 ntal crtera were used as the bass for questonnare development. The FDM technque was used to screen and ft the factors. Frst, the expert group average was calculated for the conservatve and optmstc values of each measure. Anythng outsde the two standard devatons
Sustanablty 2014, 6 2673 was elmnated. Subsequently, the mnmum (C L ), geometrc mean (C M ), and maxmum ( C U ) of the conservatve values, as well as the mnmum (O L ), geometrc mean (O M ), and maxmum (O U ) of the optmstc values, were calculated (Table 1). The values of M and Z were also calculated to determne the consstency of expert opnons. The dfferences were convergent, and the consensus value of G was calculated to screen the ndcators [60,59]. The threshold value was set to 6.0. The agreed proporton of experts was more than 80%. Based on ths prncple, the two crtera, namely, carbon governance and carbon management systems, were excluded, as shown n Table 1. These crtera were used as the bass for selectng the 10 crtera for low carbon suppler selecton n the hotel ndustry, as shown n Fgure 3. Crtera Table 1. Results of calculaton of factors wth FDM. Pessmstc value C L C U Optmstc value Geometrc mean Consensus value Energy effcency of products 3 9 7 10 5.49 7.81 0.32 7.38 > 6.0 Eco-labelng of products 5 9 7 10 6.50 8.52 0.02 7.76 > 6.0 Carbon accountng and nventory 3 7 5 10 5.10 7.61 0.51 6.16 > 6.0 Energy reducton of food processng 3 9 7 10 6.36 8.91 0.55 7.84 > 6.0 Carbon governance 1 7 3 10 4.07 6.55 1.52 5.19 < 6.0 Carbon polcy 5 9 7 10 6.10 8.41 0.30 7.65 > 6.0 Carbon reducton targets 3 7 5 10 5.21 7.63 0.42 6.19 > 6.0 Carbon and energy management systems 3 5 5 9 4.32 6.92 2.59 5.62 < 6.0 Transport effcency 3 9 7 10 6.43 8.79 0.36 7.82 > 6.0 Collaboraton of supplers 3 9 7 10 5.17 7.91 0.74 7.39 > 6.0 Measures of carbon reducton and energy conservaton 5 9 9 10 6.93 9.49 2.56 8.21 > 6.0 Food mle management 3 9 7 10 5.84 8.50 0.66 7.64 > 6.0 O L Fgure 3. The framework of low carbon suppler selecton. O U C M O M M Z G
Sustanablty 2014, 6 2674 4.2. Determnng the Relatonshps between Crtera by DEMATEL The DEMATEL method was used to examne nterdependent and nfluence relatonshps between 10 crtera usng the results of FDM. The eght experts were asked to complete the questonnares usng a fve-pont scale (.e., 0 for no nfluence, 1 for low, 2 for moderate, 3 for hgh, and 4 for very hgh) to ndcate the nfluence of each crteron on another one n ther respectve organzaton. The average ntal nfluence 10 10 matrx A (Table 2) was obtaned by parwse comparson n terms of nfluences and drectons. The normalzed ntal drect-relaton matrx D was calculated usng Equatons (3) to (5) (Table 3). The total nfluence matrx T (Table 4) was derved by Equaton (6). The NRM of the nfluental relatonshp was constructed by vectors r and c (Table 5) usng Equatons (7) and (8), as shown n Fgure 4. Table 2. The ntal nfluence matrx. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.000 3.600 2.400 0.200 2.400 2.800 2.600 2.000 2.800 2.000 C 2 3.600 0.000 3.200 1.800 3.200 3.400 3.000 2.600 3.000 2.000 C 3 3.000 2.400 0.000 1.400 3.600 3.800 3.000 3.000 3.200 2.600 C 4 1.400 2.200 1.000 0.000 1.600 1.800 1.200 1.800 1.800 2.000 C 5 3.000 2.600 3.000 1.200 0.000 3.200 3.400 3.400 3.800 2.800 C 6 3.200 2.600 3.000 0.400 3.600 0.000 3.400 3.600 3.400 2.800 C 7 1.600 2.600 2.800 0.800 3.000 3.000 0.000 2.800 3.000 2.000 C 8 2.600 2.800 3.000 1.000 2.800 3.200 2.200 0.000 2.600 2.200 C 9 3.000 2.800 3.600 1.200 3.400 3.400 2.800 3.200 0.000 2.800 C 10 1.800 2.200 2.600 2.200 2.800 2.800 2.200 3.200 2.800 0.000 Table 3. The normalzed drect-nfluence matrx. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.000 0.131 0.088 0.007 0.088 0.102 0.095 0.073 0.102 0.073 C 2 0.131 0.000 0.117 0.066 0.117 0.124 0.109 0.095 0.109 0.073 C 3 0.109 0.088 0.000 0.051 0.131 0.139 0.109 0.109 0.117 0.095 C 4 0.051 0.080 0.036 0.000 0.058 0.066 0.044 0.066 0.066 0.073 C 5 0.109 0.095 0.109 0.044 0.000 0.117 0.124 0.124 0.139 0.102 C 6 0.117 0.095 0.109 0.015 0.131 0.000 0.124 0.131 0.124 0.102 C 7 0.058 0.095 0.102 0.029 0.109 0.109 0.000 0.102 0.109 0.073 C 8 0.095 0.102 0.109 0.036 0.102 0.117 0.080 0.000 0.095 0.080 C 9 0.109 0.102 0.131 0.044 0.124 0.124 0.102 0.117 0.000 0.102 C 10 0.066 0.080 0.095 0.080 0.102 0.102 0.080 0.117 0.102 0.000 Table 4. The total nfluence matrx. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.541 0.656 0.655 0.246 0.686 0.715 0.642 0.653 0.695 0.554 C 2 0.758 0.641 0.785 0.340 0.822 0.849 0.758 0.781 0.814 0.646 C 3 0.748 0.729 0.690 0.331 0.844 0.871 0.767 0.804 0.830 0.673 C 4 0.426 0.451 0.437 0.167 0.477 0.496 0.429 0.471 0.482 0.408 C 5 0.756 0.744 0.798 0.330 0.737 0.864 0.787 0.824 0.856 0.686
Sustanablty 2014, 6 2675 Table 4. Cont. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 6 0.759 0.741 0.795 0.302 0.850 0.756 0.784 0.827 0.841 0.683 C 7 0.612 0.640 0.684 0.272 0.722 0.740 0.571 0.696 0.719 0.570 C 8 0.658 0.662 0.705 0.285 0.732 0.763 0.661 0.619 0.723 0.589 C 9 0.754 0.747 0.813 0.329 0.845 0.867 0.767 0.816 0.731 0.684 C 10 0.625 0.637 0.685 0.322 0.724 0.742 0.652 0.717 0.721 0.509 Threshold value: 0.756, the values were marked when hgher than the threshold value. Table 5. The sum of nfluences gvng and receved. Crtera r c r + c r c C 1 6.041 6.638 12.680 0.597 C 2 7.194 6.648 13.842 0.546 C 3 7.287 7.045 14.332 0.242 C 4 4.244 2.924 7.168 1.320 C 5 7.382 7.440 14.822 0.059 C 6 7.339 7.663 15.002 0.323 C 7 6.225 6.818 13.043 0.593 C 8 6.396 7.206 13.603 0.810 C 9 7.353 7.411 14.764 0.058 C 10 6.334 6.002 12.337 0.3320 Fgure 4. The causal dagram.
Sustanablty 2014, 6 2676 4.3. Fndng the Influental Weght of Crtera by DANP Ths study used the DANP to obtan the weghts of the 10 crtera based on the nfluence network of the total nfluence matrx T of DEMATEL. DANP was used to calculate an unweghted supermatrx (Table 6) and weghted supermatrx (Table 7). The lmtng power of the weghted supermatrx to confrm the supermatrx was converged and became a long-term stable supermatrx, obtanng the weghts of all crtera (Table 8). Each row represents the weghts of each crteron. Table 6. Unweghted supermatrx based on DANP. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.5409 0.6556 0.6548 0.2457 0.6857 0.7152 0.6422 0.6530 0.6946 0.5538 C 2 0.7580 0.6410 0.7847 0.3400 0.8224 0.8495 0.7576 0.7809 0.8137 0.6464 C 3 0.7480 0.7294 0.6898 0.3315 0.8443 0.8712 0.7666 0.8036 0.8296 0.6729 C 4 0.4262 0.4511 0.4366 0.1668 0.4773 0.4959 0.4293 0.4706 0.4817 0.4084 C 5 0.7560 0.7440 0.7981 0.3295 0.7374 0.8636 0.7865 0.8244 0.8561 0.6862 C 6 0.7593 0.7407 0.7954 0.3025 0.8504 0.7556 0.7841 0.8270 0.8414 0.6830 C 7 0.6122 0.6402 0.6836 0.2723 0.7218 0.7397 0.5713 0.6959 0.7186 0.5695 C 8 0.6580 0.6622 0.7047 0.2848 0.7321 0.7627 0.6612 0.6186 0.7231 0.5889 C 9 0.7544 0.7472 0.8129 0.3287 0.8453 0.8668 0.7666 0.8158 0.7315 0.6841 C 10 0.6252 0.6371 0.6845 0.3219 0.7238 0.7424 0.6524 0.7167 0.7209 0.5093 Table 7. Weghted supermatrx based on DANP. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.0895 0.1054 0.1027 0.1004 0.1024 0.1035 0.0983 0.1029 0.1026 0.0987 C 2 0.1085 0.0891 0.1001 0.1063 0.1008 0.1009 0.1028 0.1035 0.1016 0.1006 C 3 0.1084 0.1091 0.0947 0.1029 0.1081 0.1084 0.1098 0.1102 0.1105 0.1081 C 4 0.0407 0.0473 0.0455 0.0393 0.0446 0.0412 0.0437 0.0445 0.0447 0.0508 C 5 0.1135 0.1143 0.1159 0.1125 0.0999 0.1159 0.1160 0.1145 0.1150 0.1143 C 6 0.1184 0.1181 0.1196 0.1169 0.1170 0.1030 0.1188 0.1192 0.1179 0.1172 C 7 0.1063 0.1053 0.1052 0.1012 0.1066 0.1068 0.0918 0.1034 0.1043 0.1030 C 8 0.1081 0.1085 0.1103 0.1109 0.1117 0.1127 0.1118 0.0967 0.1109 0.1131 C 9 0.1150 0.1131 0.1138 0.1135 0.1160 0.1146 0.1154 0.1130 0.0995 0.1138 C 10 0.0917 0.0898 0.0923 0.0962 0.0930 0.0931 0.0915 0.0921 0.0930 0.0804 Table 8. Influental weghts of stable matrx of DANP. Crtera C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 1 0.1008 0.1008 0.1008 0.1008 0.1008 0.1008 0.1008 0.1008 0.1008 0.1008 C 2 0.1012 0.1012 0.1012 0.1012 0.1012 0.1012 0.1012 0.1012 0.1012 0.1012 C 3 0.1073 0.1073 0.1073 0.1073 0.1073 0.1073 0.1073 0.1073 0.1073 0.1073 C 4 0.0444 0.0444 0.0444 0.0444 0.0444 0.0444 0.0444 0.0444 0.0444 0.0444 C 5 0.1131 0.1131 0.1131 0.1131 0.1131 0.1131 0.1131 0.1131 0.1131 0.1131 C 6 0.1164 0.1164 0.1164 0.1164 0.1164 0.1164 0.1164 0.1164 0.1164 0.1164 C 7 0.1036 0.1036 0.1036 0.1036 0.1036 0.1036 0.1036 0.1036 0.1036 0.1036 C 8 0.1094 0.1094 0.1094 0.1094 0.1094 0.1094 0.1094 0.1094 0.1094 0.1094 C 9 0.1127 0.1127 0.1127 0.1127 0.1127 0.1127 0.1127 0.1127 0.1127 0.1127 C 10 0.0912 0.0912 0.0912 0.0912 0.0912 0.0912 0.0912 0.0912 0.0912 0.0912
Sustanablty 2014, 6 2677 After the weghts of the crtera were determned by DANP, the VIKOR method was used to evaluate the carbon performance of suppler selecton (Table 9). In ths study, fve supplers (S1, S2, S3, S4, and S5) of the hotel company n the case study were shown to assess carbon performance accordng to the 10 crtera dentfed. Three managers n the case company conductng the assessment were responsble for the felds of suppler management, procurement management, and energy management. Managers used a fve-pont scale (.e., 0 = bad, 1 = low, 2 = moderate, 3 = good, and 4 = excellent performance) to evaluate the supplers. Then, the authors evaluated these merchants by usng the hybrd MCDM model, whch combnes DANP wth VIKOR. The average performance scores of each merchant through the VIKOR method were used to obtan the performance and the deal level gaps among the supplers, as shown n Table 9. Gven the ease of use of the proposed model n the case company, n ths research, v value of VIKOR was set to 0.5 based on both maxmum group utlty and ndvdual regret n the expert opnons. As R represents the gap between the alternatve and the deal soluton, S 3 contans the smallest gap n terms of the value of VIKOR, followed by S 1, S 2, S 5, and S 4. The sum of these values for each alternatve s provded n Table 9, whch shows that S 3 s the best suppler. 5. Results and Dscusson Table 9. VIKOR results. Suppler S Q R Rankng S 1 0.536 0.087 0.312 2 S 2 0.635 0.113 0.374 3 S 3 0.474 0.085 0.279 1 S 4 0.785 0.116 0.450 5 S 5 0.641 0.116 0.379 4 We present the followng results of our proposed MCDM model that can facltate low carbon suppler selecton n the hotel ndustry. Frst, the FDM method was used to dentfy the consstency of the selecton crtera for low carbon supplers through expert opnons. The threshold value was set to 6.0; the agreed proporton of experts was more than 80%. Wth ths prncple, the two crtera, namely, carbon governance and carbon management systems were excluded, as shown n Table 1. These crtera were used as the bass for selectng the 10 crtera for the selecton of low carbon supplers n the hotel ndustry, as shown n Fgure 3. Second, the NRM of the crtera was recognzed by DEMATEL. The nfluental relatonshp wthn the 10 crtera was revealed. Consderng the sgnfcance of carbon management n suppler selecton, as presented n Table 5, the mportance s dentfed as C 6 > C 5 > C 9 > C 3 > C 2 > C 8 > C 7 > C 1 > C 10 > C 4 accordng to the degree of mportance (r + c ). Contrary to the mportance of crtera, energy reducton of food processng (C 4 ), eco-labelng of products (C 2 ), food mle management (C 10 ), and carbon accountng and nventory (C 3 ) are net causers n accordance wth the value of dfference (r c ). As ndcated n the causal relatonshps n Fgure 4 and Table 5, C 2 affects crtera C 1, C 3, C 5, C 6, C 7, C 8, and C 9 ; C 3 affects crtera C 5, C 6, C 7, C 8, and C 9. Although C 4 and C 10 are net causers, they have no nfluence on other crtera n terms of the threshold value, whch s less than 0.756. All relatonshps that met or exceeded the threshold were rendered n boldface, as shown n Table 4, matrx T.
Sustanablty 2014, 6 2678 By followng ths prncple, Fgure 4 depcts the nfluence map of the 10 mutually nterdependent crtera. One-way relatonshps are represented by dashed lnes, whle two-way relatonshps are represented by sold lnes. By understandng these nfluental relatonshps, managers can focus on the two crtera of eco-labelng of products (C 2 ) and carbon accountng and nventory (C 3 ) to determne how green supplers are exposed to carbon rsk. By followng the causal relatonshp of DEMATEL, managers can clearly understand the crteron to mprove the management of low carbon supplers. Thrd, the nfluental weghts of crtera were determned by DANP. In terms of the relatve weghts of crtera for evaluatng carbon performance of supplers, carbon reducton targets (C 6 ) (0.1164), carbon polcy (C 5 ) (0.1131), and measures of carbon reducton and energy conservaton (C 9 ) (0.1127) are the top three sgnfcant evaluaton crtera. To mprove the low carbon supply chan, settng the targets of carbon emsson reducton s mportant so supplers can montor authentc and measurable progress n addressng clmate change. To acheve the targets of carbon reducton, supplers should launch varous clmate strateges that nclude quanttatve emsson reducton targets for ther scopes 1 and 2, and occasonally even scope 3, GHG emssons [34]. Subsequent results show that measures of carbon reducton (C 9 ) s the thrd mportant crteron. The crteron of carbon polcy (C 5 ) s the second most mportant. Whle the suppler launches the carbon polcy, the company can facltate carbon management practces by establshng a carbon polcy to show ts poston on carbon emsson dsclosure, reducton targets, and emsson certfcaton, among others [12]. Consderng the sgnfcant weghts of the crtera, managers should select the best and approprate supplers through the VIKOR method of the proposed MCDM model. Fnally, S 3 s selected as best carbon performance of fve supplers. 6. Conclusons and Future Research To promote low carbon operatons n the hotel ndustry, the selecton of supplers n the feld of carbon and energy management s mportant n achevng the target of the low carbon supply chan. We presented a supply chan-based conceptual framework and an operatonal model to ncorporate carbon management nto suppler selecton n the hotel ndustry. By dentfyng the related crtera of carbon management actvtes for the proposed framework, whch s a hybrd MCDM model, an ntegraton of FDM, DANP, and VIKOR methods was appled n the emprcal analyss on a hotel company for selectng low carbon supplers. The proposed framework brngs several contrbutes to the evaluaton and selecton of low carbon supplers n the hotel ndustry. Frst, a new hybrd MCDM model for evaluatng supplers, wth emphass on carbon and energy management, was developed usng the FDM method. Such framework wth 10 crtera s rare n the prevous lterature. Second, the DEMATEL method was appled n selectng supplers n terms of carbon management. DEMATEL proved to be an approprate method to delneate the structure of a completely nterdependent suppler selecton problem model and to obtan the problem s soluton. Thrd, DANP was used to acqure consderable weghts of the 10 crtera. The three mportant crtera, namely, carbon reducton targets, carbon polcy, and measures of carbon reducton, were derved. Fnally, an emprcal study was conducted to demonstrate the applcaton of hybrd MCDM model that combnes DANP wth VIKOR. The hybrd model also consders both maxmum group utlty and ndvdual regret to measure the gaps between alternatve and deal