Ethical fashion supply chain operations: product development and moral hazards

Corporate social responsibility (CSR) is critical. As a part of CSR, fashion companies have to decide whether to be ethical or not during the product development process. Motivated by real-world practices, we conduct a gametheoretic modeling analysis and derive the firms’ optimal decisions (including ethical operations (ETO) adoption, pricing, and product greenness level) in fashion product development. We identify a key moderating factor which governs how an increase of basic market demand significantly affects the optimal product greenness level and how an increase of basic production cost influences the optimal retail price. Furthermore, we find that there is a threshold that plays a critical role in determining whether the optimal retail price and product greenness level are higher or lower with the adoption of ETO. We prove that when the fixed payment from the retailer to the manufacturer under the ETO case is set to be sufficiently small, the retailer prefers to adopt ETO and requests the manufacturer to follow. We propose three practical measures (including the use of technologies) to help encourage the supply chain members to invest in ETO willingly. We finally consider the probable occurrence of moral hazard problems and explore the managerial implications.


Background and motivation case
Socially responsible operations are very important in modern supply chain systems (Tang 2018;Waltho, Elhedhli, and Gzara 2019;Chen et al. 2020). It is known that the majority of firms have established formal corporate social responsibility (CSR) programmes over the past few years (Cruz 2013;Guo et al. 2015b;Sodhi and Tang 2018). This is especially true in the fashion apparel industry (White, Nielsen, and Valentini 2017;Choi et al. 2018). For example, the widely reported '2013 collapse of Rana Plaza' in Bangladesh hits the spotlight 1 even after many years. 2 Other instances such as some garment factories are reported to treat workers unethically, hire child labours, and many garment factories release lots of pollutants to water and air all create social awareness (Di Benedetto 2017). With an attempt to change the image and reputation, many fashion retail brands have imposed rules on their suppliers to ensure their suppliers are ethical and green with CSR considerations. 3 CSR includes two major aspects, namely the ethical operations (ETO) and environmental sustainability (i.e. product greenness level) (Sim, El Ouardighi, and Kim 2019). For example, H&M introduces a project under its established 'code of ethics', which not only requests itself but also its all business partners (including suppliers, vendors, consultants, landlords, agents, service providers and organisations) to operate ethically. 4 In addition, H&M's suppliers also have to satisfy its requirements on sustainability and greenness developments. 5 In the fashion apparel industry, it is reported that CSR is an emerging industrial practice which has a significant influence (Di Benedetto 2017). One important point to note is: Developing and producing green products is expensive and hence increases the selling price of products. Unfortunately, this may not be well-accepted by the market because consumers may find the increase 'too much' and decline to accept. This paper is motivated by a real case on Nike's supply chain. Our discussion with one of the largest factories in Asia, 6 who is also a major supplier for Nike, reveals that the factory actually has concerns about the development of green products because the green product's production cost is too high. For example, in producing one piece of sports bra by using recycled synthetic fibres mixed with organic materials, it takes the machine one and a half hour (generally, it takes less than half an hour) to complete it because of the high precision needed and the low tolerance towards tension of the selected sustainable materials. In this case, they have to face a high production cost because proper retail pricing for green products is critically important. At the same time, adoption of ETO is another important issue. On one hand, this adoption is usually treated as a symbol of success for manufacturers because they can stand out from the crowd and become the ethical suppliers (Guo, Lee, and Swinney 2015a) for top brands such as Nike, and Adidas with compliance of many important environmental and social rules. On the other hand, the ETO adoption is a big burden to apparel manufacturers.

CONTACT
Motivated by the importance of real world issues such as product greenness level, product pricing, and ETO adoption in fashion apparel product development, we explore the respective optimal decisions in a fashion apparel supply chain. First, we build a game-theoretic analytical model under the decentralised supply chain setting. We derive the optimal product greenness level and the respective product selling price for the new apparel product under the cases with and without ETO adoption. Under the decentralised setting, we reveal the role played by a factor ' DW' (representing the difference between the marginal effects of increasing the product greenness level on product demand and product wholesale price), and highlight its significance in acting as a moderating factor to govern how an increase of basic market demand affects the optimal product greenness level and how an increase of basic production cost influences the optimal retail price. We prove that if the fixed payment from the retailer to the manufacturer under the ETO case is set to be sufficiently small, then the retailer will adopt ETO and ask the manufacturer to follow. We determine the optimal decisions under the centralised supply chain scenario and prove how the supply chain system can be optimised (i.e. coordinated) by an implementable simple contract. Here, supply chain coordination means that the supply chain members under the decentralised scenario not only will choose the optimal product greenness level and optimal product pricing as the supply chain system's optimal decisions, but their ETO adoption decision will also be the same as the supply chain's optimal one. Finally, we focus on examining the situation where it is not optimal for the supply chain and its members to invest in ETO and propose implementable measures to help encourage the supply chain members to invest in ETO willingly. All results are analytically derived in closed-forms and the respective insights are discussed.

Contribution statement and paper's organization
Although the existing papers separately explore the optimal decisions regarding product greenness level, product pricing, and ETO adoption, no prior literature has analytically explored these issues together. The issue of moral hazard, which is commonly seen, is also underexplored. This paper fills these gaps by analytical studying the coordination challenge involving product greenness level, product pricing, and ETO adoption simultaneously. In this paper, in addition to finding the optimal decisions and exploring the supply chain coordination challenge, we have uncovered important insights such as the presence of different moderating factors, which govern some critical relationships between the optimal decisions and the major model parameters. Furthermore, practical measures for enticing ETO adoption and fighting the occurrence of moral hazard problems are proposed. These findings and proposals, together with insights on the moral hazard case, demonstrate the major contribution of this paper.
The rest of this paper is arranged as follows. In Section 2, we report the literature review and highlight the literature positioning of this paper. In Section 3, we present the basic model and show the structural properties of the problem. In Section 4, we derive the equilibrium decisions under the decentralised setting. In Section 5, we first explore the centralised supply chain and the respective optimal decisions. Then, we show how the supply chain can be optimised by employing an implementable coordination contract. In Section 6, focusing on the case in which it is not optimal for the supply chain and its members to adopt ETO, we propose and develop different practical measures which can entice the supply chain members to adopt ETO. In Section 7, we explore the moral hazard problem. In Section 8, we conclude the paper with a discussion on insights and future research. All proofs are placed in Appendix A1.

Literature review
This paper relates to several streams of studies, including product greenness in operations, CSR in supply chain systems, product development operations, and supply chain coordination. We review them as follows one by one.

Product greenness in operations
In the production research literature, sustainable and green operations are crucial. Chen (2001) proposes a quality-based model to analyze the strategic and policy issues regarding the development of green products with general and environmental attributes. Yenipazarli and Vakharia (2017) explore the company's strategy on green product offering. The authors consider how the market responds to product greenness and decides the respective strategy. To be specific, they establish the optimal pricing decisions under each green product alternative and investigate how the product greenness policy affects the environment. Ghosh, Shah, and Swami (2020) study the optimal product greenness and pricing decisions in the presence of consumers who are sensitive to product greenness. The authors study how government rules play a role in affecting the optimal decisions in both singlefirm and two-competing firms scenarios. They reveal that if the government increases the penalty or sponsors product greenness, the competing firm which has a smaller product greening cost will improve the product greenness level. Zhu and He (2017) study the product design challenge of green products. They consider competing supply chains and highlight how the product greenness decisions are affected by the channel structure, type of products and the specific kinds of competition. They comment on the supply chain coordination challenge with its relevance to product greenness. Sun, Sabbaghi, and Ashton (2017) study how the companies' by-product synergies affect the environment. The authors reveal that if the companies are able to share expenses associated with the by-product synergy, a higher market uncertainty will provide even stronger incentive for the development of the synergy. The authors derive the analytical conditions in which the proposed measures are efficient. Ranjan and Jha (2019) study the pricing and sale effort decisions in a dual-channel supply chain where green and non-green products are sold through the online and offline channels, respectively. Furthermore, they address the coordination issue in such type of dual-channel supply chain. Notice that there are some recent analytical studies exploring environmental greening issues from different perspectives. For instance, Wong, Wong, and Boon-Itt (2020) investigate the effects of green supply chain integration on environmental and cost performance. Zheng et al. (2019) examine the optimal strategy of design for the environment with remanufacturing in a duopoly setting. Zhang, Zhao, and Zhao (2019) investigate manufacturers' product greenness strategies when facing 'environment-conscious consumers' in the market. Li et al. (2020) study firms' green product design strategies in a competitive market environment in which firms have 'fairness concerns'. Feng et al. (2019) investigate how the emergence of secondary market platforms affects firms' pricing and product quality design strategies. The authors consider both the absence and presence of secondary market platforms cases with exogenous and endogenous product quality. Recently, Guo et al. (2020a) adopt a multi-method approach to explore green product development in fashion apparel. Shen, Cao, and Xu (2020) study the product quality strategies for green and non-green products in a supply chain, as well as its social and environmental performance. Later on, taking into account the presence of environmental taxes, Shen et al. (2021) investigate firms' green technology adoption decisions in textiles and apparel supply chains. Similar to the above reviewed papers in product greening, this paper also studies the optimal product greenness level. However, different from them, the focal point of this paper is beyond just the exploration of the optimal product greenness level decision. Instead, this paper considers several factors together (such as ETO, product greenness level, and product pricing) and aims to highlight how coordination can be achieved as well as the practical measures for encouraging fashion companies to adopt ETO. Furthermore, this paper is closely related to Chen (2001), but there are some distinct differences. First, the focal in our paper is study whether and when the manufacturer should adopt OTE while Chen (2001) focus on exploring whether the firm should implement the DfE strategy. Second, the retailer in our paper may also produce green products even if the ETO model is not adopted. However, in the work of Chen (2001), the firm may not manufacture green products when the DfE strategy is not implemented. Third, we consider a supply chain structure while there is only a monopolistic firm in Chen (2001).

CSR in supply chain systems
CSR is a timely issue as it is being advocated in almost all companies in the real world (Ngai et al. 2018;Dou and Choi 2021). For example, some works study the implementation of CSR in supply chains from the perspective of closed-loop, green or reverse supply chain management (Zhu and Sarkis 2004;Parast and Adams 2012;Wu et al. 2017;Li et al. 2019). In the literature on supply chain operations, however, a few recent studies have directly explored CSR in supply chains. Lu, Wang, and Lee (2013) empirically study how CSR affects firms' performance by collecting evidence and data from the American semiconductor industry. Arya and Mittendorf (2015) explore the role played by CSR subsidies in the supply chain. The authors uncover that with CSR subsidies, companies tend to increase product price and there is also a negative impact on the wholesale price. Letizia and Hendrikse (2016) conduct a study on the use of incomplete contracts for CSR related operations. The authors demonstrate how incentives for supporting CSR investments may be offered in the supply chain. They show that the way to divide the ownership rights of the supply chain members' production resources can be a way to incentivize the proper CSR investment decision. The authors also prove that if the CSR investment cost increases, the suppliers will be willing to invest in CSR only when the downstream supply chain structure operates in a co-operative manner. Most recently, Hou, Lu, and Hung (2019) study how CSR affects the operations of companies in the creative industry. In particular, the authors investigate how CSR investment relates to the business performance of these companies. Using empirical data and conducting a regression analysis, the authors provide evidence to support the claim that CSR has a statistically significant effect on the firms' business performance in the creative industry. As a remark, a stream of research focuses on ethical operations such as sourcing and supplier monitoring. For instance, Chen and Lee (2017) explore how the payment scheme and auditing can reduce the supplier responsibility risk in ethical sourcing. Agrawal and Lee (2019) examine how the specific procurement schemes would affect the sustainable operations of the suppliers. Similar to the above studies, we also explore the implementation of CSR for supply chain operations. However, different from all of them, we pay attention to two remarkable aspects of CSR, i.e. the ethical operations (ETO) and the environmental sustainability (product greenness level), and focus on examining how the optimal ETO adoption decision relates to the optimal product greenness level and pricing decisions. We also highlight the fact that ETO adoption need not be optimal and hence we propose incentive based methods to encourage companies in the supply chain system to invest in and adopt ETO.

Product development
This paper is related to product development operations. In the literature, Nepal, Yadav, and Johnson (2013) develop a novel method to help estimate the customer preference towards the product, which facilitates the product development process. The authors propose a Bayesian network based system to properly handle the use of information in product development. A comprehensive sensitivity analysis is also conducted to reveal further insights. Wuttke, Donohue, and Siemsen (2018) conduct a behavioural operations management study to show how the product development project relates to incentives. They find that a positive reinforcement rewarding scheme is more effective than a negative reinforcement scheme in supporting product development projects. Zhan et al. (2018) investigate how big data can be used to improve development of new products. The authors focus on revealing how big data would help to collect customer hidden requirements. Klastorin, Mamani, and Zhou (2016) study a market with two competing companies, namely the innovator and the imitator. The authors explore the research issue of whether it is wise for the innovator to pre-announce the development of a new product. By formulating a gametheoretic model, the authors interestingly identify the conditions under which it is indeed optimal for the innovator to pre-announce the product development information even though the imitator is present. Wang, Li, and Chang (2016) empirically study the product development process under the co-development scheme in emerging markets. The authors highlight the impacts brought by the seller-buyer compatibility in the product co-development operations. Chintapalli and Hazra (2015) study the optimal inventory planning and pricing problem for new product development. They focus on revealing the impact of having a hype during product shortages. Chiang and Wu (2016) investigate how early supplier involvement affects new product development. The authors derive the game-theoretic equilibrium to highlight the importance of innovation by establishing alliances in supply chains. Driessen et al. (2013) develop an integrative conceptual framework for developing new green products. The authors verify their proposed framework by conducting eight different real case studies. Shen, Cao, and Xu (2020) analytically examine the optimal product line design and quality differentiation strategies in supply chains producing green and non-green goods. Zhen et al. (2020) study the pricing decisions and government subsidy preference in a green supply chain where firms invest in development green products. Fung et al. (2021) examine the process of sustainable product development in reality and then identify the key steps and factors of the sustainable product development process in fashion industry. Similar to the above reviewed studies in product development, this paper also considers the case in which the apparel supply chain is planning to develop a green product. However, different from all of them, the focal point is on finding the optimal product greenness level and estimating the best product pricing with the considerations of ETO adoption. This is very different from the reviewed studies.

Supply chain coordination
Supply chain coordination is a typical topic that has been widely explored in the literature for many decades (Kanda and Deshmukh 2008;Guo, Cheng, and Liu 2020b). For example, Cachon (2003) explores how to coordinate a supply chain with various contracts, such as buy back contract, revenue sharing contract, quantity-flexibility contract, and sales-rebate contract. Chen (2003) further studies the supply chain coordination problem considering information asymmetry and information sharing. The author then develops some policies to incentivize information sharing between the upstream and downstream firms. Cachon and Lariviere (2005) examine how the revenue-sharing contract can be used to coordinate supply chains with random demand. They find that revenue sharing does not coordinate a supply chain with demand that depends on costly retail effort, although it works under most environments. Panda et al. (2015) investigate the coordination problem in a socially responsible three-tier supply chain. A contract-bargaining process is proposed to address the conflict in such type of supply chain. In addition, Panda, Modak, and Cárdenas-Barrón (2017) study how to coordinate a closed-loop supply chain with a revenue-sharing contract. From a collaborative decision-making perspective, Nematollahi, Hosseini-Motlagh, and Heydari (2017) explore how to coordinate a two-tier supply chain where the upstream supplier is responsible for addressing the concerns on sustainability. The authors propose a novel collaborative model which maximises the total supply chain profit. Similarly, Raza (2018) studies the coordination problem with revenue sharing contract in a supply chain in which firms invest in corporate social responsibility and face stochastic demand. Wang et al. (2020) explore the design and coordination problems in a green e-commerce supply chain where the manufacturer has fairness concerns. For more references about supply chain contracting, please refer to Kanda and Deshmukh (2008). Different to these works, we consider a twotier supply chain where the retailer acts a game leader and both members invest in corporate social responsibility. Furthermore, beside the supply chain coordination issue, our main focus is to explore firms' optimal product development strategy with the consideration of corporate social responsibility investment (i.e. ETO). The consideration of moral hazard is also not examined in the above studies.

Basic model
We consider a supply chain system with one apparel factory (called 'manufacturer') and one fashion brand (called 'retailer'). Naturally, the fashion brand is very powerful and dominates the supply chain as the factory is known to have very low bargaining power and hence the factory's life is very difficult. A specific example of this supply chain system is the one we discussed in Section 1, which involves the production of sports bras using sustainable materials for Nike. Nike is the retailer who dominates the supply chain, and the factory in Bangladesh is the manufacturer. The retailer is the Basic demand under case l ξ The marginal effect brought by the product greenness level on the demand m l Unit production cost for the green product under case l α l Standard basic cost of unit green product under case l β Coefficient of the production cost with respect to the product greenness level F l Fixed cost of the green product development under case l w l Wholesale price under case l θ Profit mark-up that the manufacturer makes by supplying the product to the retailer p Retail price of unit product K(g) Function of product design cost π R,l The profit function of the retailer under case l channel leader who makes a lot of important decisions.
To be specific, the retailer is requesting the manufacturer to produce a new green apparel product. Following the observed industrial practice, we explore the situation when the retailer requests the manufacturer to: (i) produce the green product at the product greenness level of g; (ii) adopt ETO (if it is beneficial to the retailer itself to do so). 7 All related notations are summarised by Table 1. In our analytical model, we use subscripts ETO and ETO to represent the 'ETO adoption' and 'no ETO adoption' cases, respectively. In the supply chain system, the new green apparel product's unit selling price p is decided by the retailer. Since the retailer is the fashion brand, it knows the market preference for product design, and the specific product's design and green level are decided by the retailer. To design a product with the greenness level of g incurs a product design cost K(g) = kg 2 /2, which is paid by the retailer. Observe that the quadratic form of the product design cost is common in the literature (Mukhopadhyay and Setoputro 2005;Guo, Choi, and Shen 2020a;Cai et al. 2022). In order to ensure the manufacturer is compensated in the supply chain by following the retailer's requests, the manufacturer charges the retailer by a supply contract which includes two components, namely a unit wholesale price and a fixed payment (like a 'salary' or 'sponsor' to ensure the manufacturer is well paid by offering the needed production service).
In this paper, we consider ETO that is an important decision. To be specific, we consider the model under two cases, namely one with ETO adoption and one without ETO adoption. For a given l = {ETO, ETO}, the product demand function at the price p and product greenness level g is given as follows: ( 1 ) Note that in Equation (1), the marginal effect brought by the retail price on the demand is scaled to be 1 whereas the marginal effect brought by the product greenness level on the demand is represented by ξ > 0.
The unit production cost for the green product is shown: In Equation (2), the unit production cost is a linear function of the product greenness level. This reflects the fact that the unit production cost involves a certain standard basic cost (α l ) as well as a cost which depends on the product greenness level with a coefficient β, which can be visualised as the cost involved with the use of green or remanufactured materials. Such formulation of production cost is consistent with the practices of Nike's suppliers. During the discussion with one of the largest factories in Asia, who is also a major supplier for Nike, the manager told us that the factories concerns about the development of green products because the green product's production cost is too high. In particular, in producing one piece of sports bra by using recycled synthetic fibres mixed with organic materials, it takes the machine one and a half hour (generally, it takes less than half an hour) to complete it because of the high precision needed and the low tolerance towards tension of the selected sustainable materials. Therefore, it is necessary to develop a cost function which can capture the increased production cost due to the development of green products.
For the retailer, requesting the development and production of the product incurs a fixed cost for the retailer,F l > 0, where l = {ETO, ETO}. In practice, firms always need to spend a higher cost to improve the product greenness level although consumers often prefer products with high greenness level. To capture the reality, we assume that a ETO > a ETO , α ETO > α ETO , and F ETO > F ETO .
The first condition reflects the fact that compared to the case without ETO (i.e. ETO), the basic market demand with ETO for the product is larger when everything else remains unchanged. This is very natural, and is commonly known because consumers in the market would always prefer the product to be produced by a more ethical and socially responsible manufacturer if the retail price and other details are the same. The second and third conditions show that both the per unit (i.e. variable) basic production cost and the fixed cost of green product design and development under ETO are higher than those without ETO. For the second condition, the argument is supported by industrial practices because production under ETO is known to be more costly as the workers are better paid and well-treated, the working environment is enhanced and even materials for production are sourced from responsible suppliers (Guo, Lee, and Swinney 2015a). For the third condition, it is also intuitive owing to the ETO programme's nature.
In the supply chain system, the retailer is the leader who decides the product greenness level g, the product price p, and whether to adopt ETO or ETO. The manufacturer is the follower who reacts by producing the required product and making a profit by the supply contract. As we mentioned earlier, we consider the rather widely seen and commonly adopted wholesale pricing and fixed payment (WPFP) contract in which the manufacturer charges the retailer a per unit wholesale price w l (g) as well as a fixed payment T l , where l = {ETO, ETO}. To be specific, we model w l (g) as follows: where θ ≥ 0 is the profit mark-up that the manufacturer makes by supplying the product to the retailer, and λ = (1 + θ). In addition, m l (g) represents the unit production cost which is defined in Equation (2). With the above details, for l = {ETO, ETO}, we can express the retailer's profit function below: Checking the structural properties of π R,l (p, g), we have Lemma 1.
Lemma 1: For l = {ETO, ETO}, π R,l (p, g) is a concave function of p and g if the new green apparel product's cost coefficient of design is sufficiently high (i.e. k > (ξ − βλ) 2 /2).
Proof of Lemma 1. All proofs are placed in the online Appendix. Lemma 1 provides the sufficient condition with which the retailer's profit function is concave in both the product price p and the product greenness level g. This condition reflects the fact that designing a new green apparel product is expensive which is indeed realistic and practical. This condition helps to ensure that the unrealistic actions like setting a super high price or a super high product greenness level (to the boundaries) will not happen. In the rest of this paper, we would assume this sufficient condition holds so that closed-form analysis can be conducted.

Optimal decisions
Now, we move on to explore the optimal (i.e. equilibrium) decisions 8 in the decentralised supply chain system in this section, and then the centralised supply chain system in Section 5.
Obviously, the supply chain system we proposed in this paper operates in a decentralised manner, with the downstream retailer as the Stackelberg leader and the upstream manufacturer as the follower. Based on the structural properties as revealed in Section 3, we have Proposition 1.
Proposition 1: In the decentralszed supply chain, for l = {ETO, ETO}, the equilibrium (optimal) retail pricing and product greenness level decisions are given as follows: First of all, from Proposition 1, we have the closedform expressions of the optimal (i.e. equilibrium) retail pricing and product greenness level decisions. It is also interesting to note that the optimal decisions can be expressed in the form of a linear combination of a parameters-scaled term with the basic market demand a l and a parameters-scaled term with the basic production cost α l . Note that g * R,l is positively correlated and negatively correlated with a l and α l , respectively. This shows the trade-off inherent to the problem: With ETO, the basic market demand is larger but the basic product cost is also higher. Making the optimal pricing and product greenness level decisions hence needs to consider the trade-off between them. This is natural and logical.
Second, the term ξ − βλ is critical. As it can be positive or negative (with different meanings), how the optimal pricing and product greenness level decisions vary with the changes of parameters rely heavily on it. In Lemma 2, we show some more details of ξ − βλ.
Lemma 2 shows the structural meaning behind the very important factor ξ − βλ in the decentralised supply chain. Essentially, when we increase the product greenness level: (i) ξ = ∂D l /∂g, which is positive (by definition), is a 'favorable effect' to the retailer as the new green apparel product's demand is larger, and (ii) βλ = ∂w l /∂g, which is positive (by definition), is an 'unfavorable effect' to the retailer as the unit wholesale price (which is an expense to the retailer) is higher. Based on its meaning, we denote ξ − βλ by DW, i.e. DW = ξ − βλ (P.S.: The difference between the terms related to '∂D l /∂g' and '∂w l /∂g'), and present Proposition 2.
Proposition 2: DW can be negative or positive. When DW is positive, increasing the product greenness level is relatively 'favorable' to the retailer; when DW is negative, increasing the product greenness level is relatively 'unfavorable' to the retailer.
Proposition 2 indicates the core meaning behind DW and how its positive or negative value affects the retailer. This is important as we will explore the optimal decisions under separate cases with respect to whether DW is positive, negative or zero. When DW is positive, it means the effect of increasing the product greenness level on the demand is significant while the respective effect on the product cost is weak. Therefore, increasing the product greenness level is more likely to benefit the retailer with a higher profit. However, when DW is negative, which implies the effect of increasing the product greenness level product cost is significant while the corresponding effect on the product cost is not. In this case, increasing the product greenness may hurt the retailer. To be specific, we can derive Table 2 which shows how changes of the basic market demand and the basic production cost affect the optimal retail pricing and product greenness level decisions with respect to different values of DW. From Table 2, we can see that DW is a meditating factor that plays an important role in determining the impacts brought by the changes of the respective parameters. We summarise the findings in Lemma 3.  Table 2. Effects on optimal retail pricing and product greenness level decisions when the basic market demand and the basic production cost vary in the decentralised supply chain system.

Range of DW
basic production cost (α l ) on the optimal retail price (p * R,l ), it depends on DW value: Increasing α l will (i) yield a smaller p * R,l if k/ξ < DW, (ii) lead to a larger p * R,l if k/ξ > DW. (d) For the effect brought by an increase of basic production cost (α l ) on the optimal product greenness level (g * R,l ), it depends on DW's value: Increasing α l will (i) yield a smaller g * R,l if DW > 0, (ii) have no effect on g * R,l if DW ≤ 0.
Lemma 3 shows several important findings. First of all, there are some very clear effects brought by changes of the basic market demand and the basic production cost which are true irrespective of whether DW is positive, 0, or negative. In particular, there is a positive correlation between the basic market demand and the optimal retail price, and there is a negative correlation between the basic production cost and the optimal product greenness level when DW is small, otherwise a positive correlation exists between them. Second, we find that there are some effects brought by changes of the basic market demand and the basic production cost which are moderated by DW. In particular, whether the correlation between the basic market demand and the optimal product greenness level is positive or negative or 0 depends on whether DW is positive or negative or 0, respectively. Similarly, whether the correlation between the basic product cost and the optimal retail price is positive or negative or 0 depends on whether DW is negative, positive, or 0, respectively. Part (d) of Lemma 3 shows that firms would like to set a lower product greenness level when the basic production cost increases. This result is consistent with our discussion with one of Nike's suppliers in Hong Kong. The factory has concerns about the development of green products because the production cost of green product is too high. In particular, due to the high precision needed and the low tolerance towards tension of the selected sustainable materials, it usually spends the machine one and a half hour (generally, it takes less than half an hour) in producing one piece of sports bra by using recycled synthetic fibres mixed with organic materials. With Lemma 3, we can derive Proposition 3.

Proposition 3: For the effect brought by an increase of basic market demand (a l ) on the optimal product greenness level (g * R,l ), and the effect brought by an increase of basic production cost (α l ) on the optimal retail price (p * R,l ), DW acts as the critical moderating factor
Proposition 3 reveals the managerial insights and highlights the specific effects in which DW plays as the determining factor. This is important as it helps predict how changes of the major model parameters such as basic market demand and basic production cost parameters would affect the optimal retail pricing and product greenness level decisions.
As a remark, in Appendix A2, we consider the special cases when either the product's retail price is fixed (i.e. exogenously given) or the product greenness level is fixed (i.e. exogenously given). Then, we will see that the effects brought by the changes of the basic market demand (a l ) and the of basic production cost (α l ) become monotone, which basically corresponds to the case in our general model (with both retail price and product greenness level are decisions) when DW < 0 (see Table 2, and Table A1 in Appendix A2). In addition, in the fixed retail price or fixed product greenness level cases, a larger fixed price will lead to a higher optimal product greenness level, and a higher product greenness level will yield a larger retail price. This positive correlation is intuitive and fits the real world very well. To make our research more interesting and meaningful, we focus on the case where DW > 0 in the following sections since similar analysis can be conducted for the case DW ≤ 0.
For a notational purpose, we re-express p * R,l and g * R,l in Proposition 1 as follows (P.S.: The subscripts p, g, a, and α denote retail price, product greenness level, a l and α l , respectively): g * R,l = c g,a a l − c g,α α l , where c p,a = (k + βλ Furthermore, with comparing between the optimal retail pricing and product greenness level decisions under ETO and ETO, we have Proposition 4.

R,ETO
Proposition 4 shows that the ratio between 'the difference in basic market demands (i.e. a ETO − a ETO )' and 'the difference in basic production costs (i.e. α ETO − α ETO )' plays a critical role in deciding whether the optimal retail price and product greenness level are higher or lower with the adoption of ETO (compared to the case without ETO). To be specific, if this ratio is sufficiently high, the optimal decisions (both the retail price and product greenness level) under ETO will be larger than the ones under ETO. This is a very clean result. The intuition is that when consumers prefer firms to adopt ETO and more consumers will potentially purchase the green product, that is (a ETO − a ETO )/(α ETO − α ETO ) > (c g,α /c g,a ), the retailer has more incentives to set a higher product greenness level and therefore charges a higher retail price. Otherwise, the retailer will set a lower product greenness level and then charge a lower retail price. This result is meaningful and provides guidelines for firm's ethical and greenness operations in practice. In particular, firms should do some more empirical research to predict the potential demand as well as the marginal effects brought by the product greenness level on the demand before setting the optimal product greenness level. If the potential demand is not too high and consumers are less sensitive to product greenness, firms should not set a high level of product greenness even though ethical operations are implemented. On the other hand, if the cost of producing products with a high level of greenness is too high under the case with ethical operations, the firm should not set a high greenness level, too.
Moreover, for any values of DW > 0, note that c p,α /c p,a = −λ(k − ξ(ξ − βλ))/(k + βλ(ξ − βλ)) and c g,α /c g,a = λ. Thus, a larger k means a smaller ratio c p,α /c p,a . Meanwhile, a larger ξ means a greater ratio c p,α /c p,a , and a larger λ means a larger ratio c g,α /c g,a . Therefore, these affect the achievability of the conditions in Proposition 4. Since the basic demand under the ETO is higher than that under the no ETO, the effect due to the change of the product greenness level on the retail price will be weaker under the ETO case than under the no ETO case. A larger k means a higher marginal product design cost. With the increase of k, the retailer prefers to set a lower product greenness level, and reduces the retail price as well. Therefore, with the increase of k, the retail price under the ETO becomes more likely to be higher than that under the no ETO case. A larger ξ implies that the effect of product greenness level on demand is stronger. However, with the increase of ξ , the retailer prefers to set a higher product greenness level, and increases the retail price, too. Therefore, the retail price under the ETO becomes less likely to be higher than that under the no ETO case. A larger λ means that the cost of selling a unit product is higher. With the increase of λ, the retailer prefers to reduce the product greenness level and increase the retail price. However, the increase in the product greenness level helps increase demand, and the retailer would like to increase the product greenness level. The negative effect gradually dominates the positive effect with the increase of λ. Therefore, the retailer is more likely to set a lower product greenness level under the ETO case than under the no ETO case.
In the decentralised supply chain, for a given scenario l = {ETO, ETO}, the respective optimal retail pricing and product greenness level decisions have been found in the above discussions. Then, what is the optimal ETO decision, i.e. to adopt ETO or not? To answer this question, we actually need to evaluate the retailer's profits (at the optimal retail pricing and product greenness level decisions) under the ETO and ETO cases, and then compare them. Define: It is easy to derive that for l = {ETO, ETO}, π * R,l can be expressed as follows: where Proposition 5 shows the decision rule in deciding the optimal ETO strategy in the decentralised supply chain system.

Proposition 5: (a) π R,ETO ≥ 0 if and only if T ETO ≤ (A R,ETO − A R,ETO ) + T ETO . (b) In the decentralised supply chain, if the fixed payment from the retailer to the manufacturer under ETO (i.e. T ETO ) is sufficiently small, then it is optimal for the retailer to adopt ETO.
Proposition 5 indicates that if the fixed payment under the ETO case can be set to be sufficiently small (with respect to the fixed payment under the case without ETO), then the retailer will adopt ETO and request the manufacturer to follow. When the fixed payment under the ETO scenario is low enough, for the retailer, the gain from adopting ETO can fully offset the loss from paying to the manufacturer. In the case, the retailer has more incentives to adopt the ETO. This finding is meaningful and it helps to encourage the manufacturer to request a smaller fixed payment under ETO scenario so as to entice the retailer to adopt ETO. Here, we need to note that the retailer also needs to ensure that the manufacturer can get a higher profit under the case with ETO than without, and otherwise the manufacturer may not follow the retailer's requirements. Therefore, it is necessary to identify the condition under which the manufacturer will satisfy the retailer's request.

Centralised supply chain
In Section 4, we have explored the decentralised supply chain system. In this section, we proceed to explore the supply chain systems optimisation (i.e. coordination) challenge (Arya and Mittendorf 2015). To establish the findings, we need to study the centralised supply chain and then find the situations in which the decentralised decisions will be the same as the centralised decisions. In this centralised setting, we treat the manufacturer and the retailer as a single unit, called the supply chain system. For l = {ETO, ETO}, the retailer's profit function has been derived in Section 3. We now present the manufacturer's profit function below: π M,l (p, g) = D l (w l (g) − m l (g)) + T l . We can then express the supply chain's profit function as follows: Similar to Lemma 1, it is easy to prove that π SC,l (p, g) is a concave function if the new green apparel product's cost coefficient of design (i.e. k) is sufficiently high (i.e. k > (ξ − β) 2 /2) and we will continue to consider this case. Thus, the optimal retail pricing and product greenness level decisions for the whole supply chain are given as follows: In the centralised supply chain system, the unit product cost is equal to the unit manufacturing cost m l (g), which is different from the decentralised case in which the 'product acquisition cost' is the wholesale price w l (g). 9 In the centralised supply chain, for a given scenario l = {ETO, ETO}, the respective optimal retail price and product greenness level decisions have been determined. Define: Similar to the decentralised case, it is straightforward to rewrite Equation (12) as follows: where B SC,l = (k(a l − α l ) 2 /4k − 2(ξ − β) 2 ). Define: Note that in Equation (15), by definition, we have F > 0. For the whole supply chain system, Proposition 6 shows the decision rule in deciding the optimal ETO strategy. Proposition 6 shows the situation where it is optimal for the supply chain system to adopt ETO, and this relates to the fixed cost of product design and development under ETO. This is intuitive. When the revenue from adopting the ETO is higher than that under the case without ETO and the fixed cost is lower than that respective cost. The retailer is willing to adopt the ETO, otherwise, it will not. This finding is interesting because the fixed cost F is incurred because of the need to develop the green product. As such, it relates the fact that: If the fixed cost for the new green product development under ETO is sufficiently small, it is in fact optimal to invest in ETO. However, if it is too costly, then developing the green product without adopting ETO is preferred.

Coordinating the supply chain system
In Section 4, we have explored the decentralised supply chain system and obtained the respective decentralised optimal retail pricing, product greenness level, and ETO adoption decisions. However, owing to the double marginalisation effect (Li, Choi, and Cheng 2013), the optimal decisions under the decentralised supply chain system are in general different from that under the centralised supply chain system. To achieve coordination, we have to ensure that the following two conditions are satisfied: (i) p * R,l = p * SC,l , g * R,l = g * SC,l and ETO adoption decisions for both the centralised supply chain and the retailer are the same, i.e. either both vote for ETO, or both vote for ETO. (ii) Both the retailer and the manufacturer are happy in which they can achieve their respective minimum profit requirements (suppose that J R and J M represent the respective minimum profit requirements for the retailer and the manufacturer, respectively).
To have a feasible solution, J R + J M ≤ π * SC,l for l = {ETO, ETO} must hold, otherwise the total minimum profit requirements of the supply chain members would exceed the maximum achievable supply chain profit (which does not make sense). Define: where E * M,l represents the term independent of T l in π * M,l (p * SC,l , g * SC,l ), and E * R,l denotes the term independent of F l and T l in π * R,l (p * SC,l , g * SC,l ) for l = {ETO, ETO}. Note that achieving coordination is an uneasy task for this supply chain system with most other incentive alignment contracts. Luckily, in this paper, what we have observed from the industrial practice is in fact capable of achieving supply chain coordination. The details are summarised in Proposition 7.

Proposition 7: The supply chain can be coordinated by using the WPFP contract if
Proposition 7 shows the situation in which the retailer's optimal retail pricing and product greenness level decisions under the decentralised supply chain system will be exactly the same as the ones in the centralised supply chain system. In addition, if it is optimal for the supply chain to choose ETO, the retailer will also select ETO; if it is optimal for the supply chain to choose ETO, the retailer will also choose ETO. Both the retailer and the manufacturer will be willing to implement the contract because their profit requirements are satisfied. It is easy to see that this proposed coordination method is flexible and easy to implement because there exists a range of different T l which can coordinate the supply chain. In this case, all members in the supply chain can benefit from the adoption of ETO, and the supply chain members can hence discuss the most appropriate value of T l within the bound to be implemented.
Note that even though the coordination mechanism as shown by Proposition 7 seems to be straightforward, the coordination problem in the supply chain is in fact uneasy to address. To be specific, since the coordination involves the coordination of three decisions, namely the retail pricing, product greenness level, and whether to adopt ETO or not, simple supply chain contracts (such as the ones employed in Chung, Talluri, and Narasimhan (2014), Arya and Mittendorf (2015)) are hard to coordinate these three decisions together unless we consider the complicated hybrid contracts. Thus, the relatively simple coordination mechanism as shown in Proposition 7 is probably one of the most effective and practical ways to coordinate these decisions of the supply chain.

Measures to encourage ETO adoption
From Sections 4 and 5, we note that it is not always optimal for the retailer and the supply chain to adopt ETO. In this section, we propose some measures which can encourage the supply chain members to adopt ETO under the case when ETO adoption initially looks undesirable in the absence of additional measures. As we assume that the supply chain members can coordinate between themselves (e.g. by using the method in Proposition 7), these measures are established to focus on the supply chain system, instead of the retailer.

Government supports to reduce ETO adoption cost
ETO is an important issue. If the government is willing to support companies' development of ETO by providing some sponsorships, then, companies in the supply chain may be enticed to vote for ETO. To be specific, from Proposition 6, we know that if the fixed cost for developing the green product under ETO is too high, the supply chain will not move to ETO. As such, we can prove that proper sponsorship can help. The result is summarised in Proposition 8.

Proposition 8: If originally F ETO > F ETO + B SC , which means that it is not optimal for the supply chain to invest in and adopt ETO: (a) The government can entice the supply chain to adopt ETO by sponsoring a lump-sum of money G ETO to help if and only if G ETO > F − B SC . (b) The minimum amount of required government sponsorship to guarantee ETO adoption is equal to F − B SC .
Proposition 8(a) is intuitive and it also analytically shows the magnitude of sponsorship needed in order to make the supply chain adopt ETO. Clearly, once the government offers sponsorship to the firm, it will reduce the cost to adopt ETO, and therefore the retailer is more likely to benefit from the adoption of ETO. Moreover, directly from Proposition 8(a), we can see that the minimum amount of required government sponsorship is F ETO − F ETO + (B SC,ETO − B SC,ETO ) = F − B SC , which gives Proposition 8(b). Thus, if F ETO is smaller or F ETO is larger, then the required government's sponsorship can be reduced (and hence the spending of some public money can be saved). To make F ETO smaller or F ETO larger, the government can actually adjust its taxation scheme to benefit and encourage ETO and penalise ETO. It is obvious that this proposed measure is implementable and practical.

Enhancing public awareness of ETO
For companies in the fashion supply chain, adopting ETO can bring a benefit to improve their corporate image and brand name. This is critical because an improved corporate image and brand name will lead to a bigger market size. As a result, if better promotion is made to the public for ETO adoption so that the market awareness is higher, the basic market demand for the new green apparel product when the fashion supply chain adopts ETO will be higher. This gives rise to Proposition 9, as shown below.
Proposition 9: If originally it is not optimal for the supply chain to adopt ETO, the supply chain members can promote its ETO adoption to the public. As a result, if the basic market demand a ETO is increased toâ ETO by proper promotion, and the respective B SC,ETO value is denoted bŷ B SC,ETO , then it is optimal for the supply chain to adopt ETO if and only ifB SC,ETO ≥ F + B SC,ETO .
Proposition 9 carries a good meaning in which it advocates the fashion companies to make their ETO adoption more publicly known. This is also intuitive. When the public awareness to ETO increases, consumers become more willingness to pay a higher price for the green product once the retailer implements ETO. Therefore, the adoption of ETO benefits the retailer with a higher profit. In fact, this can be achieved via many channels, including the support by not-profit making organisations. If the market response is sufficiently good to the promoted ETO adoption image, then it will naturally be optimal to adopt ETO (as the market demand is substantially enhanced). This approach helps to improve the business benefit from ETO and is in fact beneficial for the fashion companies in the long run.

Technology to reduce basic production cost under ETO
In section 6.2 above, we proposed to use 'promotion' to enhance public awareness of the fashion companies adopting ETO, which can help improve the market demand. In this subsection, we propose the use of technology to reduce the basic production cost under ETO, i.e. α ETO . Suppose that with a better technology, such as those disruptive technologies in the Industry 4.0 era (e.g., IoT, blockchain, AI), the basic production cost under ETO can be reduced from the original value of α ETO toα ETO . As a result, it will be more profitable to produce the fashion product under ETO. As such, we have Proposition 10.
Proposition 10: If originally it is not optimal for the supply chain to adopt ETO, by employing technology, if the basic production cost α ETO is decreased toα ETO , and the respective B SC,ETO value atα ETO is denoted byB SC,ETO , then it is optimal for the supply chain to adopt ETO if and only ifB SC,ETO ≥ F +B SC,ETO .
Proposition 10 follows the same logic as Proposition 9's, but from a different angle. Proposition 9 focuses on enhancing the market size, which is from the revenue improvement perspective. Proposition 10 talks about cost reduction, which is from the cost saving perspective. When the retailer can reduce the production cost by implementing new technology, it is more likely that the retailer can get higher profit from the adoption of ETO. Therefore, the adoption of ETO will become more attractive. Even though the above two measures tackle the challenge from different angles, they both serve the same purpose and aim to entice the fashion companies of the supply chain to adopt ETO.
If we look at the condition, i.e.B SC,ETO ≥ F + B SC,ETO , to make it easier to be satisfied, we would like to have a smaller F. This relates to section 6.1 arguments. In fact, if the government is willing to partially support ETO (e.g. by sponsorship or taxation benefit), which reduces F. However, if the amount is less than the minimum requirement as proposed in Proposition 6.1(b), then the government sponsor alone is insufficient. Luckily, the fashion company can employ technology to reduce the basic production cost further, which means that the conditionB SC,ETO ≥ F +B SC,ETO would be easier to achieve as ' F' has already been reduced by a partial government sponsor. This makes ETO the optimal choice. The same argument is applicable to the proposed measure in section 6.2. In fact, all the three proposed measures can be combined together to help entice the fashion companies to adopt ETO. In other words, these proposed measures are not mutually exclusive, but are in fact complementary and could be used together flexibly.

Extended model: moral hazard
In the model that we studied above, we assume that the retailer and the manufacturer are honest and they are really investing in ETO if it is optimal for them to choose the ETO mode of product development. However, this is known to be 'too ideal' because in practice, both the retailer and manufacturer may be tempted to tell lies and hence moral hazard problems arise (Cai, Choi, and Zhang 2021). In this section, we consider an extended model in which we have the moral hazard problem. Table 3 shows the scenarios.
Once the retailer and the manufacturer tell lies, their malicious actions may be discovered by consumers via various ways at a certain probability, such as mass media and news coverage. To better capture the practice in reality, we assume the probability that consumers will discover the retailer does not commit to ETO is ρ r and the manufacturer does not produce under ETO is ρ m , where 0 < ρ r , ρ m < 1. At the same time, if the retailer and the manufacturer pretend to undertake the ETO and are discovered, the retailer encounters a loss. For example, consumers may require returning their purchased products and asking for subsidies. At the same time, the retailer may penalise the manufacturer if he is discovered not to produce products under the ETO.
To be specific, if consumers in the market discover that the ETO is not undertaken by the retailer or the manufacturer, the retailer has to pay for a subsidy to every consumer, which is denoted by σ (σ > 0). If the manufacturer is found to tell lies, a penalty (denoted by μ) will be charged by the retailer while the retailer is penalised by the market, and μ > 0. If the retailer and manufacturer are found to both tell lies, the retailer is penalised by the market whereas the manufacturer will be penalised by the retailer.
Lemma 4: Considering the moral hazard issue, we can obtain the optimal pricing and greenness level decisions, as well as the payoff under different cases. Correspondingly, the optimal pricing and greenness level decisions are provided in Table 4 below, the optimal payoff is presented in Appendix A1.
Lemma 5 shows that the difference between the marginal effects of increasing the product greenness level on product demand and product wholesale price, i.e. DW = ξ − βλ, also plays an essential role in driving the retailer's optimal decisions when both the retailer and manufacturer are likely to suffer moral hazard problem. Depending on the relationship of the subsidy σ and penalty μ, the retail price under the moral hazard case can be either higher or lower than that under the basic model. In particular, if σ > μ, the moral hazard problem leads to a decreased product greenness level when DW is positive, which eventually causes a decreased consumer surplus, an increased environmental impact, and a decreased social welfare. However, if σ < μ, the moral hazard problem will induce the retailer to set an increased product greenness level when DW is positive because the retailer can acquire a positive revenue as long as the manufacturer tells lies. As a result, this will increase consumer surplus and decrease environmental impact. Furthermore, the moral hazard problem also affects the optimal selling price. Depending on the sign of k − (ξ − βλ)ξ , the optimal price under the moral hazard scenarios can be either higher or lower than that in the basic case. In the following, we aim to explore how moral hazard problems drive the equilibrium in the market. Proposition 11 summarises the result.
Proposition 11 illustrates that the retailer and manufacturer prefer to act honestly under some conditions. In particular, when the gap between the fixed cost of investing in ETO and that of investing in ETO is small, and the gap between the production cost under the ETO and that under the ETO is also small, the retailer and the manufacturer will not tell lies. In this case, it will be optimal for them to actively let the other one know the actual information. However, when these gaps are large, both the retailer and manufacturer have incentives to tell lies. In 2k − (ξ − βλ) 2 this case, TT will be the unique equilibrium. The reason for this result is clear. It is because telling lies can bring a higher profit for both the retailer and manufacturer as compared with the HH case. Finally, if the gap between the fixed cost of investing in ETO and that of investing in ETO is moderate, or the gap between the production cost under the ETO and that under the ETO is also moderate, whether or not the retailer (or manufacturer) can get a higher profit depends on the action of the manufacturer (or retailer). In these cases, either the HT or TH may be the equilibrium in the market.

Summary and insights
Ethical operations (ETO), product greenness level and retail pricing are all important decisions nowadays in new green apparel product development. In this paper, based on real world cases, we have constructed an analytical model and derived the optimal pricing, product greenness level, and ETO adoption decisions under both the decentralised and centralised supply chain settings. We have obtained some interesting results as summarized below.
Under the decentralised setting, we have uncovered the role played by DW (i.e. the difference between the marginal effects of increasing the product greenness level on product demand and product wholesale price), and highlighted its significance in acting as a moderating factor (see Proposition 4.3) to govern how an increase of basic market demand (a l ) affects the optimal product greenness level (g * R,l ), and an increase of basic production cost (α l ) influences the optimal retail price (p * R,l ). We have constructed two special cases (Appendix A2) in which the retail price and the product greenness level are fixed, respectively, and derived the corresponding optimal decisions. It is found that in the special cases, the effects brought by the changes of basic market demand (a l ) and basic production cost (α l ) on the optimal retail pricing and product greenness level decisions are monotone. We have also shown from Proposition 4.4 that the ratio between 'the difference in basic market demands (i.e. a ETO − a ETO )' and 'the difference in basic production costs (i.e. α ETO − α ETO )' plays a critical role in deciding whether the optimal retail price and product greenness level are higher or lower with the adoption of ETO. We have then proven in Proposition 5 that if the fixed payment from the retailer to the manufacturer under the ETO case (i.e. T ETO ) is set to be sufficiently small, then the retailer will be pleased to adopt ETO and request the manufacturer to follow. These results first explain why the interviewed manufacturer has concerns about Nike's development of green products and also provide some guidelines for firms like Nike and Addidas to set the optimal product greenness levels under the cases with/without ETO. In addition, we suggest that it is better for firms to do some empirical studies before the implementation of ETO in terms of the potential demand, the marginal effects of product greenness level on the demand, as well as the corresponding cost, which will be finally helpful to set the optimal product greenness level under different cases.
Then, in the centralised supply chain system, we have discovered an important factor, denoted by DM, which represents the difference between the marginal effects of increasing the product greenness level on product demand and product manufacturing cost. The role of DM in the centralised supply chain is the same as DW's role in the decentralised supply chain setting. We have also derived the analytical conditions on the fixed ETO adoption cost, under which it is optimal for the supply chain to adopt ETO. After that, we have proven how the supply chain system can be coordinated by an implementable simple WPFP contract in which the supply chain members not only will choose the optimal product greenness level and retail price as the supply chain system's optimal decisions, but their ETO adoption choice will also be the same as the supply chain's optimal decision. Furthermore, we have explored the situation in which it is not optimal for the supply chain and its members to adopt ETO, and proposed three different practical measures to help entice them to change and adopt ETO willingly. To be specific, we have proposed measures such as (i) having government's supports to reduce ETO adoption cost, (ii) enhancing market demand by promoting ETO of the company, and (iii) reducing the basic production cost by using technologies. For each measure, the respective closed-form analytical conditions are derived. Implications are discussed. Notice that these three proposed measures can be used in isolation or together, i.e. they are independent and complementary. In practice, firms can invest in either advertising to improve customers' awareness of ETO or adopting new technologies to reduce the cost of ETO, which would improve the respective demand and cost. This would hence enhance profitability. Also, to encourage firms to adopt ETO, the government can provide some subsidies to support the firms' ETO implementation so that workers can have healthy workplaces and opportunities to improve themselves, or enjoy some better health and safety. All these actions would very likely help increase the social welfare and environmental performance.
Finally, we have considered the case in which the moral hazard problem may occur between the retailer and the manufacturer. The result shows that the difference between the marginal effects of increasing the product greenness level on product demand and product wholesale price also drives the retailer's optimal pricing and greenness level decisions. The moral hazard problem will cause the retailer to reduce product greenness level. Interestingly, we find that under some conditions, to be honest will be dominant strategy for both the retailer and manufacturer. In other words, the retailer and manufacturer should actively announce the real information of their CSR actions.

Future research
This research can be extended in a number of ways. For instance, we can investigate a more complex supply chain in which there are multiple retailers and they are direct competitors in the market. How the market competition level (So 2000;Li et al. 2016) affects the optimal product greenness level and ETO decisions will be interesting to explore. Moreover, in this paper, we have not considered the probable risk associated with product development, demand, and sales. In particular, demand in this paper is assumed deterministic and as a linear function of price and product green level. In the future, operations risk for product development and supply chain operations can be examined. Last but not least, it is found in the literature that environmental awareness might stimulate innovation. Thus, future research can be conducted to explore how innovation may relate to ETO in new green apparel product development.

Data availability statement
Data available within the article. The authors confirm that the data supporting the findings of this study are available within the article.
Notes 7. Note that in this paper, we consider the case when ETO is adopted, even though the market base demand will be higher, the product's production cost and wholesale price will both increase. As a result, adopting ETO need not be desirable for the retailer. 8. In this paper, we use the term 'optimal decision' and 'equilibrium decision' interchangeably. Thus, for the decentralised supply chain system, the optimal decisions refer to the equilibrium decisions under the Stackelberg game. 9. See Appendix A3 for some more details and analyses.