Investigation of critical sustainability decisions in product recycling and remanufacturing

Transcription

1 Grdute Theses nd isserttions Iow tte University pstones, Theses nd isserttions 3 Investigtion of criticl sustinbility decisions in product recycling nd remnufcturing Wenbo hi Iow tte University Follow this nd dditionl works t: Prt of the Opertionl eserch ommons ecommended ittion hi, Wenbo, "Investigtion of criticl sustinbility decisions in product recycling nd remnufcturing" 3. Grdute Theses nd isserttions This isserttion is brought to you for free nd open ccess by the Iow tte University pstones, Theses nd isserttions t Iow tte University igitl epository. It hs been ccepted for inclusion in Grdute Theses nd isserttions by n uthorized dministrtor of Iow tte University igitl epository. For more informtion, plese contct digirep@istte.edu.

2 Investigtion of criticl sustinbility decisions in product recycling nd remnufcturing by Wenbo hi A disserttion submitted to the grdute fculty in prtil fulfillment of the requirements for the degree of OTO OF PHILOOPHY Mjor: Industril Engineering Progrm of tudy ommittee:. Jo Min, Mjor Professor Mervyn Mrsinghe Gincrlo Moschini igurdur Olfsson rh yn Iow tte University Ames, Iow 3 opyright Wenbo hi, 3. All rights reserved.

3 ii TABLE OF OTET List of Figures... iv List of Tbles... v AOWLEGEMET... vi ABTAT... vii HAPTE ITOUTIO... HAPTE LITEATUE EVIEW... 6 HAPTE 3 POUT WEIGHT A OLLETIO ATE EIIO I LOE LOOP UPPLY HAI WITH POUT EYLIG Bckground nd Objectives losed Loop upply hin Models on-ecycling cenrio entrlized upply hin with ecycling ecentrlized upply hin with ecycling ecentrlized upply hin with ecycling nd Government ubsidy/fee -G Mngeril Insights nd Policy Implictions... 3 HAPTE EMAUFATUIG EIIO WITH EMAUFATUIG OT UETAITY Bckground nd Objectives Models nd Anlyses Bsic Model Extended Model with emnufcturing ubsidy nd isposl Fee....3 umericl tudy Mngeril Insights nd Policy Implictions... 5 HAPTE 5 EMAUFATUIG A EPLAEMET EIIO UE OPEATIO A MAITEAE OT UETAITY Bckground nd Objectives Model emnufcturing ecision eplcement ecision umericl tudy Mngeril Insights nd Policy Implictions... 7

4 iii HAPTE 6 IUIO... 7 HAPTE 7 OLUIO APPEIX A. eduction of onditions in Assumption B. Proof for Proposition Proof for Proposition Proof for Proposition E. Proof for Proposition F. Proof for Proposition G. Proof for Proposition H. Proof for Propositions I. Proof for Propositions J. Proof of Eqution Proof of Proposition L. Proof of Proposition M. Proof of Proposition Proof of Proposition O. Proof of Proposition P. Proof of Proposition Q. Proof of Eqution Proof of Proposition Proof of Proposition T. Proof of Proposition 5. nd orollry U. Proof of Proposition V. Proof of Lemm 5. nd Proposition W. Proof of orollry EFEEE... 9

5 iv LIT OF FIGUE Figure. Mnufcturer-ollector losed Loop upply hin... Figure. ecentrlized upply hin with Government ubsidy nd Fee... Figure 3. Vrition of Production Weight nd ollection te with Mrginl ecycling Benefit... Figure. Vrition of the Lndfill Quntity with Mrginl ecycling Benefit... 5 Figure 5. Vrition of the Lndfill Quntity with the ollecting ubsidy... 9 Figure 6. Project Vlue vs. emnufcturing ost... 7 Figure 7. Vrition of emnufcturing ost Threshold with Voltility nd Fixed ost... 8 Figure 8. Vrition of euse Frction, Product Life, nd isposl te with nd L... 9 Figure 9. Vrition of Optiml emnufcturing ubsidy nd isposl Fee with Voltility... 9 Figure. mple Pths of the emnufcturing ost in 8 Yers... 5 Figure. Vrition of nd T with Voltility, emnufcturing ost, nd trting O&M ost Figure. mple Pths of O&M osts nd F, in 6 Months... 67

6 v LIT OF TABLE Tble. Prmeters nd Vlues... 6 Tble. umericl esults elevnt to the emnufcturing ecision... 7 Tble 3. umericl tudy Prmeter Vlues... 6 Tble. umericl esults Tble 5. Experimentl esults of Monte rlo imultion... 7

7 vi AOWLEGEMET I would like to express my deepest grtitude to my mjor professor r. Jo Min for his cdemic guidnce, ptience, s well s showing me the wy to become professionl resercher. I lso wnt to thnk my committee members r. Mervyn Mrsinghe, r. Gincrlo Moschini, r. igurdur Olfsson, nd r. r yn for their cdemic dvice nd counsel. Also, my grtitude goes to my wife Mengmeng hen, my son Yuchen hi, nd my prents. Without their love nd support, I would hve never come this fr. I would lso thnk rist Briley, Lori Bushore, onn erk, nd Lynn Frnco for helping me with dministrtive ctivities nd deling with ll my documenttions.

8 vii ABTAT In recent yers, there hve been substntil interests in sustinble processes for the endof-life products to conserve nturl resource nd reduce lndfill wstes. Among these processes, recycling nd remnufcturing provides opportunities for prctitioners from vrious industries nd government regultors to mitigte the environmentl impcts while mintining economic vibility. Menwhile, there re substntil criticl decisions for the decision/policy mkers to rep to the economic nd environmentl benefits from the recycling nd remnufcturing processes. Therefore, in this study, I investigte the product weight reduction nd usedproduct collection decisions in the context of closed loop supply chins with product recycling. Furthermore, from rel options perspective, I investigte the optiml timing of remnufcturing with the remnufcturing cost following stochstic process, nd the optiml timing of remnufcturing nd replcement with the mintennce costs following stochstic processes. Also, in this study, I investigte the influences of the governmentl economic instruments on the recycling nd remnufcturing decisions, nd exmine the environmentl vibility of these sustinble incentives. A series of mngeril insights nd policy implictions re obtined from nlyticl nd numericl nlyses.

9 HAPTE ITOUTIO In recent yers, there hve been substntil interests in the sustinble processes for the endof-life products to reduce resource consumption nd lndfill wstes. Among these processes, recycling nd remnufcturing provides opportunities for firms from vriety of industries to mitigte the environmentl impcts while mintining economic vibility. Toner crtridge recycling is 3 billion dollr industry ] nd 9% in weight of the recycled mterils re utilized in the new toner crtridges ]. Also, Pnsonic diverted thousnds of tons of cthode ry tubes T s from lndfill through recycling process, nd utilized the recycled glss, metl, nd plstics in the production of new T s 3]. within the U.., the remnufcturing industry involves more thn 73, firms nd 35, people with nnul revenue of $53 billion ]. Overll, remnufcturing progrms cn sve compnies % 65% of the mnufcturing cost ]. Xerox's remnufcturing progrm sved $ million in mteril cost in less thn five yers 5]. Therefore, from decision/policy mker s perspective, there re substntil opportunities to rep the economic nd environmentl benefits of the remnufcturing nd recycling processes, such s closed loop supply chin mngement, sustinble product design, used-product cquisition strtegy, remnufcturing timing, s well s governmentl subsidies nd fees. With this knowledge, in this study, we im to understnd the economiclly rtionl decisions on i the product weight reduction nd used-product collection efforts in the context of closed loop supply chins with recycling; ii the product remnufcturing decision with uncertin re-

10 mnufcturing cost; s well s iii the product remnufcturing nd replcement decisions with uncertin O&M cost i.e., opertions nd mintennce cost. pecificlly, Towrd the reserch objective i, we formulte three closely relted L models consisting of mnufcturer who lso recycles used products nd collector of the used products. The first L is centrlized model with the verticl integrtion of the mnufcturer nd the collector. The second L is decentrlized model with the mnufcturer s the leder nd the collector s the follower in tckelberg gme. The third L is decentrlized model with government subsidy nd fee. In ddition, we formulte non-recycling model s benchmrk for comprison. Under this frmework, the nlytic results show tht conditions e.g., the increse decrese of the mrginl recycling benefit or the collecting subsidy leding to higher lower level of the collection rte my lso result in higher lower level of the product weight, nd vice vers. Hence, decision mkers my not be ble to pursue n improvement of both fctors simultneously, nd it is dvised to consider the negtive impct on the other fctor before ny efforts on improving one fctor. As consequence, b the increse decrese of the mrginl recycling benefit or the collecting subsidy for unit weight of used-product collected my result in higher level of the lndfill quntity when the mrginl recycling benefit or the collecting subsidy is sufficiently high lower. In this sense, efforts on improving one fctor should not be supported unconditionlly, s it my mke the environment worse off. c Under the revenue-neutrlity frmework i.e., the mount of subsidies disbursed blnces the mount of fees collected, the lndfill quntity in the centrlized model cn be chieved or even further reduced in the decentrlized model by choosing proper vlue of subsidy. However, inppropri-

11 3 te vlue of subsidy my lso result in inferior environmentl performnce i.e., more lndfill quntity reltive to the non-subsidy cse. Given these findings, the pricing of subsidy/fee becomes criticl to chieve high environmentl efficiency nd void unintended negtive environmentl consequences. Towrd the reserch objective ii, we consider firm tht leses single product to service provider, nd the firm is lso responsible for the mintennce of the product. For instnce, Xerox leses photocopiers to tples which provides copying services to customers. Throughout product's lifecycle, s the mintennce cost increses due to component-deteriortion, t ny time point, the firm hs n option to terminte the life of the product nd remnufcture it. uring the remnufcturing process, depending on the physicl condition, components re either reused or replced nd disposed. After the remnufcturing process, the product is restored to like-new condition nd re-lesed to the service provider. We note tht the flexibility on remnufcturing described bove lends itself to rel options pproch. Tht is, the firm hs the right, but not the obligtion, to mke chnges to its business project under uncertinty. Moreover, in the extended scenrio, government prticiption is incorported into our model in the form of remnufcturing subsidy nd disposl fee. The criticl findings of reserch topic ii include: The uncertinty of remnufcturing cost is the driver for decision mkers to prolong the product life nd defer the exercise of remnufcturing option becuse the uncertinty increses the vlue of holding the flexibility; b Incresing the remnufcturing subsidy disposl fee incentivizes decision mkers to exercise the remnufcturing option erlier lter becuse the subsidy reduces increses the cost of exercising the option now. Tht is, the vlue of witing is diminished; c An increse of the remnu-

12 fcturing subsidy disposl fee my result in more industril wstes becuse it entils shorter product life lower reuse frction of mterils; d As the remnufcturing cost becomes more uncertin, the policy mker is dvised to increse the remnufcturing subsidy or decrese the disposl fee becuse these djustments blnce the higher level of lndfill disposl stemmed from the higher uncertinty. We hope tht these findings, s the first reserch findings on the remnufcturing decisions vi rel options pproch will stimulte relevnt discussions mong industril prctitioners, governmentl regultors, environmentl groups, s well s cdemics. Towrd the reserch objective iii, we consider firm tht leses product to service provider, nd the firm is ssumed to be responsible for the O&M costs of the product. The O&M costs for the durble prts nd the non-durble prts re ssumed to follow uncorrelted geometric Brownin motion GBM processes. Throughout the product's life-cycle, the firm hs two options for the product: remnufcturing option nd replcement option. With these options, the firm hs the right, but not the obligtion, to remnufcture or replce the product t ny point. In prticulr, the exercise of the remnufcturing option triggers the ultimte mintennce for the product by replcing the non-durble prts with new ones. The replcement option triggers the replcement of the product s whole. The nlyticl nd numericl findings in iii include tht s the O&M costs become more less voltile, it is beneficil to exercise the remnufcturing nd the replcement options lter erlier; b higher lower remnufcturing cost results in shorter longer expected product life becuse it becomes less ttrctive for the firm to conduct ultimte mintennce for the product through remnufcturing; c n increse of the remnufcturing subsidy my result in

13 5 more lndfill disposl per unit time, nd the government policy mker is dvised to crefully exmine the industril conditions before ny implementtion of new sustinble subsidy or djustment of the existing sustinble subsidy. We hope tht the questions rised nd ddressed in this study will stimulte relevnt discussions mong industril prctitioners, governmentl regultors, nd cdemics. The rest of the rticle is orgnized s follows. hpter provides comprehensive literture review relevnt to this reserch. In hpter 3, we formulte nd nlyze series of closed loop supply chin models for the product weight reduction nd the collection rte decisions with product recycling. In hpter, we formulte nd nlyze the models for the product remnufcturing decision from rel options perspective with the remnufcturing cost following GBM process. imilrly, we utilized the rel options pproch to investigte the remnufcturing nd replcement decisions when the mintennce costs re ssumed to following the GBM processes in hpter 5. Finlly, the conclusions nd future works re presented in hpter 6.

14 6 HAPTE LITEATUE EVIEW As this reserch is ssocited with severl strems of reserch topics, we will review the relevnt rticles of ech topic s follows. A growing literture in L nd reverse logistics ddresses mngement issues such s chnnel design ]3], technicl investments 7], nd pricing strtegy ]. In these ppers, the environmentl performnce is mesured by single fctor such s return rte ]3], remnufcturbility 7], nd collection quntity ]. Geyer et l. 5] demonstrted the significnce of coordinting the investments in product durbility nd collection rte. However, the study is bsed on numericl nlysis, nd nlytic results re not observble. In this pper, by considering vrible product weight nd collection rte, we exmine the interction between two fctors nd their environmentl consequences. There re numerous rticles relevnt to product tke-bck nd recycling focusing on tcticl nd opertionl issues such s inventory control 7], shop floor control 7], logistics network design 73], nd mteril resource plnning 7]. We refer reders to ] comprehensive literture review. ecently, there hs been shift in emphsis towrds the economics of product reuse nd recycle. Guide et l. ] demonstrted tht the product cquisition nd cost sving re the criticl drivers of recycling profitbility. ebo et l. 75] investigted the impct of consumer profiles nd technology investment on the recycling profitbility. Our study investigtes

15 7 the economic reltionship between the product weight nd the collection rte in the context of L s. We now proceed to review the relevnt rticles on product weight. Hung et l. 6] climed tht light weight hve been tken s the design principles of electronic products. leiner et l. 76] pointed out the significnce of lightweight design in engineering, trnsporttion, nd rchitecture industries, nd demonstrted the fesibility of reducing product weight vi utilizing lighter mterils nd/or reducing the product size. Atsu et l. ] pointed out tht, since WEEE trgets re unilterlly weight-bsed, most producers do not hve n incentive to invest in environmentl design. Hence, our study exmines the interction between investment in product weight reduction nd used product tking bck ctivities. Furthermore, Atsu et l. ] lso climed tht the cost to the environment depends on the product's physicl properties such s weight. Align with this clim, lndfill quntity i.e., totl weight of products sent to lndfill is utilized to in our models to mesure the environmentl impct of the L s. The use of txes nd subsidies s tool to improve socil welfre hs been widely studied 77]. rrro nd Top 78] nlyzed the impct of environmentl regultion in the form of txtion on the innovtion ctivity of firms. Plmer nd Wlls 38] utilized tx/subsidy combintion in which producers of intermedite goods py per-pound tx nd collectors of used products receive per-pound subsidy from the tx. Aligning with Plmer nd Wlls, under the revenue-neutrlity frmework, we implement subsidy/fee system which chrges the mnufcturer weight-bsed production fee, nd subsidies the collector for the weight of products collected. We note tht Mrozek 39] rgued tht, to chieve the mximum efficiency, deposit/refund systems must be implemented in wy tht the revenue is neutrl.

16 8 Furthermore, in recent yers, there hve been incresing interests in exmining the uncertinties in decision mking process of product remnufcturing. enizel et l. 58] investigted hybrid production plnning problem where the inputs i.e., used products re subject to different nd uncertin qulity levels. In 7], it ws demonstrted tht the remnufcturing cost is pproximtely liner to the quntified qulity deteriortion. obotics et l. 79] lso considered firm tht remnufctures products nd mkes investment decisions on incresing the reusbility of its products. Assuming norml distribution of the remnufcturing cost, they climed tht uncertinty of the remnufcturing cost my not hinder this investment, depending on the inspection cpbilities nd technologies. imilrly, by ssuming uniform distribution of product qulity, Glbreth nd Blckburn 57] investigted the optiml cquisition nd sorting policies. iddiqui nd Mrny 8] investigted the investment decision on distributed genertion unit fuelled by nturl gs for microgrid. By modeling the nturl gs genertion cost s stochstic process, they found the cost threshold tht triggers the investment. Unlike 57]79]8], this pper studies the dynmic feture of the remnufcturing cost by ssuming it follows stochstic process. Also, we demonstrte tht the uncertinty of the product qulity ctully motivtes decision mkers to defer the remnufcturing. eders my refer to Guide ] for thorough review on the uncertinties in the product remnufcturing process. The rel options nlysis, originted from finnce, hs been extended to the decision mking processes with uncertinties in expnsion, replcement, nd bndonment. Adkins nd Pxson 8] investigted the switching option between two energy sources by ssuming tht feedstock prices follow GBM. Wickrt nd Mdlener 8] investigted the optiml technology investment timing for power plnts with both fuel price nd electricity price following GBM. These rticles

17 9 differ from our pproch in the sense tht the prices re modeled s stochstic, not the cost components. On the other hnd, Ye 83] modeled the mintennce cost s GBM process nd utilized the optiml stopping method to solve the equipment replcement problem. Zmbujl- Oliveir nd uque 53] extended the replcement problem by considering mintennce cost nd equipment slvge vlue s GBM processes. More recently, Min et l. 8] investigted the entry nd the exit of the renewble power producers by modeling the operting nd mintennce cost s GBM. Adkins nd Pxson 8] studied the optiml replcement nd bndonment decisions for rel ssets. The exit decision in 8], the replcement decision in 8], nd the remnufcturing decision in this pper cn ll be considered s optiml-stopping problems. However, this pper, to our knowledge, is the first study to model remnufcturing cost s stochstic process, nd investigte the optiml remnufcturing timing from rel options perspective. The remnufcturing cost uncertinty stems from the uncertinties in end-of-lifeproduct qulity nd component replcement cost. Furthermore, in recent yers, there hs been incresing interest in utilizing the rel options pproch in engineering disciplines. Mikelin et l. 85] developed qulittive study for holistic considertion of rel options in enterprise rchitecture through mechnisms nd types. The benefit of this new development ws demonstrted over trditionl rel options nlysis in identifying broder spectrum for uncertinty mngement. rdin et l. 86] proposed fivephse txonomy of procedures to support the design nd mngement of engineering systems with uncertinty nd flexibility. In this pper, decision mkers recognize the uncertinty in remnufcturing cost, nd hve flexibility on the timing of remnufcturing. eders my refer to 87]88] for more comprehensive reviews on rel options in n engineering context.

18 HAPTE 3 POUT WEIGHT A OLLETIO ATE EIIO I LOE LOOP UPPLY HAI WITH POUT EYLIG 3. Bckground nd Objectives In recent yers, there hve been significnt interests in sustinble processing of end-of-life products such s recycling to reduce the consumption of rw mterils nd the lndfill disposl. A significnt prt of them is driven by the incresing economic motivtions nd environmentl concerns in communities in generl, nd the mnufcturers, collectors, s well s government in prticulr. A criticl fctor relevnt to the environmentl sustinbility is the product weight. In recent yers, lightweight hve been tken s the design principle of electronic products 6]. From n environmentl perspective, reducing the product weight not only brings down the consumption of nturl resources, but lso reduces the wste disposed t lndfills. From n economic perspective, lower level of the product weight implies less mterils cost nd more energy svings. Xerox climed substntil energy svings nd diverted significnt mount of mterils from the lndfill vi lightweight design of toner crtridges 7]. The product weight could be reduced vi utilizing lighter mterils 8] nd/or reducing the product size 9]]. By shrinking curved zoom lenses, onic Minolt reduces the weight of single use cmer by one third in the lst decde 9].

19 Another criticl fctor for the environmentl sustinbility is the collection rte, which is the frction of the sold products tht re collected, nd mesures the collection efficiency in the reverse chnnel. End-of-life products re tken bck to mnufcturer ], nd go through the vlue-retrievl process such s remnufcturing or recycling. Hence, the cquisition of used products is key driver for the success of product reuse nd recycle ]. A higher level of the collection rte results in both less consumption of new rw mterils nd less wste in lndfills, nd cn be chieved, for exmple, by investing in dvertising/eductionl cmpigns to enhnce the public wreness of the tke-bck progrms 3]. Given the significnce of both environmentl fctors, it is highly desirble to understnd how efforts on improving one fctor impct the other fctor nd the totl environmentl consequence. Even though numerous environmentl trgets set by governments re weight-bsed 3], only few nlyticl studies explin nd enhnce our understnding such n impct from n economiclly rtionl perspective. As first step towrd this gol of deeper nlyticl understnding, in this pper, we exmine the vrible product weight nd collection rte, s well s their environmentl consequence of the lndfill quntity vi series of strightforwrd closed-loop supply chins L s. Essentilly, our ttempts to nswer the following reserch questions Under which condition, product recycling is economiclly vible? How the mrginl recycling benefit with respect to the product weight influences the product weight, the collection rte, nd the lndfill quntity i.e., the totl weight of products disposed t lndfill? 3 How supply chin centrliztion/decentrliztion influence the product weight, the collec-

20 tion rte, nd the lndfill quntity? How government subsidy/fee influence the product weight, the collection rte, nd the lndfill quntity? 5 Under which condition, subsidizing the collecting ctivities nd txing the production ctivities re environmentlly vible? How the high level of environmentl efficiency in terms of lndfill quntity in the centrlized L model cn be chieved or outperformed in the decentrlized L model by setting pproprite subsidy nd fee? pecificlly, to nswer these questions, we formulte three closely relted L models consisting of mnufcturer who lso recycles used products nd collector of the used products. The first L is centrlized model with the verticl integrtion of the mnufcturer nd the collector. The second L is decentrlized model with the mnufcturer s the leder nd the collector s the follower in tckelberg gme. The third L is decentrlized model with government subsidy nd fee. In ddition, we formulte non-recycling model s benchmrk for comprison. Under this frmework, the nlytic results show tht i conditions e.g., the increse decrese of the mrginl recycling benefit or the collecting subsidy leding to higher lower level of the collection rte my lso result in higher lower level of the product weight, nd vice vers. Hence, decision mkers my not be ble to pursue n improvement of both fctors simultneously, nd it is dvised to consider the negtive impct on the other fctor before ny efforts on improving one fctor. As consequence, ii the increse decrese of the mrginl recycling benefit or the collecting subsidy for unit weight of used-product collected my result in higher level of the lndfill quntity when the mrginl recycling benefit or the collect-

21 3 ing subsidy is sufficiently high lower. In this sense, efforts on improving one fctor should not be supported unconditionlly, s it my mke the environment worse off. iii Under the revenue-neutrlity frmework i.e., the mount of subsidies disbursed blnces the mount of fees collected, the lndfill quntity in the centrlized model cn be chieved or even further reduced in the decentrlized model by choosing proper vlue of subsidy. However, inpproprite vlue of subsidy my lso result in inferior environmentl performnce i.e., more lndfill quntity reltive to the non-subsidy cse. Given these findings, the pricing of subsidy/fee becomes criticl to chieve high environmentl efficiency nd void unintended negtive environmentl consequences. 3. losed Loop upply hin Models For single product, we consider supply chin consisting of mnufcturer who sells directly to his/her customers nd collector who ccepts used products from the customers nd returns the collected used products to the mnufcturer for fee. The mnufcturer in turn recycles bsic mterils from the returned products in mnufcturing his/her product. We note tht, for the product in this pper, the frction of ech bsic mteril in weight is fixed e.g., 35% glss, 5% metl, 35% polymer, nd does not vry cross units of the product. We lso note tht the portion of the non-recyclble mterils in this product is ssumed to be negligible while ny used product, tht is not collected, is disposed t lndfill. We further note tht third-prty collector, who is only engged in the collection ctivities of used products nd serves s n intermediry between the mnufcturer nd his/her customers, is often observed in vrious mrkets ].

22 More quntittively, in the forwrd chnnel, the mnufcturer sells p units of the product to customers where p $/unit denotes the selling price. In the reverse chnnel, τ frction of the sold products is returned to the collector fter use, nd is in turn supplied to the mnufcturer. The frction is the collection rte while the mnufcturer s fee to the collector is denoted by b $/lb. Mnufcturer les p ustomers Mterils Flow Tke Bck τp ollector ollection τp Lndfill - τp Figure. Mnufcturer-ollector losed Loop upply hin Figure depicts such closed-loop supply chin, nd we note tht the forementioned T s is representtive product for such chin s used T s re recycled into bsic mterils such s glss, metl, nd plstics ]. In wht follows, we explin the key ssumptions of this pper. Assumption 3. The plnning horizon is sttic single period. All the decision vribles re decided in single period see e.g., ]. This ssumption enbles us to concentrte on the fundmentl reltionships of product weight vs. collection rte, centrliztion vs. decentrliztion, nd without government intervention vs. with government intervention without being distrcted by dynmic rmifictions, which re beyond the scope of this pper. Assumption 3. The investment cost to reduce the product mterils weight from w to w is w w ln w / w $ where w is the product weight while nd w w.

23 5 ecently, we observe tht lighter weight hs become mjor business inititive in electronic products 6]. For exmple, onic nd Fujifilm hve reduced the weight of single use cmer by % in the lst decde 9]]. Menwhile, the emphsis on lighter weight leds to n increse in equipment cost 5]6] s more precise nd delicte tools re required to produce the lighter nd smller-size products. Under these circumstnces, we mke simplifying ssumption tht ny product composition chnges due to weight reduction re negligible. Hence, we explined before, the frction of ech bsic mteril in weight is still fixed. A representtive exmple is the cse of Xerox, which ws ble to reduce the weight of toner crtridges without hrd composition chnge by precisely djusting the prison extrusion rte, thickness, nd distribution 7]. As for the functionl form, we note tht logrithmic function is utilized to chrcterize the diminishing returns from higher to lower level of product weight to investment. Mly 8] utilized the sme functionl form to depict the rpidly growing design cost of integrted circuits with the decrese of circuit size. Porteus 9] lso used similr functionl form to investigte the investing ctivities of setup cost reduction. For w, this threshold cn be interpreted s the given product weight of the previous genertion or physicl upper bound of the product weight tht still mintins commercil vibility to customers. Assumption 3.3 Mnufcturing cost w w $/unit where. Electronic industries e.g., single use cmers, copiers, T s re generlly equipped with highly utomted production lines 7]9], nd mterils cost is the key driver of the mnufcturing cost ]]. irect Technologies estimted the mnufcturing cost of ir conditioners by

24 6 simply doubling the mterils cost ]. Furthermore, the liner cost structure enbles us to develop first-cut nlysis nd fcilittes the nlytic results. This functionl form cn be esily extended to other structures such s functions involving qudrtic term with respect to the product weight. Assumption 3. The collector is reimbursed t rte of b $/lb by the mnufcturer for used products. This ssumption is bsed on the existing legisltions/regultions nd industril prctices. In liforni, recycler shll py collector the tndrd ttewide ombined ecovery Pyment te.6 $/lb since July, 8 for ll covered electronic products trnsferred to the recycler 3]. In onnecticut, weight-bsed price is set in E-Wste lws for collectors to get reimbursed by mnufcturers ]. As for industril prctices, odk reimburses collectors $.75/lb for the used single-use cmers regrdless of the composition nd brnd 5]. Atsu et l. 3] lso climed tht, since the WEEE recycling trget is bsed on the product weight, cost lloction between mnufcturers is currently weight-bsed, nd is mnged t best by smpling the collected products. Furthermore, the economies of scle cost structure is prticulrly pproprite in modeling drop-off collection strtegy. The collector invests in mking customers become more wre of the collection progrm so tht more customers drop off their used products t collection sites 6]. We lso mke simplifying ssumption tht other opertionl costs of collecting ctivities re negligible. Assumption 3.5 ollection rte investment function ln $ where.

25 7 The collector invests in mking customers wre of the tke-bck progrm e.g., dvertising nd/or eductionl cmpign 6], nd the investment is quntity-independent. ecll tht the collection rte τ is the frction of the sold products tht re collected. One cn think of τ s the response of customers to the investing ctivities. To incorporte the diminishing returns from lower to higher τ to investment ], we utilize n exponentil function e / where is positive scling prmeter nd is the mount of investment. imilr response functions re frequently observed in dvertising literture 7] chrcterizing customers' responsiveness to the dvertising efforts. Hence, the investment function of the collection rte cn be derived s in Assumption 3.5. Assumption 3.6 Ech returned used product is recycled for the originl purpose, resulting in benefit of w w $/unit where. In recent yers, there hs been growing emphsis on product recycling s profitble process to reduce mterils consumption nd conserve energy. ecycling process converts used products into rw mterils nd energy, nd cn be profitble in cses such s cell phones, T s, nd products with metl volume 9]. For instnce, toner crtridge recycling is 3 billion dollr industry ]. HM olution climed nerly % recycling rte for T s, nd reused the recycled glss, steel, plstics, nd rre mterils e.g., tin, gold nd plldium ]. Pnsonic lso utilized the recycled glss in producing new T s 3]. Here recycling benefit is defined s the difference between the vlue of recycled mterils nd the recycling cost. ue to the nture of recycling process, it is resonble to ssume tht the vlue of recycled mterils is proportionl to the product weight. As for recycling cost, hih et l. 8] climed tht the recycling cost of electronic products e.g., computers minly de-

26 8 pends on the disssembly time, nd estimted tht the disssembly time is proportionl to the product weight. Jeong nd Lee 9] lso pointed out tht the recycling cost of L pnels is proportionl to the product weight. Furthermore, some rticles climed tht the mterils cost is the lrgest cost driver in e-wste recycling 3], nd the recycling cost is closely llied to the product weight 3]. We note tht, Δ here represents the mrginl recycling benefit, nd will be extensively utilized in the sensitivity nlysis lter. Also, the ltter prt of this ssumption implies tht the recycling process is profitble nd the recycling benefit is less thn the mnufcturing cost due to imperfect recycling nd recycling cost. Assumption 3.7 Liner demnd function p p where p $/unit is the price nd,. Liner demnd functions re widely utilized in the supply chin ] nd economics literture 3] to fcilitte the mthemtic trctbility nd s first order pproximtion. Assumption 3.8 The uncollected used products re disposed t lndfill. All in ll, the uncollected used products i.e., collection lekge will end up t lndfills 33]3]. In this pper, we ssume tht the environmentl impct of the supply chins is mesured by the lndfill quntity i.e., the totl weight of the products sent to the lndfill L, nd L w p. Assumption 3.9 In ll the models of this pper, the optiml/equilibrium decisions re interior solutions. pecificlly, 8 nd or for model with government subsidy nd fee.

27 9 ee Appendix A for proof. by excluding pthologicl cses of boundry solutions, we intend to focus on the most relevnt nd interesting cses of our models. This ssumption implies tht the interiority of the optiml/equilibrium solutions cn be gurnteed by sufficiently lrge potentil mrket size nd recycling benefit. Also, this ssumption implies tht > ccording to Assumption 3.6. Given these ssumptions, we now proceed to the supply chin models s follows. 3.. on-ecycling cenrio Without recycling, the collector is not considered in the model. Hence, the mnufcturer decides the selling price p nd the product weight w to mximize the profit this pper, j i will denote the profit of plyer j in supply chin scenrio i. M. Throughout mx p p w ln w / w p, w M 3. The optiml solutions re p 3 8 nd w 8. ince no product is tken bck nd recycled, the lndfill quntity L w p / nd the mnufcturer's profit is ln w 3.. entrlized upply hin with ecycling In the model, we ssume tht the mnufcturer nd the collector re verticlly integrted, nd ll the decisions re mde by centrl plnner with the objective of mximizing totl supply chin profit. entrlized models hve been extensively utilized s benchmrk ginst decentrlized models for the comprtive studies of economic fctors such s price nd profit, s well s environmentl fctors such s collection rte 3]. The relevnt decision vribles for

28 the centrl plnner re the selling price p, the product weight w, nd the collection rte. The profit mximiztion problem is mx p p w] ln ln w / w p, w, 3. The optiml solutions re p 3 8, w 8, nd. Accordingly, the lndfill quntity L w p /, nd the totl supply chin profit is ln ln w 3..3 ecentrlized upply hin with ecycling In the model, we ssume tht the mnufcturer is the leder nd the collector is the follower in tckelberg gme. The ssumption of dominnt mnufcturer is frequently observed in supply chin literture nd is bsed on the belief tht downstrem supply chin members such s collectors re often smller in size nd operte in specific locl mrkets ]. Also, we re ssuming tht the core competence of the mnufcturer is not in collecting the used products ]6]. Hence, given the selling price nd the product weight, the collector optimlly determines the collection rte vi mx bw p ln 3.3 The best response function is * p, w, b bw p. onsidering the collector's best response, the mnufcturer optimlly determines the selling price, the product weight, nd the buybck price vi mx p p b w] / b ln w / w p, w, b M 3.

29 The mnufcturer's equilibrium solutions re p 3 8 8, w ], nd b. ubstituting the mnufcturer's solutions into the collector's best response function, we hve the collection rte. The lndfill quntity L w p, nd the mnufcturer's nd collector's profits re M 8 6 ln w 8 ] 3.5 ln ] ecentrlized upply hin with ecycling nd Government ubsidy/fee -G The supply chin scenrio of the -G model is exctly the sme s the model except tht we re considering the government prticiption in the form of subsidy nd fee see Figure. Assumption 3. In the -G model, under the frmework of revenue-neutrlity, the government provides subsidy of σ $/lb to the collector for the collecting of used products, nd receives fee of α $/lb from the mnufcturer for the products sold. Production Fee αwp Government ollection ubsidy σwτp Mnufcturer Tke Bck τp ollector les p ollection τp ustomers Lndfill - τp Mterils Flow Government ubsidy nd Fee Flow Figure. ecentrlized upply hin with Government ubsidy nd Fee

30 In liforni, collectors re subsidized $./lb for the collection of used video disply devices 35]. In ebrsk, mnufcturers re responsible to py the tte Agency of turl esources $.5/lb for the covered electronic devices bsed on nnul sttewide sles 36]. By revenueneutrlity, we men tht the finncil gins from fees blnce the finncil expenses of subsidies so tht government is finncilly neutrl i.e., w p w p. A revenue-neutrl crbon tx hs been implemented in British olumbi since 8, nd the tx revenue will be returned through tx reductions 37]. In the literture, Plmer nd Wlls 38] utilized tx/subsidy combintion in which producers py per-pound tx nd collectors of used products receive perpound subsidy. Furthermore, Mrozek 39] rgued tht the deposit/refund systems must be implemented in wy tht the revenue is neutrl to chieve the mximum efficiency. The entire opertion of -G model is depicted in Figure. Given the selling price nd collection rte, the collector optimlly determines the collection rte vi mx - w b p ln G 3.7 The collector's best response function is * p, w, b w b p. onsidering the best response function, the mnufcturer mximizes his profit vi p b w b/ b ln w / w 3.8 mx - p p, w, b M G The solutions re p -G 3 8 8, w -G ], nd b - G. ubstituting these solutions into the best

31 3 response function, we hve -G. Accordingly, the lndfill quntity L -G w -G -G p -G. The mnufcturer's nd the collector's profits re M - G 8 6 ln w 8 ] G ] ln 3. Bsed on the optiml/equilibrium solutions summrized in Tble, we performed the sensitivity nlysis bsed on mrginl recycling benefit nd collecting subsidy, s well s comprison nlysis mong three L scenrios. ecycling benefit is the criticl driver of recycling profitbility ], nd ccordingly, is the economic motivtion behind the mnufcturer s recycling investment nd collector s tke bck investment ]3]. Proposition 3. M nd. ee Appendix B for proof. If / bck cost nd the recycling cost. Also, if, the vlue of recycled mterils will not cover the tke /, the mnufcturer's profits in the model nd the totl supply chin profit in the model re greter thn the mnufcturer's profit in the model. Proposition 3. In the model, w,.

32 ollection te Product Mterils Weight w lb ee Appendix for proof. This proposition sttes tht the increse decrese of the mrginl recycling benefit will result in both higher lower levels of product weight nd collection rte see Figure 3 bsed on numericl exmples ecycling Benefit Δ $/lb ecycling Benefit Δ $/lb Figure 3. Vrition of Production Weight nd ollection te with Mrginl ecycling Benefit Intuitively, s increses, there is n incentive for the mnufcturer to pursue higher level of the collection rte so s to gin more benefits from recycling process. Hence, the mnufc-

33 Lndfill Quntity E7 lb 5 turer incentivizes the collector to tke bck more products vi incresing the buybck price nd reduce the investment in the product weight reduction. Hence, decision mkers my not be ble to pursue both higher level of the collection rte nd lower level of the product weight simultneously, nd it is dvised to exmine the negtive impct on the other fctor before ny efforts on improving one environmentl fctor. L Proposition 3.3 In the model, if L, nd if. ee Appendix for proof. This proposition indictes tht, if the mrginl recycling benefit is reltively low, the lndfill quntity decreses with the mrginl recycling benefit, nd if the mrginl recycling benefit is reltively high, the lndfill quntity increses with the mrginl recycling benefit see Figure bsed on numericl exmples ecycling Benefit Δ $/lb Figure. Vrition of the Lndfill Quntity with Mrginl ecycling Benefit Intuitively, when the mrginl recycling benefit is reltively high, both the product weight nd the collection rte re lso reltively high see Proposition 3.. Accordingly, the effect of diminishing returns to the collection investment domintes tht of the product weight reduc-

34 6 tion investment. As result, the mrginl increse of the product weight with respect to domintes the mrginl increse of the collection rte with respect to which cn be observed in Figure 3, which leds to higher level of the lndfill quntity. In this sense, the increse of the mrginl recycling benefit my not lwys fvor the reduction of lndfill quntity, nd hence, should not be encourged unconditionlly. This probbly unintended consequence implies tht the decision mkers must view the criticl environmentl fctors in totlity before ny effort to improve prticulr set of environmentl fctors. Proposition 3. In the,, nd models, given the sme vlues of prmeters, the product weight, the collection rte, nd the lndfill quntity re relted s follows: w w w,, L L L ee Appendix E for proof. This proposition demonstrtes tht the centrlized model domintes the decentrlized model in terms of higher collection rte nd lower lndfill quntity due to its bility to coordinte nd void double mrginliztion ]. This is becuse the centrl plnner cn improve the collection rte by coordinting the forwrd nd reverse chnnel decisions vi single two-prt triff ]. However, due to the inherent conflict between the investments in product weight reduction nd collection rte enhncement, centrliztion lso results in higher product weight reltive to the decentrliztion cse see Figure 3. Moreover, product recycling is lwys beneficil to lessening the environmentl burdens in terms of less lndfill quntity. Thus fr we hve exmined the key insights nd implictions of the supply chin models without government intervention. Let us proceed to exmine such insights nd implictions of the supply chin models with government intervention.

35 7 -G Proposition 3.5 In the -G model, w -G,. ee Appendix F for proof. imilr to Proposition 3., this proposition sttes tht the increse decrese of the collecting subsidy will result in both higher lower levels of product weight nd collection rte. Intuitively, s increses, there is n incentive for the collector to pursue higher level of the collection rte so s to gin more benefits from recycling process. From the mnufcturer's perspective, with higher level of the returned quntity, there is n incentive to reduce the investment in product weight reduction so s to tke more dvntge of the recycling benefit. Given this knowledge, policy mkers should be wre of both positive nd negtive consequences on the environmentl fctors before implementing the subsidy nd fee. Furthermore, it is extremely desirble to investigte the corresponding environmentl consequence involving these two fctors. -G Proposition 3.6 In the -G model, L if L ; -G if ee Appendix G for proof. imilr to Proposition 3.3, this proposition indictes tht, if the collecting subsidy is reltively low, the lndfill quntity decreses with the collecting subsidy, nd if the collecting subsidy is reltively high, the lndfill quntity increses with the collecting subsidy see Figure 5 bsed on numericl exmples. In this sense, the increse of the collecting subsidy my not lwys fvor the reduction of lndfill quntity, nd hence, should not be encourged unconditionlly. We refer the reders to the illustrtion of Proposition 3.3 for the essentil ides of the intuition of this proposition.

36 8 This proposition implies tht, to void unintended consequences, policy mkers must view the criticl environmentl fctors in totlity before ny finncil instrument encourging prticulr set of the environmentl fctors. ue to the similrity between the nd -G model, it is desirble to compre them in terms of the product weight, collection rte, nd lndfill quntity. Menwhile, the model is lso utilized s benchmrk to investigte whether nd how the environmentl performnce in the centrlized model cn be chieved in the decentrlized model. Proposition 3.7 In the nd -G model, given the sme vlues of prmeters, the product weight nd the collection rte re relted s follows: w w -G, -G. ee Appendix H for proof. This proposition demonstrtes tht, government prticiption in the form of collection subsidy nd production fee will result in both higher levels of the product weight nd the collection rte. This implies tht incentives for product collection lso discourge mnufcturer s efforts on product weight reduction, which my not be the originl intention of policy mkers. In this sense, it is interesting to further compre the totl environmentl consequence in terms of the lndfill quntity. Proposition 3.8 In the -G model, L-G L if ; L L L if -G or L L if -G

37 Lndfill Quntity E7 lb 9 ee Appendix I for proof. Proposition 3.8 demonstrtes the significnce of the subsidy on the totl environmentl impct under the revenue-neutrlity frmework. pecificlly, by choosing n pproprite vlue of subsidy, the lndfill quntity in the centrlized model cn be chieved or even further reduced in the decentrlized model see Figure 5. evertheless, on the other hnd, sufficiently high vlue of the subsidy my lso result in higher level of the lndfill quntity reltive to the non-subsidy cse see Figure 5. This implies tht implementing subsidy/fee instrument into the L s my not lwys be environmentlly vible. Given these findings, the pricing of subsidy/fee becomes importnt, nd it is highly dvised to exmine the whole supply chin informtion before ny finncil instrument is offered so s to relize high environmentl efficiency less lndfill quntity nd void unintended negtive environmentl consequences more lndfill quntity ubsidy σ $/lb -G Figure 5. Vrition of the Lndfill Quntity with the ollecting ubsidy

38 3 Furthermore, we note tht the selling prices in the decentrlized models re identicl to the selling price in the centrlized model. This is becuse there is no double-mrginliztion in the forwrd supply chin when the mnufcturer directly sells the products to customers. 3.3 Mngeril Insights nd Policy Implictions The nlytic results in this pper demonstrte tht: i There is n inherent conflict between product weight reduction nd collection rte improvement. Tht is, conditions e.g., increse in the mrginl recycling benefit leding to higher level of the collection rte my lso result in higher level of the product weight. ii As result, efforts on improving one environmentl fctor my negtively impct the other fctor, which, in some cses, will result in n unintended higher level of the lndfill quntity. In other words, under certin conditions, the environment is worse off s product recycling becomes more ttrctive. In this sense, decision mkers re dvised to view the impcts on both environmentl fctors s whole nd exmine the totl environmentl consequence before ny efforts. iii Menwhile, policy mker is dvised to be cutious when implementing subsidy/fee mechnism to L s under the frmework of revenue-neutrlity i.e., subsidies blnce fees. For exmple, if the subsidy is sufficiently high, the L with subsidy/fee my be inferior to the non-subsidy/fee cse with respect to the lndfill quntity. Hence, the pricing of subsidy/fee becomes criticlly importnt to chieve high environmentl efficiency nd void ny unintended negtive environmentl consequences.

39 3 HAPTE EMAUFATUIG EIIO WITH EMAUFATUIG OT UETAITY. Bckground nd Objectives Among the existing sustinble processes for end-of-life products, remnufcturing provides n opportunity for firms from vrious industries to mitigte the environmentl impcts while mintining economic vibility. From decision/policy mkers perspective, there re substntil opportunities to rep the economic nd environmentl benefits of the remnufcturing, such s reverse supply chin designs nd cquisition strtegies. Among these opportunities, remnufcturing timing is especilly importnt for firms providing services or lesing products 3] in the sense tht the product life is determined by internl process mngement nd/or lesing contrcts. We note tht, for this product life, the physicl condition e.g., component brsion deteriortes over time on verge, which complictes the decision mking process on remnufcturing timing. This is supported by the evidence tht the qulity of returned products to mnufcturers in generl vries vstly 3]. With this knowledge, it is highly desirble to understnd the economiclly rtionl decisions on the remnufcturing timing under remnufcturing cost uncertinties. Furthermore, for policy mkers, especilly in the res of governmentl subsidies nd fees, it is criticl to understnd how pproprite policies could influence the remnufcturing decisions so s to reduce the environmentl impcts. In this chpter, we consider firm tht leses single product to service provider, nd the firm is lso responsible for the mintennce of the product. For instnce, Xerox leses photocopiers to tples which provides copying services to customers. Throughout product's lifecy-