Strategic Trade-Offs in Forward and Backward Integration: Evidence of Organizational Resilience from Systemic Supply Chain Disruptions

Abstract

In an increasingly uncertain business environment, developing organizational resilience to cope with supply chain disruptions is crucial for firms aiming to achieve sustainable growth. This study investigates how forward and backward vertical integration influence organizational resilience in the face of large-scale supply chain disruptions, with particular attention to the moderating role of a firm’s position in the supply network. Drawing on a comprehensive dataset of 2931 publicly listed Chinese firms, we integrate the relational view and information processing theory to examine how integration strategies affect two key dimensions of resilience: organizational stability and flexibility. Our results show that backward integration enhances both stability (reducing the severity of loss by about 17%) and flexibility by accelerating recovery, especially benefiting downstream firms in terms of stability and upstream firms in terms of flexibility. In contrast, forward integration is associated with reduced stability (raising the severity of loss by about 7%) but enables faster recovery for firms closer to end markets. Moreover, we find that the effectiveness of vertical integration depends on organizational context and alternative resilience mechanisms. These findings highlight the importance of aligning integration direction with supply chain position to optimize resilience. By disentangling the distinct strategic trade-offs of forward versus backward integration, this study advances theoretical understanding and offers practical guidance for firms seeking to strengthen their capacity to withstand and recover from systemic shocks.

1. Introduction

Globalization and specialization have profoundly transformed supply chain management, fostering the development of complex and interconnected global supply networks [1]. Enabled by rapid advances in information technology, supply chains now span multiple countries and regions, becoming a defining feature of the modern economy [2]. While this shift from linear value chains to networked structures has improved efficiency, it has also increased firms’ dependence on inter-organizational coordination and stable external environments [3].

Over the past decade, a series of shocks—including natural disasters, geopolitical tensions, cyberattacks, climate-related events, and global pandemics—have repeatedly stressed global supply systems, posing significant threats to firms’ long-term sustainability and operational continuity [4]. Such events highlight the fragility of specialized and lean supply chains, which, although efficient in normal times, often serve as channels that amplify and propagate disruptions [5]. For instance, the 2021 fire at Renesas’s Naka plant triggered ripple effects across the global chip supply and halted downstream automotive production at companies such as Toyota and Honda. It is reported that more than 56% of companies worldwide suffer from supply chain disruptions annually, with the figure expected to rise amid intensifying external uncertainty [6].

In response, firms are rethinking supply chain architectures with resilience as a core design principle [7]. Integration strategies have re-emerged as potential buffers against volatility and supply disruptions [8,9]. Compared with horizontal integration, which primarily enhances scale and market share [10,11], vertical integration—the internalization of upstream and/or downstream activities—offers greater potential for mitigating volatility and inter-firm dependencies. Accordingly, this study concentrates on vertical integration strategies to explore their role in strengthening organizational resilience. By internalizing formerly external exchanges, vertical integration may improve coordination, enhance information flow, and reduce dependency on external partners [12,13]. Consistent ownership across supply chain stages facilitates resource sharing, strategic alignment, and response coordination [14].

While the operational benefits of vertical integration are well recognized, such as improving supply chain coordination [15], enhancing information flows [16], and reducing transaction costs [14,17], its role in enhancing organizational resilience remains underexplored. Clarifying the relationship between vertical integration and organizational resilience is critical, as it not only informs firms’ capacity to withstand disruptions but also has important implications for achieving sustainable operations and strategic viability. Though some studies suggest a positive link between integration and resilience [18,19], most do not distinguish between the differential effects of forward and backward integration. However, emerging research shows that these integration directions differ in their implications for firm behavior, performance, and risk exposure [12,17,20,21]. It is thus plausible that their impacts on organizational resilience also differ—a question that merits systematic investigation.

This study aims to bridge this gap by examining the strategic trade-offs between forward and backward integration in enhancing organizational resilience during systemic disruptions, with a particular focus on the COVID-19 pandemic—the most profound and widespread supply chain disruption in nearly a decade [22]. We argue that while vertical integration is often viewed as a resilience-enhancing mechanism, its direction—toward customers or suppliers—may entail different benefits, costs, and risks. Forward integration may improve market responsiveness but expose firms to demand volatility, whereas backward integration may enhance supply security but limit downstream adaptability. These trade-offs are particularly pronounced under high-uncertainty conditions, such as the COVID-19 pandemic, which we use as an empirical setting.

Our sample of 2931 publicly listed Chinese firms spans key industries such as manufacturing, services, transportation, and technology, and covers firms from across China’s major regions, capturing the diversity of firms and providing a broadly representative basis for our analysis. China, as the initial epicenter of COVID-19, experienced severe supply chain disruptions and stringent policy responses, offering a unique context for examining how integration strategies influenced firms’ ability to withstand and recover from systemic shocks. Building on DesJardine [23] and Sajko [24], we assess resilience through two complementary dimensions: stability (the ability to absorb disruption) and flexibility (the ability to recover).

Furthermore, drawing on contingency theory, we explore the moderating role of supply chain position. A firm’s location within the supply network—whether upstream or downstream—shapes its exposure to risk and its access to critical resources [20,25]. We hypothesize that the alignment of integration direction and supply chain position plays a critical role in shaping resilience outcomes.

Our results reveal that backward integration consistently enhances both stability and flexibility, while forward integration entails a trade-off—reducing stability but accelerating recovery, particularly for downstream firms. Our results remain robust when addressing potential self-selection bias using an IV approach. These findings suggest that resilience is not uniformly improved through integration, but rather depends on the strategic alignment between integration type and structural context.

This study contributes to the literature in several ways. First, it introduces a nuanced perspective on vertical integration by examining the distinct roles of forward and backward strategies in resilience-building. Second, it highlights the importance of contextual fit, emphasizing that the effectiveness of integration is contingent on the firm’s position within the supply network. Third, it provides a framework for managers to assess strategic trade-offs in integration decisions under high-uncertainty environments.

The remainder of this paper is organized as follows. Section 2 reviews the relevant literature and develops hypotheses. Section 3 outlines the research design and data sources. Section 4 presents the empirical results, including robustness and heterogeneity analyses. Section 5 discusses the findings and managerial implications. Section 6 concludes with limitations and directions for future research.

2. Literature Review and Research Hypothesis

2.1. Literature Review

2.1.1. Supply Chain Disruption and Organizational Resilience

In global supply chains, the growing dependence and complexity of interfirm partnerships have increased firms’ vulnerability to disruption risks arising from unexpected events such as supplier bankruptcies, natural disasters, regional conflicts, and policy changes [26]. The outbreak of the COVID-19 pandemic heightened investor concerns over supply chain security, as the crisis severely disrupted supply, demand, and logistics, triggering widespread global supply chain breakdowns [27,28]. Prior studies have shown that supply chain disruptions are costly, often resulting in significant short- and long-term losses in shareholder value, revenue, and corporate reputation [29,30]. In extreme cases, such disruptions may even lead to business failure. These experiences have reinforced the recognition that environmental uncertainty and supply chain fragility can have severe—if not catastrophic—consequences for firm operations and long-term viability.

Under turbulent market conditions and frequent supply chain disruptions, research on organizational resilience has received considerable attention among academics and organizational managers in the past decade [23,31,32]. Organizational resilience is a dynamic capability, defined as an organization’s ability to withstand and recover from adverse events that disrupt its operations and threaten its survival [33]. Unlike sustainability, which emphasizes balancing long-term growth with the preservation of vitality, organizational resilience highlights a firm’s capacity to withstand and recover from short-term shocks. Despite these differences, the two are mutually reinforcing: resilience provides the foundation for achieving sustainability, while sustainability, in turn, strengthens and enhances resilience [34,35]. Existing empirical studies on organizational resilience primarily adopt a two-dimensional perspective—stability and flexibility, which aligns with the dual capacities of enduring and recovering from shocks [23,24,36].

Research shows that stable network relationships and strong information acquisition and processing capabilities are essential for building resilience. Coordinating stakeholder networks and aligning goals are critical for maintaining continuity during external shocks [23]. These networks are broad, encompassing relationships not only with supply chain partners such as suppliers and distributors, but also with broader social and natural systems [37,38]. Firms with stable ties are better positioned to manage coordination and align incentives. At the same time, enhanced information capabilities support adaptation and recovery. For instance, Quesada [39] found that information sharing with suppliers enables real-time responses to issues such as inventory shortages, production delays, and delivery disruptions, thereby enhancing operational flexibility.

Building on this, existing research suggests that digital technologies can bridge these capabilities, allowing firms to allocate internal and external resources more flexibly and to construct agile supply chain networks capable of responding to emergencies [40,41]. Digital tools enhance real-time visibility across the entire supply chain by tracking goods flows and production processes. Moreover, digital platforms support collaborative planning and real-time information exchange among supply chain members, improving overall coordination [42]. However, while integrating digital technologies within internal organizational processes (e.g., production, procurement, logistics, and demand management) has been shown to significantly enhance resilience, extending these deployments across supply chain partners does not yield the same benefits [43].

These results suggest that effective cross-border cooperation and information sharing remain challenging in practice [44]. Customer and production information, as core competitive resources, are often difficult to share openly among supply chain partners. Self-interested parties may misuse or distort private information to maximize their benefits, leading to opportunism and future non-cooperative behavior [45]. Furthermore, in complex supply chain networks, focal firms may become more vulnerable to external shocks due to cultural differences, conflicting goals, and uneven benefit distribution among partners. Against this backdrop, understanding how firms can reconfigure governance structures to enhance resilience becomes increasingly important.

2.1.2. Literature Review on Vertical Integration

Facing potential supply disruptions, many scholars have called for companies to strengthen supply chain integration to enhance their resilience to external shocks [46,47,48]. Existing research shows that horizontal integration can expand a company’s market size and strengthen its bargaining power within the supply chain, but its ability to connect upstream and downstream supply chains is relatively weak [10,11]. Vertical integration involves a series of internal transactions that help firms avoid external market transactions [14]. While vertical integration enhances a firm’s ability to control and coordinate complex, highly interdependent activities, concerns over reduced managerial efficiency have led supply chain management practices in recent decades to favor outsourcing and “de-verticalization” [49]. However, in increasingly uncertain environments, firms appear more inclined toward vertical integration. This shift can be explained from two perspectives. First, rising external transaction costs under uncertainty incentivize firms to internalize activities, consistent with the theory of transaction cost economics. Second, ownership alignment across production stages facilitates a shared vision and consistent incentives, enhancing firms’ control over the entire system [50]. These advantages improve coordination within complex supply chains, promote better matching among supply chain components, and facilitate more effective information sharing.

Vertical integration can be categorized into forward and backward integration. Forward integration refers to a firm’s extension toward the demand side, such as merging with distributors or developing downstream operations, while backward integration involves expanding toward the supply side by acquiring suppliers or developing upstream capabilities [14]. Prior studies have identified key strategic differences between the two forms of integration. In terms of network relationships, forward integration strengthens downstream ties by expanding distribution channels, reducing dependence on intermediaries, and fostering direct interactions with consumers [51]. In contrast, backward integration reinforces upstream networks, ensuring supply stability and reducing costs associated with monopolistic inputs [52]. Regarding information acquisition, forward integration provides access to downstream or end-customer information, enhancing responsiveness to market demand [53], while backward integration offers insight into suppliers’ private capacity and delivery schedules, aiding in the management of supply uncertainties [54]. This asymmetry is ultimately reflected in the effectiveness of vertical integration strategies in reducing corporate costs [16,55,56] and alleviating the bullwhip effect [20].

Although the positive role of integration capabilities in mitigating supply chain disruptions and enhancing organizational resilience has been recognized [18,19], existing research has yet to systematically examine the differential effects of vertical integration on resilience, as well as its boundary conditions. This study addresses this gap by distinguishing between forward and backward integration and analyzing their distinct impacts on organizational stability and flexibility in the context of COVID-19. It provides a more nuanced understanding of how specific integration strategies contribute to resilience-building under disruption. Moreover, by investigating the moderating role of a firm’s supply chain position, this study clarifies the contextual effectiveness of vertical integration strategies, offering important implications for both theoretical advancement and managerial decision-making.

2.2. Research Hypothesis

2.2.1. The Relationship Between Vertical Integration and Organizational Resilience

Organizational stability reflects the ability of a firm to maintain its core organizational attributes when it is disturbed by external factors [57]. Prior studies have used the relational view to confirm that relational assets, interdependence, routines of shared knowledge, and complementary resources and capabilities are mutually beneficial and play an important role in maintaining organizational stability [58]. These studies provide theoretical arguments for the value of forward integration and backward integration in enhancing organizational stability during crises.

From the relational view, ownership consistency under a vertical integration strategy facilitates the establishment of a common vision and strategic cooperation among production departments at different levels in the supply chain [14]. In the face of potential supply chain disruptions, vertically integrated companies can reconfigure and adjust the use of resources based on common goals and shared resources to ensure the continued flow of materials, information, and funds during crises [19]. Forward integration involves expanding control over downstream activities closer to end consumers. These features enable companies to identify risks related to daily demand and develop effective strategies to manage demand fluctuations [55]. Furthermore, forward integration can foster customer loyalty by strengthening direct interactions between firms and consumers [59]. These traits can, in turn, protect firms during crises by reducing customer churn [60]. Backward integration, which extends an organization’s activities to the production side, facilitates real-time feedback on inventory, production schedules, and potential delivery delays through improved interdepartmental coordination and communication. This enhanced information flow strengthens the organization’s ability to absorb shocks and mitigate losses when facing external disruptions [39]. Therefore, we propose hypotheses H1a and H1b.

H1a. 

Firms with high levels of forward integration exhibit greater organizational stability.

H1b. 

Firms with high levels of backward integration exhibit greater organizational stability.

Organizational flexibility is defined as the ability to recover from external changes through innovative adjustments in technology, products, and so on [61]. More flexible organizations can adjust more quickly and recover faster after disturbances [62]. Information is an important factor affecting organizational flexibility. Organizations with stronger information processing capabilities often have advantages in capturing market changes and proposing solutions, which could help organizations quickly adapt to external changes and recover from disruption.

Access to information resources is an important motivation for organizations to integrate. Forward integration, which could effectively improve the information acquisition and processing capabilities of the demand side, could accelerate organizational decision-making, deployment, and resource allocation. Affected by the pandemic, market demands, such as consumer behavior, capabilities, and values, have undergone tremendous changes. Organizations that implement a forward integration strategy have greater insight into changes in the market, which facilitates the search for alternative solutions to meet customer needs [63]. On the contrary, backward integration contributes greatly to improving information exchange between organizations and suppliers, which could increase information transparency and improve information availability [64]. After the epidemic outbreak, as multiple independent production decisions are replaced by centralized decisions, organizations can enhance their competitive advantages in terms of product modification, quality improvement, cost control, and long-term relationships [65], and then recover faster than their peers. Therefore, we propose hypotheses H2a and H2b.

H2a. 

Firms with high levels of forward integration exhibit greater organizational flexibility.

H2b. 

Firms with high levels of backward integration exhibit greater organizational flexibility.

2.2.2. The Moderating Effect of the Supply Chain Position

The above analysis suggests that forward and backward integration can enhance organizational stability during crises by strengthening network relationships with consumers and suppliers, respectively. However, as Verghese [66] emphasizes, a supply chain is only as strong as its weakest link. These weak links are often associated with a firm’s position within the supply chain. Upstream firms typically supply intermediate goods to other firms for further processing, while downstream firms are primarily involved in final production and distribution, interacting more closely with consumers [20]. Consequently, upstream firms tend to have more stable access to raw materials but face limitations in market control and distribution channels [67]. In contrast, downstream firms are more consumer-facing but may lack secure upstream supply relationships.

Based on this asymmetry, we argue that integration strategies that compensate for a firm’s relational deficiencies within the supply chain are more effective in enhancing organizational stability during crises. Specifically, forward integration can help upstream firms strengthen relationships with consumers [68], thereby addressing their downstream limitations. Conversely, backward integration enables downstream firms to secure more reliable supply sources [69], mitigating their upstream vulnerabilities. Accordingly, we propose Hypotheses H3a and H3b.

H3a. 

The further upstream the firm is, the greater the positive impact of forward integration on organizational stability.

H3b. 

The further downstream the firm is, the greater the positive impact of backward integration on organizational stability.

From the perspective of information acquisition, we infer that both forward and backward integration can enhance organizational flexibility. However, drawing on organizational information processing theory (OIPT), the effectiveness with which a firm understands, absorbs, and utilizes external information depends on its existing knowledge base [67]. When newly acquired information is highly tacit or unfamiliar, it increases the cost of transforming information into actionable knowledge, thereby weakening the organization’s information processing capacity and slowing its recovery during crises. In other words, the speed of recovery depends on the relevance and usability of the information collected.

Downstream firms, being closer to end consumers, tend to focus more on marketing and distribution, adding value through precise product positioning and effective channel management [67]. As a result, these firms typically possess stronger perceptual sensitivity and higher information processing capabilities in the consumer market [70]. In contrast, upstream firms are primarily concerned with production efficiency and cost control, with their value typically embedded in intermediate goods [71]. These firms generally have deeper knowledge reserves related to input prices and production processes. Given the differences in information needs and strengths across supply chain positions, forward integration—being closer to consumers—offers greater advantages in acquiring market-related information, while backward integration facilitates access to supply- and production-related knowledge. Based on this logic, we propose Hypotheses H4a and H4b.

H4a. 

The further downstream the firm is, the greater the positive impact of forward vertical integration on organizational flexibility.

H4b. 

The further upstream the firm is, the greater the positive impact of backward vertical integration on organizational flexibility.

The conceptual model is presented in Figure 1.

3. Research Methodology

3.1. Research Context and Data Sources

This study investigates how vertical integration strategies affect organizational resilience during major disruptions, using the COVID-19 pandemic as the research context. China provides a compelling context for this research. As the first country to experience widespread lockdowns and supply chain interruptions, Chinese firms were exposed to early and severe operational shocks. Moreover, the Chinese manufacturing sector is characterized by vertically complex supply networks and strong policy incentives for integration, making it a suitable empirical setting to examine the nuanced impacts of forward and backward integration.

Our analysis focuses on publicly listed firms on China’s Shanghai and Shenzhen stock exchanges, covering a broad range of industries. For robustness checks, we separately examine firms in the manufacturing sector, classified according to the China Securities Regulatory Commission (CSRC) industry classification, given its extensive supply base, production complexity, and frequent use of vertical integration to manage risk and improve coordination. Firms in the financial sector were excluded due to their distinct operating structures and regulatory environments. We also excluded firms with special treatment (ST/PT) designations, as these companies face trading restrictions due to financial distress, potentially biasing market-based performance measures. Firms with missing financial or market data were likewise excluded.

The final sample consists of 2931 firms across a wide range of subsectors. The observation window spans from December 2019 to June 2020, capturing the pre-crisis baseline, the peak of COVID-19-induced market shocks, and the initial recovery phase. Firm-level financial and stock price data were obtained from the China Stock Market & Accounting Research (CSMAR) database. Industry-level input-output (I-O) data, published by the National Bureau of Statistics of China, were used to calculate vertical integration and supply chain position metrics.

To illustrate the representativeness of the sample,

Table 1. Industry distribution of the final firm sample (Number of firms, N= 2931).

Industry Category	Number of Firms	Percentage
Mining, Quarrying, and Gas Extraction	71	2.42%
Manufacturing
Machinery	369	12.59%
Automobile	122	4.16%
Computer and Instrument	379	12.93%
Lumber, Furniture, Paper, and Printing	68	2.32%
Electrical Equipment, Appliance, and Component	212	7.23%
Food and Beverages	131	4.47%
Textile Apparel and Leather	73	2.49%
Petroleum, Chemicals, Medicine, Rubber, and Plastic Products	552	18.83%
Glass, Minerals, and Metals	230	7.85%
Other manufacturing	34	1.16%
Electricity, Heat, Gas, and Water Production	102	3.48%
Transportation and Storage	93	3.17%
Information transmission, Software, and Information technology services	275	9.38%
Other service industries	220	7.51%
Total	2931	100.00%

presents the industry distribution across key manufacturing categories. The sample includes firms from electronics, machinery, textiles, chemicals, and other subsectors, supporting the generalizability of the findings within the broader industrial economy.

3.2. Measurement of Key Constructs

3.2.1. Organizational Resilience

Organizational resilience in this study is conceptualized as a firm’s ability to withstand and recover from large-scale external shocks. Drawing from the resilience literature in operations and risk management, we focus on two complementary dimensions: stability and flexibility. Stability refers to a firm’s ability to absorb disruption. Flexibility reflects the firm’s capability to adapt and recover swiftly from the shock once it occurs. Given the lack of real-time operational performance data across a large sample, we adopt a widely used proxy derived from the financial market to capture these resilience dimensions. This approach assumes that stock price dynamics reflect investor expectations about a firm’s operational stability, recovery prospects, and supply chain robustness during crisis periods. In studies of organizational resilience under short-term exogenous shocks, many scholars have relied on stock price changes to measure resilience, because this approach provides a relatively clean signal of the market’s expectations of how well a firm can cope with an external shock, rather than being confounded by other factors [23,24,57].

Stability (Severity of losses): We measure stability using the maximum percentage decline in a firm’s stock price from the pre-crisis (23 January 2020) to the year after the crisis. A smaller decline suggests greater stability and investor confidence in the firm’s ability to withstand disruptions.

Flexibility (Recovery Time): Flexibility is measured by the number of trading days required for a firm’s stock price to recover to its pre-crisis level. This captures the speed of perceived organizational recovery by capital markets. Firms that recover more quickly are considered more operationally adaptive and resilient in the eyes of investors.

To validate this measure and provide descriptive insight, Figure 2 presents the distribution of recovery times across all sampled firms. The variation highlights differences in how quickly firms restored perceived value, reinforcing the importance of analyzing resilience heterogeneity.

While market-based indicators are imperfect proxies for operational resilience, they offer timely, continuous, and scalable measures, particularly valuable in contexts where private data on production delays, order fulfillment rates, or logistics disruptions are unavailable. Furthermore, capital market responses are increasingly used in the supply chain and operations literature to assess firm performance under uncertainty. To strengthen validity, robustness checks using alternative recovery thresholds and shorter measurement windows are reported in Section 4. In addition, we incorporate two indicators from the operational dimension as robustness tests. Specifically, we use the downside risk of return on assets (ROA) to capture organizational stability and the average operating efficiency in the post-crisis period to measure organizational flexibility. The measurement of downside risk (DR) follows the approach of Belderbos [72], which takes the industry’s average ROA ($\mathrm{I}\mathrm{R}\mathrm{O}\mathrm{A}$) in the previous year as the target level and calculates the volatility of a firm’s ROA deviation from the target level over the current year and the subsequent two years. This method emphasizes downside fluctuations relative to the benchmark, thereby capturing the firm’s vulnerability to adverse shocks. The specific calculation method is presented in Equation (1). A higher volatility indicates lower organizational stability. Operating efficiency, defined as the ratio of operating income to total assets. The data for these indicators are obtained from the CSAMR database. These indicators, grounded in firms’ actual operating activities, complement market-based measures and help alleviate concerns regarding the absence of operational data.

$${\mathrm{D}\mathrm{R}}_{\mathrm{i},\mathrm{t}=0}=\sqrt{{\displaystyle \frac{1}{3}}{\sum }_{\mathrm{t}=0}^{\mathrm{t}=2}{\left({\mathrm{I}\mathrm{R}\mathrm{O}\mathrm{A}}_{\mathrm{t}-1}-{\mathrm{R}\mathrm{O}\mathrm{A}}_{\mathrm{i},\mathrm{t}}\right)}^2\left|{(\mathrm{I}\mathrm{R}\mathrm{O}\mathrm{A}}_{\mathrm{t}-1}>{\mathrm{R}\mathrm{O}\mathrm{A}}_{\mathrm{i},\mathrm{t}}\right)}$$

(1)

3.2.2. Vertical Integration

Vertical integration (VI) reflects the extent to which a firm expands its operations across different stages of its supply chain under unified ownership and control [73]. In the context of manufacturing firms, vertical integration may involve internalizing supply inputs (backward integration) or downstream activities such as distribution and sales (forward integration). Assessing the degree of VI at the firm level, however, presents a methodological challenge due to the lack of standardized cross-sectoral operational data. While in principle firm-level disclosures such as M&A announcements, segment reports, or textual analysis of annual reports could provide valuable information on integration strategies, such data face three key limitations: (i) coverage—only a subset of firms report M&A or integration activities, making it infeasible to construct a consistent panel for thousands of firms; (ii) heterogeneity—disclosure practices vary widely across firms and industries, reducing comparability.

To address this, we follow the established approach of using industry-level input-output (I-O) tables as a proxy for inter-sectoral linkages, as proposed by Liang [20]. The I-O tables provide structured data on the flow of goods and services between sectors, allowing us to infer the strength of vertical relationships in a firm’s operational footprint. Although indices based on I-O tables cannot fully capture firm-specific integration flows, the lack of systematic firm-level data makes this approach a widely adopted and pragmatic solution in the literature, ensuring transparent, comparable, and replicable measurement across large samples while reflecting structural interdependencies between industries [74,75,76].

In this study, we distinguish between two directions of vertical integration. Forward vertical integration (FVI) refers to the extent to which a firm internalizes downstream activities, such as distribution and sales. In contrast, backward vertical integration (BVI) captures the extent of upstream integration, where a firm internalizes activities related to input procurement or production. The computation of FVI and BVI is based on a firm’s business segment composition and the strength of inter-sector linkages derived from the I-O tables. The following equations represent the FVI and BVI measures, respectively:

$${\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\displaystyle \frac{1}{\mathrm{n}-1}}\sum _{\mathrm{k}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}\sum _{\mathrm{k}\ne \mathrm{j}}{\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}\mathrm{j}}{\mathsf{\delta }}_{\mathrm{i}\mathrm{t}}^{\mathrm{k},\mathrm{j}}$$

(2)

$${\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\displaystyle \frac{1}{\mathrm{n}-1}}\sum _{\mathrm{k}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}\sum _{\mathrm{k}\ne \mathrm{j}}{\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{j}\mathrm{k}}{\mathsf{\delta }}_{\mathrm{i}\mathrm{t}}^{\mathrm{j},\mathrm{k}}$$

(3)

In the above equation, $\mathrm{i}$ denotes firm, $\mathrm{k}$ denotes industry, and $\mathrm{t}$ denotes year. ${\mathrm{w}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}$ represents the sales weight of firm $\mathrm{i}$ in industry $\mathrm{k}$ in year $\mathrm{t}$. ${\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}\mathrm{j}}$ is the correlation coefficient between sectors in the forward integration, representing the proportion of the value of industry $\mathrm{k}$ outflow to industry $\mathrm{j}$ to the total input value of industry $\mathrm{j}$. ${\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{j}\mathrm{k}}$ is the correlation coefficient between sectors in the backward vertical, indicating the quantity of industry $\mathrm{j}$ required to produce one unit of $\mathrm{k}$ industry products. ${\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}\mathrm{j}}$ and ${\mathrm{v}}_{\mathrm{i}\mathrm{t}}^{\mathrm{j}\mathrm{k}}$ are calculated with I-O data provided by the National Bureau of Statistics of China. ${\mathsf{\delta }}_{\mathrm{i}\mathrm{t}}^{\mathrm{k},\mathrm{j}}$ is a dummy variable, which is equal to 1 if firm $\mathrm{i}$ participates in both industry $\mathrm{k}$ and industry $\mathrm{j}$ in year t, and to 0 otherwise. n is the number of industries in which the firm is active.

To operationalize these measures, we proceed as follows:

• We use I-O data from the National Bureau of Statistics of China to compute forward and backward linkage coefficients between industry pairs.
• Firm-level business segment data are obtained from the WIND database, which provides detailed revenue breakdowns categorized by the 3-digit Global Industry Classification Standard (GICS) codes.
• Each business segment is matched to its corresponding I-O industry sector.
• The linkage coefficients between these sectors are then used to compute the firm-specific FVI and BVI scores using the above formulas.

These VI metrics offer a continuous and scalable representation of a firm’s structural integration along the supply chain and allow for comparison across firms with different sectoral footprints. Higher FVI values indicate stronger downstream control (e.g., integrated sales or distribution), while higher BVI values reflect tighter upstream coordination (e.g., in-house sourcing or input production).

3.2.3. Supply Chain Position

A firm’s supply chain position (SCP) captures its relative location within the broader input-output structure of the economy, specifically whether it operates closer to the upstream (raw materials and components) or downstream (assembly and final delivery) stages of the value chain. This positioning can significantly influence how vertical integration affects organizational resilience during disruptions.

Following established methods in networked supply chain research [20,74], we compute the SCP index using the ratio of a firm’s forward linkage to its backward linkage based on industry-level input-output data. The logic is that industries with stronger forward linkages (i.e., their output flows to many downstream industries) are more upstream, while those with stronger backward linkages (i.e., depend on many inputs) are more downstream. The SCP is calculated as follows:

$${\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}\mathrm{t}}=\sum _{\mathrm{k}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}{\mathrm{P}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}$$

(4)

where ${\mathrm{w}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}$ represents firm $\mathrm{i}’\mathrm{s}$ the sales weight in industry $\mathrm{k}$ in year $\mathrm{t}$; ${\mathrm{P}}_{\mathrm{i}\mathrm{t}}^{\mathrm{k}}$ is the proportion of industry $\mathrm{k}$’s output allocated to final use, calculated as industry $\mathrm{k}$’s final use divided by industry $\mathrm{k}$’s total output; $\mathrm{n}$ is the total number of industries in which the firm is active. A higher SCP value indicates that a firm is more downstream (closer to final customers), while a lower value indicates an upstream position (closer to raw materials and suppliers).

Understanding SCP is critical because it may moderate the effectiveness of different integration strategies. For example, upstream firms may benefit more from forward integration, as it enhances access to market signals and customer demand visibility, while downstream firms may benefit more from backward integration, as it helps secure the supply of critical inputs during disruptions. We therefore incorporate interaction terms between SCP and both FVI and BVI in our empirical models to explore how supply chain positioning influences the resilience returns to vertical integration.

3.2.4. Control Variables

To enhance the robustness of our empirical analysis, we control for a comprehensive set of firm-level and regional-level factors that may influence organizational resilience during the COVID-19 disruption. These variables are selected based on theoretical relevance and prior empirical studies on firm performance under external shocks.

At the firm level, we account for six key characteristics. Firm size (size) is measured as the natural logarithm of one plus the total number of employees. Firm age (age) is defined as the number of years since establishment (i.e., 2020 minus the founding year). Operational efficiency (OE) is measured as the natural logarithm of the accounts receivable turnover ratio. Financial leverage (Lev) is calculated as the ratio of long-term debt to total assets, capturing the degree of debt burden and financial risk. Profitability (Prof) is measured by return on assets (ROA), defined as the sum of total profit and finance costs divided by total assets, reflecting pre-crisis financial performance.

We also control for macroeconomic conditions at the regional level. Regional GDP captures the overall level of economic development in the province where the firm is registered. Industrial structure (Stru) is measured as the proportion of manufacturing output in the regional economy, indicating the local economy’s dependence on industrial production.

All control variables are measured using 2019 (pre-crisis) data to avoid reverse causality and to ensure they reflect the firm’s condition before the COVID-19 shock. Descriptive statistics and a correlation matrix for these variables are reported in

Table 2. Descriptive statistics and correlation matrix.

Variable	Mean	Std	1	2	3	4	5	6
Severity of loss	0.211	0.099	1.000
Recovery time	57.880	69.131	−0.229 ***	1.000
FVI	0.014	0.04	0.038 **	0.034 *	1.000
BVI	0.014	0.032	−0.047 **	−0.000	0.271 ***	1.000
SCP	0.114	0.167	0.077 ***	0.002	−0.044 **	0.017	1.000
Size	7.654	1.212	0.086 ***	0.111 ***	0.110 ***	0.100 ***	0.045 **	1.000
Age	2.051	0.901	0.208 ***	0.073 ***	0.069 ***	0.065 ***	0.043 **	0.318 ***
OE	1.737	1.212	0.063 ***	0.192 ***	0.074 ***	0.075 ***	0.180 ***	0.254 ***
Lev	0.037	0.069	0.056 ***	0.144 ***	0.029	0.006	0.046 **	0.171 ***
Prof	0.035	0.566	0.060 ***	0.001	0.004	0.000	0.006	0.054 ***
GDP	0.060	0.033	−0.070 ***	−0.094 ***	−0.041 **	−0.100 ***	−0.033 *	−0.036 *
Stru	0.378	0.083	−0.003	−0.027	−0.027	−0.003	−0.096 ***	−0.013
Variable	7	8	9	10	11	12	13
Age	1.000
OE	0.187 ***	1.000
Lev	0.190 ***	0.048 ***	1.000
Prof	−0.051 ***	0.056 ***	−0.003	1.000
GDP	−0.191 ***	−0.102 ***	−0.085 ***	0.021	1.000
Stru	−0.009	0.025	−0.016	−0.008	0.414 ***

Note: The statistical significance is denoted by: * $p<0.1$, ** $p<0.05$, *** $p<0.01$.

, while detailed definitions and data sources are provided in Appendix A.

3.3. Empirical Models

We examine the effect of vertical integration on organizational resilience using two dimensions: stability, measured by the severity of stock price loss, and flexibility, measured by the recovery time. Each dimension requires a distinct empirical approach due to differences in data structure and distributional properties.

3.3.1. Organizational Stability: OLS Regression

To assess organizational stability, we use the severity of stock price loss during the COVID-19 crisis as the dependent variable. Since this is a continuous variable that approximates a normal distribution, we adopt an ordinary least squares (OLS) regression framework. The baseline model is specified as:

$${\mathrm{S}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{o}\mathrm{f} \mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}}_{\mathrm{i}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(5)

Here, a higher value of Severity of loss indicates a greater stock price decline and thus lower organizational stability. FVI_i and BVI_i represent the forward and backward vertical integration indices, respectively. The vector controls include the firm-level and regional control variables described in Section 3.2.4. The coefficients β₁ and β₂ capture the marginal effects of vertical integration strategies on firm resilience. A positive coefficient suggests that the corresponding type of integration is associated with greater vulnerability (i.e., lower stability), while a negative coefficient indicates enhanced stability under disruption.

To explore the moderating role of supply chain position, we extend the baseline model to include the firm’s supply chain position index (SCP) and its interactions with both FVI and BVI:

$$\begin{gathered} {\mathrm{S}\mathrm{e}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{o}\mathrm{f}\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}}_{\mathrm{i}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_3{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}\times {\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}} \\ +{\mathsf{\beta }}_4{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}\times {\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}}+{\mathsf{\beta }}_5{\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}} \end{gathered}$$

(6)

In this specification, SCP_i measures the firm’s relative downstreamness in the supply chain, with higher values indicating proximity to end consumers. The interaction terms allow us to assess whether the effectiveness of vertical integration strategies depends on a firm’s supply chain position.

3.3.2. Organizational Flexibility: Cox Proportional Hazard Model

To evaluate organizational flexibility, we analyze the recovery time, defined as the number of trading days needed for a firm’s stock price to return to its pre-crisis peak. Since a small portion of firms had not recovered by the end of the observation window, the data are subject to right-censoring, and the distribution is highly skewed. Traditional linear regression is inappropriate under such conditions. Therefore, we employ a Cox proportional hazard model, which is well-suited for time-to-event analysis with censored data. The baseline Cox model is specified as follows:

$$\mathrm{h}\left(\mathrm{t}\right)={\mathrm{h}}_0\left(\mathrm{t}\right)\exp \left({\mathsf{\beta }}_1{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}\right)$$

(7)

Here, $\mathrm{h}$(t) denotes the hazard rate, representing the instantaneous probability that a firm recovers at time t, given that it has not yet recovered. The baseline hazard $\mathrm{h}$₀(t) is unspecified, allowing flexibility in the shape of the survival distribution. A positive coefficient (β_i > 0) indicates that the corresponding variable increases the likelihood of faster recovery (i.e., higher flexibility), while a negative coefficient suggests delayed recovery. We also construct a binary event indicator equal to 1 if the firm recovered during the observation period, and 0 otherwise. This enables the Cox model to properly account for censored cases.

To examine the moderating effect of supply chain position in this context, we extend the model by incorporating SCP and its interaction terms with FVI and BVI:

$$\begin{array}{c}h\left(\mathrm{t}\right)={\mathrm{h}}_0\left(\mathrm{t}\right)\mathrm{e}\mathrm{x}\mathrm{p}({\mathsf{\beta }}_1{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}+{\mathsf{\beta }}_3{\mathrm{F}\mathrm{V}\mathrm{I}}_{\mathrm{i}}\times {\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}}+{\mathsf{\beta }}_4{\mathrm{B}\mathrm{V}\mathrm{I}}_{\mathrm{i}}\times {\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}}\\ +{\mathsf{\beta }}_5{\mathrm{S}\mathrm{C}\mathrm{P}}_{\mathrm{i}}+\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s})\end{array}$$

(8)

This extended model allows us to test whether the effects of forward and backward integration on recovery speed vary depending on the firm’s position in the supply chain. All control variables are included as covariates in both baseline and interaction models.

4. Results

4.1. Main Regression Results

Table 3. The regression results on organizational stability.

	(1)	(2)	(3)	(4)
Dependent Variable	Severity of Loss			DR
FVI	0.0385		0.0754 *	4.6262 **
	(0.0334)		(0.0340)	(2.2361)
BVI		−0.1503 **	−0.1743 ***	−8.0707 ***
		(0.0497)	(0.0515)	(2.0673)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	2931	2931	2931	2104
adj. R*	0.3901	0.3922	0.3929	0.2057

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01. The same notation applies to all subsequent tables.

presents the OLS regression results evaluating the relationship between VI and organizational resilience during COVID-19. The validity of the models has been confirmed through diagnostic tests of linearity and normality, and heteroskedasticity-robust standard errors are employed throughout to ensure reliable inference. The detailed diagnostic procedures and results are documented in Appendix B.1. Columns (1) to (3) report the effects of VI on organizational stability, using the severity of loss as the dependent variable. This set of models is used to test Hypotheses H1a and H1b. In the stability models, Column (1) includes only FVI, Column (2) includes only BVI, and Column (3) includes both. After controlling for firm-level and regional covariates as well as time-fixed effects, the full model in Column (3) shows that the coefficient on FVI is positive and statistically significant (β = 0.0754, p < 0.05). This suggests that greater forward integration is associated with more severe losses during the crisis, implying that forward integration may undermine organizational stability in highly disrupted environments. This result does not support Hypothesis H1a, which posited a positive effect of FVI on stability. Conversely, the coefficient on BVI is negative and highly significant (β = −0.1743, p < 0.01), indicating that backward integration enhances organizational stability by reducing the severity of losses. This supports Hypothesis H1b, which theorized that securing upstream inputs through internalization would reduce supply-side vulnerability during crisis periods. Furthermore, we adopt the downside risk of corporate operations as an alternative measure of organizational stability. The regression results, reported in Column (4), are consistent with those based on the maximum stock price loss indicator.

The observed finding that forward integration may increase organizational vulnerability may be driven by the specific context of the COVID-19 pandemic. Following the outbreak, the Chinese government swiftly implemented policies to resume work and production, while firms enforced strict personnel monitoring and safety protocols. Backward-integrated firms primarily focus on upstream suppliers or internal tiers, with business operations largely contained within the corporate hierarchy. Under such government and corporate controls, these firms face relatively lower uncertainty. In contrast, forward-integrated firms interact more directly with downstream consumers, whose behaviors are heterogeneous and more difficult to monitor. The complexity of managing consumer interactions under pandemic-related safety requirements increases operational uncertainty for these firms. This explanation aligns with Li [57], who finds that business activities requiring intensive customer interaction heightened firm vulnerability during the COVID-19 pandemic.

Table 4. The regression results on organizational flexibility.

	(1)	(2)	(3)	(4)
Dependent Variable	Recovery Time			Efficiency
FVI	0.1871		−0.0012	0.1411
	(0.4616)		(0.4701)	(0.1882)
BVI		0.9649 *	0.9652 *	0.4151 *
		(0.5010)	(0.5065)	(0.2430)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	2931	2931	2931	2104
adj. R2	/	/	/	0.1825
No. recovery	2759	2759	2759	/
Wald	1440.01 ***	1445.11 ***	1445.64 ***	/

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, *** p < 0.01.

presents the COX regression results evaluating the relationship between VI and organizational flexibility. The global Schoenfeld residual test (χ² = 13.74, p = 0.1318) indicates that the proportional hazards assumption is satisfied, with the detailed test results reported in Appendix B.2. Columns (1) to (3) examine the impact of vertical integration on recovery speed using the Cox model. In this framework, a positive coefficient implies a higher hazard rate, indicating faster recovery and thus greater organizational flexibility. Column (3), which includes both FVI and BVI, shows that FVI has a positive but statistically insignificant effect on recovery speed, suggesting no support for Hypothesis H2a. However, BVI is again found to be positively associated with faster recovery, with the coefficient (β = 0.9652, p < 0.10) significant at the 10% level. This finding lends partial support to Hypothesis H2b, indicating that backward integration not only stabilizes firm performance during a shock but also facilitates quicker post-crisis recovery. In addition, we use the average operating efficiency of firms over the three years following the outbreak as an alternative measure of organizational flexibility. Firms with higher flexibility can more effectively allocate resources and adjust operations in response to disruptions. Operating efficiency, defined as the ratio of operating income to total assets, serves as a continuous indicator, and we employ OLS estimation for analysis. The results, reported in Column (4) of Table 4, are consistent with those based on stock price recovery time. Specifically, backward integration enhances organizational flexibility, whereas forward integration has no significant effect.

Taken together, the results suggest an asymmetry in the resilience effects of vertical integration strategies. While backward integration consistently enhances both stability and flexibility, forward integration may be less effective or even detrimental in turbulent contexts such as the COVID-19 pandemic.

Finally, to further explore the heterogeneity in the distribution of results, we applied quantile regression and Tobit models, with the corresponding results presented in Appendix C.

4.2. The Moderating Effect

To further investigate how the effect of vertical integration on organizational resilience is influenced by a firm’s position in the supply chain, we introduce interaction terms between the supply chain position index (SCP) and both types of vertical integration. This allows us to test Hypotheses H3a, H3b, H4a, and H4b. The regression results are presented in

Table 5. The regression results of the moderating effects.

Dependent Variable	(1)	(2)
	Severity of Loss	Recovery Time
FVI	0.0789 *	0.4882
	(0.0321)	(0.4747)
BVI	−0.1488 **	1.1123 **
	(0.0461)	(0.5214)
FVI × SCP	0.2244	13.1634 ***
	(0.2348)	(4.1784)
BVI × SCP	−0.5244 *	−3.2058 *
	(0.2481)	(1.7332)
SCP	0.0340 ***	0.2567 **
	(0.0089)	(0.1188)
Controls	Yes	Yes
Industry-fixed	Yes	Yes
Time-fixed	Yes	Yes
N	2931	2931
No. recovery		2759
adj. R2	0.3964	/
Wald	/	1106.38 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

.

In Column (1), where the dependent variable is the severity of loss, we observe that the interaction term between FVI and SCP is positive but not statistically significant. This implies that the negative effect of FVI on organizational stability is more pronounced for downstream firms; however, the lack of statistical significance indicates that this moderating effect is not robust, offering no support for Hypothesis H3a. In contrast, the interaction between BVI and SCP is negative and statistically significant (β = −0.5244, p < 0.1). This suggests that the stabilizing effect of BVI is stronger for downstream firms, supporting Hypothesis H3b. These results highlight that downstream firms benefit more from internalizing upstream supply sources, especially when facing supply chain disruptions.

Column (2) examines organizational flexibility, using recovery time as the dependent variable within a Cox proportional hazard framework. The interaction term for FVI × SCP is positive and highly significant (β = 13.1634, p < 0.01), indicating that forward integration significantly improves the recovery speed of firms closer to the consumer end of the supply chain. Conversely, the BVI × SCP interaction is negative and marginally significant (β = −3.2058, p < 0.10), suggesting that backward integration is more beneficial for the recovery of upstream firms. Together, these findings support both Hypothesis H4a and H4b and emphasize the differentiated resilience benefits of FVI and BVI depending on a firm’s position within the supply network.

To further validate the moderating effect of supply chain position, we conduct subsample analyses by splitting the data into groups based on the median and quartile values of the SCP index. While such subgroup analyses reduce the sample size within each group and may consequently decrease statistical power, the results across groups still provide important evidence of the robustness of the moderating effect.

Table 6. Grouped regression results on organizational stability.

Dependent Variable	Severity of Loss
	SCP > Median	SCP < Median	SCP > 75%	SCP < 25%
FVI	0.0731	0.1036 **	0.1573 **	0.0415
	(0.0484)	(0.0444)	(0.0640)	(0.0490)
BVI	−0.2304 ***	−0.0987	−0.2623 **	0.0197
	(0.0669)	(0.0782)	(0.1071)	(0.1423)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1466	1465	732	732
adj. R2	0.3964	0.4045	0.3697	0.4056

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: ** p < 0.05, *** p < 0.01.

and

Table 7. Grouped regression results on organizational flexibility.

Dependent Variable	Recovery Time
	SCP > Median	SCP < Median	SCP > 75%	SCP < 25%
FVI	0.4264	−0.5538	3.6755 ***	−0.3857
	(0.7009)	(0.6125)	(0.8317)	(0.8604)
BVI	0.5642	1.5078 *	−0.5012	2.1902 *
	(0.6441)	(0.7765)	(0.7833)	(1.1484)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1466	1465	732	732
No. recovery	1369	1390	687	695
Wald	759.20 ***	891.488 ***	378.44 ***	852.71 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, *** p < 0.01.

report the results for organizational stability and flexibility, respectively.

Table 6 shows the grouped regression results for organizational stability. The effect of FVI is inconsistent across SCP segments. Specifically, the coefficient of FVI is significantly positive in the SCP < Median and SCP > 75% groups, but not significant in other segments. This lack of regularity suggests that Hypothesis H3a is not supported. In contrast, the coefficient of BVI is significantly negative in the SCP > Median and SCP > 75% groups, but insignificant for firms in more upstream positions (SCP < Median, SCP < 25%). These results confirm that backward integration is particularly effective in enhancing the stability of downstream firms, offering further support for Hypothesis H3b.

Table 7 reports the results for organizational flexibility. The coefficient of FVI is significantly positive only in the SCP > 75% group, suggesting that forward integration improves recovery speed specifically for downstream firms. The coefficient is insignificant in all other groups. Conversely, the coefficient of BVI is significantly positive in the SCP < Median and SCP < 25% groups, implying that backward integration enhances recovery for upstream firms, but not for downstream ones. These findings further support Hypotheses H4a and H4b, confirming the asymmetric benefits of vertical integration strategies depending on a firm’s supply chain location.

4.3. Robustness Tests

To verify the reliability of our main findings, we conduct a series of robustness checks for both the direct effects and moderating effects of vertical integration on organizational resilience. These tests address alternative model specifications, sample adjustments, and measurement variations. The results confirm the overall consistency and robustness of our conclusions.

4.3.1. Robustness of Organizational Stability Results

Table 8. Robust test on organizational stability.

Dependent Variable	Severity of Loss
	Model 1 (Shortened Window)		Model 2 (Manufacturing)		Model 3 (Alternative Measure)
FVI	0.0753 **	0.0788 **	0.0672 *	0.0656 *	0.0565 *	0.0601 **
	(0.0336)	(0.0318)	(0.0371)	(0.0336)	(0.0304)	(0.0291)
BVI	−0.1676 ***	−0.1440 ***	−0.0973 **	−0.0955 **	−0.1715 ***	−0.1412 ***
	(0.0497)	(0.0438)	(0.0436)	(0.0467)	(0.0580)	(0.0538)
$\mathrm{F}\mathrm{V}\mathrm{I}\times \mathrm{S}\mathrm{C}\mathrm{P}$		0.2063		0.3119		0.1533
		(0.2407)		(0.2191)		(0.2033)
$\mathrm{B}\mathrm{V}\mathrm{I}\times \mathrm{S}\mathrm{C}\mathrm{P}$		−0.4912 *		−0.0899		−0.6267 **
		(0.2527)		(0.1574)		(0.2482)
SCP		0.0327 ***		0.0415 ***		0.0366 ***
		(0.000)		(0.0119)		(0.0094)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2931	2931	2328	2328	2931	2931
adj. R2	0.3607	0.3642	0.3887	0.3913	0.3754	0.3795

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the robustness checks for the models using the severity of loss as the dependent variable. In Model 1, we shorten the observation window, ending the sample period in the second quarter of 2020 to ensure the results are not sensitive to post-crisis rebounds. In Model 2, we restrict the sample to manufacturing firms only, given their central role in multi-tier supply chains and frequent use of vertical integration. In Model 3, we revise the calculation of the severity of loss by using weekly closing prices instead of daily prices to smooth short-term volatility.

Across all three models, the coefficients of interest—particularly for FVI, BVI, and their interactions with SCP—remain directionally consistent with those in the baseline analysis. Specifically, BVI retains a significant negative effect on the severity of loss in all models, confirming its stabilizing role. While the interaction term BVI × SCP remains negative and significant in Models 1 and 3, the results for FVI × SCP are directionally consistent but not statistically significant. These findings reinforce the robustness of our conclusions regarding organizational stability.

4.3.2. Robustness of Organizational Flexibility Results

Table 9. Robust test on organizational flexibility.

Dependent Variable	Recovery Time
	Model 1 (Shortened Window)		Model 2 (Manufacturing)		Model 3 (Alternative Models)
FVI	−0.0998	0.3941	0.3979	0.6594	−0.0880	0.2194
	(0.5561)	(0.5275)	(0.5206)	(0.4406)	(0.4922)	(0.4850)
BVI	0.8807 *	1.0318 *	1.0534 **	1.2472 **	1.4732 **	1.6344 ***
	(0.5330)	(0.5456)	(0.5267)	(0.5514)	(0.5889)	(0.6135)
FVI $\times$ SCP		15.4508 ***		15.4738 ***		9.1041 **
		(4.0131)		(3.2621)		(4.1116)
$\mathrm{B}\mathrm{V}\mathrm{I} \times$ SCP		−3.7916 *		−3.9651 **		−3.0843 *
		(2.2915)		(1.6684)		(1.6292)
SCP		0.2809 **		0.2184		0.2406 *
		(0.1287)		(0.1472)		(0.063)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2931	2931	2931	2931	2931	2931
No. recovery	2516	2516	2198	2198	2759	2759
Wald	1382.95 ***	1006.50 ***	1332.72 ***	1324.37 ***	680.84 ***	677.19 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the robustness checks for models examining recovery time as the measure of organizational flexibility. Model 1 shortens the observation window as in the previous test. Model 2 uses a restricted sample of manufacturing firms, and Model 3 adopts the Weibull survival model instead of the Cox model to test whether our results hold under a different functional form and distributional assumption.

In all three models, BVI continues to show a positive and statistically significant effect on recovery speed, confirming its role in enhancing organizational flexibility. The interaction $\mathrm{F}\mathrm{V}\mathrm{I}\times \mathrm{S}\mathrm{C}\mathrm{P}$ remains highly significant and positive, indicating that downstream firms benefit more from forward integration. Meanwhile, BVI × SCP remains negative across specifications, consistent with the view that upstream firms benefit more from backward integration. The robustness of these interaction effects across different models further supports our earlier findings.

4.4. Endogeneity Discussion

To address potential endogeneity concerns arising from reverse causality, we employed two types of instrumental variables (IVs). First, we introduced a dummy variable indicating whether a firm operates across multiple businesses (MB). Second, we used the industry-level average of forward and backward integration in the previous year (INFVI and INBVI) as additional instruments. Business diversification reflects the degree of vertical integration, while the lagged industry average of integration is expected to affect a firm’s integration decision.

Since the baseline model for organizational stability relies on OLS regression, we conducted a standard two-stage least squares (2SLS) analysis for the endogeneity test. In contrast, the baseline model for organizational flexibility is estimated using a Cox survival model. To address potential endogeneity in this context, we applied a control function approach: we first regressed the endogenous variable on the instruments to obtain the residuals, and then included these residuals in the second-stage Cox regression. If the residual term is not significant, this would indicate that endogeneity is not a serious concern in our sample.

Table 10. Instrumental variable method.

Dependent Variable	Fist Stage	Fist Stage	Second Stage	Second Stage
	(1a)	(1b)	(2a)	(2b)
	FVI	BVI	Severity of Loss	Recovery Time
MB	0.0037 **	0.0027 **
	(0.0016)	(0.0013)
INFVI	0.8159 ***	−0.0418
	(0.1319)	(0.0591)
INBVI	−0.1092	0.5159 ***
	(0.1000)	(0.1008)
FVI			0.5019 ***	1.1093
			(0.1420)	(1.7092)
BVI			−1.1027 ***	5.7198 *
			(0.3203)	(3.3525)
vhat_fvi				−1.1772
				(1.8007)
vhat_bvi				−4.7337
				(3.3992)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
Weak identification test (Kleibergen-Paap rk Wald F statistic)			31.037
Underidentification test (Kleibergen-Paap rk LM statistic)			28.302 (0.000)
Overidentification test (Hansen J statistic)			0.560 (0.4543)

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the results. The first-stage regressions show that both the lagged industry-level average of integration and business diversification are strongly correlated with the corresponding endogenous regressors. Moreover, the Kleibergen–Paap rk Wald F statistic exceeds the conventional threshold of 10, suggesting that the instruments are not weak. The Kleibergen–Paap rk LM statistic equals 28.32 and is significant at the 1% level, confirming that the model is not underidentified. In addition, the overidentification test (Hansen J statistic = 0.391, p = 0.5315) fails to reject the null, indicating that our instruments are valid.

The second-stage results are consistent with our baseline findings. For organizational stability, forward integration exacerbates firm losses, whereas backward integration helps mitigate losses and strengthens stability. Regarding organizational flexibility, backward integration promotes recovery, while the effect of forward integration is not statistically significant. Furthermore, when we included the first-stage residuals in the Cox regression, the residual terms were insignificant, suggesting that our original Cox model estimates are robust.

4.5. Heterogeneity Analysis

While the preceding analysis establishes the general effectiveness of vertical integration strategies in enhancing organizational resilience during the COVID-19 crisis, the strength and direction of these effects may vary depending on external environmental conditions and firm-level operational capabilities. To examine these contextual influences, we conduct a series of heterogeneity analyses based on five moderating dimensions: firm size, regional market conditions, government regulation intensity, digitalization, and supplier concentration.

4.5.1. Firm Size

To explore whether the effectiveness of vertical integration varies by firm scale, we divide the sample into large-scale and small-scale enterprises based on the classification standards issued by the National Bureau of Statistics of China [77]. Table 11 (Columns 1–2) presents the regression results on organizational stability, and Table 12 (Columns 1–2) shows the corresponding results for organizational flexibility.

For large-scale firms, the coefficient of FVI is significantly positive, indicating that forward integration exacerbated performance losses during the crisis. In contrast, the coefficient of BVI is significantly negative, suggesting that backward integration contributed to greater stability. This pattern may reflect the more complex supply chains of larger firms, which amplify both the risks and benefits of vertical coordination.

In terms of recovery speed, results from Table 10 indicate that backward integration is particularly effective for small-scale firms, with a significant positive effect on recovery time. This may be due to the simpler organizational structures of smaller firms, which allow more agile implementation of backward integration strategies and faster post-shock adjustment.

4.5.2. Regional Market Conditions

Next, we examine how regional market conditions influence the performance of vertical integration strategies. We use the 2020 China Provincial Business Environment Index released by Peking University, which includes indicators on market openness, government services, legal policy, and business culture. The sample is split into high and low-market-condition subgroups based on the median index score.

Results in Table 11 (Columns 3–4) show that for firms operating in regions with weaker market environments, the negative effect of forward integration on stability is more pronounced, while the positive effect of backward integration is stronger. This suggests that under weak institutional environments, downstream coordination (via FVI) is more difficult to execute effectively, whereas upstream control (via BVI) offers greater protection against environmental uncertainty.

Table 12 (Columns 3–4) further indicates that backward integration significantly enhances recovery in low-MC regions. In such environments, external resource access is more constrained, increasing the strategic value of internalized input control. These results highlight the contingent nature of integration effectiveness under varying institutional support.

4.5.3. Government Regulation Intensity

Then, we assess how regional government regulation intensity influences the resilience outcomes of vertical integration. During COVID-19, regions with more confirmed cases experienced stricter regulatory enforcement. To capture this variation, we construct a province-level indicator based on the frequency of a region being ranked among the top ten in new infection cases each month. Firms located in regions above the sample median on this frequency metric are classified as operating under high regulatory intensity.

Table 11. The heterogeneous regression results for organizational stability (size, market conditions, and government regulation).

Dependent Variable	Severity of Loss
	(1)	(2)	(3)	(4)	(5)	(6)
Sample	Large-Scale	Small-Scale	Higher-MC	Lower-MC	Higher-GR	Lower-GR
FVI	0.0820 **	0.0228	0.0660	0.1019 **	0.0728 *	0.0780
	(0.0359)	(0.1022)	(0.0516)	(0.0448)	(0.0422)	(0.0537)
BVI	−0.1846 ***	−0.0548	−0.1565 ***	−0.1815 **	−0.1956 ***	−0.1317
	(0.0561)	(0.0962)	(0.0602)	(0.0772)	(0.0568)	(0.0928)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2116	815	1436	1495	1569	1362
adj. R2	0.3935	0.3906	0.3753	0.4134	0.4028	0.4007

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

Table 11. The heterogeneous regression results for organizational stability (size, market conditions, and government regulation).

Dependent Variable	Severity of Loss
	(1)	(2)	(3)	(4)	(5)	(6)
Sample	Large-Scale	Small-Scale	Higher-MC	Lower-MC	Higher-GR	Lower-GR
FVI	0.0820 **	0.0228	0.0660	0.1019 **	0.0728 *	0.0780
	(0.0359)	(0.1022)	(0.0516)	(0.0448)	(0.0422)	(0.0537)
BVI	−0.1846 ***	−0.0548	−0.1565 ***	−0.1815 **	−0.1956 ***	−0.1317
	(0.0561)	(0.0962)	(0.0602)	(0.0772)	(0.0568)	(0.0928)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2116	815	1436	1495	1569	1362
adj. R2	0.3935	0.3906	0.3753	0.4134	0.4028	0.4007

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

As shown in Table 11 (Columns 5–6), firms in highly regulated regions experience stronger negative effects from forward integration and greater positive effects from backward integration on stability. This likely reflects how regulation exacerbates operational disruptions (e.g., travel bans and supplier shutdowns), making vertical integration decisions more consequential.

Interestingly, Table 12 shows that backward integration significantly improves recovery only in low-regulation regions. In highly regulated areas, the benefits of BVI may be muted by external constraints on logistics and production flexibility, which could limit a firm’s ability to leverage upstream integration advantages.

Table 12. The heterogeneous regression results for organizational flexibility (size, market conditions, and government regulation).

Dependent Variable	Recovery Time
	(1)	(2)	(3)	(4)	(5)	(6)
Sample	Large-Scale	Small-Scale	Higher-MC	Lower-MC	Higher-GR	Lower-GR
FVI	0.1754	−0.8418	0.7914	−0.5075	−0.0320	0.0976
	(0.5039)	(1.1255)	(0.7964)	(0.6218)	(0.6396)	(0.6910)
BVI	0.7468	3.1035 **	1.0330	1.0146 **	0.1500	2.1763 ***
	(0.5148)	(1.4326)	(0.9419)	(0.5109)	(0.7182)	(0.6581)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2116	815	1436	1495	1569	1362
No. recovery	1982	777	1362	1397	1475	1284
Wald	1044.29 ***	450.08 ***	773.06 ***	1871.77 ***	1492.27 ***	871.92 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: ** p < 0.05, *** p < 0.01.

Table 12. The heterogeneous regression results for organizational flexibility (size, market conditions, and government regulation).

Dependent Variable	Recovery Time
	(1)	(2)	(3)	(4)	(5)	(6)
Sample	Large-Scale	Small-Scale	Higher-MC	Lower-MC	Higher-GR	Lower-GR
FVI	0.1754	−0.8418	0.7914	−0.5075	−0.0320	0.0976
	(0.5039)	(1.1255)	(0.7964)	(0.6218)	(0.6396)	(0.6910)
BVI	0.7468	3.1035 **	1.0330	1.0146 **	0.1500	2.1763 ***
	(0.5148)	(1.4326)	(0.9419)	(0.5109)	(0.7182)	(0.6581)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	2116	815	1436	1495	1569	1362
No. recovery	1982	777	1362	1397	1475	1284
Wald	1044.29 ***	450.08 ***	773.06 ***	1871.77 ***	1492.27 ***	871.92 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: ** p < 0.05, *** p < 0.01.

4.5.4. Digitalization

To further investigate the interplay between vertical integration and another resilience mechanism—digitalization—we measure firms’ digitalization levels based on the frequency of digitalization-related terms in listed companies’ annual reports. Firms are classified into high- and low-digitalization groups according to the median index score.

Table 13 (columns 1–2) shows that the negative effect of forward integration becomes insignificant among highly digitalized firms, while it remains significant among firms with low digitalization. This suggests that digitalization provides a stabilizing force that mitigates the risks associated with FVI. By contrast, the stabilizing effect of backward integration (BVI) persists across both groups. Although the coefficient for the low-digitalization group is slightly larger than that for the high-digitalization group, the evidence indicates that BVI and digitalization play a partially complementary role in strengthening supply security.

Table 14 (columns 1–2) further shows that BVI significantly enhances organizational flexibility only among firms with low digitalization. This finding implies a substitutive relationship between BVI and digitalization in fostering flexibility: as information technologies and data-driven coordination strengthen firms’ responsiveness, the marginal contribution of BVI diminishes.

Taken together, these results suggest that digitalization has partially reduced the reliance on vertical integration for enhancing organizational flexibility, but it has not displaced the enduring value of BVI in safeguarding supply stability.

Table 13. The heterogeneous regression results for organizational stability (digitalization and supplier concentration).

Dependent Variable	Severity of Loss
	(1)	(2)	(3)	(4)
Sample	Higher-Digital	Lower-Digital	Higher-SC	Lower-SC
FVI	0.0832	0.0771 *	0.1587 ***	−0.0035
	(0.0744)	(0.0431)	(0.0553)	(0.0513)
BVI	−0.1507 **	−0.1846 ***	−0.1389 *	−0.1970 ***
	(0.0663)	(0.0629)	(0.0719)	(0.0594)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1439	1439	1459	1443
adj. R2	0.4448	0.3256	0.3901	0.4075

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

Table 13. The heterogeneous regression results for organizational stability (digitalization and supplier concentration).

Dependent Variable	Severity of Loss
	(1)	(2)	(3)	(4)
Sample	Higher-Digital	Lower-Digital	Higher-SC	Lower-SC
FVI	0.0832	0.0771 *	0.1587 ***	−0.0035
	(0.0744)	(0.0431)	(0.0553)	(0.0513)
BVI	−0.1507 **	−0.1846 ***	−0.1389 *	−0.1970 ***
	(0.0663)	(0.0629)	(0.0719)	(0.0594)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1439	1439	1459	1443
adj. R2	0.4448	0.3256	0.3901	0.4075

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

4.5.5. Supplier Concentration

Finally, we examine the interplay between vertical integration and supplier diversification. Supplier diversification is measured by supplier concentration, defined as the ratio of purchases from the top five suppliers to total annual purchases. A higher value indicates greater dependence on a few suppliers and thus lower diversification. Firms are divided into two groups based on the median value of this indicator.

Columns (3) and (4) of Table 13 show that the negative effect of forward integration (FVI) is more pronounced in the high supplier concentration group, whereas the stabilizing effect of backward integration (BVI) persists in both groups and is even stronger among firms with low supplier concentration. This suggests that supplier diversification partly substitutes for the risk-mitigation function of FVI, but does not diminish the stability-enhancing advantage of BVI. In other words, BVI and supplier diversification play complementary roles in strengthening supply security.

Moreover, Columns (3) and (4) of Table 14 further indicate that the positive effect of BVI on organizational flexibility is significant only in the high supplier concentration group, but not in the low concentration group. This finding implies that supplier diversification itself provides a strong source of flexibility, thereby reducing the marginal contribution of BVI in settings with lower concentration.

In summary, the relationship between supplier diversification and vertical integration in shaping resilience is context-dependent: with respect to stability, BVI and supplier diversification are complementary, whereas in terms of flexibility, they exhibit a degree of substitution.

Table 14. The heterogeneous regression results for organizational flexibility (digitalization and supplier concentration).

Dependent Variable	Recovery Time
	(1)	(2)	(3)	(4)
Sample	Higher-Digital	Lower-Digital	Higher-SC	Lower-SC
FVI	0.3339	0.0901	−0.3963	0.6738
	(0.6689)	(0.5729)	(0.6851)	(0.6239)
BVI	1.1076	1.1037 *	1.6813 **	0.6969
	(0.7491)	(0.6673)	(0.6667)	(0.6825)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1439	1439	1459	1443
No. recovery	1378	1334	1362	1376
Wald	1555.41 ***	907.94 ***	868.73 ***	1513.93 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

Table 14. The heterogeneous regression results for organizational flexibility (digitalization and supplier concentration).

Dependent Variable	Recovery Time
	(1)	(2)	(3)	(4)
Sample	Higher-Digital	Lower-Digital	Higher-SC	Lower-SC
FVI	0.3339	0.0901	−0.3963	0.6738
	(0.6689)	(0.5729)	(0.6851)	(0.6239)
BVI	1.1076	1.1037 *	1.6813 **	0.6969
	(0.7491)	(0.6673)	(0.6667)	(0.6825)
Controls	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes
Time-fixed	Yes	Yes	Yes	Yes
N	1439	1439	1459	1443
No. recovery	1378	1334	1362	1376
Wald	1555.41 ***	907.94 ***	868.73 ***	1513.93 ***

Notes. Numbers in parentheses are the robust standard deviation, and the statistical significance is denoted by: * p < 0.1, ** p < 0.05, *** p < 0.01.

5. Conclusions and Managerial Implications

5.1. Discussion

Ensuring resilient and sustainable supply chain operations amid systemic disruptions remains a critical challenge in volatile global environments. Our findings reveal important asymmetries in the effects of vertical integration on organizational resilience. Forward integration is associated with lower organizational stability, likely due to constraints on downstream operations during crises, yet it accelerates recovery for firms closer to end markets by improving demand visibility and customer access. Backward integration provides dual benefits, enhancing both stability and flexibility by securing input availability, improving supplier coordination, and mitigating upstream volatility. These findings remain robust after addressing potential endogeneity concerns and conducting a series of robustness checks.

Building on these baseline results, our heterogeneity analyses delineate the boundary conditions under which vertical integration strategies are more or less effective, as well as their interactions with other resilience mechanisms. We find strong contextual dependence: the negative effects of forward integration are amplified in larger firms, weaker market environments, and regions with stricter regulatory intensity, whereas the stabilizing benefits of backward integration are more pronounced in such settings. Conversely, the positive impact of backward integration on recovery speed is stronger among smaller firms, in weaker market environments, and in less regulated regions. These results underscore the importance of situating integration strategies within broader organizational and institutional contexts, rather than treating them as universally applicable.

Moreover, our findings highlight the nuanced role of alternative resilience mechanisms. Both digitalization and supplier diversification act as substitutes and complements to vertical integration, depending on the resilience dimension under consideration. Specifically, they substitute for vertical integration in fostering flexibility—by providing responsiveness through digital tools and relational redundancy—but complement vertical integration in securing stability, where backward integration continues to play a critical role in safeguarding input availability. Taken together, these insights suggest that resilience is not achieved through a single dominant mechanism, but rather through a portfolio of strategies whose effectiveness depends on their mutual alignment and contextual fit.

These results extend prior studies by highlighting that vertical integration is not a one-size-fits-all strategy; its effectiveness depends on the strategic fit between integration direction and supply chain position. Compared with classical views that primarily emphasize forward integration’s market advantages or backward integration’s input security, our findings show nuanced trade-offs under high-uncertainty conditions.

5.2. Conclusions

This study examines how forward and backward vertical integration influence organizational resilience, using firm-level data from 2931 Chinese publicly listed companies. Our results demonstrate that forward integration can hinder stability but enhance recovery for firms near end markets, whereas backward integration strengthens both stability and flexibility. These findings emphasize that the effectiveness of vertical integration depends on the alignment between integration direction and supply chain context. Heterogeneity analyses further suggest that the impact of integration strategies on organizational resilience is shaped by external environmental conditions and other complementary resilience mechanisms. By distinguishing between forward and backward integration and considering supply chain position, this study offers theoretical and managerial insights for developing resilient and sustainable supply chain strategies amid systemic disruptions.

5.3. Managerial Implications

This study offers several actionable insights for managers confronting turbulent supply chain environments. First, managers should determine the direction of integration based on their resilience priorities. Our findings indicate that vertical integration can effectively mitigate supply chain disruptions, but its impact is asymmetric: backward integration enhances both stability and flexibility, whereas forward integration primarily accelerates recovery but may undermine stability. This suggests that integration should not be pursued indiscriminately. Managers need to clarify whether their foremost objective is to sustain operational stability or to achieve rapid recovery, and align their integration choices accordingly. For instance, firms operating in volatile demand environments and seeking stable operations should exercise caution with extensive forward integration, as it may exacerbate instability. By contrast, firms competing on responsiveness to shifting market demand may still find forward integration advantageous.

Second, integration strategies should be tailored to a firm’s supply chain position and organizational context. The effectiveness of integration depends not only on where a firm sits in the supply chain but also on its broader operating environment. Upstream firms can benefit from backward integration to secure input availability and stabilize operations, whereas downstream firms closer to end markets may rely on forward integration to accelerate recovery. Additionally, firm size, market strength, and regulatory intensity shape integration outcomes. Large firms in highly regulated environments may gain disproportionately from the stabilizing role of backward integration, while smaller firms can use it to enhance recovery agility. These findings highlight the need for context-sensitive strategies rather than uniform prescriptions.

Finally, managers should build resilience portfolios rather than relying on a single mechanism. Vertical integration should be complemented with alternative resilience levers such as digitalization and supplier diversification. Our results suggest that these mechanisms substitute for integration in enhancing flexibility by providing real-time visibility and relational redundancy, while at the same time complementing integration in reinforcing stability. Accordingly, integration should be viewed as a cornerstone for safeguarding critical resources, while digital technologies and diversified supplier bases provide the responsiveness needed to adapt under uncertainty.

6. Limitations and Future Research

This study has several limitations that open avenues for future inquiry. First, while COVID-19 provides a powerful context for studying systemic disruption, it is distinctive due to its health-related impacts and the rapid, large-scale policy interventions it prompted. Unique characteristics such as government-mandated lockdowns, workplace safety protocols, and resumption-of-production policies may have influenced the observed effects of forward and backward integration, thereby limiting the generalizability of our conclusions. Future research should examine whether the observed patterns hold in other disruptions—such as geopolitical conflicts, environmental crises, or digital outages—to test the robustness of our findings. Second, although our dataset spans diverse firms and industries, it focuses on Chinese listed companies. The results may therefore be context-dependent, reflecting China’s institutional and cultural environment, where vertical organizational models and collective goals often take precedence over individual initiative. Cross-country comparisons could offer insights into the role of institutional and regulatory environments. Third, our conceptualization of resilience—stability and flexibility—captures core recovery dynamics but omits other dimensions such as adaptability or strategic renewal. Moreover, while we complement stock market–based measures with operational indicators such as downside risk of ROA and post-crisis operating efficiency, these proxies still do not fully capture firms’ direct operational responses (e.g., production recovery, order fulfillment, or logistics continuity). Future research could incorporate more granular operational data to provide a more comprehensive assessment of organizational resilience. Our measures of vertical integration and supply chain positions are constructed from industry-level input–output (I-O) data. While this method is transparent and widely adopted, it assumes that industry-level interdependencies sufficiently capture firm-level integration. This assumption may not hold for diversified or conglomerate firms and may oversimplify the networked nature of supply chains. Future research could integrate firm-level data—such as intra-firm transaction ratios, textual analysis of annual reports, network centrality, or value-added complexity measures—to improve the construct validity of the network once appropriate data become available.

To extend this work, future studies could investigate how vertical integration strategies perform across different disruption types and under what conditions forward or backward integration is more effective. Expanding the resilience framework to include growth, innovation, and transformation would offer a fuller picture of post-crisis trajectories. Scholars may also construct more comprehensive indices of resilience that integrate financial, operational, and organizational measures. Qualitative case studies of firms that pursued integration before or during crises could help unpack contextual enablers and implementation challenges.

Additionally, future research should explore alternative resilience mechanisms beyond structural integration. Digital coordination technologies, platform-based supply ecosystems, and dynamic capabilities such as adaptive learning or real-time sensing may serve as either complements or substitutes to vertical integration. Exploring these tools can deepen our understanding of how firms can build resilience in more fluid, digitally enabled networks.

Collectively, these directions can help build a more holistic and forward-looking understanding of how firms can strengthen their resilience and thrive amid increasing uncertainty and systemic disruption.