Green supply chain management is defined as a strategic capability consisting of strategies, practices and policies that concentrate on managing the environmental impact of supply chain operations (Rauer & Kaufmann 2015). It may also refer to the integration of ecological thinking into supply chain management, including product design, material sourcing and selection, manufacturing process, delivery of the final product to the consumers as well as the end-of-life management of a product after its useful life (Srivastava 2007). These definitions implicitly include an ecosystem philosophy of decreasing externalities (waste and pollution) and materials recovery while still focusing on the economic benefits of environmental responsibility (Griggs et al. 2013; Zhu & Sarkis 2004). For many organisations, GSCM is a way to demonstrate their sincere commitment to sustainability (Bacallan 2000).

Much research on GSCM has focused on the functional areas of supply chain management such as purchasing and supply management (Carter & Dresner 2001; Mollenkopf et al. 2010). However, according to Sarkis (2012), the integration of environmental issues and supply chain management has become a thriving subfield over the past few decades. The potential benefits associated with GSCM include enhanced reputation, increased efficiency, effectiveness, differentiation and revenue growth (Golicic & Smith 2013; Rao & Holt 2005; Wu & Pagell 2011). Furthermore, being ‘green’ is also important from a long-term economic perspective because without natural resources, both business and the consumption of goods are severely stifled (Bell, Mollenkopf & Stolze 2013). Golicic and Smith (2013) found evidence of positive tangible economic and ecological outcomes associated with GSCM practices. Also, according to Lamming and Hampson (1996), GSCM practices can be categorised into green purchasing, eco-design (ED), reverse logistics (RL) and legislation and regulation (LR). The implementation of these practices has an impact on the operations of businesses throughout the supply chain.

Green purchasing refers to the practice of cooperating with suppliers to develop products that are environmentally sustainable (Carter & Carter 1998; Zhu, Sarkis & Lai 2008). According to Lee (2008), a buying organisation with a GSCM initiative will pay attention to green practices of their suppliers by deploying collaboration-based activities that include training, environmental information sharing and joint research. Other organisations may opt for a less collaborative approach by merely demanding that their suppliers make use of environmental systems such as ISO 14001. External motivators, particularly customer pressure, are key drivers of the adoption of ISO 14001 (Heras-Saizarbitoria, Landin & Molina-Azorin 2011; Vachon & Klassen 2007). Other aspects of green purchasing include the facilitation of recycling, reuse and resource reduction (Diabat & Govindan 2011). There is also evidence that some organisations adopt compliance and evaluative approaches to the green purchasing practices of their suppliers (Large & Thomsen 2011). This involves evaluation of suppliers based on environmental criteria and a requirement for suppliers to develop and maintain some form of the environmental management system (EMS) (Large & Thomsen 2011; Min & Galle 2001; Sarkis 2012; Zhu, Sarkis & Geng 2005). These insights lead to the following hypothesis:

H1: Implementation of green purchasing improves SME environmental performance.

Eco-design is concerned with the design of products that minimise the consumption of materials and energy, which facilitate the reuse, recycle and recovery of component materials and parts, and that avoid or reduce the use of hazardous products within the manufacturing process (Zhu et al. 2008). The importance of ED was identified by Buyukozkan and Cifci (2012) when they revealed that about 80% of product-related effects on the environment emanate from the design. Eco-design practices fall into two main categories: product-related design and packaging-related design. Concerning product design, Min and Galle (2001) suggested that cost-saving opportunities at the beginning of the supply chain tend to be greater and that buying organisations need to seek for opportunities to utilise recycled and reused components actively. However, Wu and Pagell (2011) stressed that the environmental impacts of a product occur at all stages of its life-cycle and identified life-cycle assessment as a commonly used attribute of GSCM.

Building on the theme of life-cycle impacts, Field and Sroufe (2007) noted that one of the sources of recycled materials is post-consumer waste, while Zhu et al. (2005) suggested that it is possible to sell or reuse products or their contents. The implication is that it is important for organisations to ensure that their products comprise contents that can be reused or recycled. With respect to packaging-related ED, a discussion of GSCM practices by Zhu et al. (2005) suggested that organisations and their suppliers should collaborate to ensure that they use green packaging for their products. Other studies have identified elements of green packaging to include ensuring that packaging is reusable and recyclable (Carter & Carter 1998; Large & Thomsen 2011), minimising waste by reducing packaging (Walker, Di Sisto & McBain 2008) and avoidance of hazardous material (Buyukozkan & Cifci 2012). These perspectives invoke the following hypothesis:

H2: Implementation of ED improves SME environmental performance.

Reverse logistics refers to a flow of products or goods back from the consumer to an earlier stage of the supply chain (Seroka-Stolka 2014). This is performed for the purpose of recycling, reusing, remanufacturing, repairing, refurbishing or ensuring safe disposal (Carter & Ellram 1998; Younis Sundarakani & Vel 2016). Previous literature indicates that RL activities include product returns and remanufacturing (Olorunniwo & Li 2010), recovery, recycling and reuse (Field & Sroufe 2007) and redistribution (Das 2012). Reverse logistics practices concern final products, their components (Das 2012) and packaging material (Field & Sroufe 2007), which apply to both upstream and downstream of the supply chain (Lau & Wang 2009). Just like other GSCM initiatives, RL plays an important role in enhancing the organisation’s operational efficiency, improves competitiveness and reduces system-wide costs (Lau &Wang 2009). Greening the RL networks is considered a strategic tool as it can provide lucrative economic benefits and improve organisational competitiveness (Buyukozkan & Cifci 2012; Murphy & Poist 2003). Reverse logistics therefore focuses on saving money and adding value to organisational performance by fostering the reuse or resell of materials in order to recover the loss of profits and reduce operational costs (Xie & Breen 2012). These views lead to the following hypothesis:

H3: Implementation of RL improves SME environmental performance.

Legislation and regulatory practices refer to laws, policies and rules that are usually promulgated by government and other regulatory bodies for the purposes of promoting the preservation of the environment (Björklund, Martinsen & Abrahamsson 2012). Both proactive and reactive organisations have developed effective green regulation programmes as a result of government pressure and other regulatory bodies (Chung & Wee 2011; Sheu 2011). Chen and Sheu (2009) suggest that relevant public policies are essential in substantiating the greening of the supply chain. Lu, Wu and Kuo (2007) further assert that many organisations give up their old practices to steadily engage in efforts to green their supply chains in response to stringent LR. These efforts include proactively addressing environmental and social issues in advance of regulation (Zailani et al. 2012). Zhu et al. (2015) stress that regulations and related programmes tend to promote energy efficiency and reduction of pollution. This leads to the following hypothesis:

H4: Adherence to LR improves SME environmental performance.

Environmental performance may be perceived as the measure of reduction of substances and emissions that reduce environmental impacts caused by business organisations (Kumar, Chattopadhyaya & Sharma 2012). Environmental performance is usually measured in terms of reduction in air emission, energy consumption, hazardous material, material usage and compliance to environmental standards (Zhu, Tian & Sarkis 2012). There has been an upsurge in the environmental awareness of consumers in general, together with a growing number of corporations developing company-wide environmental programmes and green products sourced from markets around the world (Delmas, Nairn-Birch & Lim 2015; Min & Galle 1997; Zobel 2015). Lindenberg and Steg (2007) suggest that some firms pursue ecological initiatives because their primary decision-makers have an innate concern for the environment and want to do as much good for the environment as possible. In these firms, top management is responsible for their environmental management leadership, which idealise rather than rationalise the best course of action (Hoejmose, Brammer & Millington 2012). As such, GSCM is now recognised as a strategy for protecting the environment as well as a source of motivation which focuses on improving employee morale and individual satisfaction (Govindan et al. 2014). Besides, environmental issues have become a source of competitiveness, which also helps to improve efficiency and synergy amongst business partners and enhance environmental presence, minimise waste and achieve cost savings and firm reputation (Rao & Holt 2005). This leads to the following hypothesis:

H5: The better the SME environmental performance, the greater the performance of the supply chain.

Supply chain performance is the operational measures that contribute to the improvement of the entire supply chain as a result of collaborative relationships amongst supply chain members (Gagalyuk, Hanf & Hingley 2013; Nyaga et al. 2013; Odongo et al. 2016; Whipple, Lynch & Nyaga 2010). It has also been defined as a systematic process of measuring the effectiveness and efficiency of supply chain operations (Neely, Gregory & Platts 1995). Literature suggests that supply chain relationships create opportunities for firms to experience improved performance (Fynes, de Búrca & Mangan 2008; Wu, Choi & Rungtusanatham 2010). The perceived contribution of a supply chain member to SCP is typically measured using the four constructs of efficiency, responsiveness, quality and supply chain balance. Efficiency is a measure of how well resources are utilised and includes logistics costs and profits (Aramyan et al. 2007). Responsiveness is a measure of speed or rate of providing the requested products and is measured in terms of lead time and customer complaints (Aramyan et al. 2007). Quality consists of products and process quality, with product quality consisting of safety and attractiveness, while process quality is measured by environmental friendliness (Aramyan et al. 2007; Chen & Paulraj 2004; Neely et al. 1995). Supply chain balance is defined as the distribution of risks and benefits as well as supply chain understanding. Thus, SCP promotes collaboration amongst members of the supply chain and ensures continual improvement of effectiveness and efficiency of processes (Anand & Grover 2015; Sundram, Chandran & Bhatti 2016).

In general, most SMEs are confronted with various challenges that prevent them from engaging in GSCM. Some of these challenges include the lack of financial and human resources, limited innovation capability and limited operational know-how (Abbasi & Nilsson 2012; Wang 2016). As observed by Chin et al. (2012), the majority of SMEs are financially constrained to implement GSCM initiatives and lack formalised organisational structures required to implement GSCM action programmes. A previous study by Preuss (2011) noted that some SMEs are of the view that a formalised corporate structure limits flexibility in decision-making. Also, implementing GSCM requires innovative technological capabilities in the realms of green design, managerial competencies in supplier evaluation and negotiation skills, which are inherently scarce in most SMEs (Mohanty & Prakash 2014). It also requires employee training and development to instil the green corporate culture needed to roll out and sustain GSCM practices (Diabat & Govindan 2011). However, although these challenges may prove overwhelming, GSCM remains a viable strategy in the long term through cost savings and product differentiation (Mohanty & Prakash 2014).

Figure 1 outlines the conceptual model that directed this study.

Figure 1 shows that the implementation of GSCM practices through green purchasing, ED, RL and legislation and regulatory practices (predictor variables) has a positive influence on the environmental performance (mediating variable), which in turn has a positive impact on SCP (outcome variable).

The sample used in this study was composed of 65 SMEs based in Gauteng province. There was no sampling frame in this study as it was difficult to obtain a single list of all SMEs operating in South Africa. In view of this, a non-probability convenience sampling technique was used to select owners and managers of SMEs in Gauteng. The convenience sampling technique considered respondents who were readily available, and who formed part of the target population. To determine the adequacy of the sample size (n = 65) used in this study, two considerations were made. The first was the historical reference technique, which gave reference to previous but similar studies in supply chain management (Huang, Boon & Xiaoming 2015; Lee 2008; Mafini & Muposhi 2017; Younis et al. 2016). These studies used sample sizes ranging between 50 and 600 respondents. The second consideration was to use the standard put forward in a survey conducted by Wolf et al. (2013) that examined sample size requirements for common applied structural equation modelling (SEM). In that study, it was discovered that sample sizes ranging between 30 and 460 respondents produced meaningful patterns of association between parameters. Therefore, the total sample size of n = 65 cases was accepted for data analysis in this study. The profile of the SMEs that participated in the study is presented in Table 1.

An analysis of the statistics in Table 1 indicates that manufacturing SMEs were the majority in the sample (64.6%; n = 42), followed by SMEs in the transport and electricity, gas and water industry (10.8%; n = 7) with construction SMEs (6.2%; n = 6) coming third. Further analyses show that 27.7% (n = 18) of participating SMEs employed between 21 and 50 people, while 20.0% (n = 13) employed less than 20 employees. Another 20.0% (n = 13) of the SMEs employed between 51 and 100 people. Concerning annual turnover, 26.2% (n = 17) of the participating SMEs had an annual turnover of between R10 and R20 million, while 24.6% (n = 16) had annual turnovers of less than R10 million. Another 24.6% (n = 16) of SMEs had annual turnovers of less than R50 million.

In this study, 95 copies of the survey questionnaire were distributed to owners and managers of SMEs in Gauteng, South Africa in September 2017. A combination of e-mails and the drop and collect method was used in distributing the questionnaires to respondents. Initially, a week was given to respondents to complete the questionnaires. However, this deadline was extended to two weeks as a result of a prolonged return rate of the questionnaires. A confidentiality agreement accompanied each questionnaire to facilitate the data collection. All of the respondents completed the same questionnaire as they represented their respective SMEs. After the data collection process, a total of 67 questionnaires were returned, and 2 were discarded because they had errors. This culminated in 65 questionnaires used in the final data analysis, providing a response rate of 68%. Measurement scales used in the study were operationalised using previous studies. Table 2 shows these questions and their sources.

To achieve the aim of the study, statistical analyses were performed using the Statistical Packages for Social Sciences (SPSS version 24.0) and the Analysis of Moments Software (AMOS version 24). Relationships between the research constructs were tested using SEM.

Inferential statistics refer to statistical analyses that allow the researcher to draw inferences about large populations from relatively small samples (Leedy & Ormrod 2015). Inferences are drawn based on tests conducted to determine the relationship between green practices and environmental performance, and the relationship between environmental performance and SCP. A two-way formula recommended by Anderson and Gerbing (1988) was used, which involves performing a CFA followed by testing the proposed research hypotheses utilising the path analysis approach.

A CFA was performed to test the reliability, validity and determine the model-fit requirements. The results of the CFA are presented in Table 3.

The statistical results from Table 3 show that the reliability was ascertained using the Cronbach’s alpha test and the composite reliability (CR). Mitchell and Jolley (1996) suggest that the values of Cronbach’s alpha test that are greater than 0.7 are considered acceptable as they fulfil the requirements of internal consistency. In Table 3, the values of the Cronbach’s alpha test lie between 0.866 and 0.966, which demonstrates that the requirements for internal consistency were met. Regarding the CR test, Fornell and Lacker (1981) assert that the minimum threshold of 0.7 should be attained to confirm the acceptability of the reliability. In this study, the value of CR ranged between 0.84 and 0.95, which confirms that the minimum threshold was attained. Therefore, all measurement scales used for this study were deemed to be reliable.

Convergent validity was determined through factor loadings and average variance extracted (AVE). According to Karatepe (2006), factor loadings greater than 0.5 are deemed to be acceptable. Table 3 shows that the values of factor loadings ranged between 0.637 and 0.942, which implies that the requirements of conversion validity were met. In addition, AVE values were above the recommended cut-off value of 0.5 (Fraering & Minor 2006), which further confirms the adequacy of convergent validity within the scales. To check for discriminant validity, two criteria were used. Firstly, as suggested by Fornell and Larcker (1981), the AVE values for each construct should be higher than the highest shared variance (HSV) for that construct. This parameter was satisfied, as AVE values were higher than the corresponding HSV for all of the constructs. Secondly, correlations between constructs should be below a unit value of 1.0. (Fraering & Minor 2006). As indicated in Table 5, the minimum inter-construct correlation was 0.535. Therefore, discriminant validity was deemed acceptable in this study.

Upon completion of the assessment of reliability and validity, the model-fit assessment was conducted for both CFA and SEM to determine if the data were fitting both the measurement and structural models (Westland 2015). Acceptable model-fit was determined by indices that include the chi-square value over degree of freedom (c2/d.f.) of a value ≤ 3, with the values of goodness-of-fit index (GFI), comparative fit index (CFI), incremental fit index (IFI) and Tucker–Lewis index (TLI) equal or greater than 0.90, and the root mean square error of approximation (RMSEA) value to be equal to or less than 0.08. Table 4 provides the results of the model-fit analysis.

The analysis of the results in Table 4 shows that the model-fit for both CFA and SEM were acceptable as they satisfied the suggested thresholds for each indicator.

Table 5 displays the correlation between constructs.

Model-fit analysis for the structural model revealed that all model fits were within the acceptable range and are as follows: (d/f) = 2.476; (IFI) = 0.940; (TLI) = 0.915; (CFI) = 0.937; (NFI) = 0.931; and (RMSEA) = 0.052. Individual hypotheses test results are presented in Table 6. The study tested hypotheses using the path analysis procedure. The results are reported in Table 6.

The path coefficients of the proposed hypotheses (H1, H2, H3, H4 and H5) are as follows: 0.116; −0.006; 0.390; 0.721; and 0.611, respectively. Except for H1 and H2, all of the other path coefficients were significant at a confidence level of 0.001. This implies that the results from H1 and H2 do not support the proposed research hypotheses, while H3, H4 and H5 support them.

The first hypothesis (H1) of the study suggested that the implementation of GP leads to improved SME environmental performance. This hypothesis was not supported in this study because it showed statistical insignificance and an almost negligible positive relationship (β = 0.116; p = 0.168). This result implies that there is no relationship between GP and environmental performance. It contradicts previous studies (Carter & Ellram 1998; Min & Galle 2001) where positive relationships between GP and environmental performance were observed. However, the results of the study are supported by a survey conducted by Large and Thomsen (2011), which found no relationship between GP and environmental performance. The results of this study could be attributed to the inability of SMEs in South Africa to incorporate green practices into their organisational strategy and, hence, they cannot pursue the route of sustainability during purchases. Furthermore, engaging in green initiatives requires substantial capital investments that SMEs usually do not possess. This lack of financial resources may limit their affordability. The results could also be attributed to the view that the owners and managers in SMEs are yet to familiarise themselves with the benefits of implementing green practices. It could further be that SMEs do not have trained and experienced personnel committed to green initiatives and, hence, they are yet to realise the importance of these initiatives.

The second hypothesis (H2) suggested that the implementation of ED leads to improved SME environmental performance. This hypothesis was not supported in this study as there was a negative and insignificant relationship between ED and environmental performance (β = −0.006; p = 0.946). These results challenge previous studies conducted by Laosirihongthong et al. (2013) who assert that there is a positive relationship between ED and environmental performance. However, the present study finds support in previous studies (Montabon, Sroufe & Narasimhan 2007; Richey et al. 2005) where evidence of a negative relationship between ED and environmental performance was found. The results of the current study could be attributed to the view that ED is too costly to implement, making it difficult for SMEs to consider it in their operations. This is because the implementation of ED requires expensive technologies coupled with extensive research and development processes, many of which are beyond the capacity of South African SMEs. For the few SMEs that have adequate financial resources, it could be that they perceive that ED is impractical for them to adopt or they merely fail to foresee any gain in implementing it.

The third hypothesis (H3) suggested that the implementation of RL leads to improved SME environmental performance. The hypothesis was supported in this study as the results revealed a moderate, positive and significant relationship between RL and environmental performance (β = 0.390; p = 0.000). A previous study conducted by Diabat, Khodaverdi and Olfat (2013) to explore the relationship between RL and environmental performance found that RL positively impacts environmental performance. This moderate relationship observed in the current study may imply that some SMEs perceive RL to be very expensive to implement. It could also be that some SMEs do not understand precisely the role of recycling in the success of their businesses or they have just decided to discard RL practices and are more inclined to other green practices. Lau and Wang (2009), in their study, found that the embracing of RL is often negatively affected by the lack of law enforcement, high costs and low public awareness on environmental sustainability. This study suggests that these adverse factors may also apply to SMEs in South Africa.

The fourth hypothesis (H4) suggested that adherence to LR leads to higher SME environmental performance. This hypothesis was supported in this study, as justified by the strong positive and significant relationship between the two variables (β = 0.721; p = 0.000). This implies that government-imposed LR contribute to improving the environmental performance of SMEs. A study conducted by Laosirihongthong et al. (2013) supports this result as it also reached a similar conclusion. Perhaps, the respect for environmental regulations could be attributed to the fear of consequences related to non-compliance. This study suggests that to hedge themselves against reprisals by the long arm of the law, SMEs implement defensive and reactional strategies through their adherence to existing environmental laws. Adherence to these laws leads to better environmental performance. It is also noteworthy that LR scored the highest beta when compared to the other GSCM practices. This top beta score confirms the high level of importance that SMEs attach to LR regarding the pollution of the environment.

The fifth hypothesis (H5) suggested that the better the SME environmental performance, the higher the performance of the supply chain. A strong positive and significant relationship between these two constructs was observed (β = 0.611; p = 0.000) to support the hypothesis. This result corresponds with previous studies (Green, Whitten & Inman 2011; Vachon & Klassen 2007) which suggest that integrating environmental sustainability into organisational strategy positively impacts on the performance of the supply chain. This result may imply that SMEs understand the benefits of environmental performance and are willing to manipulate this positive effect to stimulate the performance of the supply chain. In the view of these results, it is confirmed that embracing environmental sustainability leads to improved SCP.

Figure 2 is the resultant structural model of the relationships between GSCM, environmental performance and supply chain performance.

The comprehensive structural model shows the extent of the relationships between the constructs related to this study as well as the factor loadings for each item in the constructs. Green purchase (GP) practices have a positive but weak and insignificant relationship with environmental performance (r = 0.12; p = 0.168). Eco-design has a negative and insignificant relationship with environmental performance (r = −0.01; p = 0.946). Reverse logistics has a positive and significant relationship with environmental performance (r = 0.39; p = 0.000). Legislation and regulation have a strong positive and significant relationship with environmental performance (r = 0.72; p = 0.000). Finally, environment performance and supply chain performance have a strong positive and significant connection (r = 0.72; p = 0.000). A glance at this model demonstrates that legislation and regulatory practices have the highest influence on environmental performance (r = 0.72).