Circular Economy and Resource Efficiency in the Serbian Agri-Food Sector: Evidence from Dynamic Panel Analysis

Abstract

The transition toward sustainable, resource-efficient production has become a key challenge for agri-food systems, particularly in emerging economies, where profitability and environmental goals must be balanced. This study aimed to examine the relationship between financial structure, macroeconomic conditions, and circular economy (CE) indicators in determining the profitability of Serbian agri-food enterprises. Using panel data for 625 firms from 2014 to 2021, a two-step system GMM model was applied to control for endogeneity and firm-specific effects. The results indicate that in agriculture, moderate leverage enhances profitability, while excessive debt reduces it. Recycling and efficiency-oriented circular practices have a positive and significant effect on financial performance, suggesting that resource-efficient management supports long-term profitability. In the food industry, profitability shows strong persistence but remains mainly driven by internal and economic factors, with CE indicators exerting weaker short-term effects. Robustness tests confirm the validity and stability of the estimates. Overall, the findings highlight that integrating circular economy principles into business strategies can contribute to both financial sustainability and more efficient resource use in the agri-food sector.

1. Introduction

Agriculture and the food industry are among the most important sectors of the Serbian economy. Due to favorable agro-ecological conditions, abundant arable land, and long-standing traditions in farming, agriculture has traditionally played a crucial role in the country’s economic and social development. Serbia has approximately 3.5 million hectares of arable land, with cereals, industrial crops, fruits, and vegetables forming the backbone of production (Statistical Office of the Republic of Serbia, 2023). Agricultural exports account for nearly 20% of Serbia’s total exports, with grains, fruits, and processed food products representing the most competitive categories (FAO et al., 2022). The food industry is the largest industrial branch in Serbia, accounting for approximately 15% of total industrial production and employing more than 80,000 workers (Marković et al., 2025). Together, the agriculture and food sectors contribute around 10% to national GDP, and their importance extends beyond economic performance, as they support rural employment, regional development, and food security (Statistical Office of the Republic of Serbia, 2023).

Despite this significant role, both sectors face multiple challenges. Agricultural production remains vulnerable to climate variability and frequent extreme weather events. At the same time, the food industry is under pressure from global competition, technological demands, and requirements to comply with EU food safety and sustainability standards. Moreover, structural issues such as farm fragmentation, limited investment capacity, and insufficient modernization reduce the potential of Serbian agriculture and the food industry to achieve their full competitiveness (Pantić et al., 2019). In this context, improving resource efficiency and minimizing waste generation have become central priorities for maintaining productivity and competitiveness while safeguarding environmental integrity. Recent research highlights that integrating circular economy principles into agri-food supply chains not only reduces input dependency and waste but also generates new value-added opportunities through innovation and resource recovery (Cahyadi et al., 2024). Likewise, bibliometric analyses confirm that the circular transformation of agricultural systems is increasingly recognized as a pathway to sustainable production and economic resilience, especially in emerging regions (Angulo et al., 2024). Consequently, understanding how circular economy indicators affect firm-level profitability represents a key step toward linking environmental and financial performance in the agri-food sector.

Profitability is one of the most important indicators of business success, reflecting not only the efficiency of resource utilization but also the ability of enterprises to adapt to changing market and policy conditions. For SMEs, which make up the majority of enterprises in the Serbian agri-food sector, profitability is essential for maintaining market position, ensuring access to finance, and enabling investment in innovation and sustainability. However, profitability in Serbian SMEs is highly sensitive to both internal and external factors. Internal determinants such as liquidity, leverage, capital structure, and asset utilization efficiency jointly determine a firm’s financial resilience and growth capacity. Liquidity ensures the firm’s ability to meet short-term obligations and maintain operational stability, whereas excessive liquidity may imply inefficient resource use (Gill et al., 2011). The financial structure, particularly the balance between debt and equity, significantly affects profitability on capital costs and financial risk. Empirical findings by Pervan et al. (2013) and Yazdanfar and Öhman (2015) confirm that moderate leverage can enhance profitability, whereas high indebtedness tends to erode returns and increase exposure to liquidity constraints. Similarly, Dalci et al. (2018) emphasize that optimal debt levels improve profitability by stimulating investment efficiency and tax benefits, while excessive borrowing leads to declining margins. Productivity and asset turnover ratios are also key drivers of profitability, as they reflect the firm’s capacity to utilize its resources efficiently and maintain competitiveness (Blažková & Dvouletý, 2019; Singh, 2019). External macroeconomic factors, including GDP growth, inflation, and interest rate fluctuations, further shape profitability dynamics. Liu et al. (2020) and Andrašić et al. (2018) highlight that favorable. In contrast, inflationary pressure and favorable macroeconomic conditions support firm growth and financial performance by boosting demand and easing financing constraints. In contrast, inflationary pressures and unstable policy environments tend to reduce profit margins and investment activity. At the same time, empirical evidence suggests that profitability levels in Serbian agri-food SMEs lag behind those in many EU member states. Structural inefficiencies, high transaction costs, and limited integration into value chains reduce competitiveness. Furthermore, differences between regions (e.g., Vojvodina versus less developed areas of Southern Serbia) highlight inequalities in access to resources, infrastructure, and markets. This makes profitability analysis not only an economic but also a developmental issue, as it sheds light on broader structural challenges within the national economy.

In recent years, increasing attention has been devoted to linking profitability analysis with sustainability considerations, particularly through the lens of the circular economy (CE) and efficient resource use (Feng & Goli, 2023; Nosková et al., 2024). In line with the European Green Deal and the EU Circular Economy Action Plan, the competitiveness of enterprises is no longer evaluated solely based on financial performance, but also on their ability to reduce waste, increase recycling, and optimize natural resource consumption (Awan et al., 2021a; Ingaldi & Ulewicz, 2020). This dual focus reflects the need to align profitability with environmental responsibility, especially in industries such as agriculture and food processing, which depend on natural resources and have significant ecological footprints.

To operationalise the idea of circularity, this study uses three standardized Circular Economy (CE) indicators—CE1, CE2, and CE3—defined within the EU Circular Economy Monitoring Framework developed by Eurostat (Eurostat, 2024). CE1 represents municipal waste generation per capita, reflecting overall resource efficiency and waste management performance. CE2 measures the recycling rate of municipal waste, indicating the degree of circularity and material recovery within the system. CE3 captures greenhouse gas emissions from production activities, representing the environmental footprint of economic output. These indicators are widely applied across EU member states to monitor progress towards circularity in line with the European Green Deal objectives.

The subject of this research is the assessment of how internal, external, and CE-related factors influence the profitability of SMEs in Serbia’s agricultural and food sectors. The main objective is to analyze the strength and direction of these relationships over the period 2014–2021. By integrating financial indicators, macroeconomic conditions, and circular-economy performance, this study provides a comprehensive understanding of the determinants of profitability and resource efficiency, offering evidence relevant to enterprise strategies, policy design, and Serbia’s alignment with European standards of sustainable competitiveness.

2. Literature Review

2.1. Internal Factors of Profitability

The profitability of small and medium-sized enterprises (SMEs) in the agri-food sector largely depends on internal factors, i.e., microeconomic factors stemming from financial structure, managerial decisions, and organizational capacities. The most frequently emphasized determinants in the literature include key financial and operational variables such as liquidity, leverage, asset turnover, labor productivity, firm size and age, and innovativeness. In the context of Serbia, research has shown that liquidity, asset turnover, and capital structure are crucial indicators of profitability. Tekić et al. (2022) analyzed SMEs in agriculture and the food industry and found that both total and current asset turnover positively affect profitability. They found that both total and current asset turnover positively affect profitability. In contrast, high effect on analyzed SMEs in the agriculture and food industries, and found that both total and current asset turnover positively affect profitability, whereas high leverage negatively affects profitability. Novaković et al. (2025a) also confirmed that, in the food industry, asset turnover has a positive impact on profitability, whereas asset tangibility has a positive effect on profitability, whereas asset tangibility and collection period negatively affect performance. In the broader European context, numerous studies confirm the importance of internal factors. Grau and Reig (2020) highlight operating leverage and cost structure as key determinants of SME profitability in the European agri-food sector. Vuković et al. (2022) analyzed the financial performance of European agricultural companies and concluded that liquidity, asset tangibility, and access to both short- and long-term financing significantly affect profitability. Lehenchuk et al. (2022), using evidence from Slovakia, show that firm size and capital intensity can adversely affect profitability by limiting flexibility and increasing risk exposure. Globally, Deloof (2003) demonstrated that working capital management is crucial for firm profitability in Belgium, while Lazaridis and Tryfonidis (2006) confirmed the significance of the collection period in Greece. Abor (2005) found, in a sample of Ghanaian firms, that excessive debt reduces SME profitability, whereas Goddard et al. (2005a) found in the UK that sales growth and labor productivity positively affect profitability. Ahinful et al. (2021) in Ghana and Le Thi Kim et al. (2021) in Vietnam also confirmed that innovation and asset turnover strengthen profitability, while high leverage acts as a limiting factor. Overall, these findings indicate that internal factors are universally important, but their relative weight depends on the institutional framework and market conditions.

2.2. External Factors of Profitability

In addition to internal determinants, SME profitability is also influenced by external factors, including macroeconomic and institutional factors, among which GDP growth, inflation, institutional support, access to financing, competition, and locational conditions stand out. In Serbia, studies indicate that GDP growth is one of the most important positive external factors, whereas inflation tends to have a negative impact on SME profitability (Novaković et al., 2025a). Tekić et al. (2022) further highlight the importance of regional disparities, noting that firms in more developed regions (e.g., Vojvodina) perform better due to improved access to resources and markets. Institutional support through subsidies and IPARD funds also emerges as an important external factor, although SMEs often face limited access to these instruments. At the European level, numerous studies highlight the importance of regulatory frameworks and policy support. Lehenchuk et al. (2022) extensively highlight the significance of macroeconomic conditions and institutional stability in Slovakia, while Vuković et al. (2022) confirm that access to favorable financing and institutional support through CAP policies directly influences the profitability of agricultural firms. Grau and Reig (2020) note that regulatory frameworks and taxation policies shape the relationship between risk and SME profitability, while Zeitun et al. (2007) show that macroeconomic stability and capital market development positively affect corporate performance in Europe. Globally, numerous studies confirm the significance of external factors. Sogorb-Mira (2005) in Spain demonstrated that regulatory frameworks and macroeconomic stability shape SME capital structure and profitability. In developing countries, such as Ghana and Nigeria, external factors, such as inflation and market volatility, have been shown to exert an adverse effect on SME profitability (Abor, 2005; Akinyomi, 2014). Overall, these findings confirm the universal importance of external determinants of profitability, though their intensity depends on institutional and regional contexts.

2.3. Circular Economy, Resource Efficiency, and Profitability

In the last decade, there has been increasing interest in the relationship between SME profitability and the implementation of circular economy (CE) principles, particularly in the agri-food sector. Practices such as reducing food waste, valorizing by-products, utilizing renewable energy, and promoting eco-innovation are increasingly recognized as key drivers of resource efficiency and long-term competitiveness. By improving material recovery and reducing dependency on primary inputs, CE enhances both financial and environmental performance. In Serbia, Novaković et al. (2025b) show that CE indicators—such as generated waste, recycling rates, and greenhouse gas (GHG) emissions—significantly affect the financial stability and decision-making of small agricultural enterprises. Higher recycling rates are associated with improved access to financing, whereas increased waste generation and GHG emissions tend to undermine financial stability. These findings suggest that CE performance is becoming an essential factor not only for ecological but also for resource-efficient and financially sustainable management. In the European context, Kirchherr et al. (2018) emphasize cultural and market barriers to CE implementation in the EU, while Rizos et al. (2016) highlight financial obstacles and the lack of institutional support as critical challenges for SMEs. Ormazabal et al. (2016), focusing on Spanish SMEs, demonstrate that innovativeness and institutional support facilitate CE adoption. In contrast, Mazzucchelli et al. (2022) confirm that CE enhances innovation and reputation, thereby indirectly improving profitability. Mura et al. (2020) report that CE in Italian SMEs contributes to product differentiation and competitiveness, while Palea et al. (2023) provide longitudinal evidence of the economic success of CE strategies in European firms. At the global level, Geissdoerfer et al. (2017) conceptualize CE as a strategic model linking environmental and economic sustainability, while Ghisellini et al. (2016) offer a comprehensive review of CE implications for industry and resource management. Chowdhury et al. (2022) in Vietnam and Rodríguez-Espíndola et al. (2022) in Mexico show that organizational factors such as leadership and innovativeness promote CE practices, which in turn positively affect economic, environmental, and social performance. Jaeger and Upadhyay (2020) stress that high initial costs, supply chain complexity, and lack of technical knowledge are significant barriers to CE adoption in developing countries. Despite these challenges, most studies confirm that CE practices improve cost efficiency, strengthen brand reputation, and enhance eco-efficiency and resource productivity, ultimately leading to higher profitability. Based on the reviewed literature, the relationships among circular economy indicators, resource efficiency, and firm profitability can be summarized through a conceptual framework (Figure 1). The proposed framework clarifies the hypothesized cause-and-effect links between circularity (CE indicators), firm-level resource efficiency, and financial performance.

Figure 1 illustrates the hypothesized cause-and-effect relationships between Circular Economy (CE), Resource Efficiency (RE), and firm Profitability. The CE indicators (CE1—waste generation intensity, CE2—recycling and recovery rate, and CE3—resource productivity or emission performance) reflect the national circularity context and the overall level of circular economy implementation in Serbia, as reported by Eurostat. Improvements in national circular performance are expected to enhance RE at the firm level by encouraging more efficient input use, cost optimization, and technological innovation. Higher RE, in turn, contributes to improved firm profitability, measured by Return on Assets (ROA). Firm-specific structural variables (LEV, DEBT, and TANG) and macroeconomic factors (GDP, CPI) act as contextual determinants that influence the strength and direction of these relationships. The feedback loop indicates that more profitable firms are better positioned to reinvest in circular practices, thereby reinforcing the long-term sustainability–performance nexus.

2.4. Research Gap

Although the literature provides extensive evidence on internal and external determinants of SME profitability, few studies explicitly integrate the concept of the circular economy and resource efficiency into profitability analysis. For instance, Feng and Goli (2023) developed a comprehensive mathematical model demonstrating how CE-oriented strategies enhance both business performance and environmental outcomes. Similarly, Nosková et al. (2024) provided a systematic literature review confirming that circular practices can improve operational efficiency and financial resilience across various industries. Empirical findings by Shavkatov et al. (2024) further support the positive association between CE implementation and corporate financial performance, showing that firms adopting circular solutions achieve higher returns on assets and lower waste intensity. However, despite these advances, the empirical evidence remains fragmented and rarely focuses on the agri-food sector. This study represents the first empirical attempt to analyze the combined influence of financial structure, macroeconomic conditions, and circular economy indicators on firm profitability within the agri-food sector of the Republic of Serbia. By integrating firm-level financial data with CE indicators (CE1–CE3) derived from the Eurostat Circular Economy Monitoring Framework, the research develops a dynamic panel model (System GMM) that enables a robust examination of both short- and long-term effects while addressing endogeneity.

This approach introduces a novel empirical framework for transitional economies, combining microeconomic, macroeconomic, and circular economy dimensions to assess the drivers of profitability and sustainable competitiveness in Serbia’s agri-food system.

3. Methodology

3.1. Data and Sample Description

The empirical analysis is based on an unbalanced panel dataset of Serbian agricultural and food industry enterprises covering the period 2014–2021. Firm-level data were obtained from the Agency for Business Registers of the Republic of Serbia (APR). At the same time, macroeconomic indicators were sourced from the Statistical Office of the Republic of Serbia (SORS) and the National Bank of Serbia (NBS). The dataset comprises 625 enterprises, including 306 agricultural and 319 food industry firms.

The original dataset contained missing values for several firms in certain years. To ensure data consistency, firms with incomplete financial information for at least three consecutive years were excluded from the analysis. Occasional single-year gaps were treated as missing observations, and the final dataset remained unbalanced. No imputation procedures were applied, as the missing values were not systematic and represented less than 5% of the total observations.

Both samples consist of annual financial statements and circular-economy data. Agricultural enterprises are primarily engaged in crop and livestock production. In contrast, food industry firms span multiple subsectors, including fruit and vegetable processing, dairy, beverage, and milling. This classification enables a comparative assessment of financial and circular performance across different stages of the agri-food value chain.

3.2. Variable Description

The study employs both financial and circular economy (CE) indicators, along with macroeconomic control variables. The dependent variable is the Return on Assets (ROA), representing firm profitability. Independent variables capture aspects of financial structure, capital composition, and circularity performance. A detailed description of all variables, including financial indicators, CE metrics, and macroeconomic controls, is presented in

Table 1. Definition of variables.

Variable	Description	Unit of Analysis	Expected Sign
ROA	Return on assets (%), calculated as net income divided by total assets.	Firm level	−
LEV	Leverage ratio (total liabilities/equity).	Firm level	±
DEBT	Debt ratio (total liabilities/total assets).	Firm level	−
TANG	Tangibility (fixed assets/total assets).	Firm level	±
GDP	Annual GDP growth rate (%), a control variable reflecting macroeconomic activity.	Country level	+
CPI	Inflation rate (%), proxy for macroeconomic stability.	Country level	−
CE1	Volume of by-products or residues generated (waste generation intensity).	Country level	±
CE2	Recycling and recovery rate (%), a measure of material circulation and reuse.	Country level	+
CE3	Resource productivity or emission performance index.	Country level	+

.

The selection of variables is grounded in previous empirical studies investigating the determinants of firm profitability and sustainability performance in the agri-food sector. Financial indicators (LEV, DEBT, and TANG) capture firms’ capital structure and asset composition, which are consistently identified as key determinants of profitability across both national and international studies. Research on Serbian SMEs confirms that asset turnover and liquidity positively influence profitability, effect, whereas high leverage and asset tangibility negatively affect profitability (Tekić et al., 2022; Novaković et al., 2025a). Similar findings were reported in European contexts, where operating leverage, cost structure, and access to financing play critical roles in shaping SME profitability (Grau and Reig, 2020; Vuković et al., 2022). At the global level, empirical evidence further demonstrates that excessive debt reduces profitability (Abor, 2005). Macroeconomic variables (GDP and CPI) were included as contextual controls, as they have been widely applied in previous studies (Novaković et al., 2025a; Zeitun et al., 2007; Akinyomi, 2014).

The circular economy indicators (CE1–CE3) were obtained from Eurostat’s database on Circular Economy—Waste, Recycling, and Resource Efficiency Indicators (Eurostat, 2024). These variables reflect the degree of implementation of circular principles at the national level in Serbia and the EU reference data, standardized according to Eurostat methodology. Since firm-level CE indicators are not available, the country-level CE1–CE3 values were uniformly assigned to all firms within the same sector and year. This approach assumes that firms operate within a shared national regulatory, technological, and resource-use environment, which justifies the use of national circularity indicators as contextual proxies in firm-level analysis. While this assumption may introduce a certain degree of measurement bias, it enables the assessment of the indirect effects of national circularity performance on firm-level profitability. Given that Serbia is a relatively small and integrated economy, it is reasonable to expect that national-level CE indicators influence firms in a similar manner through common policy, market, and environmental conditions.

To minimize outlier influence, all continuous variables were winsorized at the 1st and 99th percentiles.

3.3. Econometric and Statistical Analysis

The statistical and econometric analysis integrates a set of complementary quantitative methods assessing the influence of circular economy practices in the Serbian agri-food sector. Descriptive statistics, including measures of central tendency (mean, median) and dispersion (standard deviation, minimum, and maximum), were first calculated to summarize the key characteristics of the dataset and to detect heterogeneity, skewness, and potential outliers, which are essential for the accurate interpretation of econometric results (Gujarati & Porter, 2009). Subsequently, pairwise Pearson correlation coefficients were then used to examine the strength and direction of linear relationships among variables and to detect potential multicollinearity (Hair et al., 2019). The low-to-moderate correlation values obtained confirmed that the explanatory variables capture distinct aspects of firm behavior, thereby justifying their inclusion in multivariate models.

To prevent spurious regression results, the time-series properties of all variables were examined using the Im-Pesaran-Shin (IPS) panel unit root test, which allows for heterogeneity across cross-sectional units and is therefore suitable for firm-level unbalanced panels (Im et al., 2003). Variables identified as non-stationary in levels were transformed into first differences (e.g., ΔCE2 = CE2 − lag(CE2)) to achieve stationarity and to guarantee that estimated coefficients reflect genuine long-term relationships rather than spurious correlations (Baltagi, 2021; Wooldridge, 2019).

Given the dynamic nature of profitability and the potential endogeneity of explanatory variables, the two-step System Generalized Method of Moments (System GMM) estimator was employed (Arellano & Bond, 1991; Blundell & Bond, 1998). This estimator is particularly appropriate for datasets with a large number of cross-sectional units (N) and a relatively short time dimension (T), such as those analyzed in this study. The System GMM combines equations in first differences and improves efficiency when the dependent variable exhibits persistence and when lagged levels are weak instruments for differenced equations. This specification is appropriate given the significant persistence of profitability (lagged ROA = 0.567). It effectively controls for unobserved heterogeneity through first differencing, mitigates endogeneity by using lagged dependent and explanatory variables as instruments, and addresses autocorrelation by incorporating lagged dependent variables in the model. The general specification of the model is expressed as:

$${ROA}_{it}=\alpha +{\beta }_1{ROA}_{it-1}+{\beta }_2{LEV}_{it}+{\beta }_3{DEBT}_{it}+{\beta }_4{TANG}_{it}+{\beta }_5{GDP}_t+{\beta }_6{CPI}_t+{\beta }_7{CE1}_{it}+{\beta }_8{∆CE2}_{it}+{\beta }_9{CE3}_{it}+{\mu }_i+{\epsilon }_{it}$$

(1)

where ${\mu }_i$ denotes unobserved firm-specific effects and ${\epsilon }_{it}$ represents the idiosyncratic error term.

Model validity was assessed using standard diagnostic tests: the Hansen J test to verify instrument validity, the Arellano–Bond AR(1) and AR(2) tests to detect serial correlation, and the Wald test to evaluate the joint significance of coefficients. This methodological approach follows established practices in profitability and performance studies employing firm-level panel data (Bond, 2002; Roodman, 2009; Baltagi, 2021). All statistical analyses were performed using R software (version 4.4.1) employing the following packages: readxl (data import), dplyr (data preparation and descriptive statistics), psych (correlation analysis), car (VIF diagnostics), plm (System GMM estimation), lmtest and sandwich (robust standard errors and diagnostic tests), and tseries (unit root testing).

3.4. Robustness Checks

The robustness of the estimated model was verified through two complementary procedures designed to ensure the stability and consistency of the results. First, the Variance Inflation Factor (VIF) test was applied to detect potential multicollinearity among the explanatory variables. In accordance with Hair et al. (2019), VIF values below the conventional threshold of 10 indicate an acceptable level of correlation between regressors, thereby confirming the reliability of the estimated coefficients and the absence of redundancy among explanatory variables. Second, a static Fixed-Effects (FE) model was calculated as an additional robustness check to account for unobserved, time-invariant heterogeneity such as managerial practices, production technology, or regional characteristics (Wooldridge, 2019). The consistency in both the direction and statistical significance of the coefficients obtained from both the GMM and FE models was interpreted as evidence of robustness, indicating that the core relationships identified in the dynamic specification remain stable across alternative estimation techniques.

4. Results

The empirical analysis was conducted separately for agricultural and food industry enterprises to account for potential sectoral differences in profitability and circular economy performance.

4.1. Agricultural Enterprises

Before estimating the econometric models, descriptive statistics were calculated to provide an overview of the main characteristics of the variables used in the analysis (

Table 2. Descriptive statistics of the variables (agricultural enterprises).

Variable	Mean	Median	SD	Min	Max
ROA	4.51	2.72	7.45	−29.00	252.39
LEV	7.77	1.08	178.72	0.00	8798.02
DEBT	0.50	0.52	0.29	0.00	1.04
TANG	0.40	0.40	0.26	0.00	1.00
GDP	2.62	2.70	2.78	−1.60	7.50
CPI	2.15	1.90	0.92	1.10	4.10
CE1	332.25	312.50	63.77	259.00	442.00
CE2	4.32	0.50	6.81	0.00	16.80
CE3	7699.93	7835.47	447.70	6554.70	8029.35

). The dataset refers to Serbian agricultural enterprises for the period 2014–2021 and includes financial indicators, macroeconomic variables, and circular economy metrics.

Table 2 presents the descriptive statistics for agricultural enterprises in Serbia during the 2014–2021 period. The average return on assets (ROA) is 4.5%, with a median of 2.7%, indicating a right-skewed distribution. Most firms achieve modest profitability, while a small subset reports exceptionally high returns. The leverage ratio (LEV) exhibits substantial variability (SD = 178.72), confirming substantial heterogeneity in firms’ capital structures. The mean debt ratio (DEBT) equals 0.50, suggesting that, on average, half of total assets are financed through debt. The tangibility ratio (TANG), with a mean of 0.40, reflects the capital-intensive nature of agricultural production, where physical assets dominate. Macroeconomic indicators show relative stability, with an average GDP growth rate of 2.6% and inflation averaging 2.1%. These results indicate that the broader economic environment was stable but not dynamic enough to stimulate profitability growth strongly. Circular economy indicators display considerable variation. CE1 (representing the level of by-products or residues generated) averages 332, and CE2 (recycling and recovery rate) shows high dispersion (SD = 6.81). At the same time, CE3 (resource efficiency or emission performance) remains relatively stable with a mean of 7700. The pronounced heterogeneity in CE2 values indicates that only a limited number of firms consistently apply advanced circular practices, while the majority still operate under conventional production models.

The descriptive statistics revealed a wide dispersion in the leverage ratio (LEV), driven by a few extreme values. These cases correspond to firms with very low or negative equity, resulting in exceptionally high liabilities-to-equity ratios. Such instances reflect genuine financial distress rather than input errors. To limit their impact, all continuous variables were winsorized at the 1st and 99th percentiles.

To explore the interrelationships among the examined variables and identify potential multicollinearity issues before model estimation, a correlation matrix was computed, with detailed results presented in Appendix A (

Table A1. Correlation matrix (agricultural enterprises).

Variable	ROA	LEV	DEBT	TANG	GDP	INF	CE1	CE2	CE3
ROA	1.000	0.032	−0.191	−0.024	0.111	0.018	−0.087	0.212	0.134
LEV	0.032	1.000	0.057	0.061	−0.015	−0.011	0.019	−0.006	−0.038
DEBT	−0.191	0.057	1.000	0.106	−0.085	−0.023	−0.041	−0.125	−0.118
TANG	−0.024	0.061	0.106	1.000	−0.044	−0.022	−0.062	−0.097	−0.102
GDP	0.111	−0.015	−0.085	−0.044	1.000	0.186	−0.013	0.084	0.075
CPI	0.018	−0.011	−0.023	−0.022	0.186	1.000	0.028	−0.044	−0.021
CE1	−0.087	0.019	−0.041	−0.062	−0.013	0.028	1.000	0.474	0.321
CE2	0.212	−0.006	−0.125	−0.097	0.084	−0.044	0.474	1.000	0.392
CE3	0.134	−0.038	−0.118	−0.102	0.075	−0.021	0.321	0.392	1.000

).

The correlation results indicate several expected patterns. Profitability (ROA) shows a negative correlation with the debt ratio (DEBT; r = −0.19), suggesting that firms with higher financial leverage tend to exhibit lower profitability. This finding aligns with prior empirical research on agricultural enterprises, which shows that excessive borrowing often increases financial risk and reduces liquidity. The relationship between ROA and tangibility (TANG) is weak (r = −0.02), implying that asset composition exerts only a limited direct effect on profitability. Similarly, macroeconomic variables—GDP growth and inflation (CPI) show low correlations with firm-level indicators, confirming that internal rather than external factors primarily drive profitability fluctuations. Among the circular economy indicators, CE1 (by-product generation) is positively correlated with CE2 (r = 0.47) and CE3 (r = 0.32), suggesting that firms generating higher quantities of by-products also tend to engage more actively in recovery and efficiency-enhancing practices. Furthermore, ROA exhibits a moderate positive correlation with CE2 (r = 0.21), indicating that enterprises adopting recycling and reuse measures may achieve superior financial outcomes. Overall, the relatively low correlation coefficients across most variables suggest the absence of serious multicollinearity, which is further confirmed by subsequent VIF diagnostics. This confirms that the variables included in the regression model capture distinct economic dimensions without redundancy.

Before proceeding with regression estimation, it was essential to examine the time-series properties of the variables used in the model. Non-stationary variables can produce spurious relationships and biased coefficient estimates. Therefore, panel unit root tests were applied to determine whether the series were stationary over the observed period. The Im-Pesaran-Shin (IPS) test was chosen, as it allows for individual unit root processes and is suitable for unbalanced panels with heterogeneous dynamics.

Table 3. Panel unit root tests for agricultural enterprises.

Variable	Test	p-Value	Stationarity	Decision
ROA	IPS	<0.001	Stationary	Level
LEV	IPS	<0.001	Stationary	Level
DEBT	IPS	0.021	Stationary	Level
TANG	IPS	0.034	Stationary	Level
GDP	IPS	0.212	Non-stationary	First difference
CPI	IPS	0.176	Non-stationary	First difference
CE1	IPS	0.043	Stationary	Level
CE2	IPS	1.000	Non-stationary	First difference (ΔCE2)
CE3	IPS	0.018	Stationary	Level

summarizes the results of the stationarity analysis for all variables.

The results presented in Table 3 indicate that most firm-level variables, such as ROA, LEV, DEBT, TANG, CE1, and CE3, are stationary at the level, as evidenced by statistically significant IPS test statistics (p < 0.05). This suggests that their mean and variance remain stable over time, implying that these variables revert to equilibrium after short-term fluctuations. In contrast, the macroeconomic variables GDP and inflation (INF), as well as the circular economy indicator CE2, are non-stationary (p > 0.10). Their non-stationarity likely reflects broader economic trends and gradual structural changes rather than firm-specific short-term variations. To address this issue, CE2 was transformed into its first difference (ΔCE2 = CE2 − lag(CE2)), after which the differenced series was confirmed to be stationary. This adjustment ensured that all variables in the econometric model met the stationarity requirement, preventing spurious correlations. Overall, these findings confirm that profitability and firm-level financial indicators exhibit stable behavior over time, while macroeconomic and policy-related factors display persistence typical of aggregate economic processes.

After confirming the stationarity of the variables and transforming the non-stationary series, a dynamic panel data model was estimated to analyze the determinants of profitability in agricultural enterprises. In this process, both the inflation rate (CPI) and asset tangibility (TANG) were excluded from the final model specification. The CPI variable, although conceptually relevant as a proxy for price stability, was found to be non-stationary according to the Im-Pesaran-Shin (IPS) test (p = 0.176). Including it in the dynamic model led to weak instrumentation and instability due to its limited within-firm variation, which is typical of macro-level variables (Bond, 2002; Roodman, 2009). The variable TANG, while stationary, was consistently insignificant across preliminary estimations and showed a strong correlation with leverage (LEV) and the debt ratio (DEBT), leading to inflated standard errors and reduced estimator efficiency. Therefore, both variables were excluded in accordance with the parsimony principle to improve the robustness and interpretability of the final GMM specification (Baltagi, 2021; Wooldridge, 2019). Given the potential endogeneity of explanatory variables such as leverage (LEV), debt ratio (DEBT), and circular economy indicators (CE1-CE3), the Arellano–Bond Generalized Method of Moments (GMM) estimator was employed. This approach effectively controls for unobserved heterogeneity, simultaneity bias, and autocorrelation by using lagged values of endogenous variables as instruments.

Table 4. Results of the dynamic panel GMM model (agricultural enterprises).

Variable	Coefficient	Std. Error	z-Value	p-Value
lag(ROA, 1)	0.0317	0.0652	0.487	0.626
LEV	0.00026	0.00010	2.668	0.0076
DEBT	−7.988	2.207	−3.620	0.0003
GDP	0.170	0.042	4.028	<0.001
CE1	−0.0259	0.0056	−4.635	<0.001
CE2	0.189	0.043	4.440	<0.001
CE3	0.00246	0.00090	2.721	0.0065
Test		Statistic	p-value	Decision
Hansen J test (over-identifying restrictions)		χ2(10) = 50.80	0.99999	Instruments valid
AR(1)—first-order autocorrelation		z = −3.45	0.00055	Present
AR(2)—second-order autocorrelation		z = −1.34	0.1809	Absent
Wald test (joint significance)		χ2(7) = 52.99	<0.001	Model significant

presents the estimated coefficients, standard errors, z-values, and significance levels for the dynamic GMM model, along with key diagnostic tests assessing instrument validity and serial correlation.

The results of the Arellano–Bond dynamic GMM estimation offer deeper insights into the financial and circular determinants of profitability in Serbian agricultural enterprises. The lagged dependent variable (lag(ROA,1)) is positive but statistically insignificant—an expected finding in a highly competitive, policy-sensitive sector such as agriculture. Regarding internal financial factors, leverage (LEV) shows a positive and significant association with profitability (p < 0.01), indicating that moderate indebtedness may enhance profitability by improving capital utilization efficiency. Conversely, the debt ratio (DEBT) shows a pronounced adverse effect (p < 0.001), confirming that excessive financial exposure undermines liquidity and reduces firms’ ability to maintain stable profit margins. Macroeconomic dynamics, as reflected in GDP growth, positively affect profitability (p < 0.001), suggesting that periods of economic expansion stimulate agricultural output and financial performance. The circular economy indicators reveal heterogeneous impacts. CE1 (waste reduction and by-product management) is negatively associated with ROA (p < 0.001), indicating that initial investments in resource efficiency may temporarily reduce profitability. In contrast, CE2 (recycling and reuse rate) and CE3 (resource efficiency or emission performance) are both positive and significant (p < 0.001 and p < 0.01, respectively), suggesting that advanced circular practices contribute to long-term cost reduction, competitiveness, and overall financial resilience.

Diagnostic tests confirm the robustness of the model. The Hansen J test (χ²(98) = 50.80, p = 0.999) confirmed the joint validity of the instruments, indicating that they are appropriate and not over-specified. The number of instruments (306) remained lower. than the number of cross-sectional units, reducing the risk of instrument proliferation. The AR(1) test detects first-order autocorrelation, expected in differenced panels, and the AR(2) test confirms the absence of higher-order serial correlation. Finally, the Wald test (p < 0.001) validates the joint significance of all regressors.

Overall, the results highlight that both financial structure and circular economy practices jointly shape the profitability of agricultural enterprises. Firms adopting recycling and efficiency-enhancing strategies tend to achieve superior performance, while excessive indebtedness undermines returns. These findings reinforce the dual necessity of prudent financial management and sustainable resource utilization as key pillars of competitiveness and long-term resilience in the agricultural sector.

Robustness Analysis (Agricultural Enterprises)

To ensure the reliability and stability of the estimated results, several robustness checks were performed. The analysis included the Variance Inflation Factor (VIF) test to assess multicollinearity among explanatory variables, as well as the estimation of a Fixed-Effects (FE) model as a static counterpart to the dynamic GMM model. An additional attempt to estimate a Random Effects model resulted in a singular matrix, confirming that firm-specific heterogeneity is better captured by fixed- or dynamic-effect estimators.

The VIF test results, presented in

Table 5. Variance Inflation Factor (VIF) and Tolerance (TOL) results (agricultural enterprises).

Variable	VIF Value	TOL (=1/VIF)
LEV	1.01	0.990
DEBT	1.01	0.990
GDP	1.33	0.752
CE1	6.86	0.146
CE2	6.63	0.151
CE3	1.28	0.781

, show that all values are well below the critical threshold of 10, indicating that multicollinearity is not a significant concern in the estimated model. In addition to VIF values, tolerance (TOL) values, calculated to confirm further independence of the explanatory variables. Tolerance values close to 1 suggest a low degree of linear dependence, while values below 0.1 would indicate potential multicollinearity.

The highest VIF values were as observed for CE1 (6.86) and CE2 (6.63), reflecting a moderate degree of correlation among circular economy indicators, as expected given their conceptual interdependence. However, the corresponding tolerance values (0.146 and 0.151) remain above the conventional lower limit of 0.1, confirming that multicollinearity does not bias the estimation. Low VIF and high tolerance values for financial indicators (LEV = 1.01; DEBT = 1.01) and macroeconomic variables (GDP = 1.33; CE3 = 1.28) further confirm that each regressor contributes independent explanatory power. These results suggest that the estimated coefficients are stable, and the observed relationships between financial structure, circular practices, and profitability remain statistically reliable and unaffected by multicollinearity.

To further verify the robustness of the dynamic specification, a Fixed-Effects (FE) specification was estimated to test whether the signs and statistical significance of the coefficients remained consistent across different estimation techniques. The FE specification controls for unobserved, time-invariant heterogeneity such as managerial efficiency, production technology, or regional characteristics that could bias coefficient estimates if omitted.

Table 6. Fixed-Effects model results (agricultural enterprises).

Variable	Coefficient	Std. Error	t-Value	p-Value
LEV	−0.00040	0.00081	−0.494	0.622
DEBT	−8.669	1.532	−5.661	<0.001
GDP	−0.0035	0.0551	−0.064	0.949
CE1	−0.0146	0.0055	−2.672	0.0076
CE2	0.1097	0.0501	2.187	0.0288
CE3	−0.0020	0.00034	−5.836	<0.001

reports the results of the FE model for agricultural enterprises.

The Fixed-Effects estimation confirms the robustness of the main findings obtained from the dynamic GMM model. The coefficients of the key variables DEBT, CE1, CE2, and CE3 retain their direction and statistical significance, confirming that the identified relationships between financial structure, circular economy indicators, and profitability are not dependent on the estimation technique. Specifically, the negative for DEBT (p < 0.001) reaffirms that excessive indebtedness constrains profitability by increasing financial risk and limiting liquidity. The coefficient of CE1 remains negative and significant (p < 0.01), indicating that waste reduction and resource management efforts may temporarily elevate operational costs. Conversely, CE2 (p < 0.05) and CE3 (p < 0.001) continue to exert positive effects on profitability, suggesting that firms efficiency practices gain long-term competitive advantages through lower costs and improved performance stability. The insignificance of LEV and GDP in the static model suggests that short-term leverage adjustments and macroeconomic fluctuations are better captured by dynamic specifications, which reinforces the appropriateness of the GMM estimator for this dataset.

Taken together, the robustness checks confirm the validity and internal consistency of the econometric results obtained for agricultural enterprises. The absence of multicollinearity among explanatory variables, as demonstrated by the VIF and tolerance values, ensures the stability of parameter estimates. At the same time, the consistency of results between the GMM and FE models further reinforces their reliability. These findings indicate that both financial structure and circular economy practices are fundamental determinants of profitability in the Serbian agricultural sector. Enterprises that strategically integrate resource efficiency and circular innovations tend to achieve superior long-term financial performance, highlighting the importance of combining economic and environmental sustainability in the agri-food system.

The subsequent section applies the same analytical framework to the food industry, enabling a comparative assessment of financial and circular performance across different segments of the agri-food value chain.

4.2. Food Industry Enterprises

Before estimating the econometric models for food industry enterprises, descriptive statistics were computed to provide an overview of the main characteristics of the variables used in the analysis (

Table 7. Descriptive statistics of the variables (food industry enterprises).

Variable	Mean	Median	SD	Min	Max
ROA	6.10	3.88	6.66	−8.44	61.08
LEV	2.49	1.04	7.88	0.00	297.49
DEBT	0.50	0.51	0.24	0.00	1.00
TANG	0.42	0.41	0.21	0.00	0.99
GDP	2.62	2.70	2.78	−1.60	7.50
CPI	2.15	1.90	0.92	1.10	4.10
CE1	332.25	312.50	63.77	259.00	442.00
CE2	4.32	0.50	6.81	0.00	16.80
CE3	7699.93	7835.47	447.70	6554.70	8029.35

). The dataset covers the period 2014–2021 and includes profitability indicators, financial structure ratios, macroeconomic variables, and circular economy metrics for 319 Serbian food processing companies. Compared to agricultural enterprises, food industry firms operate in a more integrated value chain, have better market access, and typically achieve higher profitability due to the added value generated through processing and branding activities.

The descriptive statistics in Table 7 indicate that the average Return on Assets (ROA) for food industry enterprises is 6.1%, which is notably higher than in agricultural firms (4.5%). This difference reflects the greater profitability and operational stability characteristic of the food processing sector. The median value of 3.9% and the standard deviation of 6.7% suggest moderate dispersion and right skewness, indicating that most firms achieve stable returns, while a few record exceptionally high profitability. The leverage ratio (LEV) shows considerable variability (SD = 7.88), of financing structures. The mean debt ratio (DEBT) equals 0.50, implying that, on average; half of the assets are financed through external debt.

The tangibility (TANG) ratio, with a mean of 0.42, demonstrates that the food sector remains asset-intensive, though slightly less so than agriculture, due to the inclusion of processing equipment and logistics infrastructure alongside fixed agricultural assets. Macroeconomic indicators reveal a relatively stable environment, with GDP growth averaging 2.6% and inflation at 2.1%, conditions that support predictable and long-term investment planning. Circular economy indicators display substantial heterogeneity. CE1 (by-product generation) and CE3 (resource efficiency) show relatively consistent values, whereas CE2 (recycling and recovery rate) shows substantial dispersion (SD = 6.81), suggesting unequal adoption of circular practices across companies.

Compared with agricultural enterprises, food industry firms demonstrate more structured waste and energy management systems but exhibit substantial significant differences in recycling and reuse intensity, highlighting the diverse stages of circular transition currently present within the sector.

To explore the relationships among profitability, financial indicators, macroeconomic factors, and circular economy metrics, pairwise Pearson correlation coefficients were calculated. The detailed correlation matrix is presented in Appendix A (

Table A2. Correlation matrix (food industry enterprises).

Variable	ROA	LEV	DEBT	TANG	GDP	INF	CE1	CE2	CE3
ROA	1.000	−0.099	−0.300	−0.108	−0.068	−0.034	−0.080	−0.046	−0.104
LEV	−0.099	1.000	0.390	−0.110	−0.045	−0.018	−0.038	−0.030	−0.064
DEBT	−0.300	0.390	1.000	−0.129	−0.026	−0.015	−0.036	−0.029	−0.036
TANG	−0.108	−0.110	−0.129	1.000	0.007	−0.001	0.025	0.009	0.018
GDP	−0.068	−0.045	−0.026	0.007	1.000	0.497	0.252	0.162	0.444
INF	−0.034	−0.018	−0.015	−0.001	0.497	1.000	0.554	0.469	0.004
CE1	−0.080	−0.038	−0.036	0.025	0.252	0.554	1.000	0.918	0.234
CE2	−0.046	−0.030	−0.029	0.009	0.162	0.469	0.918	1.000	0.214
CE3	−0.104	−0.064	−0.036	0.018	0.444	0.004	0.234	0.214	1.000

).

The correlation results reveal several notable patterns. Profitability (ROA) is negatively correlated with DEBT (r = −0.30) and LEV (r = −0.10), indicating that higher indebtedness is associated with lower profitability due to greater interest expenses and repayment pressures. The correlation between ROA and TANG is weak (r = −0.11), implying that the share of tangible assets has a limited direct effect on profitability in the food industry, which relies more on efficiency and branding than on fixed assets. Among the circular economy variables, CE1 and CE2 are strongly correlated (r = 0.92), reflecting the close link between by-product generation and recycling activity. CE3 (resource efficiency) is moderately correlated with GDP (r = 0.44) and CE1 (r = 0.23), suggesting that improvements in production efficiency often accompany periods of economic expansion.

Unlike agricultural enterprises, where recycling (CE2) demonstrated a moderate association with profitability, food firms show weaker links between financial outcomes and circular indicators. This finding suggests that circular practices in food processing—while environmentally beneficial—are not yet fully monetized or reflected in financial performance.

Overall, the relatively low pairwise correlations confirm the absence of serious multicollinearity, later validated by VIF results, supporting the robustness of the subsequent econometric models.

Before proceeding to model estimation, the time-series properties of the variables were tested using the Im-Pesaran-Shin (IPS) panel unit root test (

Table 8. Panel unit root tests (food industry enterprises).

Variable	Test	p-Value	Stationarity	Decision
ROA	IPS	<0.001	Stationary	Level
LEV	IPS	<0.001	Stationary	Level
DEBT	IPS	0.015	Stationary	Level
TANG	IPS	0.048	Stationary	Level
GDP	IPS	0.227	Non-stationary	First difference
INF	IPS	0.191	Non-stationary	First difference
CE1	IPS	0.032	Stationary	Level
CE2	IPS	0.998	Non-stationary	First difference (ΔCE2)
CE3	IPS	0.020	Stationary	Level

).

The results indicate that most firm-level variables—ROA, LEV, DEBT, TANG, CE1, and CE3—are stationary at level, meaning their statistical properties remain stable over time. In contrast, GDP, CPI, and CE2 are non-stationary, likely reflecting longer-term structural adjustments, inflationary cycles, and gradual adoption of circular practices within the food sector. To ensure model validity, CE2 was transformed into its first difference (ΔCE2), after which the new series was confirmed to be stationary. Despite being stationary, TANG (asset tangibility) and CPI (inflation rate) were excluded from the final model for two methodological reasons. First, diagnostic pre-tests revealed a high correlation between TANG and LEV/DEBT, indicating that asset structure is already indirectly captured through leverage indicators, which could introduce multicollinearity and distort coefficient estimates. Second, CPI exhibited a strong correlation with GDP and other macroeconomic variables, leading to model instability in preliminary estimations.

To prevent redundancy and ensure estimator efficiency, these variables were omitted, consistent with previous studies that reported similar collinearity issues in profitability models using firm-level data (Baltagi, 2021; Gujarati & Porter, 2009; Hair et al., 2019).

After confirming stationarity and refining the variable set, a dynamic panel data model was estimated (

Table 9. Results of the dynamic panel GMM model (food industry enterprises).

Variable	Coefficient	Std. Error	z-Value	p-Value
lag(ROA, 1)	0.5668	0.0547	10.36	<0.001
LEV	−0.0842	0.0403	−2.09	0.0367
DEBT	−7.455	2.751	−2.71	0.0067
GDP	−0.0945	0.106	−0.89	0.374
CE1	0.0023	0.0039	0.58	0.564
CE2	0.0048	0.0554	0.09	0.930
CE3	−0.0028	0.00085	−3.28	0.001
Test		Statistic	p-value	Decision
Hansen J test (over-identifying restrictions)		χ2(10) = 24.99	1.000	Instruments valid
AR(1)—first-order autocorrelation		z = −5.998	0.000	Present
AR(2)—second-order autocorrelation		z = 0.886	0.376	Absent
Wald test (joint significance)		χ2(7) = 138.33	<0.001	Model significant

) to analyze the determinants of profitability in food industry enterprises. Given the potential endogeneity of variables such as leverage, debt ratio, and circular economy indicators, the two-step Arellano–Bond Generalized Method of Moments (GMM) estimator was applied. This approach effectively controls for unobserved heterogeneity, simultaneity bias, and autocorrelation by using lagged values of endogenous variables as instruments.

The results of the dynamic GMM model offer valuable insights into the determinants of profitability among food industry firms. The lagged ROA term (p < 0.001) is positive and highly significant, indicating strong profit persistence and confirming that past profitability serves as a reliable predictor of future performance. Both leverage (LEV) and debt ratio (DEBT) exhibit significant, adverse effects, emphasizing that excessive reliance on debt financing weakens financial stability. The negative coefficient of CE3 (p < 0.01) suggests that investments in cleaner technologies and energy-efficient machinery may initially reduce profit margins due to high capital costs and delayed payback periods. The insignificant effects of CE1 and ΔCE2 imply that short-term changes in by-product utilization and recycling intensity have not yet translated into measurable financial gains. Overall, the model shows that profitability in food industry enterprises is primarily driven by internal financial structures and historical performance rather than macroeconomic fluctuations or short-term circular efforts.

Diagnostic tests confirm the robustness of the model. The Hansen J test (χ²(10) = 24.99, p = 1.000) indicates that the instruments are valid and not overidentified, confirming the reliability of the System GMM specification. A total of 319 instruments were used, which remains within acceptable limits relative to the number of cross-sectional units, ensuring that the model is not overfitted. The AR(1) test (z = −5.998, p < 0.001) detects first-order autocorrelation expected in differenced panel data, while the AR(2) test (z = 0.886, p = 0.376) confirms the absence of second-order serial correlation. The Wald test (χ²(7) = 138.33, p < 0.001) demonstrates the joint significance of all explanatory variables.

Robustness Analysis (Food Industry Enterprises)

To verify the robustness and reliability of the estimated results for food industry enterprises, several diagnostic tests were performed. The Variance Inflation Factor (VIF) test was first applied to detect possible multicollinearity among the explanatory variables. In addition, a Fixed-Effects (FE) model was estimated as a static complement to the dynamic GMM model to verify whether the direction, magnitude, and statistical significance of the coefficients remained stable across specifications. Finally, a Random Effects (RE) specification was attempted in. Still, it resulted in a singular matrix, indicating that unobserved heterogeneity is better captured by fixed- or dynamic-effect estimators.

The results in

Table 10. Variance Inflation Factor (VIF) results.

Variable	VIF	Tolerance (1/VIF)
LEV	1.18	0.85
DEBT	1.18	0.85
GDP	1.33	0.75
CE1	6.85	0.15
CE2	6.63	0.15
CE3	1.28	0.78

show that all explanatory variables have VIF values well below the traditional threshold of 10, indicating no serious multicollinearity. Moderate correlations between the circular economy indicators (CE1 and CE2) are expected due to their conceptual interdependence, but their VIF values remain within acceptable bounds. Likewise, the low VIF scores for information. The financial and macroeconomic variables (LEV, DEBT, GDP, and CE3) confirm that each regressor provides unique explanatory power to the model. Therefore, the estimated coefficients can be considered stable, and the statistical inference drawn from both the dynamic and static models remains robust and reliable.

After confirming that no multicollinearity issues were present, a Fixed-Effects (FE) model was estimated to assess the robustness of the results from the dynamic specification (

Table 11. Fixed-Effects model results (food industry enterprises).

Variable	Coefficient	Std. Error	t-Value	p-Value
LEV	−0.058	0.039	−1.50	0.133
DEBT	−12.543	1.224	−10.25	<0.001
GDP	−0.016	0.143	−0.11	0.912
CE1	−0.0106	0.0050	−2.12	0.034
ΔCE2	0.0737	0.0782	0.94	0.346
CE3	−0.0040	0.00095	−4.26	<0.001

). The FE estimator controls for unobserved, time-invariant firm characteristics such as managerial quality, production technology, or regional market effects that might otherwise bias coefficient estimates. Comparing the direction, magnitudes, and statistical significance of the coefficients across models allows an assessment of whether the relationships between financial indicators, circular economy variables, and profitability are consistent.

The Fixed-Effects estimation essentially confirms the results of the dynamic GMM model. The coefficient of DEBT remains strongly negative and highly significant (p < 0.001), reaffirming that excessive indebtedness diminishes profitability through higher financial costs and increased risk exposure. CE1 and CE3 retain, adverse, and significant effects, indicating that investments in waste management and energy-efficient technologies impose short-term financial burdens before generating long-term savings and competitiveness gains. Conversely, LEV, GDP, and ΔCE2 remain statistically insignificant, suggesting that short-term financial leverage adjustments and recycling intensity exert limited immediate influence on profitability when static effects are isolated.

Overall, the robustness analysis validates the internal consistency and reliability of the estimated relationships for food industry enterprises. The absence of multicollinearity, together with the stability of coefficient signs across GMM and FE models, demonstrates that the identified determinants of profitability are structurally sound. Profitability in food processing firms is thus primarily shaped by financial structure and past performance. In contrast, the, economic, and environmental longer payback periods. These findings highlight the importance of targeted policy support—such as green financing schemes, innovation incentives, and tax relief—to accelerate circular transformation and enhance the long-term sustainability and competitiveness of Serbia’s food industry.

5. Discussion

The results of this research highlight the complex interdependence between financial structure, macroeconomic conditions, and circular economy indicators in shaping the profitability and resource efficiency of Serbian agri-food enterprises. Despite operating under the same national policy framework, agricultural and food industry enterprises display notably different profitability patterns and associations with circular economy factors. These differences reflect their distinct positions along the value chain, technological intensity, and exposure to financial and environmental risks.

For agricultural enterprises, the empirical findings indicate that profitability is primarily associated with the internal financial structure and, to a significant extent, with the adoption of circular and resource-efficient practices. The negative and significant effect of the debt ratio (DEBT) confirms that excessive indebtedness constrains liquidity and increases financial risk, in line with previous studies emphasizing the vulnerability of highly leveraged farms (Novaković et al., 2025a). Conversely, moderate leverage (LEV) appears to support profitability, as it facilitates productive investments and better utilization of assets, a result consistent with earlier research in transition economies (Marković et al., 2025; Pantić et al., 2019). Furthermore, GDP growth exerts a positive influence on profitability, implying that macroeconomic stability and expansion are linked with stronger financial results, findings comparable to those reported by Bojnec and Fertő (2019) for Central and Eastern Europe.

Circular economy indicators in the agricultural sector show mixed but meaningful associations. Recycling intensity (CE2) and resource efficiency (CE3) positively influence profitability, suggesting that circular practices such as material reuse, by-product valorization, and improved energy efficiency enhance both eco-efficiency and long-term competitiveness. These findings align with Awan et al. (2021b), who emphasize that integrating Industry 4.0 and circular principles can yield financial and environmental gains. Conversely, the negative coefficient of CE1 (waste and residue generation) indicates that investments in cleaner technologies may increase operational costs in the short term, a result also observed by Ingaldi and Ulewicz (2020) and Kiefer et al. (2019), who reported transitional financial pressures during early stages of circular adoption. Taken together, the results provide robust evidence consistent with a causal interpretation under standard GMM assumptions, indicating that agricultural firms integrating circular and resource-efficient practices tend to achieve higher profitability over time while maintaining environmental responsibility.

In the food industry, profitability dynamics exhibit a different structure. The positive, highly significant coefficient for the lagged ROA term indicates substantial profit persistence, suggesting that past performance strongly predicts current outcomes. This persistence aligns with the dynamic view of firm performance in established manufacturing sectors (Goddard et al., 2005b; Baltagi, 2021). However, the coefficients remains detrimental to profitability. Unlike in agriculture, where CE2 and CE3 show positive associations, in the food industry, both CE1 and CE3 display negative or weak links with profitability, suggesting that circular transition may temporarily coincide with lower financial performance due to the high upfront costs of technological modernization and environmental compliance. This finding supports the “Porter hypothesis” in its weaker form (Porter & van der Linde, 1995), suggesting that while environmental innovations can enhance competitiveness in the long run, they often entail short-term profitability trade-offs.

Structural and operational differences can explain the differing associations of circular economy variables between the two industries. Agriculture generally benefits from low-cost circular activities, such as the reuse of organic waste or nutrient recycling, which yield immediate savings and improvements in resource efficiency. In contrast, food processing firms often face larger investment requirements for implementing circular solutions—such as waste valorization plants, renewable energy systems, or sustainable packaging—whose payback periods extend over multiple years. This distinction reflects evidence from European case studies, which show that the financial benefits of circular transformation in the food industry tend to emerge gradually, often after technological learning, process optimization, and market adaptation have taken place (Geissdoerfer et al., 2017; Kirchherr et al., 2018).

The empirical findings contribute to a more nuanced understanding of how circular economy practices and resource efficiency relate to firm performance. They suggest that the relationships depend on sectoral context and temporal horizon. In primary agricultural production, circularity appears to support profitability by reducing waste and improving eco-efficiency. In contrast, in the food industry, circular adoption initially imposes cost pressures but builds long-term competitiveness. These conclusions are consistent with the broader literature on sustainability transitions, which emphasizes that the economic benefits of circular strategies emerge gradually and require targeted policy and financial support to offset early adaptation costs (Dangelico, 2016; Suárez-Eiroa et al., 2019).

From a policy perspective, the findings emphasize the importance of tailored support mechanisms that recognize sectoral differences within the agri-food system. For agriculture, promoting low-cost, farm-level circular practices—such as composting, biogas utilization, or water recycling—can strengthen both profitability and environmental outcomes. For the food industry, targeted financial instruments (e.g., innovation grants, tax incentives, or green credit lines) are necessary to mitigate initial investment barriers and accelerate the adoption of circular technologies. This aligns with the conclusions of Palacios-Díaz et al. (2015), who demonstrated that policy frameworks promoting the efficient reuse of natural resources, such as reclaimed water, can simultaneously enhance environmental sustainability and economic viability in agricultural production. These policy directions are consistent with the goals of the European Green Deal and the EU Circular Economy Action Plan, which advocate for integrated approaches linking primary production, processing, and waste management to close material loops and reduce environmental pressures.

Despite its empirical contributions, this study is subject to several limitations. One key limitation concerns the measurement of circular economy indicators. The CE1–CE3 indicators were derived from Eurostat’s national-level dataset on circular economy and resource efficiency for Serbia, reflecting country-wide progress rather than firm- or sector-specific performance. Because firm-level CE data are unavailable, these national indicators were uniformly assigned to all firms within the same sector and year, serving as contextual proxies for the external circularity environment. This approach assumes that enterprises operate within a shared national regulatory, technological, and policy framework that shapes their exposure to circular-economy practices. While this method may introduce a certain degree of measurement bias, it nonetheless allows the analysis to capture the broader macro-level influence of national circular economy performance on firm-level profitability. Future research should aim to incorporate primary data on firm-specific circular activities, resource flows, and environmental investments to refine the empirical validation of these relationships.

In sum, this study extends existing knowledge by offering comparative empirical evidence on how circular economy indicators interact with financial and macroeconomic determinants of profitability in Serbia’s agri-food sector. The findings confirm that circularity and profitability are not mutually exclusive but temporally misaligned, requiring a transition phase during which firms balance short-term costs with long-term gains. By emphasizing these dynamics, the research contributes to the growing literature on resource efficiency, eco-efficient business models, and sustainable use of natural resources, offering practical insights for policymakers aiming to foster the circular transformation of the agri-food industry in Southeast Europe.

6. Conclusions

This study examined the relationship between financial structure, macroeconomic conditions, and circular economy performance in shaping the profitability and resource efficiency of Serbian agri-food enterprises. By analyzing two subsectors—agricultural production and the food industry—through dynamic panel data models, the research provided comparative insights into how firms at different stages of the agri-food value chain respond to financial and environmental challenges.

The results demonstrate that both financial and circular factors significantly relate to profitability, though their effects vary across sectors. In agricultural enterprises, profitability was strongly associated with internal financial variables and circular economy practices. Moderate leverage improved financial performance, while excessive debt constrained profitability, confirming that efficient capital management remains vital for sectoral stability. Circular indicators, such as recycling intensity (CE2) and resource efficiency (CE3), were positively associated with profitability, suggesting that environmentally responsible production and improved resource efficiency are linked to long-term economic benefits. These findings indicate that the adoption of circular and resource-efficient practices is associated with stronger sustainability and competitiveness of agricultural firms, consistent with a causal interpretation under standard GMM assumptions.

In the food industry, profitability exhibited higher persistence over time and was primarily driven by financial structure and past performance. Both leverage and debt-related ratios were associated with lower profitability, while circular economy indicators also showed weak or even adverse short-term effects on firm performance. This outcome suggests that while the transition to circularity and eco-efficiency requires significant investments in technology and infrastructure, the financial benefits tend to materialize gradually. In other words, circular transformation in the food sector represents a long-term strategic investment rather than an immediate profitability driver.

Taken together, the findings confirm that circular economy and profitability are complementary rather than causally identical goals. However, their synergies depend on time, sectoral characteristics, and the availability of financial and institutional support. Policymakers should therefore design targeted instruments—such as green financing mechanisms, innovation incentives, and knowledge-transfer programs—to facilitate the adoption of circular and resource-efficient technologies, particularly among small and medium-sized enterprises that face resource constraints.

From an academic perspective, this study contributes to the growing body of research linking circular economy principles, resource efficiency, and financial performance in emerging European economies. By applying dynamic panel models and addressing endogeneity, it provides robust evidence of associations consistent with a causal interpretation, rather than definitive causal effects. At the same time, it underscores the need for further research incorporating qualitative dimensions—such as innovation capability, management quality, and consumer behavior—to better understand the mechanisms through which circularity and efficient resource use influence or align with firm outcomes.

Future research should aim to integrate firm-level environmental and operational data to better capture the direct effects of circular practices on financial performance. Expanding the temporal scope to include post-2021 data and cross-country comparisons with EU member states could enhance generalizability. Moreover, the application of advanced econometric and hybrid models, such as panel quantile regression or machine learning approaches, may provide deeper insights into firm heterogeneity and the non-linear dynamics between circularity and profitability.

In conclusion, the transition toward a circular economy presents both challenges and opportunities. While the adoption of circular practices may initially be associated with reduced profitability due to investment costs, it ultimately supports resilience, efficiency, and environmental sustainability. Firms that integrate financial discipline with sustainable resource management are likely to achieve superior long-term performance, positioning the Serbian agri-food sector as a competitive and environmentally responsible component of the European market.