Investment Efficiency–Risk Mismatch and Its Impact on Supply-Chain Upgrading: Evidence from China’s Grain Industry

Abstract

This study examines how investment efficiency and risk jointly shape sustainable grain supply-chain upgrading. Using firm-level panel data for 25 listed grain supply-chain firms in China from 2015 to 2023, this study examines efficiency–risk structures and their heterogeneity across upstream, midstream, and downstream segments. A three-stage data envelopment analysis (DEA) is applied to measure investment efficiency while controlling for environmental heterogeneity and statistical noise, and a multidimensional investment risk index is constructed using principal component analysis (PCA), with an emphasis on sustainability metrics. The results reveal a clear supply-chain gradient: downstream firms exhibit the highest mean third-stage investment efficiency (crete = 0.633) and scale efficiency (scale = 0.634), midstream firms are intermediate (crete = 0.308; scale = 0.326), and upstream firms remain lowest (crete = 0.129; scale = 0.138). This ordering is also visible year by year, while risk profiles indicate higher exposure upstream and pronounced volatility midstream. Efficiency decomposition shows that upstream inefficiency is mainly driven by scale inefficiency rather than insufficient pure technical efficiency. Overall, efficiency–risk mismatch—manifested as persistent low scale efficiency and elevated risk exposure in upstream, volatility in midstream, and stability in downstream—constitutes a key micro-level barrier to long-term and resilient upgrading. The study thus offers policy-relevant insights for segment-specific interventions that align with sustainable agricultural development: facilitating land consolidation and integrated risk management for upstream scale inefficiency, promoting supply-chain finance and digital integration for midstream risk volatility, and leveraging downstream stability to drive coordinated upgrading and sustainable value creation through market-based incentives.

1. Introduction

Global grain supply chains face compounded shocks from climate change, geopolitical tensions, and market volatility, making upgrading a critical lever for sustainable food security and long-term agricultural competitiveness [1,2]. In China, the grain supply chain encompasses a vast array of firms ranging from primary production to retail. The investment decisions of these firms—specifically how capital, labor, and technology are allocated across different segments—fundamentally shape the chain’s structure and long-term evolutionary path [3,4]. In this context, the investment efficiency reflects firms’ ability to convert inputs into value, while the investment risk captures the financial and operational uncertainties surrounding these decisions. When efficiency and risk are examined in isolation, however, the resulting diagnosis may be incomplete. Their joint interaction directly influences firms’ incentives to reallocate resources toward higher-value activities, which lies at the core of supply-chain upgrading. Yet this efficiency–risk interaction and its potential to generate upgrading bottlenecks remain inadequately understood. Nevertheless, even with the increasing amount of research concerning efficiency and risk in investment, empirical research that addresses both efficiency and risk at the same analytical level and makes comparatively uniform analysis across segments of the supply chain is rare. The available literature has also been either concerned with enhancing the technical performance within an agricultural or manufacturing firm [5,6,7] or in understanding how the financial and operational risks of firms and their financial results are linked to each other [8], with the two strands of study normally emerging independently of each other. Simultaneously, macro-level policy perspectives or industry-structure approaches [9,10] have been by far the predominant point-of-view in agricultural supply-chain upgrading studies, and there have been fewer applications of firm-level microdata to describe segment-specific efficiency gaps and risk or exposure. Consequently, there is a significant shortage of systematic data on the effect of efficiency–risk mismatches in determining supply-chain upgrading dynamics. This gap impairs theoretical knowledge and also restricts the structuring of more specific and efficient policies aimed at supply-chain upgrading.

Crucially, investment efficiency and investment risk do not operate in isolation; they jointly shape corporate investment behavior. Much of the existing literature, however, examines these dimensions separately, which may yield an incomplete diagnosis of upgrading constraints. A segment may display high efficiency yet remain underinvested when it is associated with elevated financial or operational risk. Conversely, a relatively low-risk segment may fail to attract resources if it suffers from persistent scale inefficiency. Such misalignment between efficiency and risk distorts firms’ incentives for resource reallocation. This efficiency–risk mismatch suggests that the key constraint on supply-chain upgrading may lie not in low efficiency or high risk per se but in their interaction at the segment level. When the expected efficiency gains fail to compensate for the perceived risk exposure, firms are less likely to commit the long-horizon investments required for process and functional upgrading, potentially locking the supply chain into a suboptimal structure.

Against this background, this study introduces introduce the concept of efficiency–risk mismatch to examine micro-level constraints on grain supply-chain upgrading. Here, upgrading is understood as segmental movement toward higher value-added and more resilient value creation, primarily captured in this study through process upgrading and functional upgrading. In this study, supply-chain upgrading is not modeled as a directly observed dependent variable; rather, it is interpreted through segment-level patterns of investment efficiency, investment risk, and their alignment or mismatch over time. Accordingly, this paper address three research questions:

RQ1: Do upstream, midstream, and downstream firms exhibit systematic differences in investment efficiency?

RQ2: Do these segments differ in investment risk, and how does the risk structure—including its level and volatility—vary across segments?

RQ3: How does the alignment or mismatch between efficiency and risk at the segment level shape the potential for coordinated supply-chain upgrading?

To address these questions, this study adopts an integrated efficiency–risk assessment approach that jointly characterizes firms’ investment performance and risk exposure across upstream, midstream, and downstream segments. By mapping segment-specific efficiency–risk configurations and their evolution over time, the analysis identifies where persistent mismatches weaken the incentives and capacity for sustained factor reallocation, thereby constraining supply-chain upgrading. This approach enables a segment-comparable diagnosis of micro-level bottlenecks and supports targeted policy-relevant interventions. There are three key contributions in this research as compared to previous research. First, it combines investment efficiency and risk in a single assessment scheme and formally addresses the concept of the efficiency–risk mismatch. Second, it offers a new micro-level explanation of the upgrading of agricultural supply chains because it highlights behavior at the firm level. Third, it provides a comparative efficiency–risk evaluation level across a supply-chain segmentation, which provides quantitative evidence that can be used to support more focused policy interventions.

The remainder of the paper is organized as follows. Section 2 reviews the related literature; Section 3 describes the research design, including efficiency measurement, risk index construction, and data sources; Section 4 presents the empirical results; Section 5 provides a discussion of key findings and policy implications; and Section 6 concludes with a summary of the study and its contributions.

2. Literature Review

2.1. The Theoretical Basis and Measurement Methods of Enterprise Investment Efficiency

At the firm level, investment efficiency reflects how effectively firms allocate capital and other inputs to generate output and value under given constraints [11]. In grain supply chains, where segments face distinct asset specificity and exposure to shocks, investment efficiency provides a segment-comparable lens to diagnose the process-upgrading capacity. In particular, it summarizes both internal governance/managerial discipline and external operating constraints, thereby offering a micro-foundation for process upgrading and functional upgrading.

The efficiency of corporate investment is the subject of research traceable to the input–output theory that revolves around measuring whether companies could efficiently utilize their inputs of capital, costs, and labor for productive outputs [11]. DEA and stochastic frontier analysis (SFA) have become the two most popular methodological frameworks in the global literature in the field of empirical applications. DEA is especially appreciated, because it can measure the relative efficiency without imposing a priori assumptions on the production function and, therefore, has been extensively used in manufacturing, food processing, energy, and other industries [12,13,14,15]. The DEA models, however, cannot identify the managerial inefficiency and exogenous environmental heterogeneity. In order to overcome this drawback, Fried et al. [16] established the three-stage DEA model that uses the SFA regressions to eliminate the environmental impacts and statistical errors in the input slacks, increasing the comparability of efficiency estimates among firms and industries. Due to these benefits, the three-stage DEA method has been increasingly implemented in the research of corporate efficiency in investment or other similar topics, such as green efficiency and digital efficiency [17,18,19,20]. In addition, the evidence suggests that digital technology management improves resource efficiency in agricultural production and may, therefore, support resilience-oriented upgrading by reducing input redundancy and enhancing adaptive capacity [21].

Regarding the determinants of corporate investment efficiency, the literature has generally approached the analysis of firm behavior in two extensive dimensions: internal governance systems and external operating conditions [22,23,24]. This analytical framework has been applied to other categories of firms within the context of the grain supply chain, where empirical studies have evaluated efficiency performance and its drivers at the production stage, processing stage, distribution stage, and retail stage. In particular, the research on the upstream segments, like cultivating and primary processing, has highlighted the implications of how the natural conditions, production cost, and input constraint influence the efficiency of the investment made by the firms [25,26,27]. Conversely, studies have been conducted to identify the significance of mid-stream operations such as processing and storage–logistics to making efficiency gains, and the results in this area have emphasized economies of scale, technological changes, and structuring of production [28,29]. In the downstream segments that touch on branding and retail, the literature has put more emphasis on market-channel development, consumer preference, and the ability to manage brands as key determinants of investment returns and efficiencies of operations [30,31].

However, existing micro-level evidence remains fragmented in two respects. First, many studies examine investment efficiency within one supply-chain stage (e.g., production, processing, storage/logistics, branding/retail), with limited effort to systematically compare upstream, midstream, and downstream firms from a vertically integrated perspective. Second, the literature on efficiency and the literature on risk have largely evolved along separate tracks, and relatively few studies integrate investment efficiency with investment risk within a unified analytical framework. As a result, the micro-level mechanisms through which efficiency–risk alignment (or mismatch) shapes firms’ investment behavior and, ultimately, coordinated upgrading potential along the grain supply chain are not yet well understood. This gap motivates an integrated efficiency–risk perspective and segment-comparable analysis across upstream, midstream, and downstream firms.

2.2. Multi-Dimensional Construction and Comprehensive Assessment of Enterprise Investment Risks

Corporate investment risk is multidimensional, dynamic, and systemic by nature. Its sources include not only general financial risks, including capital structure weakness, liquidity measures, and earnings variations, but also those peculiar to agricultural and food-related companies, like price fluctuations, seasonality, unpredictability of supply and seasonal demand, and natural and biological risks [32,33,34,35]. Conventional research has mainly used one financial variable to rank firms’ risk exposure, which is either leverage ratios or liquidity ratios. But the accumulating literature reveals that it is not possible to effectively capture several aspects of risk, including leverage intensity, solvency capacity, earnings stability, and quality of operations [36,37,38] simultaneously.

In order to overcome the issue of the high inter-correlation of risk indicators and subsequent “indicator fragmentation”, recent studies have been using PCA to generate composite risk indices to a larger extent [39,40,41]. Through the extraction of shared latent factors of various risk factors, PCA minimizes the dimension and explanatory variance of the risk. It has, therefore, exhibited evident benefits regarding empirical research on financial risk, systemic risk, and supply-chain risk [42,43,44].

2.3. The Driving Mechanism for the Upgrading of the Grain Supply Chain

Supply-chain upgrading is most often analyzed in the realms of industrial economics and development economics in terms of global value chain (GVC) analysis. The conceptualizations of the seminal literature of upgrading present upgrading as a network of mutually connecting pathways, the most prominent of which are process upgrading, product upgrading, and functional upgrading, where the firms advance their place in the value chain through the improvement of technological advancement, streamlined organizational structure, or an enhancement of relevant capacities [45,46,47,48]. Within this framework, supply-chain upgrading is generally considered a key tool in adding value through enhancing industrial competitiveness and has been used extensively in empirical studies of manufacturing, agri-food processing, and the food industry, in general [49,50,51].

Studies examining the supply-chain upgrading drivers have also been conducted mainly on macro and meso levels, specifically technological innovation, government policy, foreign direct investment, industrial agglomeration, and institutional environments [49,52,53,54,55,56]. Even though these studies present useful contributions on the upgrading pathways, in most instances, they interpret firm behavior as an exogenous reaction, and consequently, they do not capture the agency of the firm as the major driver of the upgrading process. At the micro-level, it is up to the companies to invest in their supply-chain upgrading. Working capital, labor, and technological inputs are distributed among various supply-chain units by the firms and, thus, directly define the structure of the channel and its evolutionary path. An impressive amount of research suggests that the efficiency of firms in terms of productivity and investment is the basic precondition for value-chain upgrading [56,57,58]. In industries where resource limits are constraining, like in the grain supply chain, where the risk exposure is high, however, the willingness to invest in more or upgrading segments by firms is not only limited by the aspects of efficiency but also largely limited by the multidimensional financial and operational risks [59,60].

Based on the argumentation above, it is claimed in this paper that joint structure of investment efficiency at the firm and investment risk is the determinant of different segments of a supply chain to be more attractive to investment and subsequently the need to reallocate factors to influence the course of upgrading a supply chain. When a particular segment simultaneously has high efficiency and manageable risk, it is more attractive to get the endogenous momentum of long-term capital, labor, and technological inputs. Recent agribusiness research also emphasizes that strategic management and ESG-oriented impact assessment can shape sustainable upgrading pathways by strengthening governance discipline and aligning firms’ investment decisions with long-term value creation [61]. Conversely, an efficiency–risk mismatch can lead to the distribution of factors in the economy being distorted, with potentially upgrading segments being continuously underinvested. Although the relevance of this mechanism has been recognized, the prevalent studies have provided no systematic evidence of efficiency and risk integration within a single framework and compared the structural heterogeneity of these variables across the vertical supply chain. Accordingly, based on GVC upgrading theory and adopting a firm investment–behavior perspective, this research creates an analytical framework in connection to investment efficiency, investment risk, and supply-chain upgrading. This model allows a methodical description of the efficiency–risks segment-specific patterns and provides new empirical findings on the micro-level efficiency towards upgrading dynamics in the grain supply chain.

3. Research Design

3.1. Research Method

Following the approach outlined in the Introduction, this study implements a two-step empirical design to quantify investment efficiency, construct a composite investment-risk index, and then examine segment-specific efficiency–risk configurations using firm-level panel data.

3.1.1. Three-Stage DEA Method

This study takes a three-stage DEA approach to evaluate the efficiency of investments made by companies in various strands of the Chinese grain supply chain. In contrast to traditional models of DEA, which define all the deviations of the efficiency frontier by managerial inefficiency, the three-stage DEA breaks down the efficiency deviations into three categories, namely, managerial inefficiency, external environmental factors, and noise [16]. This has been important in the case of supply-chain studies, as firms in the upstream, midstream, and downstream of the grain supply chain are often highly heterogeneous in terms of financing capacity, market rivalry, support of policies and economies of scale. In the absence of isolating these environmental influences, there would be systematic bias with regard to differences in efficiency between firms at various stages of the supply chain, which results in the wrongful depiction of the actual managerial performance by the firms. The approach has three distinct stages, as discussed below.

Stage 1: Traditional DEA Model. The first stage of the study involves the application of an input-based DEA model to estimate the initial efficiency measures of every sample firm. The stage offers preliminary efficiency estimates and determines the input slack values. These slack values are the total possible items of inputs, which are a combination of managerial inefficiency, environmental factors, and noise.

Stage 2: SFA Model. The second stage views the slack variables with regard to efficiency. Such slack variables are mainly due to managerial inefficiency, environmental, and statistical noise, which is a reflection of the initial efficiency scores. The focus of this stage is to break these slack variables into three effects, namely, the environmental effect, managerial effect, and noise effect. The model is to be specified in the following way:

$${\mathrm{S}}_{\mathrm{n}\mathrm{i}}={\mathrm{f}}^{\mathrm{n}}({\mathrm{Z}}_{\mathrm{i}};{\mathsf{\beta }}^{\mathrm{n}})+{\mathrm{v}}_{\mathrm{n}\mathrm{i}}+{\mathrm{u}}_{\mathrm{n}\mathrm{i}},$$

where ${\mathrm{S}}_{\mathrm{n}\mathrm{i}}$ represents the slack value for firm $\mathrm{i}$ in the $\mathrm{n}$th input; ${\mathrm{f}}^{\mathrm{n}}({\mathrm{Z}}_{\mathrm{i}};{\mathsf{\beta }}^{\mathrm{n}})$ is the deterministic impact of environmental variable $\mathrm{Z}$ on input slack, typically modeled as a linear function ${\mathrm{Z}}_{\mathrm{i}}{;\mathsf{\beta }}^{\mathrm{n}}$; ${\mathrm{v}}_{\mathrm{n}\mathrm{i}}$ denotes statistical noise, assumed to follow a normal distribution $\mathrm{N}(0,{\mathsf{\sigma }}_{\mathrm{u}\mathrm{n}}^2)$; and ${\mathrm{u}}_{\mathrm{n}\mathrm{i}}$ signifies managerial inefficiency, assumed to follow a truncated normal distribution ${\mathrm{N}}^{+}(\mathsf{\mu },{\mathsf{\sigma }}_{\mathrm{u}\mathrm{n}}^2)$.

The purpose of the SFA regression is to eliminate the impact of environmental factors and random disturbances on the measurement of technical efficiency, ensuring that all decision-making units are evaluated under the same external conditions. Fried et al. (2002) [16] proposed an adjusted formula:

$${\mathrm{X}}_{\mathrm{n}\mathrm{i}}^{\mathrm{A}\mathrm{d}\mathrm{j}}={\mathrm{X}}_{\mathrm{n}\mathrm{i}}+[\mathrm{m}\mathrm{a}\mathrm{x}(\mathrm{f}({\mathrm{Z}}_{\mathrm{i}};{\widehat{\mathsf{\beta }}}_{\mathrm{n}}))-\mathrm{f}({\mathrm{Z}}_{\mathrm{i}};{\widehat{\mathsf{\beta }}}_{\mathrm{n}})]+[\mathrm{m}\mathrm{a}\mathrm{x}({\mathrm{v}}_{\mathrm{n}\mathrm{i}})-{\mathrm{v}}_{\mathrm{n}\mathrm{i}}],$$

where ${\mathrm{X}}_{\mathrm{n}\mathrm{i}}^{\mathrm{A}\mathrm{d}\mathrm{j}}$ is the adjusted input; ${\mathrm{X}}_{\mathrm{n}\mathrm{i}}$ is the original input; $[\mathrm{m}\mathrm{a}\mathrm{x}(\mathrm{f}({\mathrm{Z}}_{\mathrm{i}};{\widehat{\mathsf{\beta }}}_{\mathrm{n}}))-\mathrm{f}({\mathrm{Z}}_{\mathrm{i}};{\widehat{\mathsf{\beta }}}_{\mathrm{n}})]$ adjusts for external environmental factors; and $\left[\mathrm{m}\mathrm{a}\mathrm{x}({\mathrm{v}}_{\mathrm{n}\mathrm{i}})-{\mathrm{v}}_{\mathrm{n}\mathrm{i}}\right]$ ensures that all decision-making units are placed under the same external conditions.

Stage 3: Adjusted DEA Efficiency Estimation. Lastly, the same DEA-BCC model is rerun using adjusted input data ${\mathrm{X}}_{\mathrm{n}\mathrm{i}}^{\mathrm{A}\mathrm{d}\mathrm{j}}$ and original output data. The obtained efficiency values at this stage will reflect purified measures of managerial efficiency. These scores are used to show how much a firm is capable of transforming its inputs, taking into account the heterogeneity of fields of operation and random statistical noise, and with this, it is a better basis on which to compare the intrinsic investment efficiency of the upstream, midstream, and downstream links of the grain supply chain.

As conventional DEA requires non-negative input and output variables, special treatment is applied to the output variable scale output, defined as the operating revenue plus operating profit. In a small number of firm-year observations, this variable takes negative values due to temporary operating losses. To ensure the feasibility of the DEA estimation while preserving relative performance comparisons, a translation transformation is applied by adding a constant equal to the absolute value of the minimum observed output plus a small positive constant, so that all output values are strictly non-negative. This linear translation does not affect the relative distances between decision-making units and therefore does not distort the efficiency rankings, as DEA efficiency scores are invariant to such monotonic shifts. Similar treatments of negative outputs have been widely adopted in DEA applications dealing with financial performance data.

3.1.2. PCA Method

In developing a composite index of the investment risk of companies within the grain supply chain, this paper uses PCA. The problem of investment risk is multidimensional in nature, as it includes financial stability of the firm, operational pressure, and earnings volatility. A single financial measure of risk usually measures only a single dimension of risk, and the ones that are measured are frequently highly multicollinear. PCA, as a powerful dimensionality reduction method, captures orthogonal principal components of a group of highly correlated variables, in a systematic presentation of the commonality of risk profile structures across firms [38].

As far as operational steps are concerned, PCA standardizes all risk-related indicators first to remove the effect produced by dimensional differences on the results. The data are then normalized, and a correlation matrix is constructed, which undergoes eigenvalue decomposition to derive the eigenvalues and eigenvectors of each principal component. Based on the criterion proposed by Kaiser [62] and the cumulative variance contribution principle, the key elements to describe the larger part of the original information are chosen. Lastly, the chosen major factors are multiplied and added up following their weightings, which generates a composite risk score, reflecting the total risk level of each company.

The risk index, which was developed based on PCA, has a number of benefits. First, it includes various risk facets, including the capital structure, the capacity to repay debt, and profitability, as well as asset turnover, in one measure, enabling the comparison of risk levels among various segments of the supply chain. Second, PCA removes the multicollinearity issue between the indicators, and so the measurement of risk becomes more stable. Third, the composite index of risk supplements the DEA index of investment efficiency, allowing for further investigation of the effect of efficiency–risk compliance on the upgrading of supply chains, to contribute to this paper.

Prior to PCA, all risk indicators were standardized to z-scores to eliminate scale effects. The suitability of the data for factor analysis was confirmed by a Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity. In determining the number of principal components to retain, a dual ex ante criterion was specified. Specifically, principal components were retained if they satisfied either (i) the Kaiser rule (eigenvalue > 1) or (ii) contributed to achieving a cumulative explained variance of at least 80%, ensuring sufficient coverage of the underlying risk information.

3.1.3. Efficiency–Risk Quadrant Classification

In order to examine the nature of the joint distribution of the investment efficiency and investment risk by various segments of the grain supply chain, this research uses a two-dimensional quadrant-based classification framework. Based on the multidimensional performance evaluation scheme that is widely applicable in both operations management and supply-chain management [63], the study positions the firms in a two-dimensional coordinate structure characterized by the DEA investment efficiency and the composite risk score produced by PCA. In this way, the coincidence/inconsistency between efficiency and risk can be determined, and the effect of that on the upgrading of the supply chain can be considered. In particular, all sample firms are classified according to the mean investment efficiency and the investment risk, which forms the threshold of classification and the coordinate plane consists of four quadrants, as indicated in Figure 1.

Quadrant 1: High Efficiency and High Risk. Firms in this quadrant exhibit high investment efficiency but are also exposed to higher levels of risk. This typically indicates that these firms employ aggressive capital strategies or face significant external pressures, resulting in unstable efficiency improvements.

Quadrant 2: High Efficiency and Low Risk. Firms in this quadrant demonstrate both high efficiency and low risk. These firms are the most resilient and capable of driving supply-chain upgrading, serving as the core source of momentum within the supply chain.

Quadrant 3: Low Efficiency and High Risk. This quadrant represents the most concerning area, characterized by the coexistence of low efficiency and high risk. Firms in this quadrant are the weak links in the supply chain and potential points of failure, susceptible to disruption under external shocks.

Quadrant 4: Low Efficiency and Low Risk. Firms in this quadrant exhibit low risk but insufficient efficiency, often adopting conservative or risk-averse strategies. However, their limited innovation capacity may hinder the pace of supply-chain upgrading.

The main direction of analytical approach of this research is to project all the sample firms on this matrix and differentiate them based on their location in various supply-chain segments. The resulting scatter plot will provide a graphical depiction of the clustering of each segment and provide an analysis of the structural mismatches of the supply chain, which will directly answer the research questions.

3.2. Variable Definition and Source

3.2.1. Sample Selection

This study examines investment efficiency across different segments of the grain supply chain using A-share listed firms in China as the research sample. Firm-level financial and market data are obtained from the Wind and CSMAR databases and cross-validated with companies’ annual reports. The initial sample includes all A-share listed firms over the period 2015–2023. Following the Guidelines for the Industry Classification of Listed Companies, firms operating in sectors closely related to the grain industry—namely “Agriculture, Forestry, Animal Husbandry, and Fishery” (A01), “Agricultural and By-Product Processing” (C13), and “Food Manufacturing” (C14)—are selected to form the core sample of the grain supply chain.

Firms are segmented according to their dominant grain-related business activities, as described in annual reports and segment disclosures. The “main business share exceeding 50%” criterion is used as a practical screening rule to identify a firm’s primary position along the supply chain, rather than as an exact revenue-share calculation. This classification is further cross-validated using Wind industry classification codes to ensure consistency and replicability. Then, the companies that were marked as ST or *ST during the time interval were excluded to reduce the effects of financially anomalous companies. Incomplete observations regarding important monetary variables were also eliminated. The risk index calculations and balance of the panel data are to be ensured; this is why firms listed later than 2015 were excluded. Therefore, the overall dataset consists of a balanced panel of 25 firms, with a year of 2015 to 2023, and 225 firm-year observations. The list of the 25 sample firms and their corresponding supply-chain segment classification is provided in Table S1 in the Supplementary Materials.

To enable the comparative analysis between the segments of the supply chains, the sample firms would be operationalized into three segments of the grain supply chain, upstream, midstream, and downstream, according to the GVC theory and functional attributes of the core value-creating activities of the firms [46,64]. In particular, the upstream segment consists of companies that are either involved in grain cultivation, breeding, the production of agricultural inputs, including fertilizers and pesticides, or primary processing, which is a modification of the physical shape of products only. Midstream segment includes companies engaging in grain storage, logistics, milling, oil extraction, and feed production, among other advanced processing and distribution services, where the major focus is to add value to products by intensifying processing and relocating the location spatially. The downstream segment involves companies that deal directly with end consumers, including branded food production, snack food processing, retail and food service supply-chain management, where value realization depends mostly on brand building, developing the channel and advancing on final goods.

3.2.2. Variable Definitions

• Selection of Input and Output Variables

Using the available literature on economic principles of agriculture and the efficiency of industries as well as taking into account the realities operating in the grain industry among listed companies, this paper develops an input indicator system based on three dimensions: capital, costs and labor. An output variable will be taken as a scale output, and the size of the firm and its performance in operations are reflected, which is a more complete parameter to individualize the efficiency of investment in grain industry firms [65,66,67,68]. The input variables will be capital investment, which is directly proportional to the accumulation of long-term asset and new productive investments, the operating costs input denoting the direct cost of production incurred by the firm to earn revenue, period cost input denoting the consumption of indirect resources in activities, that is, sales, management, and R&D, and lastly, labor input, which captures the scale of human resource in the firm. The output variable is the scale output, which not only indicates the level of scale of operation and potential level of functionality of the firm in the markets but also represents the overall outcome of the ultimate profits of the firm in the core of its operations, which is a holistic expression of the level of value creation in the firm.

• Environmental Variable

Based on mainstream research paradigms in the analysis of corporate efficiency and productivity and taking into account the circumstances of the unique Chinese institutional and market organization, environmental variables, like the size of firms, the year of IPO, per capita GDP, ownership concentration (OWNC), and market competition are introduced in the second-stage regression of SFA, to control the exogenous variables [69,70]. The size of firms indicates how the firm is endowed with resources and managerial powers, which are generally thought to play a large role in scale effects and technological choices in the production process. It has been widely researched on its relations with efficiency in agricultural enterprises [71]. The listing year is used as a measure of the maturity of a firm in capital market and stability of its governance, which in most cases influences the managerial effectiveness and strategic decision-making of the firm. The business location of the firm in relation to its external conditions has critical importance in influencing the operational efficiency of a firm, whose regional per capita GDP measures the level of economic development of the firm in which it is located. The concentration of ownership, being one of the characteristics of corporate governance system, is important in the areas of internal control, investment policy, and efficiency in distributing resources [72]. Lastly, the market rivalry indicates external rivalry pressures on the firm, which shape the decision and performance of the firm in relation to its behavior patterns. The stiffer the competition, the higher the probability of the firms being pushed into enhancing their operational efficiency [73]. Having these environmental variables included in the three-stage DEA model also eradicates biases in firm efficiency due to external environmental differences, thus giving more precise estimates of pure technical efficiency.

• Investment Risk Variable

In order to formulate an all-encompassing and sound composite investment risk index, this paper constructs a multidimensional risk integration model building on contemporary financial theory, including four main dimensions [37]. The leverage risk is calculated on the basis of the asset–liability ratio (ALR) and the equity multiplier (EM) that assess the level of emissions of the financial structure of firms and their possible pressure to repay the debt. These two are basic to the measurement of financial risk and probability of bankruptcy, where ALR is the ratio between the total debt and total assets, and EM is the ratio between the assets and equity magnification [74]. The current ratio (Liquid) and cash flow ratio (CashFlow) are used to estimate the liquidity risk at any given time, as the former shows the overall capacity of a firm to meet its short-term debts by a given short-term asset, and the latter is the most immediate cash asset available in a firm to uniformly meet its financial obligations [75]. The profitability and the volatility risk consist of the return on equity (ROE), the net profit growth rate (NPG), and return on assets (ROA). While the ROE and ROA evaluate the efficiency of profitability using the shareholder and asset perspectives, respectively, NPG is an indicator of the stability of profits, with increased volatility itself being a risk [76]. The operational risk is determined through the accounts receivable ratio (Rec) and inventory ratio (Inv) that comprise the efficiency in working capital management. High ratios are potentially a warning of slow collections, inventory accumulation, and cash flow problems [77]. Direct aggregation may be biased because of the overlap/multicollinearity of these indicators. Thus, the dimensionality reduction is performed by PCA, which has been extensively proven to be an objective and strong method of data compression in recent risk integration works [78]. The last composite risk index is a continuous proxy variable that measures the overall risk of the firms; the higher the index values, the higher the risk, which can be analyzed later as the relationship with the efficiency of investment.

3.2.3. Descriptive Statistics

Table 1. Descriptive statistics of main variables.

Category	Variable Name	Symbol	Variable Description	Unit	Mean	Sd	Min	Max
Input Variables	Capital Investment	CPA	Net fixed assets + capital expenditures	100 million CNY	19.203	26.009	0.875	164.014
	Operating Cost Investment	OC	Operating revenue × (1 − gross profit margin)	100 million CNY	282.152	890.018	0.902	5266.734
	Period Cost Investment	PE	Operating revenue × (operating expense ratio + management expense ratio)	100 million CNY	10.370	39.094	0.001	253.652
	Labor Investment	LAB	ln(number of employees)	-	7.761	1.160	4.615	10.566
Output Variables	Scale Output	SO	Operating revenue + operating profit	100 million CNY	301.003	916.158	−2.202	5429.613
Environmental Variables	Firm Size	SIZE	ln(total assets)	-	22.426	1.110	20.725	25.589
	Listing Year	AGE	Current year − year of establishment + 1	Years	18.840	6.243	4.000	32.000
	Per Capita GDP	PGDP	Annual per capita GDP of the region	10,000 CNY	6.812	3.307	2.595	20.028
	Ownership Concentration	OWNC	Shareholding ratio of the largest shareholder	%	33.788	15.951	9.131	64.143
	Market Competition Level	HHI	Herfindahl–Hirschman Index of the industry	-	0.174	0.167	0.016	1.000
Leverage Risk	Asset–Liability Ratio	Lev	Total debt/total assets	%	0.481	0.214	0.059	1.290
	Equity Multiplier	EM	Total assets/total equity	-	3.538	13.353	1.063	187.114
Liquidity Risk	Current Ratio	Liquid	Current assets/current liabilities	-	1.832	1.296	0.285	9.978
	Cash Flow Ratio	CashFlow	Cash equivalents/current liabilities	-	0.072	0.069	0.000	0.528
Profitability and Volatility Risk	Return on Equity	ROE	Net profit/shareholder equity	%	0.318	3.034	0.001	45.551
	Net Profit Growth Rate	NPG	(Current period net profit − previous period net profit)/previous period net profit	%	1.836	4.896	0.004	48.322
	Return on Assets	ROA	Net profit/total assets	%	0.049	0.045	0.001	0.301
Operational Risk	Accounts Receivable Ratio	Rec	Accounts receivable/operating revenue	%	0.054	0.036	0.001	0.182
	Inventory Ratio	Inv	Inventory/operating revenue	%	0.207	0.112	0.022	0.484

Note: A translation transformation is applied to the output variable in the DEA estimation to ensure non-negativity.

shows the descriptive statistics of the important variables to enable the comprehension of the basic features of sample firms in specific dimensions of inputs, outputs, external environment, and risk. On the whole, one can demonstrate considerable differences between companies in regard to the indicators of operations such as capital investment, operating costs, period costs and scale output, which indicates the fact of the large heterogeneity of the grain supply chain in resource distribution and the efficiency of output. Regarding the environmental variables, the sample firms seem to have a significant difference in terms of the firm size and GDP per capita in different regions, implying that the sample firms enjoy dissimilar external development conditions. The different risk indicators are also highly dispersive, especially when it comes to the operating leverage, cash flow ratio, and net profit volatility, which have high variance in the financial health of companies. The given variation also gives the needed foundation to construct the risk composite index on the basis of PCA. Overall, the descriptive statistics indicate that there is significant heterogeneity in the sample, and this fact supports the further empirical research of the efficiency assessment, measurement of risk, and its effects on supply-chain upgrading. All monetary variables in this study are reported in CNY. No currency conversion is applied, and all analyses are conducted using the original CNY values.

4. Data Analysis and Empirical Results

4.1. Investment Efficiency Analysis

4.1.1. Stage I: Initial BCC Efficiency Estimation

Using input–output data, this study estimates the investment efficiency of 25 listed firms in the grain supply chain with DEAP 2.1, and the results are presented in

Table 2. The investment efficiency of each link in the first stage of the grain industry chain and its decomposition index values.

Supply-Chain Segment	Indicator	2015	2016	2017	2018	2019	2020	2021	2022	2023	Mean
Upstream	crete	0.262	0.297	0.319	0.296	0.030	0.285	0.395	0.448	0.500	0.315
	vrete	0.824	0.837	0.837	0.786	0.065	0.780	0.810	0.816	0.845	0.733
	scale	0.315	0.357	0.382	0.375	0.041	0.372	0.482	0.537	0.589	0.383
Midstream	crete	0.302	0.356	0.401	0.482	0.528	0.570	0.608	0.569	0.553	0.485
	vrete	0.789	0.789	0.801	0.779	0.736	0.790	0.814	0.785	0.802	0.787
	scale	0.362	0.434	0.482	0.573	0.652	0.663	0.701	0.672	0.633	0.575
Downstream	crete	0.739	0.865	0.879	0.655	0.649	0.684	0.766	0.814	0.797	0.761
	vrete	1.000	0.996	0.945	0.865	0.824	0.864	0.907	0.922	0.935	0.918
	scale	0.739	0.867	0.915	0.716	0.732	0.754	0.814	0.853	0.819	0.801
Overall Mean	crete	0.434	0.506	0.533	0.478	0.402	0.513	0.590	0.610	0.617	0.520
	vrete	0.472	0.553	0.593	0.555	0.475	0.596	0.666	0.687	0.680	0.586
	scale	0.871	0.874	0.861	0.810	0.542	0.811	0.844	0.841	0.861	0.813

Note: crete denotes investment efficiency, vrete denotes pure technical efficiency, and scale denotes scale efficiency.

.

All in all, there is an apparent efficiency gradient across grain supply-chain segments during 2015–2023, with upstream firms showing the least investment efficiency, the midstream firms suffering a consistent improvement, and downstream firms being the most efficient among all. On the aggregate level, the investment efficiency (crete) grows slowly from 0.434 to 0.617, which is a sign of the long-term improvement in resource allocation throughout the industry. Yet, there is a large heterogeneity across supply-chain stages: the upstream firms continue to be inefficient (average 0.315), indicating the lack of input–output matching; the midstream firms achieve the largest efficiencies, reaching the level of 0.553 in 2023, which corresponds to the scale effect and technological upgrading; the downstream firms sustain the most efficient level in the entire period of the sample, which reaches 0.761 in 2023, which is consistent with the effects of scale expansion and technological modernization.

The efficiency breakdown indicates that both technical and scale elements make contributions to this gradient. Downstream firms have the highest level of pure technical efficiency (vrete) (0.918), with midstream companies at (0.787) and upstream companies at (0.733), indicating that technological capability disparities are an important factor in investment efficiency determination. Conversely, the scale efficiency (scale) is lowest in upstream firms (average 0.383), indicating persistent deviations from the optimal production scale, whereas the downstream firms run more efficiently at scale. On the whole, the closer the company is to the final market, the more efficient it is in terms of its investments, and the upstream segments are more limited by the lack of scale effect, which gives an empirical foundation for the future analysis of supply-chain upgrading.

4.1.2. Stage II: SFA Regression on Input Slacks

The second-stage analysis was used to determine how external environmental factors influence input redundancy. Specifically, this study employs an SFA model to regress the input slacks obtained from stage two (see

Table 3. Second-stage SFA regression results for input slacks.

Variables	Input(CPA) Slack	Input(OC) Slack	Input(PE) Slack	Input(LAB) Slack
Constant	−86.507 *** (−4.86)	−54.918 * (−1.670)	−0.248 (−0.359)	−0.407 (−0.208)
Env(SIZE)	3.700 *** (4.447)	2.502 (1.523)	−0.014 (0.404)	0.035 (0.365)
Env(AGE)	0.363 ** (2.4853)	0.412 (1.4489)	−0.004 (−0.6732)	0.060 *** (3.7277)
Env(PGDP)	−0.000 *** (−5.476)	−0.000 *** (−2.696)	−0.001 * (−1.808)	−0.000 *** (−5.211)
Env(OWNC)	26.148 *** (5.901)	18.669 * (1.923)	0.442 ** (2.342)	1.996 *** (3.351)
Env(HHI)	−3.003 (−0.892)	−5.124 (−0.608)	−0.123 (−0.703)	0.031 (0.065)
${\mathsf{\sigma }}^2$	432.309 *** (3.057)	24,810.371 *** (52.056)	34.640 *** (3.559)	5.076 *** (3.291)
$\mathsf{\gamma }$	0.962 *** (71.687)	0.991 *** (1068.488)	0.997 *** (1232.763)	0.933 *** (44.309)
Log	−681.153 ***	−996.982 ***	−127.645 ***	−237.825 ***
LR	304.022 ***	748.355 ***	860.574 ***	259.669 ***

Notes: Values in parentheses are t-statistics; ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

). The results show that the total variance parameter ${\sigma }^2={\sigma }_u^2+{\sigma }_v^2$ and the inefficiency-share parameter $\mathsf{\gamma }={\sigma }^2/({\sigma }_u^2+{\sigma }_v^2)$ are significantly different from zero in all the slack equations. This indicates that input redundancy is generated by both random disturbances $\left(v\right)$ and persistent inefficiency effects $\left(u\right)$, and the magnitude of $\gamma$ further implies that a substantial share of the variation in input redundancy is attributable to inefficiency rather than pure noise. Interpreted in the supply-chain context, this supports subsequent segment-level discussion: differences observed across upstream, midstream, and downstream segments are more likely to reflect structural and managerial frictions than purely transitory shocks. Therefore, the SFA specification is statistically valid for controlling environmental heterogeneity in the three-stage DEA framework. In the context of the individual determinants, the size of firms (SIZE) has a substantively positive relationship with input slacks, and the most significant impact is associated with the capital input equation (CPA) (p < 0.01), implying that the growth of the scale can be subject to an increase in the ineffective behavior regarding capital allocation. OWNC also generates a positive and significant consistent impact on all the slack equations, especially on capital and labor input (p < 0.01); thus, an ownership structure, whose concentration is high, can increase the inefficiency in resource allocation. On the contrary, regional per capita GDP (PGDP) imposes a negative influence on all input slack variables, which means that the more developed a region is, the higher the alleviation of the input redundancy in firms. There is a positive relationship between the listing age of firms (AGE) and capital and labor input slacks, which implies that the longer firms survive, the more resources will be wasted as a result of path dependence. Meanwhile, market competition (HHI) shows no statistically significant impact on any slack equation. This insignificance may reflect that an industry-level HHI does not fully capture the effective competitive pressure faced by listed grain firms operating across multiple product and regional markets. Moreover, procurement arrangements, price-stabilization policies, and long-term contracting practices in grain-related sectors may weaken the role of contemporaneous competition intensity in shaping short-run input redundancy. Finally, the limited listed-firm sample may reduce cross-sectional variation in HHI exposure, lowering statistical detectability. Taken together, this result suggests that reducing input redundancy may rely more on structural coordination and governance improvements than on competition intensity alone.

In general, the findings of the second-stage show that external variables such as the level of the regional economy, firm size, and the ownership structure contribute to the input redundancy systematically. This could, therefore, result in biased estimates if firm efficiency is directly compared without considering these exogenous differences, hence the importance of environmental adjustment in an efficiency estimation exercise in the future to provide a closer estimate of the individual efficiency levels of firms.

4.1.3. Stage III: Adjusted Efficiency After External Factor Correction

Following the elimination of the effects of the environmental heterogeneity and random noise using the second-stage SFA regressions, this research results in the environment-adjusted investment efficiency estimates along with their breakdown of firms in the three segments of the grain supply chain. All in all, the efficiency levels are significantly lower than the ones achieved in the first-stage DEA, meaning that external environmental factors play an important role in the input redundancy of businesses, and the three-stage DEA correction process is required. The average value of the investment efficiency of the whole sample is 0.357, as stated in

Table 4. Investment efficiency and its decomposition across grain supply-chain segments in the third stage.

Supply- Chain Segment	Indicator	2015	2016	2017	2018	2019	2020	2021	2022	2023	Mean	Change Rate (%)
Upstream	crete	0.102	0.108	0.275	0.093	0.108	0.100	0.107	0.121	0.146	0.129	−59.05
	vrete	0.960	0.968	0.870	0.948	0.934	0.934	0.930	0.929	0.917	0.932	27.15
	scale	0.106	0.113	0.291	0.093	0.115	0.107	0.118	0.134	0.168	0.138	−63.97
Midstream	crete	0.228	0.255	0.277	0.328	0.344	0.349	0.349	0.325	0.316	0.308	−36.49
	vrete	0.939	0.933	0.930	0.909	0.904	0.901	0.895	0.894	0.892	0.911	15.76
	scale	0.236	0.266	0.288	0.348	0.368	0.371	0.372	0.348	0.336	0.326	−43.30
Downstream	crete	0.525	0.552	0.635	0.633	0.642	0.655	0.682	0.698	0.676	0.633	−16.82
	vrete	0.993	1.000	0.994	1.000	0.990	0.987	0.985	0.988	0.988	0.992	8.06
	scale	0.526	0.552	0.636	0.633	0.643	0.657	0.685	0.700	0.678	0.634	−20.85
Overall Mean	crete	0.285	0.305	0.396	0.351	0.365	0.368	0.379	0.381	0.379	0.357	−37.45
	vrete	0.964	0.967	0.931	0.952	0.943	0.941	0.937	0.937	0.932	0.945	16.99
	scale	0.289	0.310	0.405	0.358	0.375	0.378	0.392	0.394	0.394	0.366	−42.71

, which means that there is an observed possibility of learner–environment regulation, since investment efficiency has reduced significantly compared to the stages of the first-stage estimates. By contrast, the mean pure technical efficiency stands at a comparatively high level (0.945), whereas the mean scale efficiency is 0.366, which means that scale inefficiency is the most dominant cause of overall inefficiency, as opposed to shortcomings with respect to the functionality of the pure technical performance.

There is a clear analysis of segmental gradients when the supply-chain position is evaluated. Downstream companies have the highest adjusted investment efficiency (mean 0.633) and the largest drop, compared to the first-stage outcomes (−16.82%), which suggests a rather sound standing in resource allocation and scale growth. Their technical efficiency is kept aimed at the frontier, and the scale efficiency is affected with diminishing effect after the adjustment, but it stands at the best position among all segments. Midstream companies are in a moderate stage, and they show a trend of sluggishness before showing a downturn and then leveling off. Though they retain a rather high level of pure technical efficiency (0.911), the scale efficiency deteriorates considerably (−43.30%), indicating that inefficiency in this segment is primarily caused by the incongruities of the scale structure. On the other hand, upstream companies have the lowest adjusted investment efficiency (mean 0.129) and the highest decline (−59.05%). The most significant efficiency loss is manifested in their scale efficiency (−63.97%), which becomes the main point of loss of efficiency. Even though the upstream firms have a relatively high degree of pure technical efficiency (0.932), this merit cannot convert to a higher degree of efficiency, pointing to serious limitations in terms of input allocation and scale of production, as well as the capacity to utilize resources.

From a temporal perspective, all three supply-chain segments exhibit a short-term rebound around 2017, followed by a pronounced decline during 2018–2019, coinciding with heightened external shocks. After 2020, investment efficiency across upstream, midstream, and downstream firms shows varying degrees of stabilization.

In an attempt to offer a more intuitive comparison of how the investment efficiency of the various segments of the grain supply chain evolved and differed over the study period, this study plots the time trends of the environment-adjusted investment effectiveness and decomposition indicators of each segment, as illustrated in Figure 2. The level of investment efficiency of downstream firms, as Figure 2a shows, is always above the average level of the supply chain, with an overall upward direction and rather regular variability. The midstream segment also stabilizes in terms of efficiency of investment after 2017, with the upstream firms continuing to work at very low levels of efficiency, and the volatility is high. Specifically, the accelerated growth in 2017 is preceded by a steep fall, which could be a result of policy adjustments or external shocks. As Figure 2b shows, the pure technical efficiency is high and constant across all segments, and the downstream firms are always being operated near the frontier, which means that technological and managerial strengths are not the main sources of efficiency loss. Conversely, the upward tendency in scale efficiency shown in Figure 2c is also similar to that in the overall investment efficiency, which once again suggests that inappropriate scale is a dominant structural limitation to the scaling up of efficiency along the supply chain, with the upstream segment in particular.

In order to conduct further analysis of the transformation of investment efficiency among supply-chain segments in a firm-level view, three points in time, including the first year (2015), the intermediate point (2019), and the last year (2023), are chosen. Figure 3 is drawn by creating scatter plots of the listed firms with pure technical efficiency (vrete) in the horizontal axis and scale efficiency (scale) in the vertical axis.

The downstream firms in 2015, to a large extent, are concentrated within the upper-right sector, which has relatively high levels of pure technical and scale efficiency, with most of the observations being on the efficiency frontier. Midstream firms are featured by relatively high vrete and lower scale, whereas the upstream firms are heavily dispersed in both directions, with few observations of efficiency in the scale, perceiving 0.1, which signifies a high degree of scale inefficiency.

By 2019, the distribution of vrete becomes more concentrated within the range of 0.85–1.00 across all segments, suggesting an overall improvement in managerial and technical efficiency. However, substantial heterogeneity in scale efficiency persists. Midstream firms experience the most notable improvement in scale, whereas upstream firms remain clustered in the low-scale-efficiency region. Downstream firms continue to exhibit the highest efficiency levels, with most observations located in the upper-right quadrant.

In 2023, differences in pure technical efficiency across segments further narrow, with a marked increase in observations approaching vrete = 1, particularly among midstream firms, indicating near-complete technical efficiency. Nevertheless, structural divergence in scale efficiency remains evident: upstream firms largely continue to operate at low scale efficiency, midstream firms exhibit the widest dispersion in scale, and downstream firms consistently maintain a dominant position in the high-efficiency region.

4.2. Investment Risk Analysis

4.2.1. Principal Component Risk Structure and Enterprise Risk Distribution Characteristics

PCA is applied to the nine investment risk indicators, and the results are reported in

Table 5. Principal component load matrix, characteristic roots and contribution rate of investment risk indicators.

Variable	PC1	PC2	PC3	PC4	PC5
Lev	0.762	−0.542	−0.064	0.063	0.074
EM	0.581	−0.236	0.212	−0.481	−0.462
Liquid	−0.608	0.526	0.203	−0.071	−0.312
CashFlow	0.079	0.189	−0.627	−0.513	0.431
ROE	0.846	0.325	0.056	−0.038	−0.135
NetProfitGrowth	0.492	0.497	0.038	0.372	0.118
ROA	0.438	0.751	−0.048	0.224	0.028
Rec	0.052	−0.198	0.764	0.025	0.498
Inv	−0.036	−0.475	−0.375	0.591	−0.161
Eigenvalue	2.448	1.844	1.215	1.044	0.808
Variance Explained (%)	27.198	20.487	13.501	11.605	8.979
Cumulative Variance Explained (%)	27.198	47.685	61.186	72.790	81.769

Notes: PC1–PC5 denote the first five principal components extracted from the nine investment risk indicators.

. The data were confirmed to be suitable for PCA, with a KMO value of 0.72 and Bartlett’s test of sphericity significant at the 1% level (p < 0.001). As reported in Table 5, five principal components were ultimately retained in accordance with the component-retention criteria specified in Section 3.1.2. The eigenvalues of the first five principal components are 2.448, 1.844, 1.215, 1.044, and 0.808, accounting for 27.20%, 20.49%, 13.50%, 11.61%, and 8.98% of the total variance, respectively. Together, these components explain 81.77% of the overall risk variation, indicating that the dominant risk structure in the sample is effectively captured. A study of the loading patterns indicates obvious economic explanations. PC1 puts a great burden on Lev, EM and ROE, which is a systemic risk, formed by both the leverage intensity and profitability. PC2 has high loadings on NPG and ROA, reflecting disparities in earnings volatility and profit stability. The main drivers of PC3 are Rec and CashFlow, demonstrating risks associated with the efficiency of operations and short-term solvency. The strongest relationship with Inv is demonstrated in PC4, which is inventory management–related risk, and PC5 is not linked with one prevalent risk factor.

The PC1–PC2 risk projection presented in Figure 4 also implies that two dimensions of corporate investment risk, leverage burden and earnings volatility, are the major drivers of corporate investment risk. The profitability-related metrics (ROE and ROA) have fairly high loadings on the two principal components, hence indicating that they cause the highest variation among the risk profiles observed. Conversely, the indicators related to liquidity will move in the opposite direction to leverage risk, meaning that the indicators partially buffer against financial vulnerability. A pattern of central clustering and peripheral dispersion of sample firms in the two-dimensional space results in a relatively low overall risk heterogeneity, while a small group of firms is defined by high leverage or nearly extreme earnings dispersion.

4.2.2. Investment Risk Evolution and Segment-Specific Risk Profiles

Once the main component structure of investment risk and the distribution features of the firm-level risk are identified, this section discusses the temporal development of investment risk in each of the many parts of the grain supply chain. As shown in Figure 5, the investment risk of upstream, midstream, and downstream companies shows a typical aspect of fluctuation followed by a gradual convergence in the 2015–2023 time period. Nevertheless, significant inter-segment differences can be distinguished.

In 2015, midstream firms had the highest level of risks (1.004), which decreased rapidly and then leveled off to a low level in 2016. The risk of upstream firms is moderate, relatively stable, and oscillates around the zero line. Conversely, downstream firms only temporarily increase in 2016 (0.615) and subsequently are consistently negative, indicating that it is under a low-risk regime in the vast majority of the sample period, a consequence of the stabilizing impact of the demand-side benefits.

Figure 6 shows how the investment risk is distributed between listed firms in the chain of grain supply. These findings indicate that upstream corporations demonstrate a strong risk heterogeneity, and there are both strongly risky types and significantly low-risk types of firms simultaneously, which spreads the range of risk levels widely. Midstream companies show a similarly spread risk profile, though there are no low-risk companies, as in the case of the upstream sector, which implies that the segment is a risk-intensive region in the supply chain. Conversely, the downstream firms have the best risk profile, with all the firms within the moderate risk bracket, which means that this segment has a significant ability to absorb both financial and operational risks.

4.3. Joint Impact Analysis

The analysis of the investment risk and efficiency of investments in a unified analytical framework to identify where the upstream, midstream, and downstream segments of the grain supply chain lie relative to one another in terms of efficiency–risk space, as well as how these various segments change over time, is mutually analyzed in this section. In the joint efficiency–risk analyses presented in Figure 3 and Figure 7, the quadrant thresholds are constructed using the global sample mean values of investment efficiency and the composite investment risk index calculated across all firm-year observations in the balanced panel. Mapping the observation of firm-year on a scatter plot with investment efficiency as the horizontal axis and investment risk as the vertical axis (Figure 7), we investigate which parts of the supply chain are most appealing to investors. This will help us to evaluate the existence of inter-segment variation in efficiency and risk as a binding constraint to grain supply-chain upgrading. Using global thresholds provides a unified benchmark that allows for a direct comparison of relative positioning and transitions across upstream, midstream, and downstream segments within a common efficiency–risk space, which is consistent with the study’s comparative and diagnostic objective.

In Figure 7a, upstream firms also constantly report investment efficiency, which is lower than supply-chain efficiency (0.272), with the majority of the observations placed in Quadrant IV (low efficiency and low risk) or Quadrant III (low efficiency and high risk). Despite the slow convergence of the risk levels after 2016 and efficiency showing a low growth trend improvement, the fact remains that upstream firms never reach the efficiency threshold and are not able to attain the desired and high efficiency and low risk configuration. As shown in this trend, during the sample period, the exposure to risks in upstream firms remains comparatively manageable, although their efficiency continues to be low, which limits their capabilities to create long-term investment systems to encourage the upward supply-chain adjustments.

The midstream segment has a significantly different structure, as evident in Figure 7b. With regard to the efficiency of investments, the results for all years are higher than the threshold, which means that this level of efficiency is usually higher. Risk levels, however, are of a wide range, with a strong intertemporal differentiation. In particular, some of the observations, such as in 2021, fit in Quadrant II (high efficiency and low risk) configuration, which is not present in the upstream segment. This implies that the midstream firms can more frequently attain the synergy between risk control and efficiency improvement. In addition, in the initial years (2015–2019), the volatility of midstream firms is relatively higher, suggesting that high risk will decrease, resulting in more stable operations throughout the years. The midstream segment of the supply chain is being upgraded through an increase in efficiency and a reduction in risk concurrently, with the effect being of material benefit.

As shown in Figure 7c, downstream observations fall in Quadrant II (high efficiency and low risk); the level of efficiency is significantly higher than the upstream and midstream segments. With the exception of 2015 and 2016, investment risk is also consistently low during the sample period. This delegation suggests that the downstream companies have developed a long-term matching of efficiency improvement and risk mitigation. Risk-taking no longer forms their development, but rather their development is based upon long-lasting competitive advantages obtained through branding, channels of distribution, and supply-chain management capabilities. The downstream segment, in comparison to the upstream and midstream segments, has been in a steady and endogenous growth pattern and is now a major point of stabilization of risk and value creation in the grain supply chain.

In summary, the empirical results reveal a clear and persistent efficiency–risk gradient across grain supply-chain segments. Downstream firms consistently exhibit the highest investment efficiency and the lowest risk exposure, upstream firms remain constrained by severe scale inefficiency and elevated risk, and midstream firms occupy an intermediate position characterized by improving efficiency but pronounced risk volatility. The joint efficiency–risk analysis further indicates that the efficiency–risk mismatch constitutes a key micro-level barrier to coordinated supply-chain upgrading.

5. Discussion and Implications

5.1. Discussion of Main Findings

This paper is based on firm-level data of listed firms in the grain supply chain of China, used to conduct a systematic analysis of the China grain supply chain in terms of its efficiency–risk structure and its development over time upstream, midstream, and downstream. As for the results, there is a strong gradient differentiation in the manner in which efficiency and risk are input collectively among the supply-chain segments. Such structural heterogeneity forms a fundamental micro-level basis of explaining the driving forces as well as the constraining breaks of upgrading supply chains in the grain industry. These gradients are documented in the segment-level summary statistics (Table 4 and Table 5) and the time-series trajectories of efficiency and risk (Figure 2, Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7).

First, the outcomes show that downstream firms in the grain supply chain are stable and constantly more efficient when it comes to their investments with comparatively low average investment risk, and upstream firms are featured by the constant low efficiency and significant risk exposure, with midstream firms taking a middle position. Specifically, the segment-level comparisons show that downstream firms rank highest in average investment efficiency and lowest in average risk exposure, whereas upstream firms rank lowest in average efficiency and highest in composite risk exposure. Consistent with Table 4, the mean third-stage investment efficiency (crete) is 0.633 for downstream firms, 0.308 for midstream firms, and 0.129 for upstream firms; the corresponding mean scale efficiency (scale) is 0.634, 0.326, and 0.138, respectively. Moreover, the time-series evidence indicates that the downstream performance is comparatively more stable, while the upstream risk indicators exhibit larger volatility over the sample period (Figure 5, Figure 6 and Figure 7). This gradient is also visible in year-to-year comparisons (e.g., in 2023, crete is 0.676 for downstream firms versus 0.146 for upstream firms; Table 4).This trend is in tandem with the results provided by Reardon and Timmer [79], who claim that the latter and lower branches of value chains tend to enjoy larger benefits of superior resource allocation. It coincides as well with Swinnen and Kuijpers [80], who underline that midstream and downstream segments in food value chains are likely to realize higher efficiency gains with the help of marketization and scale expansion processes. This segmental gradient is also consistent with the SFA variance decomposition in Table 3: a high $\gamma$ indicates that input redundancy is largely driven by inefficiency rather than random noise, reinforcing that the observed cross-segment differences reflect persistent structural frictions along the supply chain.

Second, this paper provides evidence that one of the major micro-level phenomena that limits long-term growth and stability in the grain supply chain is the existence of a lack of accordance between investment efficiency and investment risk. This issue coincides with the previous research, which highlights the beneficial purpose of firm efficiency or productivity in enhancing the value-chain upgrading [56,81], as far as additional investment efficiency is mostly associated with an enhanced capacity of resource distribution and the ensuing upgrading potential. Notably, this study’s findings also imply that a high efficiency in investment is not enough to maintain high-level continuous reallocation of resources by the firms in either more value addition or upgrading sectors. In cases where high efficiency goes hand in hand with increased financial and operational risk, the investment incentives of firms can be significantly lowered so that the prospective upgrading segments cannot experience steady growth. This efficiency–risk mismatch is also reflected in the joint segmental patterns over time, where upstream firms are more frequently characterized by low-efficiency–high-risk states, midstream firms show more pronounced temporal switching in risk conditions, and downstream firms cluster more persistently in high-efficiency–low-risk states. In that regard, the research results are consistent with the findings of Wang and Zhu [82], as well as Zhang and Wang [83], which point to the binding role of firm-level risk when it comes to influencing investment decisions.

Lastly, the findings show that scale inefficiency is the main cause of the losses of efficiency in the upstream segment, and the investment risk is multidimensional and transmissive in the supply chain. The three-stage DEA decomposition demonstrates that low pure technical efficiency does not affect upstream firms; instead, the general investment inefficiency is mainly driven by the excessively strong scale inefficiency. Consistent with the decomposition results, upstream firms exhibit markedly lower scale-efficiency scores than midstream and downstream firms in the segment-level comparisons (Table 4). Such a result confirms the findings of Rada and Fuglie [84] and Eder [85], who state that the scale of production is still restrictive due to structural aspects of land fragmentation and scattered farm activities, which also lead to significant scale efficiency losses. Meanwhile, the outcome of the PCA shows that investment risk is not a single-dimensional phenomenon. Upstream firms are also subjected to natural risks and market risks and thus exposed to compounded uncertainty. This multidimensionality is also visible in the segment-specific risk profiles and their co-movements over time (Figure 6 and Figure 7). This fact further generalizes the studies by Xue et al. [32] and Bellemare and Bloem [86] on the diffusion of agricultural risk through value chains, indicating that the upstream segment of the grain supply chain is not just an efficiency point but also serves as a cluster point of risks and their transfer to other participants in the chain.

5.2. Theoretical Significance

This study makes three specific theoretical contributions to the literature on agri-food supply chains and global value chain upgrading. First, this paper advances existing upgrading theories by developing an integrated efficiency–risk mismatch framework that moves beyond the traditional separation between investment efficiency and investment risk. While prior studies often examine efficiency improvement or risk exposure in isolation, this analysis shows that supply-chain upgrading depends critically on the alignment between efficiency and risk at the segment level. The empirical evidence demonstrates that high efficiency alone is insufficient to sustain upgrading when it is accompanied by elevated or unstable risk. By linking firm-level investment decision-making to segmental upgrading outcomes, this framework provides a bottom-up micro-mechanism through which upgrading dynamics can be more accurately interpreted. Second, the study contributes to the literature on firm behavior and resource allocation by identifying efficiency–risk mismatch as a key micro-level constraint on structural change within supply chains. Rather than treating upgrading barriers as the result of uniformly low efficiency or high risk, this study’s findings highlight how misalignment between these two dimensions distorts investment incentives and limits the reallocation of capital, labor, and technology toward higher value-added activities. This perspective refines existing conceptualizations of upgrading by emphasizing the joint structure of efficiency and risk as a determinant of sustained transformation. Third, by focusing on the Chinese grain supply chain, this paper provides context-specific theoretical insights into upgrading in transition economies. The empirical results reveal that persistent scale inefficiency—rather than insufficient pure technical efficiency—constitutes a structural source of entrapment for upstream firms. This finding underscores the importance of institutional and structural constraints in shaping upgrading trajectories. By explicitly incorporating these constraints into the analysis, the study enriches theoretical understanding of how supply-chain upgrading unfolds under transitional institutional settings.

5.3. Practical Significance

To make explicit how the proposed measures are derived from the empirical findings,

Table 6. Evidence-based link between empirical findings and segment-specific policy measures.

Key Empirical Result	Identified Bottleneck	Segment-Specific Measure	Expected Mechanism	Primary Target Audience
Upstream firms exhibit persistently low investment efficiency, mainly driven by scale inefficiency rather than pure technical inefficiency (Table 4; three-stage DEA results).	Structural scale inefficiency combined with high-risk exposure discourages long-term factor inflows.	Promote scale-enhancing land consolidation, cooperative/contract farming, and scale-oriented service platforms, combined with risk-buffer instruments.	Improve scale organization and reduce efficiency–risk mismatch, thereby restoring incentives for sustained capital, labor, and technology investment.	Policy makers; upstream producers; agricultural service providers
Midstream firms show improving efficiency but pronounced risk volatility over time (Figure 5, Figure 6 and Figure 7), limiting the sustainability of efficiency gains.	Efficiency–risk mismatch caused by unstable risk expectations.	Develop risk-stabilizing supply-chain finance, inventory and warehouse-receipt financing, and digitally enabled contract coordination.	Stabilize risk expectations, allowing efficiency gains to translate into long-horizon upgrading investments.	Financial institutions; midstream processors; supply-chain managers
Downstream firms consistently operate in a high-efficiency and low-risk regime and cluster in favorable efficiency–risk configurations (Table 4; Figure 5, Figure 6 and Figure 7).	Stability is not effectively transmitted upstream and midstream.	Leverage downstream lead firms through market-based coordination mechanisms (long-term procurement contracts, quality standards, and data-sharing incentives).	Transmit stability upstream, reduce system-wide efficiency–risk mismatch, and promote coordinated upgrading along the supply chain.	Lead firms; regulators; platform operators

summarizes the evidence-based link between the key results and the corresponding segment-specific policy implications. This paper then elaborates on each segment in Section 5.3.1, Section 5.3.2 and Section 5.3.3, following the evidence–bottleneck–measure logic summarized in Table 6.

5.3.1. Upstream Segment

Upstream firms exhibit persistently low investment efficiency, primarily driven by scale inefficiency, and elevated compounded risk exposure (Table 4; Figure 5, Figure 6 and Figure 7). This suggests a structural bottleneck in scale organization and risk buffering, which discourages long-term factor inflows and sustains an efficiency–risk mismatch. In the context of sustainable supply-chain development, the lack of proper risk management and inefficient scale organization can lead to long-term underinvestment and reduced resilience. Accordingly, upstream interventions should prioritize scale-enhancing strategies, such as land-use rights transfer, cooperative/contract farming, and service platforms, combined with risk-buffering instruments (e.g., insurance–credit linkages). These measures can reduce mismatch by improving scale efficiency and stabilizing investment expectations, thereby fostering sustainable upgrading-oriented investment in the long term.

5.3.2. Midstream Segment

Midstream firms show improving efficiency but pronounced risk volatility over time (Figure 5, Figure 6 and Figure 7), suggesting that efficiency gains may not translate into sustainable upgrading when risk expectations remain unstable. Thus, the key bottleneck is risk volatility, and the priority is to build risk-stabilizing coordination. Transaction- and inventory-based supply-chain finance (e.g., warehouse-receipt and advance-payment arrangements) and digitally enabled coordination can improve liquidity, reduce information asymmetry, and dampen risk fluctuations. This helps convert efficiency gains into stable, upgrading-oriented investments, fostering sustainable economic development and ensuring the long-term resilience of midstream operations.

5.3.3. Downstream Segment

Downstream firms consistently operate in a high-efficiency and low-risk regime and cluster in favorable efficiency–risk configurations (Table 4; Figure 5, Figure 6 and Figure 7), implying that downstream can function as a stability anchor for the entire supply chain. However, the bottleneck is that this stability is not sufficiently transmitted upstream and midstream. To promote sustainable growth across the entire supply chain, policy should leverage downstream lead firms through market-based coordination mechanisms—especially risk-sharing long-term procurement contracts, enforceable quality standards, and data-sharing incentives—to transmit stability along the chain. These mechanisms can reduce the system-wide efficiency–risk mismatch and promote coordinated upgrading. This approach ensures that resilience and sustainability are embedded in the supply chain’s development.

5.4. Limitations and Prospects

This study systematically analyzes the efficiency–risk structure across the grain supply chain and its implications for supply-chain upgrading but faces limitations.

First, the data are primarily drawn from listed firms, with a small sample of only 25 public companies, which may limit both the representativeness of non-listed agricultural operators and the statistical power of segment-level comparisons. Future research could incorporate firm-level surveys or farm-household microdata to extend the conclusions to a broader range of entities. Second, the relatively static analytical framework does not fully capture the dynamic interactions between efficiency and risk or policy intervention lagged effects. Future work could expand data sources to construct mixed samples encompassing multiple types of economic agents and apply dynamic panel models or longitudinal case-tracking approaches to more deeply explore the evolutionary paths and driving mechanisms underlying efficiency–risk alignment over time.

6. Conclusions

This study set out to investigate how the joint structure of investment efficiency and investment risk shapes the upgrading dynamics of the grain supply chain in China. Drawing on firm-level panel data and an integrated efficiency–risk analytical framework, three main conclusions can be drawn, corresponding directly to the research questions.

First (addressing RQ1), this study documents a pronounced segmental gradient in investment efficiency across the grain supply chain. Downstream firms consistently operate in a high-efficiency regime, upstream firms remain characterized by persistent inefficiency—largely driven by scale inefficiency—and midstream firms occupy an intermediate position. The downstream segment plays a key role in ensuring the long-term resilience and sustainability of the grain supply chain.

Second (addressing RQ2), this study finds that investment risk exhibits strong structural heterogeneity across segments. Upstream firms face multifaceted and compounded risk exposure, midstream firms are characterized by heightened risk volatility, while downstream firms operate under comparatively stable and lower-risk conditions. Sustainable risk management is crucial to improve the stability and adaptability of the entire supply chain.

Most importantly (addressing RQ3), the analysis shows that the alignment or mismatch between investment efficiency and risk constitutes a critical micro-level constraint on supply-chain upgrading. High investment efficiency alone is insufficient to attract sustained factor inflows when accompanied by elevated or unstable risk, thereby weakening firms’ incentives for the resource reallocation required for process and functional upgrading. This efficiency–risk mismatch can lock the supply chain in a suboptimal structure, undermining its sustainability in the long term.

Overall, these findings contribute to the literature on agri-food value chains by providing micro-level evidence that the alignment of efficiency and risk, rather than considering them in isolation, is critical for sustainable segmental upgrading. From a practical perspective, they underscore the importance of segment-specific strategies that address not only efficiency gaps but also the risk profiles that distort investment incentives, with a particular focus on promoting sustainable resource allocation. A key limitation of this study lies in its focus on listed firms; future research incorporating non-listed firms and smallholder data could further enhance the generalizability of the efficiency–risk mismatch framework and its applicability to more inclusive and resilient agricultural systems.