The relationship between supply chain integration and sustainability performance: a meta-analysis

ABSTRACT

Purpose

Supply chain integration (SCI) has been considered an antecedent of sustainability performance (SP) by some studies; however, empirical findings on the nexus between SCI and SP in the existing literature are inconsistent. The study aims to provide an extensive literature review on the SCI–SP relationship.

Design/methodology/approach

We propose a conceptual research framework to investigate the SCI–SP relationship with consideration of the effects of contextual moderators. A meta-analysis is adopted to analyse the sample of 60 selected empirical papers focusing on this relationship.

Findings

The results provide evidence of a positive relationship between SCI and SP, which is affected by the two types of moderating variables: operationalisation of SP construct and control variables (industry sector, economic development, and the country's logistics capabilities). We find that SCI has the greatest impact on economic SP and the lowest impact on social SP. The SCI–SP relationship is more robust in the service industry than in the manufacturing industry. When a country's logistics capabilities increase, SCI's influence on SP becomes weaker.

Limitations/implications

The results of this study are drawn from extant empirical research, and thus the validity and reliability of these findings rely on the validity and reliability of the original studies.

Originality/value

This study is the first meta-analysis to examine the vital relationship between SCI and SP and adopted a country's logistics capabilities as a moderator. The results offer managerial insights regarding integration dimensions that positively affect various measures of SP.

KEYWORDS:

• Meta-analysis
• supply chain integration
• sustainability performance

1. Introduction

Sustainability has received much attention from industry and academia in the global economic context (Govindan et al. 2020). Pursuing economic growth is no longer the sole focus; seeking environmental and social performance has also become critical (Shen et al. 2017). Given the complexity of achieving sustainability performance (SP), both academia and industry have recognised the importance of supply chain integration (SCI) (Chaudhuri, Boer, and Taran 2018). In the extant literature, growing evidence implies that SCI can effectively increase a firm's SP in a dynamic environment (Huo, Zhao, and Lai 2014; Um and Kim 2019). Based on data from 308 Chinese manufacturing companies, Yu, Zhang, and Huo (2019) argued that cooperation between consumer and supplier have positive impacts on environmental performance; similar conclusions are also drawn by other studies (Ateş et al. 2012; Carter and Carter 1998; Du, Zhang, and Feng 2018; Han and Huo 2020). In the fashion business, Li et al. (2016) stated that supplier integration has a significantly positive impact on a sustainability performance. Internal integration positively affects sustainability performance (Pakurár et al. 2019). Across the world, the importance of SCI has been further heightened by the shock of COVID-19. During the pandemic, transportation was restricted and deliveries inconsistent. Most suppliers were unable to fulfil orders and manufacturers were not able to continue production. Despite the difficult situation, some exceptional companies exhibited great SCI ability; for example, LVMH updated its perfume factories to produce hand sanitiser and Gucci reformed its designer clothing factory to make medical supplies (Bain & Company 2020). This rapid business transformation requires a high SCI degree, including fast raw-material sourcing, product development and testing, and efficient logistics and transportation service.

Though many scholars agree that SCI has a positive impact on SP, different opinions have arisen over the years. For instance, Danese and Romano (2011), after analyzing data from 200 global manufacturing plants, argued that customer integration has no positive effect on efficiency in general. Mao, Zhang, and Li (2017) revealed that internal integration is only partially significant in terms of financial performance, while Laari et al. (2016) even argued that internal environmental collaboration affects sustainability performance insignificantly—a result also supported by other studies (Du, Zhang, and Feng 2018; Shin, Park, and Park 2019).

In addition, several meta-analyses have been conducted investigating the relationship between SCI and sustainability performance (Chang et al. 2016; Leuschner, Rogers, and Charvet 2013; Mackelprang et al. 2014). Of these, only Chang et al. (2016) considered the function of contextual variables in moderating the relationship between SCI and performance. Context refers to external environmental contingencies and internal contingencies (Danese, Molinaro, and Romano 2020). In this paper, we study the moderating effect of three contextual variables: the economic development level of the sample country, industry type, and the country's logistics capabilities, as suggested by Danese, Molinaro, and Romano (2020).

Driven by these historically mixed findings on the relationship between SCI and SP and the lack of research on the role of contextual variables, we conduct a meta-analytical study to aggregate all the related empirical research to uncover the generalised interaction between SCI and SP. Sustainability performance may contain environmental, social, and economic performance. Through a comprehensive literature review, we identified 60 articles published between 1998 and 2021 that surveyed a total of 15,332 companies. Our study differs from the most recent meta-analyses (Ataseven and Nair 2017) in two ways. First, our study focuses on the relationship between SCI and SP, rather than being confined to economic or operational aspects. Second, we carefully adopt a country's logistics capabilities as a moderator.

The remainder of the paper is structured as follows. Section 2 includes literature review, coding process, research framework development and hypotheses proposition. Section 3 focuses on the details of the research method; publication bias and heterogeneity issues are discussed and resolved. Section 4 presents the meta-analytical results, and Section 5 contains further discussion.

2. Literature review and research framework

2.1. Sampling and literature review

In this research, a meta-analysis approach is adopted to reveal the relation between SCI and SP. Meta-analysis has been prevalently conducted for analysing the quantitative findings of prior published research. Based on the comprehensive evaluation, the research results’ validity can be examined to give an overview of the existing research and propose insights on both theory and practice for future research. Our study focuses on the empirical study in which supply chain integration indicators have functioned as an independent variable, and those of supply chain SP are dependent variables. All the procedures follow the instructions and procedures of Borenstein et al. (2009) and Hunter and Schmidt (2014).

Researches published by peer-reviewed journals in English from 1987 until December 2021 are our primary focuses, in line with the fact that sustainability was initially proposed by the UN Brundtland Commission in 1987. The selected databases are ABI/INFORM Complete, Business Source Complete (EBSCO), Emerald, Scopus, and Web of Science. We link the search strings to ‘supply chain,’ ‘integration,’ ‘sustainability,’ and ‘performance,’ consistent with prior literature (Chen et al. 2017). ‘Supply chain,’ ‘supplier,’ and ‘customer’ are commonly used key search terms for the supply chain. The keywords for ‘integration’ include ‘integration,’ ‘collaboration,’ and ‘cooperation.’ For sustainability, we adopt ‘green,’ ‘sustainable*,’ ‘corporate social responsibility,’ and ‘environment*’ as search terms, consistent with prior research on sustainable supply chain management (Carter and Rogers 2008). The search terms used for performance are ‘performance,’ ‘benefit,’ ‘return,’ ‘profit*,’ ‘firm performance,’ ‘green performance,’ and ‘sustainability performance.’

The search terms above were combined to collect titles and keywords of articles from the databases. Further sifting allowed the removal of duplicate and irrelevant articles. An initial sample of 2,823 journals was obtained. Next, we identified valid samples via three selection criteria. First, the articles should use an empirical methodology in the study of SCI's effect on SP. Second, the articles should report Pearson's correlation coefficient between SCI and SP. If such coefficient was not provided, other statistics such as the F-statistic should be reported to obtain the corrected correlation coefficient via the calculation procedures shown in Appendix A. Third, the articles should have distinct data sets. If several papers used the same data set, we selected only one of them. After applying these three criteria, 60 papers remained. Finally, records of effect size were extracted from these 60 papers. The distribution of the selected papers is shown in Table 1.

Table 1. The distribution of selected papers.

Journal Name	Count	Percentage	Time Period
International Journal of Production Economics	10	17%	2009 - 2020
Supply Chain Management: An International Journal	7	12%	2013 - 2020
Decision Sciences	3	5%	2017 - 2019
Sustainability	3	5%	1998 - 2012
The International Journal of Logistics Management	3	5%	2016 - 2018
International Journal of Production Research	3	5%	2016 - 2020
Industrial Management & Data Systems	2	3%	2016 - 2020
Industrial Marketing Management	2	3%	2017 - 2019
International Journal of Operations & Production Management	2	3%	2006 - 2017
Journal of Cleaner Production	2	3%	2009 - 2013
Journal of Operations Management	2	3%	2012 - 2014
Others	19	32%	2011 - 2020
Total	60	100%	1998 - 2020

In Table 2, the theoretical lens adopted in the sampled articles is summarised. 26 papers (43%) (did not specify the applied theory; in the remainder, a variety of theories are evident, including the resource-based view (RBV), the natural resource-based view (NRBV), resource dependence theory (RDT), the knowledge-based view (KBV), and social capital theory, with the RBV (17%) and the NRBV (8%) the most widely applied. Table 2 also summarises the analysis methods of the sampled articles. Most of the reviewed papers adopted either structural equation modelling (SEM) (60%) or regression analysis (32%) for data analysis.

Table 2. Distribution of theoretical lens and analysis methods.

Theoretical lens	Count	Percentage
Not specified	26	43%
RBV	10	17%
NRBV	5	8%
Organizational capability theory	3	5%
Social capital theory	2	3%
Knowledge-based theory	2	3%
Contingency theory	2	3%
Trust theory	1	2%
Social capital theory & Resource dependence theory	1	2%
Coordination theory	1	2%
Market-oriented environmental sustainability theory	1	2%
Social exchange theory	1	2%
Institutional theory	1	2%
Resouce dependence theory	1	2%
Relational view	1	2%
Media richness theory	1	2%
RBV & Dynamic capabilities view	1	2%
Total	60	100%
Analysis Methods	Count	Percentage
SEM	36	60%
Regression analysis	19	32%
PCA	2	3%
CFA	2	3%
ANOVA	1	2%
Total	60	100%

2.2. Coding

2.2.1. Coding process

Prior research includes a variety of constructs for SCI and SP. To reach an agreement, two authors independently collected essential information (e.g. effect size, sample size) from the selected papers. If there was any disagreement, the two authors reviewed the paper and had an in-depth discussion until they reached an agreement. Otherwise, a third expert was invited to give their opinion on the disputed point. To ensure that the concepts had been accepted and applied in prior research, this research followed a strict coding process (Wang, Lai, and Shou 2018; Govindan et al. 2020). The inter-coder reliability was 100%, implying our coding results, as shown in Table 3, did not lose consistency or continuity.

Table 3. Coding results.

No.	Study	Data	Analysis method	Theoretical lens	Country	Development	LPI	LPI Level	Industry	Integration type	Sustainability performance	Sample size	Effect size
1	Abdallah, Abdallah, and Saleh (2017)	Survey	SEM	Relational view	Jordan	Developing	2.69	Below average	Service	Supplier	Economic	152	0.562
2	Afum et al. (2020)	Survey	SEM	Not specified	Ghana	Developing	2.46	Below average	Manufacturing	Supplier	Environmental	178	0.356
3	Agyabeng-Mensah et al. (2020)	Survey	SEM	Not specified	Ghana	Developing	2.46	Below average	Manufacturing	Internal	Economic	139	0.184
4	Ahmed et al. (2020)	Survey	SEM	Not specified	Pakistan	Developing	2.92	Below average	Manufacturing	Internal	Environmental	126	0.312
5	Alfalla-Luque, Marin-Garcia, and Medina-Lopez (2015)	Survey	SEM	RBV	International	Mixed	2.89	Average	Manufacturing	Customer	Social	266	0.700
6	Ateş et al. (2012)	Survey	SEM	RBV	Turkey	Developing	3.51	High	Manufacturing	Internal	Environmental	96	0.690
7	Carter and Carter (1998)	Survey	SEM	Organizational theory	United States	Developed	3.84	High	Manufacturing	Internal	Environmental	437	0.452
8	Chavez et al. (2015)	Survey	Regression analysis	Social capital theory	Ireland	Developed	3.79	High	Manufacturing	Customer	Economic	228	0.263
9	Cheng, Chaudhuri, and Farooq (2016)	Survey	SEM	Not specified	International	Mixed	2.89	Average	Manufacturing	Internal	Economic	606	0.260
10	Cho et al. (2019)	Survey	SEM	Not specified	United States	Developed	3.89	High	Service	Supplier	Social	795	0.376
11	Danese and Romano (2011)	Archival Data	Regression analysis	Not specified	International	Mixed	2.87	Average	Mixed	Customer	Economic	200	0.244
12	Danese (2013)	Survey	Regression analysis	Not specified	International	Developed	2.89	Average	Manufacturing	Supplier	Economic	186	0.316
13	Delic, Eyers, and Mikulic (2019)	Survey	SEM	RBV	Europe	Developed	3.40	Average	Manufacturing	Customer	Social	124	0.524
14	Droge, Jayaram, and Vickery (2004)	Survey	Principal component analysis	Not specified	North America	Developed	3.88	High	Manufacturing	Supplier	Social	57	0.294
15	Du, Zhang, and Feng (2018)	Survey	Regression analysis	Resource dependence theory	China	Developing	3.61	High	Manufacturing	Customer	Environmental	176	0.260
16	Eltantawy, Giunipero, and Fox (2009)	Survey	SEM	RBV	United States	Developed	3.86	High	Mixed	Supplier	Economic	152	0.720
17	Errassafi, Abbar, and Benabbou (2019)	Survey	SEM	Not specified	Morocco	Developing	2.54	Below average	Manufacturing	Internal	Economic	75	0.791
18	Feng et al. (2017)	Survey	Regression analysis	RBV & DCV	China	Developing	3.61	High	Manufacturing	Customer	Economic	126	0.440
19	Gimenez, Sierra, and Rodon (2012)	Survey	Regression analysis	NRBV	International	Mixed	2.87	Average	Manufacturing	Supplier	Social	519	0.235
20	Graham and Potter (2015)	Survey	Regression analysis	NRBV	United Kingdom	Developed	4.07	High	Manufacturing	Customer	Economic	149	0.200
21	Han and Huo (2020)	Survey	SEM	NRBV	China	Developing	3.61	High	Manufacturing	Supplier	Environmental	206	0.249
22	Hassan, Abindin, and Nordin (2018)	Survey	SEM	Not specified	Malaysia	Developing	3.22	Average	Manufacturing	Customer	Environmental	107	0.501
23	He et al. (2014)	Archival Data	SEM	Trust theory	International	Mixed	2.89	Average	Manufacturing	Customer	Social	726	0.230
24	Huo et al. (2013)	Survey	SEM	Institutional theory	China	Developing	3.53	High	Manufacturing	Supplier	Economic	617	0.217
25	Jayaram and Xu (2013)	Survey	SEM	Coordination theory	China	Developing	3.53	High	Manufacturing	Supplier	Social	197	0.258
26	Kim (2009)	Survey	SEM	RBV	Japan	Developed	3.97	High	Manufacturing	Supplier	Economic	137	0.338
27	Lai, Wong, and Venus Lun (2014)	Survey	SEM	Not specified	China	Developing	3.53	High	Manufacturing	Customer	Economic	134	0.190
28	Laari et al. (2016)	Survey	Regression analysis	NRBV	Finland	Developed	3.92	High	Service	Customer	Economic	311	0.126
29	Li et al. (2016)	Survey	Regression analysis	RBV	China	Developing	3.66	High	Manufacturing	Supplier	Economic	90	0.263
30	Li, Zhou, and Wu (2017)	Survey	SEM	Market-oriented environmental sustainability theory (MES)	China	Developing	3.61	High	Manufacturing	Customer	Economic	305	0.180
31	Liu and Lee (2018)	Survey	SEM	RBV	Taiwan	Developed	3.60	High	Service	Customer	Economic	161	0.429
32	Liu et al. (2018)	Survey	SEM	RBV	China	Developing	3.61	High	Manufacturing	Internal	Environmental	102	0.498
33	Longoni, Golini, and Cagliano (2014)	Survey	Regression analysis	Not specified	International	Mixed	2.89	Average	Manufacturing	Internal	Social	377	0.220
34	Mandal (2017)	Survey	SEM	Not specified	India	Developing	3.18	Average	Service	Supplier	Social	192	0.386
35	Mao, Zhang, and Li (2017)	Survey	SEM	NRBV	China	Developing	3.66	High	Manufacturing	Internal	Environmental	115	0.384
36	Munir et al. (2020)	Survey	SEM	Not specified	International	Mixed	2.89	Average	Manufacturing	Internal	Social	931	0.345
37	Pakurár et al. (2019)	Survey	Exploratory factor analysis	Organizational capability theory	Jordan	Developing	2.69	Below average	Service	Internal	Economic	249	0.823
38	Pan et al. (2020)	Survey	Regression analysis	Not specified	China	Developing	3.66	High	Manufacturing	Supplier	Environmental	162	0.193
39	Panahifar et al. (2018)	Survey	SEM	Not specified	Thailand	Developing	3.41	Average	Mixed	Internal	Economic	189	0.794
40	Rajaguru and Matanda (2019)	Survey	SEM	Not specified	Australia	Developed	3.75	High	Service	Supplier	Economic	302	0.473
41	Ralston et al. (2015)	Survey & Archival Data	Confirmatory factor analysis	Not specified	United States	Developed	3.99	High	Manufacturing	Customer	Social	220	0.250
42	Rehman et al. (2021)	Survey	SEM	Not specified	Malaysia	Developing	3.22	Average	Manufacturing	Internal	Environmental	354	0.385
43	Schoenherr and Swink (2012)	Archival Data	ANOVA	RBV	International	Mixed	2.87	Average	Manufacturing	Internal	Economic	403	0.138
44	Shin, Park, and Park (2019)	Survey	SEM	Social capital theory & RDT	South Korea	Developed	3.61	High	Mixed	Supplier	Economic	423	0.373
45	Singh and Power (2014)	Survey	SEM	KBV	Australia	Developed	3.81	High	Manufacturing	Internal	Economic	418	0.509
46	Song, Feng, and Jiang (2017)	Survey	Regression analysis	Not specified	China	Developing	3.61	High	Manufacturing	Customer	Economic	58	0.445
47	Sundram, Chandran, and Awais Bhatti (2016)	Survey	Regression analysis	RBV	Malaysia	Developing	3.43	Average	Manufacturing	Supplier	Economic	156	0.872
48	Swink, Narasimhan, and Kim (2005)	Survey	Regression analysis	Not specified	North America	Developed	3.88	High	Manufacturing	Supplier	Social	57	0.280
49	Terjesen, Patel, and Sanders (2012)	Survey	Regression analysis	Contingency theory	United States	Developed	3.93	High	Manufacturing	Internal	Economic	261	0.196
50	Thai and Jie (2018)	Survey	Regression analysis	Supply chain management	Singapore	Developed	4.00	High	Service	Internal	Economic	159	0.516
51	Thomas (2013)	Survey	SEM	Media richness theory	United States	Developed	3.92	High	Manufacturing	Supplier	Social	157	0.138
52	Vereecke and Muylle (2006)	Archival Data	Principal component analysis	Not specified	Europe	Developed	3.26	Average	Manufacturing	Customer	Economic	374	0.176
53	Wiengarten et al. (2019)	Archival Data	Regression analysis	Not specified	International	Mixed	2.87	Average	Manufacturing	Supplier	Economic	293	0.213
54	Willis, Genchev, and Chen (2016)	Survey	SEM	KBV	India	Developing	3.42	Average	Manufacturing	Supplier	Social	92	0.431
55	Wong, Boon-Itt, and Wong (2011a)	Survey	Regression analysis	Contingency theory	Hong Kong	Developed	4.12	High	Service	Customer	Economic	196	0.206
56	Wong et al. (2020)	Survey	SEM	Organizational capability theory	Hong Kong	Developed	4.12	High	Mixed	Customer	Economic	192	0.204
57	Yu, Zhang, and Huo (2019)	Survey	SEM	Not specified	China	Developing	3.61	High	Manufacturing	Customer	Environmental	308	0.347
58	Yu, Zhang, and Huo (2019)	Survey	SEM	Not specified	China	Developing	3.61	High	Manufacturing	Supplier	Economic	296	0.323
59	Yuen and Thai (2017)	Survey	SEM	Not specified	Singapore	Developed	4.00	High	Mixed	Internal	Economic	312	0.938
60	Zhang et al. (2020)	Survey	Regression analysis	Social capital theory	China	Developing	3.61	High	Manufacturing	Supplier	Environmental	206	0.381

2.2.2. Independent variable

In this research, SCI was chosen as our independent variable. In prior research, numerous definitions of SCI have been given. Some define SCI as seamless interaction built with suppliers, manufacturers, and customers for strategic alignment and global rather than local optimisation (Devaraj, Krajewski, and Wei 2007), which implies that SCI is a strategically external integrated process or practice among supply chain parties. In this study, SCI is defined as ‘the degree to which a manufacturer strategically collaborates with its supply chain partners and collaboratively manages intra – and inter-organisation processes’ (Flynn, Huo, and Zhao 2010, 58). Following Ataseven and Nair (2017), we characterise SCI into three types: 1. Customer integration, 2. supplier integration, and 3. internal integration, defined as follows:

1. Customer integration (CI) refers to the ‘close collaboration and information sharing activities with key customers that provide the firm with strategic insights into market expectations and opportunities, ultimately enabling a more efficient and effective response to customer needs’ (Schoenherr and Swink 2012). Famiyeh et al. (2018) pointed out that manufacturing companies have begun implementing supply chain management in response to customer demand for environmentally sustainable products and services, with the potential to improve business performance. CI helps a firm arrange its internal operations in close alignment with market demands (Ataseven and Nair 2017). The coordination of inventory control, demand forecasting, and delivery tracking between a firm and its customers is the sustainable example of CI (Wong, Lai, and Cheng 2011b).

2. Supplier integration (SI) represents the ‘coordination and information sharing activities with key suppliers that provide the firm with insights into suppliers’ processes, capabilities and constraints, ultimately enabling more effective planning and forecasting, product and process design, and transaction management’ (Schoenherr and Swink 2012). The greenness of manufacturing product depends directly on raw materials provided by suppliers (Pan et al. 2020). At the same time, on the basis of cooperation with suppliers, manufacturing enterprises promote the application of green technology innovation and improve the sustainable environmental performance of enterprises (Pan et al. 2020). Critical SI practices include joint product development, joint inventory management and quality control, and the long-term sustainable buyer–supplier relationship (Li et al. 2016).

3. Internal integration (II) is the ‘degree to which a firm can structure its organisational practices, procedures and behaviours into collaborative, synchronised and manageable processes in order to fulfil customer requirements’ (Zhao et al. 2011). II provides a platform for sharing information within internal functions and provides capabilities for processing, absorbing, analyzing, and implementing information gathered from supplier and customer integration (Flynn, Huo, and Zhao 2010). II enables better intra-firm alignment and utilisation of each department's strengths and competencies to pursue excellence (Schoenherr and Swink 2012), and includes data and information sharing, strategic collaboration for logistics, continuous process improvement and joint new product development (Feng et al. 2017; Wong, Lai, and Cheng 2011b).

Several studies on the relationship between SCI and SP have identified the impact of CI, SI, and II. For example, Cheng, Chaudhuri, and Farooq (2016) argued that II is not significantly related to performance, but that the relationship is mediated by external integration. Lai, Wong, and Venus Lun (2014) revealed that higher CI would help a company achieve better market performance. In our sample, 18 of 60 records focus on internal SCI practices, while the remainder are related to CI (19 papers) and SI (23 papers).

2.2.3. Dependent variable

In this meta-analysis, SP is the dependent variable. According to the definition of sustainability, essential aspects of SP include the expectations of the environment, society, and stakeholders in the long term. Elkington (1998) defined SP in terms of the ‘triple bottom line’ (TBL) of economic, environmental, and social performance; this was also empirically applied in Shen’s (2014) work. The editorial by Bentahar and Benzidia (2018) and the literature review conducted by Chen et al. (2017) also suggested measuring SP in terms of these three aspects (economic, environmental, and social):

1. Economic performance refers to profitability, efficiency, and productivity in general and serves as the most critical reason for integration practices in the supply chain (e.g. Abdallah, Abdallah, and Saleh 2017; Chavez et al. 2015; Cheng, Chaudhuri, and Farooq 2016; Kim 2009). It is possible for firms to pursue new economic objectives when they have strong connections, which generates positive externalities, which are conducive to improving economic performance (Lee and Kim 2011). The closer the connection between the enterprise and the supplier, the greater the likelihood of cooperation between the two parties. Measurement of economic performance in the reviewed literature includes growth in sales, profits, market share, and return on investment (Flynn, Huo, and Zhao 2010).

   In terms of environmental performance, companies limit and maintain control over environmental cooperation activities with suppliers through the use of contracts, risk sharing arrangements, and management systems to reduce environment hazards (Hassan, Abindin, and Nordin 2018). At the same time, these companies and suppliers have developed consistent cultures and goals through the exchange and sharing of environmental knowledge (Zhang et al. 2020), promoting a willingness on both sides to share business in partnerships to better understand shared values (Du, Zhang, and Feng 2018; Yu, Zhang, and Huo 2019).

2. Social performance is related to the evaluation of an organisation's effects on employees, the local community, and the whole society in terms of creating social welfare, including ethical programmes, employee education and training, health and safety, and human resource management (e.g. Alfalla-Luque, Marin-Garcia, and Medina-Lopez 2015; Delic, Eyers, and Mikulic 2019; Gimenez, Sierra, and Rodon 2012; Longoni, Golini, and Cagliano 2014; Mandal 2017). Businesses can improve their social relationships by integrating various green practices to strengthen collaboration (Longoni, Golini, and Cagliano 2014). With frequent interactions, mutual trust and interdependence increase, social networks gradually take shape, facilitating access to more valuable resources for both parties (Delic, Eyers, and Mikulic 2019; Gimenez, Sierra, and Rodon 2012).

2.2.4. Control variables

Control variables in a meta-analysis are those variables that the primary relationship is conditional on; these can change the direction or magnitude of the main effect (Wang, Lai, and Shou 2018). Consistent with prior research (Graham and Potter 2015; He et al. 2014; Liu et al. 2018), we extracted industry type (manufacturing or service) and the country's economic development level (developed or developing) as control variables in our meta-analysis. For instance, Liu et al. (2018) found a difference between China and Western firms in the automobile industry, such that, in the Chinese context, the relationship between internal SI and successful implementation of an environmental strategy is significant, but not in Europe. Schoenherr and Swink (2012) and Singh and Power (2014) found that II can contribute to SP in the manufacturing industry. However, Mao, Zhang, and Li (2017) and Laari et al. (2016) failed to prove the significance of the impact of II on SP in the same industry. In the selected articles, the service industry (9 articles), including exporting, hotels, hospitals, logistics, and retailing, received less attention than manufacturing sectors (45 articles). We identified three types of industry: manufacturing, service, and mixed. When multiple industries are investigated, the research is recognised as ‘mixed,’ such as Panahifar et al. (2018) and Yuen and Thai (2017). A country's economic development level is coded as developed, developing or both; ‘mixed’ refers to international surveys in both developing and developed regions. To clarify economic region, we use criterion from World Bank high-income economies to separate the regions into the group of developed or developing.

In addition to economic development level and industry type, we also introduced a country's logistics capabilities as a control variable. In general, efficient and effective logistics operation, as a dynamic capability, is of great importance to achieve competitive advantage (Sandberg and Abrahamsson 2011). However, based on the observation by Danese, Molinaro, and Romano (2020), few articles have investigated the moderating effect of such capabilities on the relationship between SCI and SP, especially from the perspective of country. Thus, thus study also included country's logistics capabilities into control variables. The aggregated logistics performance index (LPI) from the World Bank was adapted to measure each country's logistics capabilities. LPI is also used in Wiengarten et al. (2014) to explore logistics capabilities in external SCI. In terms of aggregated LPI ranking, Germany has the highest score (4.19 on a scale of 1–5) among the 167 countries, while Somalia has the lowest (2.0). To see each country's logistics capabilities intuitively, we define this capability as high if the score is more than 3.5, average if it is less than 3.5 but larger than 3.0, and below average if less than 3.0.

2.3. Research framework and hypotheses

2.3.1. The SCI–SP relationship

RDT is adopted as the theoretical lens to develop the research hypotheses in our study. RDT argues that external and internal resources are key to companies to gaining and sustaining competitive advantages (Esfahbodi, Zhang, and Watson 2016). RDT helps to understand how and to what extent SCI creates mutual benefits among players in the chain (Shin, Park, and Park 2019). Jointly possessed or controlled resources among supply chain partners are often viewed as the focal firm's source of capability development. In a collaborative context, suppliers and customers share mutual goals with the focal firm, which creates a collaborative atmosphere for knowledge and information sharing. With joint planning and decision-making, the focal firm can align with partners, adopt unified standards, share advanced technologies, and obtain the newly added competitive advantages (Vachon and Klassen 2008). Within the firm, integrating multiple departments (e.g. logistics, production, marketing, and R&D) can attain strategic alliance of product/service design (Braunscheidel and Suresh 2009) and maintain and enhance competitive capability (Kim 2006), which has been proved to positively affect sustainability performance (Flynn, Huo, and Zhao 2010; Liu and Lee 2018; Zhao et al. 2011). As demonstrated by empirical tests, cross-functional coordination and supplier involvement may also contribute to eco-innovation of new product design and environmental logistics, subsequently reflected in improved profits, efficiency, competitiveness, market share, and environmental brand image; namely, excellent economic, environmental, and social performance (Mitra and Datta 2014). Therefore, we propose the following hypothesis:

H1: SCI is positively associated with SP.

2.3.2. Constructs of SCI

In our study, SCI is categorised into the CI, SI, and II. It has been a research area of active interest to study supplier and customer relationships as well as internal integration and sustainability performance in recent years and a large amount of literature emphasises the importance of the SCI concept (Flynn, Huo, and Zhao 2010). There is evidence that customer collaboration may be more effective than supplier collaboration in improving environmental performance (Ralston et al. 2015). Three SCI dimensions (CI, SI, and II) have been shown to produce different environmental, cost, and economic benefits (Chavez et al. 2015; Cheng, Chaudhuri, and Farooq 2016; Kim 2009). Meanwhile, the association between SCI and sustainability tends to differ depending on whether internal or external integration is being considered (Braunscheidel and Suresh 2009). For instance, Zhao et al. (2011) found that II is positively related to business and operational performance, while CI is only directly associated with operational performance, and SI is not a direct contributor to either. Cheng, Chaudhuri, and Farooq (2016) argued that external integration significantly affects operational performance of a networked plant, whereas II does not. Mao, Zhang, and Li (2017) argued that internal environmental integration improves environmental performance but hinders a firm's financial performance, whereas external green integration improves both environmental performance and financial performance. Consequently, both internal and external integration may be beneficial for SP, but RDT theory implies that external (customer or supplier) relationships lead to tight interconnection with other partners, which also reflect the superior status of the cross-functional relationship. Therefore, we propose the following hypotheses:

H2a: Each type of integration—customer, supplier, and internal—is positively associated with SP.

H2b: The effect of external (customer or supplier) integration is stronger than that of II on SP.

2.3.3. Operationalisation of SP

SP is a multidimensional construct and has multiple measurements. Most studies focus on SCI's effect on economic performance by measuring operational performance, cost performance, quality performance, business performance, and market performance (Flynn, Huo, and Zhao 2010; Thai and Jie 2018; Wong, Lai, and Cheng 2011b). Some have emphasised SCI's green performance (Ateş et al. 2012; Liu et al. 2018; Yu, Zhang, and Huo 2019), but seldom measure social performance in the relationship between SCI and SP (Gimenez, Sierra, and Rodon 2012; Hassan, Abindin, and Nordin 2018). Gimenez, Sierra, and Rodon (2012) argued that implementing collaborative practices with supply chain partners has a statistically significant impact on the TBL. However, internal social initiatives contribute to better social and environmental performance, but not to economic benefits regarding unit manufacturing cost. However, some findings suggest that enhanced environmental performance can increase economic returns (Liu et al. 2018). Hence, measurements of SP lead to differing conclusions on the SCI–SP relationship. Therefore, we propose the following:

H3: The SCI–SP relationship varies with the measurements of SP.

2.3.4. Control variables

In the reviewed literature, control variables are added to ensure reliability and help explore the main relationship between key variables. Our sampled research collected data from different industrial backgrounds across various economic regions. The influence of control variables in terms of industry type and the country's economic development level and logistics capabilities conceivably affect SCI implementation and performance (Devaraj, Krajewski, and Wei 2007; Kim 2006; Mao, Zhang, and Li 2017; Yu, Zhang, and Huo 2019). Hence, we consider industry type, economic development level, and a country's logistics capabilities as contextual factors in the relationship between SCI and SP.

Firm activities are strongly related to industry type, and different industries are subject to different market or regulatory pressures regarding sustainability issues. Because of the different regulations and standards, firms may conduct various proactive or reactive SCI schemes, leading to different sustainability outcomes (Mao, Zhang, and Li 2017). Except for institutional pressure from authorities and other stakeholders, industry-related characteristics, such as product and market complexity (Wong, Lai, and Bernroider 2015), foreign investment (Bardy, Drew, and Kennedy 2012), and pollution level (Golgeci et al., 2018), also affect firms’ collaboration practices and, consequently, performance. Most extant studies on SCI and SP are based on manufacturing industries (Errassafi, Abbar, and Benabbou 2019; He et al. 2014; Liu and Lee 2018); much less consideration has been given to service sectors, such as logistics (Laari et al. 2016; Liu and Lee 2018; Yuen and Thai 2017) and hospitals (Abdallah, Abdallah, and Saleh 2017; Cho et al. 2019; Mandal 2017). In light of the diversity in the industries in our sample and the different characteristics from the cross-industry perspective, we have:

H4: The SCI–SP relationship varies with industry type.

It has been noted that developed and developing countries vary in terms of culture and education, infrastructure, and institutions (Wiengarten et al. 2014), which formulate the contextual environment for SCI practices. In developed countries, firms tend to have abundant available resources and knowledge for cross-organisational alliances, such as ICT. Interpersonal relationships and social connections are crucial for business collaboration in developing countries, such as China (Yang et al. 2010). These differences between economic regions also exert an impact on SCI implementation and therefore SP. Cho et al. (2019) supported a positive effect of inter-firm integration on sustainability in developed countries (the US). Errassafi, Abbar, and Benabbou (2019) argued that cross-firm supply chain collaboration could not benefit a firm's operational performance in Morocco, while Liu et al. (2018) proved that both internal and external integration practices in China indirectly contribute to environmental and economic outcomes. Consequently, we propose the following:

H5: The SCI–SP relationship varies with economic development level.

As suggested by Danese, Molinaro, and Romano (2020), a country's logistics capabilities is a good contextual factor to consider when investigating the relationship between SCI and sustainability performance. Shen, Xu, and Guo (2019) also stated the importance of logistics capabilities for short life cycle products in the global supply chain. Unfortunately, few studies have investigated the moderating effect of a country's logistics capabilities. One exception is Wiengarten et al. (2014), who showed that plants located in areas with a high level of logistics capabilities adopt a significantly lower level of external SCI. To contribute to this literature, we investigate the moderating effect of a country's logistics capabilities on the SCI–SP relationship, and propose the following:

H6: The SCI–SI relationship is stronger when the level of a country's logistics capabilities is lower.

3. Research method

3.1. Meta-analysis process

In this study, the meta-analytic approach is adopted to analyse the effect of SCI practice on SP. As mentioned in Hunter and Schmidt (2014), meta-analysis provides a comprehensive and generalised conclusion of extant empirical studies considering the differentiation across individual research projects. With limited available resources and accessible sample size, each research project can only focus on segmented fields, and fails to generalise the findings (Rosenbusch, Brinckmann, and Bausch 2011). Meta-analysis can calculate the population effect size, based on the correlation coefficient reported in each sampled empirical study. In our case, if the correlation was not given, we collected the values of Student's $t$, $F$-ratios, Chi-square, and Cohen's $d$, which can be converted into effect size via the formula given by Hunter and Schmidt (2014). The population effect size captures the pooled sampled studies’ mean effect size and reflects the generalised correlation between independent variables and dependent variables.

We first obtained the sampled studies via the procedures discussed in Section 2 and present the coding results in Table 3. Comprehensive Meta-Analysis (CMA) version 3.0 was used to obtain the aggregated effect size and conduct the sample tests for publication bias and heterogeneity based on the coding results.

3.2. Publication bias

In our study, confidence intervals are chosen since they are used in a frequentist confidence interval prior to data collection. Credibility intervals in a Bayesian credible interval are used after data is collected (Hunter and Schmidt 2014). Furthermore, z-scores were calculated in order to determine whether the effect sizes within the different groups were statistically significant.

Publication bias occurs in the academic publishing arena as a result of supporting rather than rejecting hypotheses contributing to the probability of acceptance for publication; this tendency implies that the resulting published studies then fail to truly represent the research field's overall characteristics (Borenstein et al. 2009). To examine the possibility of publication bias, we estimated the sampled studies before meta-analysis by the fail-safe $N$ (Borenstein et al. 2009). The fail-safe $N$ ($N_{\mathrm{fs}}$) is calculated to identify the required number of studies with zero-effect size to achieve an insignificant $p$-value; it can be obtained via the formula:$N_{\mathrm{fs}}={\displaystyle \frac{N\left(d-d_c\right)}{d_c}}$ (Orwin 1983), where N is the number of selected studies, $d$ is the average effect size, and $d_c$ the criterion value. To maintain consistency with Wang, Lai, and Shou (2018), we adopted a threshold of $N_{\mathrm{fs}}$ at the 95% confidence level: $N_{\mathrm{fs}}>5\times \mathrm{sample}\mathrm{size}+10$, where $N_{\mathrm{fs}}$ should be larger than the threshold. In this research, the fail-safe $N$ is 1,5532 $\left(P<0.001,\alpha =0.05\right)$, which significantly exceeds 310 ($5\times 60+10$). The results of the fail-safe N imply that publication bias is under control in our sample.

3.3. Heterogeneity

Heterogeneity represents the significance of difference among sampled studies in our research. We conduct heterogeneity analysis with the two most frequently used methods—the $Q$-test and the $I^2$index—to statistically detect the ‘significance and proportion of observed variance in effect size attributed to systematic cross-samples variability’ (Fang and Zhang 2018, 1073). As claimed by Huedo-Medina et al. (2006), the $Q$-test can check the existence of heterogeneity, while the $I^2$ index can point out directly the degree of heterogeneity. If the $Q$-value is significant, heterogeneity in the sample can affect the results. According to Higgins and Thompson (2002), the $I^2$ index can be classified into three levels—25%, 50%, and 75%—reflecting a low, moderate, and high level of heterogeneity. According to Borenstein et al. (2009), the fixed-effect model should be applied on identifying studies with similar conditions. If high heterogeneity is detected, the random-effect model should be applied (Borenstein et al. 2009).

Table 4 presents the results of the heterogeneity analysis. The $Q$-value 1,327.64 is significant at the confidence level of 0.05, which implies that the results of the meta-analysis would be affected by the sampled studies’ heterogeneity. Further, the corresponding $I^2$ value is 94.26%, which quantitively confirms the heterogeneity in our sample. Hence, a random-effect model is applied in this meta-analysis, considering the significant systematic difference between the sample articles.

Table 4. The overall meta-analysis results.

				95% Confidence Interval
Random-effect model	n	K	r	Lower bound	Upper bound	Z	P	Q	I^2 (%)	SE
Overall effect
Supply chain integration	15332	60	0.44	0.32	0.50	8.99	0.00	1327.64	94.26	0.03
Differential effect
Customer	4361	19	0.32	0.24	0.42	6.98	0.00	128.37	92.45	0.01
Internal	5349	18	0.54	0.39	0.75	4.82	0.00	693.54	95.35	0.10
Supplier	5622	23	0.34	0.32	0.49	7.21	0.00	246.57	94.32	0.02

*Notes: n is the total sample size; K is the total studies; r is the effect size; Z and P are statistics to check the significance; Q and I² are values to determine the heterogeneity; SE is the standard error.

4. Results

4.1. The SP effect of SCI

The meta-analysis results on the correlation between SCI and SP are shown in Table 4. We evaluated the correlation effect using the guidelines given by Cohen et al. (2003). A weak correlation is obtained with an effect size less than 0.10, while the moderate correlation is 0.10–0.30, and a strong correlation is higher than 0.30.

Our meta-analysis results show that the overall SCI–SP relationship is positive, strong (r = 0.44), and significant (Z = 8.99, P < 0.001) at the 95% confidence level. The corresponding confidence interval of (0.32, 0.50) does not include 0, which supports SCI's overall positive effect on firm SP; therefore, hypothesis H1 is accepted.

The effects of customer (r = 0.32, P < 0.001), supplier (r = 0.34, P < 0.001) and internal (r = 0.54, P < 0.001) integration are all positive and significant. All three confidence intervals exclude 0, which confirms the positive impact of CI, SI, and II on SP; therefore, H2a is supported. However, our results show that II has a stronger impact on SP than either external integration component (customer or supplier); hence, H2b is rejected.

4.2. Moderating effects

4.2.1. Operationalisation of SP

In our study, moderating effects are investigated via three control variables; namely, the measurement of SP, industry type, economic development level, and the country's logistics capabilities, measured by the LPI. The meta-analysis results for the control variables are presented in Table 5. SP is measured in terms of three categories (economic, environmental, and social) with three effect sizes (r = 0.54; r = 0.45; r = 0.36) respectively. The results are significant at the 95% confidence level (Z = 7.25, P < 0.001; Z = 8.13, P < 0.001; Z = 7.56, P < 0.001). SCI has strong effects on all three aspects of SP, among which the effect on economic performance is the strongest, followed by environmental and social performance. Therefore, hypothesis H3 is supported; the SCI–SP relationship does vary with the measurement of SP.

Table 5. Meta-analytic results of the moderating results.

				95% Confidence Interval
Random-effect model	Factors	Subsamples	n	K	r	Lower bound	Upper bound	Z	P	Q	I^2 (%)	SE
Operationaliaation	SP Measurements	Economic	7443	33	0.535	0.325	0.587	7.245	0.000	955.535	98.452	0.056
		Environmental	2573	13	0.453	0.233	0.527	8.134	0.000	54.256	72.466	0.012
		Social	5316	14	0.364	0.268	0.464	7.563	0.000	153.626	92.562	0.017
Moderators	Industry	Mixed	1468	6	0.725	0.365	0.903	4.256	0.003	463.635	98.356	0.310
		Manufacturing	11347	45	0.356	0.246	0.465	8.596	0.000	462.635	93.562	0.014
		Service	2517	9	0.456	0.256	0.646	4.683	0.000	189.385	95.485	0.046
	Region's development	Mixed	4321	9	0.287	0.189	0.484	4.853	0.000	193.536	94.362	0.020
		Developed	5808	23	0.443	0.241	0.544	6.252	0.000	837.456	97.205	0.047
		Developing	5203	28	0.489	0.395	0.595	7.284	0.000	459.593	93.648	0.045
	LPI Level	Average	6191	18	0.483	0.224	0.583	6.945	0.000	429.463	95.385	0.038
		Below average	919	6	0.645	0.464	0.824	5.395	0.000	34.625	96.986	0.101
		High	8222	36	0.385	0.274	0.489	8.385	0.000	835.395	96.492	0.030

*Notes: n is the total sample size; K is the total studies; r is the effect size; Z and P are statistics to check the significance; Q and I² are values to determine the heterogeneity; SE is the standard error.

4.2.2. Control variables

Three subgroups in terms of industry type were examined to provide evidence for hypothesis H4. Results for the manufacturing sector (r = 0.36, P < 0.001), service sector (r = 0.46, P < 0.001), and mixed industries (r = 0.73, P = 0.003) show that industry type positively moderates the SCI–SP relationship. In investigations of the manufacturing sector, SCI's effect on SP is comparatively weaker than for the service or mixed industry. The strongest correlation is captured in the mixed industry, followed by the service industry and then the manufacturing industry. These statistical results confirm that the relationship between SCI and SP varies by industry; therefore, H4 is supported.

In addition, in the three subgroups of economic region (developing, developed, and mixed), the statistics show that the effects of SCI on SP are positive and significant (r = 0.5, P < 0.001; r = 0.44, P < 0.001; r = 0.29, P < 0.001) at the 95% confidence level. Based on these statistics, the SCI–SP relationship is the strongest in developing countries (r = 0.50) than in developed countries (r = 0.44), but the difference is not large. In mixed regions (developed and developing), the effect of SCI on SP is rather modest (r = 0.29). It can be said that the SCI–SP relationship varies with the country's economic development levels; hence, H5 is accepted.

Last, we present the moderating effect of the country's logistics capabilities on the SCI–SP relationship. As discussed in Section 2.2, logistics capabilities are categorised as high, average, and below average. In Table 5, we can see that the SCI–SP relationship is strongest in countries with logistics capabilities below average (r = 0.65, P < 0.001), less strong in countries with average logistics capabilities (r = 0.48, P < 0.001), and weakest in countries with a high level of logistics capabilities (r = 0.39, P < 0.001). In general, the SCI–SP relationship is stronger when a country's logistics capabilities are lower; therefore, H6 is supported.

5. Discussion

5.1 Theoretical contributions

Reviewing the prior research, findings on the effect of SCI on SP are inconsistent and even contradictory. This meta-analytical study selects all the extant empirical findings to provide a synthesis, integration, and interpretation of SCI and SP's nexus. Besides, the moderating effects of operationalisation and control variables on the relationship have also been considered in our research.

First, the meta-analysis confirms the significant and positive relationship between SCI and SP, consistent with some previous conclusions (Ataseven and Nair 2017; Chen et al. 2017; Geng, Mansouri, and Aktas 2017). Our results refute several studies’ conclusions (Laari et al. 2016; Mao, Zhang, and Li 2017; Willis, Genchev, and Chen 2016). Willis, Genchev, and Chen (2016) analyzed survey data from 92 Indian companies and found that II has an insignificant effect on firms’ flexibility performance, which can be classified as economic performance (Killick 1995). Mao, Zhang, and Li (2017) proposed that the effects of internal low-carbon integration is only partially significant for economic performance. Their findings are refuted by our results, which confirm the differential effect of SCI's three constructs. Both internal and external (customer or supplier) integration are positively and significantly associated with SP. We even found that II has a more substantial average correlation with SP than CI or SI. One possible reason is that a company should integrate internally before seeking external integration (Ataseven and Nair 2017). Consequently, the influence of II on SP should be stronger than CI or SI.

Second, our meta results confirm that SCI has overall strong positive influences on all three facets of SP and that the SCI–SP relationship varies with SP measurement. The positive relationship between SCI and the TBL has been examined by Gimenez, Sierra, and Rodon (2012). The nature of SCI as a goal-achieving tool for mutual strategic schemes may explain this. SCI focuses on global optimisation rather than local optimisation within one department or firm. With different strategic foci, SCI enables focal firms to promote different facets of SP. Consistent with RDT, both cross-departmental alliances and supply chain partnerships can break boundaries of information, knowledge, and resources to gain sustainable competitive advantages. Darmawan, Wong, and Thorstenson (2021), for example, emphasised the importance of supply chain design with coordinated inventory control. Shou et al. (2021) also argued that matching traceability and supply chain collaboration can help firms achieve superior operational performance. Therefore, a solid upstream or downstream integration fosters relational resources and capability, which are hard to imitate or replace (Chang et al. 2016). II also yields optimal utilisation of resources among different departments. Hence, the outcome of SCI initiatives (e.g. resource complementarity, coordination of production process, information sharing and knowledge creation, and mutual compliance) contribute to shared strategic goals (e.g. sustainability). As shown in our results, SCI has a stronger effect on economic SP than environmental or social SP. This finding emphasises the economic aspect of SCI. However, the reliability of this conclusion remains to be examined due to the limited sample of articles focusing on environmental performance and social performance.

The context of SCI implementation may affect the consequent SP. In the sampled articles, industry and economic development levels are often adopted as control variables, explaining the variance in the SCI–SP relationship. As for industry, the results confirm that the SCI–SP relationship is stronger in the mixed industry sector than in the service or manufacturing industries. The different effects of SCI may be related to industry characteristics, such as average market margins, inventory levels, concentration ratio of the industry and competition intensity (Geng, Mansouri, and Aktas 2017) or the different nature of product/service process characteristics (Ataseven and Nair 2017). It would be worth examining why the service rather than the manufacturing industry is more likely to gain SP from SCI practices. Yuen and Thai (2017) compared SCI's impact on operational performance between manufacturing and service supply chains and found that service supply chain receives more influences from external SCI than manufacturing supply chains, but the difference among internal SCI is not significant. Our results partially support their first finding but refute their second finding.

SCI practices in varying economic environments are found to have differentiated impacts on SP. In our findings, SCI implementation in a global business environment (mixed regions) shows weak SP relative to a single economic region (developing or developed region). However, mixed regions represent multiple data sources, which would provide a more comprehensive understanding for scholars on the effects of SCI implementation. Moreover, country-level differences, such as national culture (Chang et al. 2016) and logistics capabilities (Wiengarten et al. 2014), would affect SCI implementation and the corresponding outcomes. That may be the reason why the SCI–SP relationship is stronger in developing than in developed countries.

The most innovative part of our study is the examining of the moderating effect of a country's logistics capabilities, measured by LPI. We find that the SCI–SP relationship is relatively weak (r = 0.39, P < 0.001) in countries with a high LPI, but relatively high (r = 0.65, P < 0.001) in countries with below-average LPI. The relationship becomes weaker when the LPI level increases. This finding may be surprising, but is consistent with those of the study by Wiengarten et al. (2014). As stated in Wiengarten et country's (2014) research, strong logistics capabilities reduce the resources for supply chain partners to coordinate, but the situation does not necessarily mean that high logistics capabilities would not benefit the SCI–SP relationship. Indeed, those reduced resources can be used on other more idiosyncratic elements of the relationship to create a truly innovative SCI.

5.2. Managerial implications

Our meta-analysis results also yield several practical implications for participants in collaborative practices. First, it is beneficial for managers to adopt SCI practices because SCI practices positively influence SP. In summary, the SCI–SP relationship varies with economic development level and industry type. However, the SCI–SI relationship is stronger when the level of a country's logistics capabilities is lower. However, it might be better to build II before pursuing external integration, as we found that II has a stronger effect on SP than CI or SI. Nevertheless, II is an inside-out concept that is instrumental for external cooperation, which requires more resources at a higher implementation cost. It is not guaranteed that adopting SCI ensures better SP, which depends on problem-solving in light of uncertainty and contextual factors.

Second, this study provides managers with insights into different effects on SP from each aspect of SCI. Our results show a strong correlation between environmental and economic integration and SP. Therefore, companies that adopt a sustainability strategy are encouraged to conduct green integration initiatives, in addition to regular collaboration based on efficiency and profitability. Social integration, such as monitoring suppliers’ ethical compliance, would not hinder the general sustainable benefits, though it seems directly costly and non-financially beneficial. As for measurement of SP, our results imply that SCI activities yield better influence on environmental SP than on social aspects. In this regard, SCI is an efficient tool for green supply chain performance (e.g. waste and pollution reduction, water utilisation). The multiple measurements of SP empower companies to understand and evaluate the significance and advantages of SCI implementation.

Third, managers may usefully put more effort into developing SCI when their companies do business in both the manufacturing and service sectors, where SCI has the most effective influence. Manufacturing industry could adopt the servitization strategy. Compared with the manufacturing industry, managers of companies in the service industry could contribute to companies’ SP after developing SCI, as the effect of SCI on SP is stronger in the service industry than in the manufacturing industry. In addition, managers should also consider the country's logistics capabilities (measured by LPI) when implementing SCI. For the companies in low LPI countries, managers should put more effort into SCI than companies in high LPI countries, because the SCI–SP relationship is the most profound in countries with low LPI. However, managers do not need to balance SCI effort when their companies are located either in developing or developed countries, as the effect of SCI on SP is significantly high in both such regions. Therefore, no matter where firms are located, managers should plan their logistics and transportation properly. According to McKinsey & Company’s (2020) strategic report, one of the actions firms should take to combat COVID-19 is to identify and secure logistics capacity.

6. Conclusion

With 60 reviewed empirical studies, we conducted a meta-analysis of the relationship between SCI and SP. In the subgroup analysis, the differentiated effects of internal and external (customer or supplier) integration are confirmed. The moderating effects of control variables (i.e. industry, economic development level, and country's logistics capabilities) and operationalisation of SP are also examined. Our meta-analysis results support all our hypotheses except for H2b. Our study extends Ataseven and Nair (2017) by examining the moderating effect of control variables. One of the novelties is the investigation of the SCI–SP relationship considering logistics capabilities; to our best knowledge, this is the first meta-analytic research paper that adopted a country's logistics capabilities as a moderator.

Despite the contributions there remain limitations and research opportunities. First, we examined the linear relationship between SCI and SP, but there may be nonlinear relationships still to be explored. Second, although we included social performance in the analysis, we have not gone far in this field. In extant studies, few papers have explored the nexus between SCI and social SP considering ethical issues (e.g. labour, fairness, and safety). The trend of supply chain integration and sustainability performance study could focus on specific domain and green practices. Our future work could identify socially sustainable initiatives through SCI and explore the measurement of social outcomes. Last, we only verified the moderating effects of selected contextual variables (industry type, economic development level, and a country's logistics capabilities), but other control variables could be further explored, including firm size, firm age, and product customisation level. We advocate future research to examine the mediating and moderating effects of these variables, since the existing research on such mechanisms is far from comprehensive or conclusive.

Acknowledgements

The authors would like to thank the editors and anonymous reviewers for their valuable and constructive comments.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors gratefully acknowledge the financial support by Natural Science Foundation of China Young Scientist Fund (no. 71902159).

Appendix A:

Formulae to transform other statistics

Table

Other statistics to be transformed	Transforming formula	Notes
Student's t	$r=\sqrt{\left(t^2\right)/\left(t^2+\mathrm{df}\right)}$	Can be used in either paired or unpaired $t$ test
F-ratio	$r=\sqrt{\left(F\right)/\left(F+\mathrm{df}\left(\mathrm{error}\right)\right)}$	Only aviable in one-way ANOVA
${\chi }^2$	$r=\sqrt{{\chi }^2/n}$	$n$ is the sample size; can be used with $\mathrm{df}=1$
$d$	$r=d/\sqrt{d^2+4}$	$d=$Cohen's $d$
$\text{β}$	$r=0.98\times \text{β}+0.05$, if $\text{β}\ge 0$; $r=0.98\times \text{β}$, if $\text{β}<0.$	$\text{β}$ is the beta coefficient from the regression models, $\text{β}\in \left(-0.5,0.5\right)$

*Sources: Adapted from Wang, Lai, and Shou (2018)