Determinants of Capital Structure Under Financial Constraints: Debt Composition in Moroccan Agricultural SMEs

Abstract

This study investigates the determinants of capital structure in Moroccan agricultural SMEs, with particular emphasis on the distinction between interest-bearing debt and non-interest-bearing liabilities in a context characterized by persistent credit constraints. While traditional capital structure theories typically treat debt as a homogeneous aggregate, such an approach may obscure important financing dynamics in financially constrained environments. Using a panel dataset of 52 agricultural SMEs observed over the period 2017–2022, the analysis employs a correlated random effects model to control for unobserved heterogeneity. The results indicate a negative relationship between profitability and both total and short-term debt, consistent with the predictions of the Pecking Order Theory. Liquidity, asset tangibility, and firm size are negatively associated with non-interest-bearing current liabilities, suggesting that trade-based financing may serve as an adjustment mechanism when access to formal credit is limited. In contrast, long-term debt is only weakly explained by firm-level characteristics, pointing to potential supply-side constraints in agricultural credit markets. Overall, the findings suggest that financing patterns in agricultural SMEs appear to be more closely associated with credit market imperfections than with optimal trade-off considerations. By distinguishing between different debt components, this study contributes to the literature by highlighting the importance of debt composition when analyzing capital structure in emerging and financially constrained economies.

1. Introduction

Capital structure remains one of the most extensively debated issues in corporate finance, particularly in the context of small and medium-sized enterprises, whose financing decisions are shaped by information asymmetries, limited access to capital markets, and institutional frictions. The trade-off theory (Kraus & Litzenberger, 1973; Myers, 1984) posits the existence of an optimal leverage ratio balancing the tax advantages of debt against expected financial distress costs. In contrast, the pecking order theory (Myers & Majluf, 1984) argues that firms follow a financing hierarchy driven by asymmetric information, prioritizing internal funds over external debt and resorting to equity only as a last option. While empirical evidence for SMEs often supports the pecking order framework, findings remain heterogeneous, particularly in emerging economies characterized by structurally imperfect credit markets. Moreover, much of the empirical literature treats debt as a homogeneous aggregate, potentially masking fundamental differences between interest-bearing bank debt and non-interest-bearing liabilities such as trade credit.

These theoretical debates are especially salient in financially constrained environments. In Morocco, SMEs are widely recognized as key drivers of economic growth and employment; yet their development remains significantly hindered by limited access to finance. Empirical evidence indicates that structural characteristics of the Moroccan financial system—namely stringent collateral requirements, pronounced information asymmetries, and credit rationing practices—systematically restrict SMEs’ access to formal lending (Boutfssi & Quamar, 2024). A substantial proportion of firms are either denied bank credit or granted financing under unfavorable conditions due to perceived risk and insufficient transparency (Abdelghani et al., 2024; Merroun & Hamiche, 2023). These constraints are further reinforced by the predominance of traditional banking as the primary source of external finance and the limited development of alternative instruments such as equity financing, leasing, or participatory finance (Didi Seddik et al., 2024; Abdelghani et al., 2024).

Within this broader SME financing landscape, the agricultural sector occupies a strategic position in the Moroccan economy. According to the High Commission for Planning, agriculture contributes approximately 12% of GDP, 20% of total exports, and nearly 40% of national employment (HCP, 2024, 2025). Agricultural SMEs and structured family farms play a central role in value creation, rural employment, and food security. However, their financial constraints are amplified by sector-specific risks, including income seasonality, climate variability, recurrent drought exposure, and price volatility (African Development Bank Group, 2019; Louali, 2019). The interaction between structural credit market imperfections and sectoral risk factors thus generates a persistent financing gap that constrains long-term investment and modernization.

In response, public authorities have prioritized agricultural transformation through the “Green Generation” strategy, which builds upon the achievements of the Plan Maroc Vert and emphasizes sustainability, value-chain integration, youth entrepreneurship, and technological upgrading (Département de l’Agriculture, 2020; MAPMDREF, 2020). However, the effectiveness of such policies largely depends on the availability of appropriate and diversified financing mechanisms capable of supporting agricultural investment. Understanding how agricultural SMEs structure their liabilities under institutional and market imperfections is therefore essential both for assessing the effectiveness of public policies and for evaluating the empirical relevance of capital structure theories in financially constrained environments.

Despite the extensive literature on SME financing, most empirical studies—particularly in emerging economies—rely on aggregate leverage indicators. Short-term leverage is commonly measured as total current liabilities relative to total assets, without distinguishing between interest-bearing bank debt and non-interest-bearing liabilities such as trade credit, accruals, or shareholders’ current accounts. Yet these financing instruments differ substantially in terms of contractual rigidity, renegotiation flexibility, cost structure, and dependence on formal credit markets. Aggregating them into a single leverage indicator may therefore obscure important substitution mechanisms and lead to ambiguous interpretations regarding the predictions of the trade-off and pecking order theories.

Although a growing body of research examines capital structure determinants in Morocco, existing empirical evidence remains limited and largely focused on listed firms or broad cross-sectoral SME samples. For instance, Boumlik et al. (2025) provide valuable insights into the financing behavior of Moroccan family-owned SMEs. However, sector-specific evidence remains scarce, particularly for agricultural SMEs. This limitation largely reflects the restricted availability of firm-level accounting data for non-listed agricultural enterprises, which are not accessible through centralized public databases and must typically be collected individually from official filings. Consequently, the financing patterns of agricultural SMEs—especially the relative roles of bank debt and non-interest-bearing liabilities—remain largely unexplored in the Moroccan context.

The Moroccan agricultural sector therefore provides a particularly informative setting to reassess the empirical relevance of competing capital structure theories. Agricultural firms face pronounced income seasonality, climatic shocks, and biological production cycles, all of which increase earnings volatility and reduce cash-flow predictability. Under such conditions, fixed-interest obligations may amplify expected financial distress costs, potentially weakening optimal leverage targeting as predicted by the trade-off theory. At the same time, asset specificity and limited collateral liquidity may restrict access to formal long-term credit, thereby reinforcing reliance on relationship-based and non-interest-bearing financing mechanisms, consistent with the hierarchical financing behavior emphasized by the pecking order theory.

Agricultural SMEs thus represent a valuable empirical context in which firm-level optimization incentives and structural credit supply constraints may diverge more clearly than in less volatile industries. By explicitly distinguishing between interest-bearing and non-interest-bearing liabilities, this study does not merely disaggregate leverage; rather, it examines whether debt composition reflects rational financial adaptation to sectoral risk or evidence of segmented credit markets.

This study contributes to the literature in three main ways. First, it provides novel firm-level evidence on the determinants of debt composition in Moroccan agricultural SMEs, a sector that remains largely underexplored in empirical capital structure research. Second, by distinguishing between interest-bearing and non-interest-bearing liabilities, the analysis moves beyond aggregate leverage measures and highlights the mechanisms through which firms adjust their financing structure under credit constraints. Third, the study offers new insights into the relative explanatory power of the trade-off and pecking order theories in a context characterized by structural credit market imperfections.

Accordingly, three central research questions guide the analysis: (i) What firm-level determinants explain the use of interest-bearing versus non-interest-bearing debt among Moroccan agricultural SMEs? (ii) To what extent do observed financing patterns align with the predictions of the trade-off and pecking order theories? and (iii) Does debt composition provide evidence of adaptive financial flexibility or of binding credit supply constraints in a segmented financial system?

Using a balanced panel dataset of 52 agricultural SMEs over the period 2017–2022 and a correlated random effects specification, this study examines the influence of profitability, asset tangibility, liquidity, growth opportunities, firm size, and firm age on different debt components. The empirical results indicate that internal financial capacity is negatively associated with the use of interest-bearing debt, consistent with pecking order predictions. Asset tangibility and firm size are negatively associated with reliance on non-interest-bearing liabilities, suggesting that firms with greater collateral capacity and organizational scale tend to depend less on informal financing channels. Sales growth is positively associated with short-term bank debt, consistent with higher working capital requirements during expansion phases. Non-interest-bearing liabilities therefore appear to function as adjustment mechanisms in contexts characterized by credit constraints. By contrast, long-term bank debt is only weakly associated with firm-level characteristics, pointing to possible supply-side limitations in agricultural credit markets. Overall, the findings suggest that capital structure patterns in Moroccan agricultural SMEs appear to be more closely associated with institutional credit constraints than with active leverage targeting, underscoring the importance of debt composition in financially segmented environments.

The remainder of the paper is organized as follows. Section 2 reviews the theoretical and empirical literature and develops the research hypotheses. Section 3 presents the data and empirical methodology. Section 4 reports the empirical results. Section 5 discusses the findings and provides robustness analyses. Section 6 concludes and outlines implications for policy and future research.

2. Literature and Development of Hypotheses

2.1. Theoretical Perspectives

The financing behavior of small and medium-sized enterprises (SMEs) has traditionally been analyzed through two dominant frameworks in capital structure research: the Trade-Off Theory (TOT) and the Pecking Order Theory (POT). These theoretical perspectives provide contrasting explanations of how firms determine their financing choices and how leverage evolves over time.

The Trade-Off Theory (Kraus & Litzenberger, 1973; Myers, 1984), building upon the foundational contributions of Modigliani and Miller (1958, 1963), argues that firms determine an optimal capital structure by balancing the tax advantages of debt against the expected costs of financial distress and agency conflicts (Jensen & Meckling, 1976; Titman & Wessels, 1988). Within this framework, debt financing increases firm value up to the point where the marginal costs associated with bankruptcy risk and agency problems offset the fiscal benefits of leverage. However, the empirical relevance of a clearly defined target leverage ratio becomes less evident in the case of SMEs. Small firms often operate under conditions of higher earnings volatility, limited collateral, and greater exposure to idiosyncratic risk. These characteristics are particularly pronounced in the agricultural sector, where income seasonality, climatic shocks, and asset specificity substantially increase the expected costs of financial distress. Under such circumstances, the adjustment mechanisms predicted by the Trade-Off Theory may become weaker or less observable.

In contrast, the Pecking Order Theory (Myers, 1984; Myers & Majluf, 1984) rejects the notion of an optimal capital structure and instead explains financing choices through information asymmetries between firm insiders and external investors. According to this framework, firms follow a hierarchical financing order: internal funds are preferred first, followed by debt, while external equity is used only as a last resort. For SMEs—whose financial transparency is often limited and whose access to capital markets is constrained—this hierarchy implies that leverage levels are primarily determined by the availability of internally generated resources. Consequently, more profitable firms are expected to rely less on external debt, leading to a negative relationship between profitability and leverage.

Beyond these canonical theories, a growing body of research highlights the importance of institutional imperfections and credit market segmentation in shaping SME financing behavior. In many emerging and financially underdeveloped economies, limited financial intermediation and credit rationing restrict firms’ access to formal bank financing, encouraging the development of alternative financing arrangements.

Fisman and Love (2003) demonstrate that in countries with underdeveloped financial systems, firms substitute trade credit for bank financing. Industries more reliant on supplier credit exhibit relatively stronger growth in financially constrained environments, suggesting that inter-firm credit serves as a second-best financing mechanism when formal financial intermediation is limited.

These empirical observations are consistent with the contract-theoretic model developed by Burkart and Ellingsen (2004). Their framework suggests that suppliers may extend credit more readily than banks because physical inputs are more difficult for borrowers to divert than cash loans. As a result, trade credit may arise even when bank lending is rationed. Importantly, their model predicts that trade credit can act either as a complement to or a substitute for bank debt, depending on monitoring incentives and market conditions.

Similarly, Petersen and Rajan (1994, 1997) document that small firms increase their reliance on trade credit when access to institutional finance becomes constrained. While trade credit may initially serve to reduce transaction costs, its persistent use as a financing instrument often signals underlying bank credit rationing. Moreover, suppliers implicitly limit the maturity of such credit through early-payment discounts and penalties, indicating that trade credit cannot fully substitute for formal bank lending.

Additional evidence on this substitution mechanism is provided by García-Teruel and Martínez-Solano (2010), who document a negative relationship between access to bank financing and the use of trade credit among UK SMEs. Firms with stronger internally generated cash flows or better access to short- and long-term bank financing tend to rely less on supplier credit. Conversely, firms facing stronger financing constraints—particularly those with significant growth opportunities—tend to increase their reliance on trade credit to support expansion.

Complementing these perspectives, the financial growth cycle framework (Mac An Bhaird & Lucey, 2011; Berger & Udell, 1998) emphasizes that SME capital structures evolve over the firm lifecycle. Younger firms typically depend more heavily on owner contributions, personal guarantees, and external debt, whereas more mature firms gradually accumulate retained earnings and build stronger relationships with financial institutions. This dynamic view partly supports the pecking order logic while incorporating lifecycle effects and relational financing dynamics.

Robb and Robinson (2014) challenge the traditional view that early-stage firms rely primarily on informal financing. Their empirical evidence shows that bank debt represents a substantial component of start-up capital structures, suggesting that entrepreneurial financing decisions are highly sensitive to the availability of credit supply.

More broadly, Ang (1991) argues that capital structure theories developed for large corporations cannot be directly applied to SMEs without considering their institutional context. Small firm financing is often relationship-based, limited liability is frequently mitigated by personal guarantees, and financial decisions are strongly influenced by the risk preferences of owner-managers. Consequently, SME capital structures reflect not only theoretical trade-offs but also institutional constraints, relational dynamics, and the personal wealth exposure of entrepreneurs.

Despite these theoretical advances, most capital structure models implicitly treat debt as a homogeneous financing instrument. In practice, however, different forms of debt—such as bank loans, trade credit, or shareholder current accounts—operate under distinct contractual arrangements and institutional constraints. In credit-constrained environments, these financing instruments may respond differently to firm characteristics and market conditions.

Distinguishing between interest-bearing debt (primarily bank loans) and non-interest-bearing liabilities (such as trade credit or shareholder current accounts) therefore provides a more informative perspective on SME financing behavior than aggregate leverage indicators. When leverage is measured as a single aggregate ratio, potential substitution or complementarity mechanisms between formal bank financing and relationship-based liabilities may remain concealed.

This distinction becomes particularly relevant in the context of agricultural SMEs. Agricultural production is characterized by pronounced income seasonality, exposure to climatic shocks, and biological production cycles, all of which contribute to volatile cash flows and uncertain collateral values. Under such conditions, fixed-interest obligations increase firms’ exposure to financial distress, potentially limiting the use of bank debt as predicted by the Trade-Off Theory. At the same time, uncertainty and limited collateral liquidity may reinforce hierarchical financing behavior consistent with the Pecking Order Theory.

Consequently, debt composition in agricultural SMEs reflects the interaction between financial flexibility and contractual rigidity under stochastic income conditions. The asymmetric effects of firm-level characteristics across different debt components may therefore reveal whether observed leverage patterns stem from optimal capital-structure choices or from binding credit supply constraints. Examining debt composition rather than aggregate leverage thus provides a valuable empirical framework for reassessing the applicability of traditional capital structure theories in segmented and risk-intensive financial environments.

2.2. Determinants of Capital-Structure Decisions

In agricultural SMEs exposed to recurrent climate shocks and pronounced income seasonality, capital-structure decisions cannot be interpreted solely through the lens of static trade-off considerations or hierarchical financing arguments. Instead, financing behavior emerges from the interaction between income volatility, contractual rigidity, collateral liquidity, and segmented credit supply.

Agricultural production is inherently characterized by weather variability, biological production cycles, and price fluctuations, all of which generate highly volatile earnings streams. Under such conditions, financial distress costs depend not only on leverage levels but also on the degree of mismatch between fixed repayment obligations and uncertain operating income. Consequently, financial flexibility and renegotiation capacity become central considerations in financing decisions.

In addition, Moroccan agricultural SMEs are predominantly family-controlled, with ownership structures characterized by high wealth concentration and limited diversification. Such governance arrangements tend to increase risk aversion among owner-managers and heighten sensitivity to downside financial risk. This may affect not only firms’ borrowing capacity but also their willingness to assume rigid fixed-interest commitments.

Within this context, traditional determinants of capital structure—such as profitability, asset tangibility, liquidity, and growth opportunities—must be examined in relation to debt composition and institutional credit segmentation. Distinguishing between interest-bearing bank debt and non-interest-bearing liabilities allows for a more precise assessment of whether observed financing patterns reflect rational risk-adjusted optimization or binding credit constraints.

2.2.1. Profitability and Capital Structure

Profitability plays a particularly important role in agricultural SMEs operating under climate-related income volatility and seasonal production cycles. Beyond representing internal financing capacity, profitability also acts as a form of self-insurance against ad-verse shocks. Agricultural revenues are inherently stochastic due to weather variability, biological production lags, and fluctuations in commodity prices. In such an environment, retained earnings help firms absorb temporary liquidity shortfalls and reduce the probability of financial distress during unfavorable production seasons.

From a Trade-Off Theory perspective (Kraus & Litzenberger, 1973; Modigliani & Miller, 1963), higher profitability should increase leverage by enhancing debt capacity and increasing the value of interest tax shields. However, when earnings volatility is high, the expected costs of financial distress become more sensitive to leverage levels. Fixed-interest repayment schedules may exacerbate cash-flow mismatches during droughts or production shocks, potentially weakening the empirical relevance of a stable target leverage ratio.

In contrast, the Pecking Order Theory (Myers & Majluf, 1984) predicts that more profitable firms rely primarily on internally generated funds and therefore reduce their reliance on external borrowing. This prediction has received substantial empirical support in the SME literature. A negative relationship between profitability and leverage has been documented in several contexts, including Ecuador (Gutiérrez-Ponce, 2024), GCC countries (Abdalla et al., 2025), Ghana (Agyei et al., 2020), Portugal (Serrasqueiro et al., 2016), Croatia (Šarlija & Harc, 2016), France (Adair & Adaskou, 2015), Sweden (Öhman & Yazdanfar, 2017), and Morocco (Boumlik et al., 2025). Evidence specific to Moroccan family-owned SMEs further confirms that internal financing tends to dominate in environments characterized by informational opacity and constrained credit supply (Boumlik et al., 2025).

Hypothesis 1.

Profitability is negatively associated with debt usage in Moroccan agricultural SMEs, with a stronger negative effect on interest-bearing debt than on non-interest-bearing liabilities.

In agricultural SMEs, hierarchical financing behavior may be further reinforced by concentrated family ownership. Owner-managers typically face limited wealth diversification and therefore exhibit greater aversion to downside financial risk. Under such governance conditions, profitability affects not only firms’ capacity to borrow but also their willingness to assume rigid interest-bearing obligations.

Importantly, the effect of profitability may differ across debt components. Interest-bearing bank debt involves fixed repayment commitments and formal contractual enforcement. By contrast, non-interest-bearing liabilities—such as trade credit and shareholder current accounts—often incorporate relational flexibility and renegotiation capacity. In environments characterized by climate risk and seasonal income fluctuations, profitable firms may therefore reduce their reliance on rigid bank financing while maintaining operational liabilities linked to commercial transactions.

2.2.2. Assets Tangibility and Capital Structure

Asset tangibility is traditionally considered a key determinant of firms’ borrowing capacity due to its role as collateral. Within the Trade-Off framework (Jensen & Meckling, 1976; Titman & Wessels, 1988), tangible assets reduce expected bankruptcy costs and mitigate agency conflicts between lenders and shareholders, thereby facilitating access to debt financing, particularly at longer maturities. Empirical evidence generally supports a positive association between asset tangibility and long-term leverage across various institutional contexts (Abdalla et al., 2025; Agyei et al., 2020; Rao et al., 2019; Saarani & Shahadan, 2013; Šarlija & Harc, 2016). However, some studies report weaker or even negative relationships between tangibility and short-term debt or aggregate leverage (Czerwonka & Jaworski, 2021; Hacini et al., 2022; Öhman & Yazdanfar, 2017; Serrasqueiro et al., 2016), suggesting that the collateral effect may depend on institutional conditions and debt maturity structure.

In agricultural SMEs, the collateral channel operates under specific constraints. Agricultural assets—including land, biological assets, and specialized machinery—often exhibit limited secondary market liquidity, valuation uncertainty, and price volatility. Climatic exposure and production risk may further reduce the predictability of collateral recovery values. Consequently, lenders may apply significant valuation discounts, weakening the effective borrowing capacity associated with nominal asset tangibility.

Hypothesis 2.

Asset tangibility is positively associated with interest-bearing debt in Moroccan agricultural SMEs, while its effect on non-interest-bearing liabilities is expected to be weaker or insignificant.

In the Moroccan financial system, where bank lending remains largely relation-ship-based and credit allocation tends to be conservative, collateral may represent a necessary but not sufficient condition for obtaining formal credit. Under such circum-stances, higher asset tangibility is expected to facilitate access to interest-bearing bank debt—particularly long-term loans—while its effect on non-interest-bearing liabilities is likely to be limited. Trade credit and shareholder current accounts generally depend more on relational trust and transactional continuity than on pledged collateral.

2.2.3. Liquidity and Capital Structure

Liquidity reflects the stock of internal resources that can be mobilized immediately to absorb short-term shocks. Within the Pecking Order framework, firms with higher liquidity rely less on external financing because available internal funds reduce the need for debt issuance. Consequently, liquidity is generally expected to be negatively associated with leverage. Empirical evidence largely supports this prediction across SME contexts (Abdalla et al., 2025; Agyei et al., 2020; Gutiérrez-Ponce, 2024; Öhman & Yazdanfar, 2017; Serrasqueiro et al., 2016). Nevertheless, some studies report a positive relationship between liquidity and long-term debt (Poutziouris et al., 2022; Serrasqueiro et al., 2016), which is interpreted within a Trade-Off framework whereby higher liquidity improves solvency and enhances borrowing capacity.

Hypothesis 3.

Liquidity is negatively associated with debt usage in Moroccan agricultural SMEs, with a stronger negative effect on interest-bearing debt than on non-interest-bearing liabilities.

In agricultural SMEs, liquidity plays a structurally more critical role due to pronounced income seasonality and exposure to climatic shocks. Cash inflows are typically concentrated around harvest periods, whereas operating expenses and input purchases occur throughout the production cycle. This temporal mismatch increases vulnerability to short-term liquidity shortages. Under such conditions, liquid assets serve as a precautionary buffer against seasonal volatility rather than merely representing residual in-ternal funds.

Consequently, firms with higher liquidity may deliberately limit their reliance on rigid interest-bearing debt in order to preserve financial flexibility. This precautionary behavior is further reinforced in family-owned agricultural SMEs, where downside risk directly affects concentrated household wealth. In contrast, non-interest-bearing liabilities—particularly trade credit—often remain linked to operational transactions and supplier relationships and may therefore decline less proportionally with increases in liquidity.

2.2.4. Sales Growth and Capital Structure

The relationship between sales growth and leverage remains theoretically ambiguous. From a Trade-Off and agency perspective (Myers, 1977), firms experiencing high growth may limit debt in order to avoid underinvestment problems and maintain financial flexibility. Conversely, the Pecking Order Theory predicts that growing firms may rely more heavily on debt when internal funds are insufficient to finance expanding investment opportunities. Empirical findings remain mixed, with studies documenting both positive and negative associations between growth and leverage (Agyei et al., 2020; Czerwonka & Jaworski, 2021; Poutziouris et al., 2022; Rao et al., 2019; Saarani & Shahadan, 2013; Šarlija & Harc, 2016; Serrasqueiro et al., 2016).

In agricultural SMEs, however, growth may intensify rather than reduce risk exposure. Expansion typically requires additional working capital to finance inputs such as seeds, fertilizers, and labor, as well as investments in specialized equipment whose resale value may be limited. Because climatic and price shocks affect expanded production proportionally, growth does not necessarily stabilize earnings; instead, it may increase cash-flow volatility.

Hypothesis 4.

Sales growth is negatively associated with interest-bearing debt in Moroccan agricultural SMEs, while its association with non-interest-bearing liabilities is expected to be weaker or potentially positive.

In a credit-segmented environment such as Morocco, financial institutions may interpret rapid expansion in volatile agricultural activities as an increase in risk rather than as a signal of improved performance. Conservative lending practices and strict collateral requirements may therefore constrain access to additional interest-bearing debt during growth phases. Under such conditions, growing firms may avoid rigid bank commitments in order to preserve financial flexibility and mitigate underinvestment risk.

At the same time, expansion mechanically increases transaction volumes and input purchases, which may strengthen reliance on supplier-based financing. Trade credit may therefore operate as an adjustment margin during expansion when access to formal bank lending remains limited.

2.3. Control Variables

2.3.1. Firm Size and Capital Structure

Firm size is commonly associated with higher leverage due to improved diversification, reduced information asymmetry, and stronger access to financial markets. Larger firms are generally perceived as less risky by lenders because they possess more diversified activities, greater organizational capacity, and more transparent financial re-porting. Empirical evidence consistently documents a positive relationship between firm size and leverage across a wide range of institutional contexts (Abdalla et al., 2025; Agyei et al., 2020; Gutiérrez-Ponce, 2024; Šarlija & Harc, 2016; Serrasqueiro et al., 2016). Evidence from Morocco also supports this pattern, as Boumlik et al. (2025) report a significant association between firm size and leverage among unlisted family SMEs. Nevertheless, results may vary depending on debt maturity structure and institutional conditions.

Hypothesis 5.

Firm size is positively associated with debt usage in Moroccan agricultural SMEs, with a stronger positive effect on interest-bearing debt than on non-interest-bearing liabilities.

In agricultural SMEs, firm size may play an even more important role in mitigating exposure to climate-related income volatility. Larger firms may benefit from crop di-versification, broader customer bases, and stronger bargaining power within supply chains, all of which contribute to stabilizing revenue streams. In segmented credit markets, firm size also serves as a credibility signal for lenders, reflecting organizational maturity, financial transparency, and repayment reliability.

Consequently, larger agricultural SMEs are expected to have improved access to formal bank financing, particularly interest-bearing debt. In contrast, non-interest-bearing liabilities such as trade credit are primarily transaction-driven and may be less sensitive to firm size beyond operational scale.

2.3.2. Age and Capital Structure

The relationship between firm age and leverage remains theoretically ambiguous. Within the Pecking Order framework, older firms may accumulate retained earnings over time, which reduces their reliance on external financing (Adair & Adaskou, 2015; Agyei et al., 2020; Boumlik et al., 2025; Gutiérrez-Ponce, 2024). Conversely, firm age may enhance reputation, reduce informational opacity, and strengthen long-term relationships with financial institutions, thereby facilitating access to external credit (Abdalla et al., 2025; Rao et al., 2019).

Hypothesis 6.

Firm age is positively associated with interest-bearing debt in Moroccan agricultural SMEs, while its effect on non-interest-bearing liabilities is expected to be weaker or ambiguous.

In the context of agricultural SMEs, firm age may also capture accumulated sector-specific experience and the development of stable relationships with suppliers, buyers, and local financial institutions. In relationship-based lending environments such as Morocco, longer operating histories may therefore reduce perceived credit risk and improve eligibility for formal bank financing.

At the same time, the accumulation of internal reserves over time may attenuate firms’ dependence on external debt. However, given the importance of reputational capital and relational banking in segmented financial systems, the credibility channel is expected to dominate in the case of formal lending.

3. Methodology and Research Design

3.1. Data Collection

The empirical analysis is based on a balanced panel of Moroccan agricultural SMEs covering the period 2017–2022. Financial data were extracted from financial statements filed with the Moroccan Office of Industrial and Commercial Property (OMPIC) and accessed through the “Directinfo” platform. This source provides official legal and accounting documents, including certified balance sheets and income statements. In the absence of a centralized public database dedicated to research on Moroccan SMEs, each financial file was individually obtained, thereby ensuring the authenticity and traceability of the data used.

The period 2017–2022 was selected for both institutional and sectoral reasons. First, it corresponds to the final phase of the Plan Maroc Vert and the launch of the Green Generation 2020–2030 strategy, which marked a strategic shift in Moroccan agricultural policy toward greater SME integration, financial inclusion, and value-chain modernization. Second, this timeframe encompasses major macroeconomic and environmental shocks, including the COVID-19 pandemic and successive drought episodes that significantly affected agricultural production, cash flows, and credit conditions. Analyzing capital-structure decisions during this transitional period therefore provides insights into firms’ financing behavior under both policy reform and heightened risk exposure. In addition, limitations in data availability constrain consistent firm-level financial reporting prior to 2017 and after 2022, making this interval the most suitable for constructing a balanced panel.

The initial dataset contained financial statements for 73 agricultural SMEs. The sample construction followed a rigorous selection procedure designed to ensure data consistency and comparability. Only firms meeting the national definition of SMEs and operating in agricultural, horticultural, or livestock activities were retained. The analysis was restricted to limited liability legal forms (SARL, single-member SARL, and GIE) to ensure institutional and accounting homogeneity.

Additional filtering criteria were applied to enhance the statistical quality of the panel. Firms were excluded when: (i) financial statements were incomplete or contained missing data over the study period; (ii) major accounting inconsistencies were detected that could distort the computation of financial ratios; or (iii) financial reports indicated a lack of genuine economic activity, notably firms reporting structurally negative revenues over several consecutive years. After applying these filters, 21 firms were excluded, resulting in a final sample of 52 SMEs.

The final balanced panel consists of 52 firms observed over six consecutive years, yielding 312 firm-year observations. This structure allows the exploitation of both cross-sectional and time-series variation. In line with standard methodological recommendations in panel data econometrics (Greene, 2012; Wooldridge, 2010), such a configuration supports consistent estimation and reliable statistical inference even in moderately sized panels.

3.2. Variables and Definitions

This study measures the capital structure of Moroccan agricultural SMEs using four dependent variables, each capturing a specific dimension of financing decisions. Unlike part of the literature that equates short-term debt with total current liabilities, this paper explicitly distinguishes between interest-bearing short-term debt and non-interest-bearing short-term liabilities, in line with studies emphasizing SME financing and the role of trade credit in financially constrained environments (McGuinness et al., 2018; McGuinness & Hogan, 2016).

The first dependent variable is non-interest-bearing current liabilities (NIBCL), defined as the ratio of non-interest-bearing current liabilities to total assets. This measure primarily includes trade credit granted by suppliers as well as short-term internal financing through shareholders’ current accounts. Trade credit represents a major source of short-term external financing for SMEs, particularly when access to bank financing is limited (Fisman & Love, 2003; García-Teruel & Martínez-Solano, 2010; Martínez-Sola et al., 2014; Petersen & Rajan, 1997). Shareholders’ current accounts constitute a hybrid form of internal financing combining features of debt and equity, and play a critical role in funding SMEs facing liquidity constraints (Ang, 1991; Berger & Udell, 1995, 1998; Mac An Bhaird & Lucey, 2011; Myers & Majluf, 1984; Robb & Robinson, 2014; Sogorb-Mira, 2005).

The three remaining dependent variables are: (i) short-term debt (StLEV), measured as the ratio of short-term bank debt to total assets; (ii) long-term debt (LtLEV), defined as the ratio of long-term bank debt to total assets; and (iii) total leverage (TLEV), calculated as the ratio of total bank debt to total assets. These indicators allow for differentiation across debt maturities and enable the analysis of potential substitution effects between formal and informal financing sources (Gutiérrez-Ponce, 2024; Öhman & Yazdanfar, 2017; Poutziouris et al., 2022; Rao et al., 2019; Serrasqueiro et al., 2016).

The explanatory variables used to examine the determinants of capital structure include profitability (ROA), liquidity (LIQUID), asset tangibility (TANG), and sales growth (SGR). These variables are widely employed in the capital structure literature (Abdalla et al., 2025; Agyei et al., 2020; Amoa-Gyarteng & Dhliwayo, 2023; Gutiérrez-Ponce, 2024) and are directly associated with the core predictions of the Trade-Off Theory and the Pecking Order Theory.

Finally, two control variables are incorporated into the empirical models: firm size (SIZE) and firm age (AGE), in order to account for structural heterogeneity among SMEs that may influence financing decisions (Abdalla et al., 2025; Degryse et al., 2012). A detailed definition and calculation method for all variables are summarized in

Table 1. Definitions and measurements of Variables.

Variable	Symbol	Definition and Measurement
Dependent Variable: Capital Structure
Non-interest-bearing current liabilities	NIBCL	Short-term financial obligations that do not accrue interest divided by total assets: (Current liabilities − StLEV)/Total assets
Short-term Debt	StLEV	Short-term debt divided by total assets
Long-term Debt	LtLEV	Long-term debt divided by total assets
Total leverage	TLEV	The sum of short-term debt and long-term debt divided by total assets
Independent Variables
Profitability	ROA	Earnings Before Interest and Taxes (EBIT) divided by total assets
Assets tangibility	TANG	Firm’s physical assets (property, plant, and equipment) relative to its total assets
Liquidity	LiQUID	Current assets/Current Liabilities
Sales growth (Growth opportunities)	SGR	Percentage increase in sales: ((Current Sales − Previous Sales)/ Previous Sales) × 100%
Control Variables
Firm size	SIZE	Natural logarithm of total assets
Firm age	AGE	Natural logarithm of firm age (number of years since establishment)

.

3.3. Empirical Models

To examine the influence of financial and operational determinants on the capital structure of Moroccan agricultural SMEs, this study adopts a panel data regression framework. This choice is motivated both by the structure of the dataset, which combines a cross-sectional dimension (52 firms) and a time dimension (2017–2022), and by the need to account for unobserved firm-level heterogeneity (Ngatno et al., 2021).

Compared to purely cross-sectional approaches or time-series analyses, panel data offer several important methodological advantages. First, tracking the same firms over multiple periods makes it possible to control for unobservable, time-invariant firm-specific characteristics, thereby reducing omitted variable bias and improving the consistency of the estimators (Baltagi, 2008; Wooldridge, 2010). Second, combining cross-sectional and time dimensions increases the total number of observations, enhancing degrees of freedom and statistical efficiency. Third, panel models capture temporal variations in financing behavior, which is particularly relevant in the agricultural sector, characterized by strong seasonality, exposure to climatic shocks, and sensitivity to public policy interventions.

Consistent with the empirical literature on SME capital structure, several standard linear panel specifications are considered. As an exploratory step, pooled ordinary least squares (pooled OLS), fixed effects (FE), and random effects (RE) estimations were conducted. The selection of the most appropriate econometric specification relies on conventional statistical tests, including the Breusch–Pagan Lagrange Multiplier (LM) test to discriminate between pooled OLS and random effects models, and the Hausman test to choose between fixed and random effects specifications. The results of these tests are presented and discussed in Section 3.5.

On this basis, four distinct empirical models are estimated, each corresponding to a dependent variable capturing a specific dimension of capital structure: non-interest-bearing current liabilities (NIBCL), short-term bank debt (StLEV), long-term bank debt (LtLEV), and total leverage (TLEV). The general regression model is specified as follows:

$${CS}_{it}= {\mathsf{\beta }}_0+ {\mathsf{\beta }}_1{X}_{it}+ {\mathsf{\beta }}_2{C}_{it}+ {\mu }_i+ {\mathsf{\epsilon }}_{it}$$

(1)

where ${CS}_{it}$ denotes the debt component of firm i in year t, $X_{it}$ represents the main firm-level explanatory variables, $C_{it}$ is the vector of control variables, ${\mu }_i$ captures unobserved firm-specific effects, and ${\mathsf{\epsilon }}_{it}$ is the idiosyncratic error term.

Following the general specification in Equation (1), four separate regression models are estimated, each corresponding to a distinct component of capital structure. This disaggregated approach allows us to examine whether firm characteristics affect different types of liabilities asymmetrically. The estimated equations are defined as follows:

• Model 1—Non-interest-bearing current liabilities (NIBCL)

$${NIBCL}_{it}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{ROA}_{it}+{\mathsf{\beta }}_2{TANG}_{it}+{\mathsf{\beta }}_3{LiQUID}_{it}+{\mathsf{\beta }}_4{SGR}_{it}+{\mathsf{\beta }}_5{SIZE}_{it}+{\mathsf{\beta }}_6{AGE}_{it}+{\mu }_i+{\mathsf{\epsilon }}_{it}$$

(2)

• Model 2—Short-term bank debt (StLEV)

$${StLEV}_{it}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{ROA}_{it}+{\mathsf{\beta }}_2{TANG}_{it}+{\mathsf{\beta }}_3{LiQUID}_{it}+{\mathsf{\beta }}_4{SGR}_{it}+{\mathsf{\beta }}_5{SIZE}_{it}+{\mathsf{\beta }}_6{AGE}_{it}+{\mu }_i+{\mathsf{\epsilon }}_{it}$$

(3)

• Model 3—Long-term bank debt (LtLEV)

$${LtLEV}_{it}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{ROA}_{it}+{\mathsf{\beta }}_2{TANG}_{it}+{\mathsf{\beta }}_3{LiQUID}_{it}+{\mathsf{\beta }}_4{SGR}_{it}+{\mathsf{\beta }}_5{SIZE}_{it}+{\mathsf{\beta }}_6{AGE}_{it}+{\mu }_i+{\mathsf{\epsilon }}_{it}$$

(4)

• Model 4—Total leverage (TLEV)

$${TLEV}_{it}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{ROA}_{it}+{\mathsf{\beta }}_2{TANG}_{it}+{\mathsf{\beta }}_3{LiQUID}_{it}+{\mathsf{\beta }}_4{SGR}_{it}+{\mathsf{\beta }}_5{SIZE}_{it}+{\mathsf{\beta }}_6{AGE}_{it}+{\mu }_i+{\mathsf{\epsilon }}_{it}$$

(5)

To account for potential correlation between unobserved firm-specific effects and the explanatory variables, this study employs the correlated random effects (CRE) approach proposed by Mundlak (1978) and further developed by Wooldridge (2010). The CRE model allows disentangling within-firm and between-firm effects while retaining the efficiency advantages of random effects estimation (Schunck, 2013).

Formally, the model is specified as follows:

$${CS}_{it}= \mathsf{\alpha } + {\mu }_i+ {\mathsf{\beta }}_w{X}_{it}+ {\mathsf{\beta }}_b{\overline{X}}_i+ {\mathsf{\epsilon }}_{it}$$

(6)

where ${CS}_{it}$ denotes the capital structure measure of firm i at time t. $X_{it}$ is a vector of time-varying firm-specific characteristics, including profitability (ROA), asset tangibility (TANG), liquidity (LIQUID), firm size (SIZE), and sales growth (SGR). ${\overline{X}}_i$ represents the firm-specific averages of these explanatory variables over the sample period, capturing between-firm variation. The term ${\mu }_i$ denotes unobserved firm-specific heterogeneity assumed to be uncorrelated with the idiosyncratic error term ${\mathsf{\epsilon }}_{it}$, conditional on the included regressors, ${\mathsf{\beta }}_w$ and ${\mathsf{\beta }}_b$ are within and between estimates, respectively.

A joint Wald test on the coefficients of ${\overline{X}}_i$ is conducted to assess the validity of the random effects assumption. Failure to reject the null hypothesis that these coefficients are jointly equal to zero supports the orthogonality condition underlying the random effects estimator, whereas rejection indicates correlation between regressors and unobserved heterogeneity, justifying the use of the CRE specification.

As part of the robustness analysis, we re-estimate the CRE specification by incorporating year-fixed effects to control for common macroeconomic shocks and unobserved time-specific factors affecting all firms over the 2017–2022 period—including the COVID-19 pandemic, global economic volatility, and recurrent drought episodes. These time dummies absorb aggregate disturbances common across firms and mitigate potential omitted variable bias arising from temporal heterogeneity The augmented specification is given by:

$${CS}_{it}=\mathsf{\alpha }+{\mu }_i+{\mathsf{\beta }}_w{X}_{it}+{\mathsf{\beta }}_b{\overline{X}}_i+{\mathsf{\delta }}_t+{\mathsf{\epsilon }}_{it}$$

(7)

where ${\mathsf{\delta }}_t$ denotes year-specific fixed effects capturing time-variant common shocks.

3.4. Diagnostic Tests

3.4.1. Unit Root Test

Prior to estimating the panel data models, it is necessary to assess the stationarity of the variables in order to avoid spurious regression relationships that could bias the empirical results. To this end, a panel unit root test was conducted, namely the Levin–Lin–Chu (LLC) test, which is commonly used in panels with a relatively short time dimension. The LLC test is based on the null hypothesis of a unit root, indicating non-stationarity of the series.

The results of the LLC test, reported in

Table 2. Stationarity check of variables.

Variables	Panel Unit Root Test (Levin, Lin & Chu t)
	Statistic	Prob.
NIBCL	−7.52615	0.0000
StLEV	−5.82270	0.0000
LtLEV	−1.80876	0.0352
TLEV	−7.42946	0.0000
LiQUID	−17.2410	0.0000
TANG	−5.52057	0.0000
SIZE	−3.81837	0.0001
AGE	−28.2129	0.0000

, indicate that all variables in the study can be considered stationary in levels, as the null hypothesis of a unit root is rejected at conventional significance levels. For the long-term bank debt variable (LtLEV), the LLC statistic yields a p-value of 0.0352, indicating marginal stationarity at the 5% significance level.

To reinforce this finding and ensure the robustness of the stationarity diagnosis for this variable, two complementary tests were applied: the Phillips–Perron (PP) and the Augmented Dickey–Fuller (ADF). The results of these additional tests confirm the stationarity of LtLEV in levels, with p-values below 0.01.

Overall, these findings suggest that the variables used in the analysis do not exhibit non-stationarity issues, allowing the estimation of the panel models without requiring further data transformations.

3.4.2. Multicollinearity Test

To assess the potential presence of multicollinearity among the explanatory variables, tolerance statistics and the Variance Inflation Factor (VIF) were computed, in line with standard practices in applied econometrics (Gujarati & Porter, 2009; Wooldridge, 2013).

The results, reported in

Table 3. Tolerance and variance inflation factors.

	Collinearity Statistics
	Tolerance	VIF
ROA	0.819	1.221
LiQUID	0.766	1.305
TANG	0.661	1.513
SIZE	0.696	1.437
AGE	0.698	1.433
SGR	0.958	1.044

, show that all VIF values are well below the conventional threshold of 10, ranging between 1.044 and 1.513. Similarly, all tolerance values exceed the critical threshold of 0.1. These findings indicate the absence of severe multicollinearity among the explanatory variables, suggesting that coefficient estimates are not affected by excessive redundancy of information and that statistical inferences derived from the regression models remain reliable.

3.4.3. Heteroscedasticity Test

The presence of heteroscedasticity, defined as non-constant variance of the error terms, was evaluated using the Breusch–Pagan test, the modified Breusch–Pagan test, and the F-test, which are commonly employed in panel data settings (Breusch & Pagan, 1979; Wooldridge, 2013). The results, presented in

Table 4. Heteroscedasticity diagnostics.

Model	Modified B. Pagan Test		Breusch-Pagan Test		F Test
	Chi-Square	Sig.	Chi-Square	Sig.	Chi-Square	Sig.
NIBCL	18.232	0.006	42.375	0.000	3.286	0.005
StLEV	33.595	0.000	49.021	0.000	6.816	0.000
LtLEV	44.814	0.000	116.173	0.000	10.009	0.000
TLEV	34.815	0.000	26.878	0.000	7.134	0.000

Note: These test the null hypothesis that the variance of the errors does not depend on the values of independent variables.

, indicate that the null hypothesis of homoscedasticity is rejected at the 1% significance level across all tests and for all estimated models (p-values < 0.01).

These results reveal the presence of significant heteroscedasticity in the data. As emphasized by Greene (2012) and Wooldridge (2013), such a violation of classical assumptions may lead to biased standard errors and invalid statistical inference, even if the coefficient estimators themselves remain consistent.

Accordingly, to ensure the validity of statistical significance tests, all subsequent estimations rely on heteroscedasticity-robust standard errors, following the recommendations of the econometric literature on panel data models (Arellano & Bond, 1991; Cameron & Trivedi, 2005; Wooldridge, 2013).

3.4.4. Serial Correlation Test

Serial correlation in the idiosyncratic error term was examined to assess the reliability of statistical inference. Given the panel structure of the data, we first computed the Durbin–Watson (DW) statistic for each specification. The DW values are 0.515 for NIBCL, 0.656 for StLEV, 0.531 for LtLEV, and 0.618 for TLEV.

According to the critical bounds proposed by Savin and White (1977), adjusted for the number of regressors and sample size, these statistics fall below the lower critical threshold, suggesting the presence of positive first-order serial correlation.

Although the Durbin–Watson test was originally developed for time-series regressions, its use here provides an initial diagnostic indication of residual dependence. To account for this potential serial correlation, all models are estimated using robust standard errors clustered at the firm level, which correct for both heteroskedasticity and intra-firm serial dependence.

3.5. Model Specification

The selection of the appropriate econometric specification follows the standard panel data estimation protocol outlined by Dougherty (2011), which involves three sequential steps. The first step consists of determining whether unobserved individual heterogeneity is statistically significant, in order to assess the relevance of a panel data model relative to pooled ordinary least squares (pooled OLS). The second step compares fixed effects (FE) and random effects (RE) specifications using the Hausman test. Finally, the Lagrange Multiplier (LM) test is employed to discriminate between a random effects specification and pooled OLS estimation.

The results of the LM test, reported in

Table 5. Results of specification tests (Hausman and LM tests).

Model	Lagrange Multiplier Test		Hausman Test
	Chi-Sq. Statistic	Prob.	Chi-Sq. Statistic	Prob.
NIBCL	300.6660	0.0000	4.778899	0.5725
StLEV	249.1671	0.0000	12.189136	0.0579
LtLEV	256.3138	0.0000	8.228065	0.2219
TLEV	265.5118	0.0000	8.485069	0.2047

, are statistically significant for all estimated models (p-value = 0.0000). This leads to the rejection of the null hypothesis of no individual effects, indicating that pooled OLS estimation is inappropriate and that a panel data specification is required. The presence of firm-specific heterogeneity therefore justifies the use of models incorporating individual effects.

The comparison between fixed and random effects models is subsequently conducted using the Hausman test. For all models, the null hypothesis of no correlation between the regressors and the unobserved individual effects cannot be rejected, supporting the random effects (RE) specification over the fixed effects (FE) alternative. This result suggests that the RE estimator is both consistent and more efficient than the FE estimator in the context of this study.

It should nevertheless be noted that, for Model 2, the Hausman test yields a marginal p-value (p = 0.0579), providing limited evidence against the null hypothesis of no systematic difference between FE and RE estimates. In this context, and in line with the recommendations of Mundlak (1978) and Wooldridge (2010), the analysis is subsequently complemented by the estimation of a correlated random effects (CRE) model, which allows for a more refined test of the orthogonality assumption underlying the RE framework (Schunck, 2013). This approach is implemented as a robustness check and does not alter the initial choice of the RE estimator as the baseline specification.

Moreover, the diagnostic tests applied to the panel models reveal violations of certain classical assumptions underlying ordinary least squares estimation, notably the presence of heteroscedasticity and intra-individual autocorrelation of the error terms. These issues are common in micro-level panel datasets and, if unaddressed, may result in biased standard errors and invalid statistical inference.

To address these econometric limitations and enhance the reliability of the results, all random effects (RE) and correlated random effects (CRE) models are estimated using robust standard errors that are consistent in the presence of heteroscedasticity and autocorrelation. As emphasized by Wooldridge (2010, 2013) and Cameron and Trivedi (2005), the use of robust standard errors in panel data models constitutes an appropriate strategy for obtaining valid inference under non-spherical error structures. This approach ensures that statistical significance tests rely on correctly specified variance–covariance matrices, without affecting the consistency of the point estimators.

4. Analysis and Findings

4.1. Summary Statistics

Descriptive statistics for the main variables are reported in

Table 6. Descriptive statistics.

	Minimum	Maximum	Mean	Std. Dev.	Skewness		Kurtosis
	Statistic	Statistic	Statistic	Statistic	Statistic	Std. Error	Statistic	Std. Error
NIBCL	0.043	3.271	0.557	0.537	2.332	0.194	6.891	0.386
StLEV	0.000	0.611	0.138	0.151	0.999	0.194	0.303	0.386
LtLEV	0.000	0.486	0.057	0.104	2.469	0.194	5.651	0.386
TLEV	0.000	0.629	0.188	0.183	0.519	0.194	−1.033	0.386
ROA	−0.413	0.298	0.038	0.087	−1.253	0.194	7.380	0.386
LiQUID	0.001	5.260	0.865	0.921	2.678	0.194	7.928	0.386
TANG	0.000	0.845	0.262	0.231	0.820	0.194	−0.428	0.386
SIZE	13.380	22.980	17.962	2.124	−0.127	0.194	−0.288	0.386
AGE	1.099	4.304	2.897	0.676	0.058	0.194	−0.292	0.386
SGR	−1.000	7.714	0.155	0.840	5.771	0.194	45.230	0.386

. The results reveal a financing structure largely dominated by non-interest-bearing short-term liabilities, reflecting the limited access to formal financing faced by Moroccan agricultural SMEs.

Non-interest-bearing current liabilities (NIBCL) exhibit a high mean value of 55.7%, with a standard deviation of 53.7%, and values ranging from 4.3% to 327.1%. This elevated average, combined with substantial dispersion, indicates a pronounced yet heterogeneous reliance on informal financing sources, notably trade payables and shareholders’ current accounts. The presence of values exceeding 100% suggests that, for some SMEs, such liabilities constitute the primary source of financing for operating activities, in some cases exceeding invested capital. By contrast, bank financing remains relatively limited. Short-term bank debt (StLEV) has a mean of 13.8%, while long-term bank debt (LtLEV) displays an even lower average of 5.7%, confirming the weak integration of agricultural SMEs into formal banking channels, particularly with respect to long-term funding.

Descriptive statistics indicate that a non-negligible proportion of firms report zero interest-bearing debt in certain years, reflecting SMEs that rely exclusively on internally generated funds and non-interest-bearing liabilities. The presence of these zero observations does not indicate data truncation but rather represents a corner solution consistent with hierarchical financing behavior and potential credit constraints. Examination of the distribution of total leverage (TLEV) further supports this interpretation: the variable exhibits moderate right skewness (0.519) and negative kurtosis (−1.033), indicating a relatively platykurtic distribution without excessive concentration at zero.

Regarding the explanatory variables, economic profitability (ROA) shows a mean of 3.8%, with considerable dispersion, indicating strong heterogeneity in performance. Liquidity (LIQUID) presents a high average level (86.5%) but with substantial variability, reflecting contrasting financial positions across firms. Asset tangibility (TANG) remains moderate (26.2% on average), suggesting limited collateral capacity. The sales growth rate (SGR) is characterized by high volatility, illustrating the instability of growth trajectories among agricultural SMEs.

The distribution of several variables, particularly sales growth (SGR) and non-interest-bearing liabilities (NIBCL), exhibits noticeable skewness and high kurtosis, suggesting the presence of extreme observations. To ensure that these values do not unduly influence the empirical analysis, additional estimations were conducted using winsorized versions of these variables at the 1st and 99th percentiles.

Finally, the skewness and kurtosis statistics indicate that the variables do not follow a normal distribution. However, consistent with Greene (2012) and Wooldridge (2013), non-normality does not undermine the validity of regression-based estimations, provided that classical assumptions are adequately addressed and robust standard errors are employed.

4.2. Pearson Correlation Analysis

Table 7. Matrix of correlations.

Pearson Correlation	NIBCL	StLEV	LtLEV	TLEV	ROA	LiQUID	TANG	SIZE	AGE	SGR
NIBCL	––
StLEV	−0.439 **	––
	(0.000)
LtLEV	−0.326 **	0.105	––
	(0.000)	(0.194)
TLEV	−0.519 **	0.865 **	0.589 **	––
	(0.000)	(0.000)	(0.000)
ROA	−0.289 **	−0.125	−0.146	−0.176 *	––
	(0.000)	(0.121)	(0.069)	(0.028)
LiQUID	−0.253 **	−0.242 **	−0.069	−0.230 **	0.268 **	––
	(0.001)	(0.002)	(0.389)	(0.004)	(0.001)
TANG	−0.148	−0.121	0.411 **	0.111	−0.212 **	−0.372 **	––
	(0.065)	(0.132)	(0.000)	(0.167)	(0.008)	(0.000)
SIZE	−0.621 **	0.400 **	0.182 *	0.416 **	0.228 **	−0.200 *	0.199 *	––
	(0.000)	(0.000)	(0.023)	(0.000)	(0.004)	(0.012)	(0.013)
AGE	−0.264 **	0.280 **	−0.123	0.159 *	0.175 *	−0.037	−0.320 **	0.357 **	––
	(0.001)	(0.000)	(0.126)	(0.047)	(0.029)	(0.647)	(0.000)	(0.000)
SGR	0.080	0.075	0.022	0.074	−0.080	−0.130	0.145	−0.059	−0.125	––
	(0.322)	(0.350)	(0.786)	(0.356)	(0.318)	(0.105)	(0.071)	(0.464)	(0.119)

Note: Values in parentheses indicate significance levels. ** and * denote correlation significance at the 0.01 (two-tailed) and 0.05 (two-tailed) levels, respectively.

presents the Pearson correlation matrix among the study variables and provides preliminary evidence on the relationships between firm characteristics and financing choices.

Non-interest-bearing current liabilities (NIBCL) are negatively and significantly correlated at the 1% level with short-term bank debt (−0.439) and long-term bank debt (−0.326). These relationships suggest a substitution effect between informal short-term financing and bank debt, confirming that agricultural SMEs rely more heavily on non-interest-bearing liabilities when access to formal credit is constrained.

Moreover, NIBCL is negatively and significantly correlated (at the 1% level) with profitability (ROA) (−0.289), liquidity (LIQUID) (−0.253), firm size (SIZE) (−0.621), and firm age (AGE) (−0.264). These findings indicate that less profitable, less liquid, smaller, and younger firms tend to depend more heavily on non-interest-bearing liabilities, consistent with the presence of more severe financial constraints among these firms.

Short-term bank debt (StLEV) is negatively correlated with liquidity (−0.242), suggesting that firms with tighter liquidity positions are more likely to resort to short-term bank credit. In contrast, StLEV is positively and significantly correlated with firm size and age (at the 1% level), reflecting improved access to bank financing among larger and more mature SMEs.

Finally, long-term bank debt (LtLEV) is positively and significantly correlated with asset tangibility (0.411, at the 1% level) and firm size (0.182, at the 5% level). These associations confirm the central role of tangible assets and organizational scale in securing long-term bank financing, which is typically contingent upon the availability of collateral.

4.3. Regression Results

4.3.1. Random Effects Estimation

The estimation results for the four models are reported in

Table 8. Random effects estimations with robust standard errors.

RE	NIBCL		StLEV		LtLEV		TLEV
	Coef.	t-Stat.	Coef.	t-Stat.	Coef.	t-Stat.	Coef.	t-Stat.
C	3.539 ***	3.126	−0.289	−1.445	−0.276	−1.406	−0.571 *	−1.709
ROA	−0.646	−1.662	−0.087 **	−2.656	−0.085	−1.694	−0.159 ***	−2.862
LiQUID	−0.119 ***	−2.835	−0.010	−1.598	0.001	0.200	−0.008	−0.950
TANG	−0.500 ***	−3.549	0.054	0.507	0.121	0.799	0.167	0.770
SIZE	−0.147 **	−2.618	0.024	1.578	0.022	1.568	0.045 **	2.068
AGE	−0.028	−0.370	−0.002	−0.053	−0.030	−0.452	−0.026	−0.309
SGR	0.013	1.070	0.007 **	2.684	−0.001	−0.339	0.007 **	2.079
R-squared	0.437		0.088		0.124		0.162
Adj. R-sq.	0.415		0.052		0.089		0.128
F-statistic	19.309		2.404		3.509		4.796
Prob. F	0.000		0.030		0.003		0.000

Note: ***, **, and * represent levels of significance of 1%, 5%, and 10%, respectively.

. In accordance with the specification tests, the equations were estimated using the random effects estimator with robust standard errors, in order to account for the heteroscedasticity and autocorrelation detected in the panel data.

For the model explaining non-interest-bearing short-term liabilities (NIBCL), liquidity (LIQUID) is negatively and statistically significantly associated with NIBCL at the 1% level (β = −0.119). A one-percentage-point increase in liquidity is associated with a 0.119 percentage-point lower ratio of non-interest-bearing short-term debt. This pattern is consistent with the interpretation that firms with stronger liquidity positions rely less on informal financing to meet short-term funding needs. Asset tangibility (TANG) is also negatively and significantly associated with NIBCL at the 1% level (β = −0.500). In addition, firm size (SIZE) exhibits a negative association significant at the 5% level (β = −0.147), suggesting that smaller SMEs tend to display higher reliance on informal financing sources. The explanatory power of the model is relatively strong, with an adjusted R² of 41.5%.

Regarding the model for short-term bank debt (StLEV), profitability (ROA) is negatively and significantly associated with short-term bank leverage at the 5% level (β = −0.087), indicating that higher profitability is systematically related to lower reliance on short-term bank financing. In economic terms, the estimated coefficient implies that a five-percentage-point increase in profitability is associated with a reduction of approximately 0.44 percentage points in short-term bank leverage. Relative to the sample mean of 13.8% and a standard deviation of 15.1%, this represents a modest but economically meaningful adjustment in firms’ short-term financing structure. Conversely, the sales growth rate (SGR) is positively and significantly associated (5%) with short-term debt (β = 0.007), consistent with increased working capital needs among expanding firms.

For the long-term bank debt model (LtLEV), none of the explanatory variables exhibits a statistically significant association under the RE specification.

Finally, in the model explaining total leverage (TLEV), profitability (ROA) displays a negative and highly significant association at the 1% level (β = −0.159). In contrast, firm size (SIZE) and sales growth (SGR) are positively and significantly associated with total leverage at the 5% level, with coefficients of 0.045 and 0.007, respectively.

4.3.2. Correlated Random Effects Estimation

Although the Hausman test suggests that the random effects (RE) estimator is ap-propriate, this study employs the correlated random effects (CRE) framework as the preferred estimation strategy (Mundlak, 1978; Schunck, 2013; Wooldridge, 2010). The CRE approach extends the standard RE model by explicitly allowing potential correlation between time-varying regressors and unobserved firm-specific heterogeneity through the inclusion of firm-level means of the explanatory variables.

By relaxing the strict orthogonality assumption underlying the conventional RE estimator, the CRE specification provides a more flexible panel-data framework. It combines the efficiency of random effects estimation with a treatment of unobserved heterogeneity comparable to fixed-effects models, while preserving between-firm variation. The within coefficients of the CRE model correspond to fixed-effects estimates, whereas the inclusion of firm-level means allows explicit modeling of between-firm differences.

The CRE estimation also provides an empirical reinforcement of the Hausman test through a joint Wald test on the individual means of the explanatory variables. This procedure constitutes a more robust alternative to the standard Hausman test, which may be sensitive to heteroscedasticity (Schunck, 2013).

The results of the CRE estimation are presented in

Table 9. Correlated random effects estimations with robust standard errors.

CRE (Mundlak)	NIBCL		StLEV		LtLEV		TLEV
	Coef.	t-Stat.	Coef.	t-Stat.	Coef.	t-Stat.	Coef.	t-Stat.
C	4.213 ***	3.741	−0.500 **	−2.395	−0.128	−1.092	−0.610 **	−2.323
ROA (w)	−0.646	−1.670	−0.066 *	−1.733	−0.093 *	−1.763	−0.147 **	−2.386
LiQUID (w)	−0.113 **	−2.477	−0.007	−1.060	0.001	0.132	−0.006	−0.530
TANG (w)	−0.525 ***	−3.635	0.138	1.267	0.108	0.644	0.234	1.026
SIZE (w)	−0.144 *	−2.016	0.013	0.450	0.056	1.487	0.065	1.299
AGE (w)	−0.035	−0.433	−0.011	−0.244	−0.034	−0.517	−0.038	−0.422
SGR (w)	0.013	1.067	0.007 **	2.598	−0.001	−0.325	0.006	1.596
ROA (b)	0.906	0.810	−0.856 **	−2.200	−0.245	−0.911	−1.066 **	−2.102
LiQUID (b)	−0.222 *	−1.945	−0.015	−0.558	0.020	0.751	0.004	0.083
TANG (b)	0.023	0.096	−0.458 ***	−2.794	0.035	0.179	−0.416	−1.574
SIZE (b)	−0.031	−0.560	0.031	1.036	−0.043	−1.070	−0.009	−0.177
SGR (b)	−0.068	−0.340	0.048	0.952	−0.024	−0.456	0.029	0.496
Wald Test
Chi-square	8.574		15.189		3.242		9.295
Probability	0.127		0.010		0.663		0.098

Note: ***, **, and * represent levels of significance of 1%, 5%, and 10%, respectively. (w) denotes the within, (b) denotes the between estimate. Wald test p-value refers to the joint significance of the time-series means.

. The Wald test yields mixed outcomes across models. For Model 2 (short-term bank debt), the null hypothesis of no correlation between individual effects and regressors is rejected (p = 0.010), indicating that the standard RE estimator may produce biased estimates in this specification. By contrast, for Models 1, 3, and 4 (NIBCL, LtLEV, and TLEV), the null hypothesis cannot be rejected (p = 0.127; 0.663; 0.098, respectively), suggesting that the orthogonality conditions underlying the random effects framework are broadly satisfied.

For reasons of methodological consistency and to ensure robustness against potential unobserved heterogeneity across all specifications, this study adopts the CRE framework as the preferred specification. The CRE model remains valid regardless of whether the strict assumptions of the RE model hold. It provides a unified framework encompassing both fixed and random effects approaches and ensures that the results are not solely driven by cross-sectional differences across firms but remain consistent when accounting for within-firm longitudinal variation (Schunck, 2013).

Overall, the CRE estimates broadly confirm the RE results in terms of coefficient signs and magnitudes. In Model 1, liquidity (β = −0.119), asset tangibility (β = −0.525), and firm size (β = −0.144) remain negatively associated with non-interest-bearing short-term liabilities (NIBCL), with minor adjustments in statistical significance. The liquidity coefficient implies that a ten-percentage-point increase in liquidity is associated with an approximate 1.2 percentage-point reduction in non-interest-bearing liabilities relative to total assets, suggesting that firms with stronger internal financial buffers rely less on informal financing sources.

Similarly, in Model 2, profitability (ROA) remains negatively associated with short-term bank debt, and sales growth (SGR) retains a positive and statistically significant association (β = 0.007, p < 0.05), although the statistical significance of ROA is slightly attenuated. The estimated coefficient indicates that a ten-percentage-point increase in sales growth is associated with an increase of approximately 0.07 percentage points in short-term bank leverage, reflecting higher working capital requirements during periods of sales expansion.

For Model 3, the CRE estimation reveals a negative association between ROA and long-term bank debt (β = −0.093), significant at the 10% level—an association not detected under the RE specification. Finally, in Model 4, the negative association between ROA and total leverage remains statistically robust (β = −0.147). The estimated magnitude implies that a five-percentage-point increase in profitability is associated with a reduction of approximately 0.74 percentage points in total leverage. By contrast, SIZE and SGR no longer display statistically significant associations in this specification.

Beyond confirming the overall stability of the estimates, the CRE decomposition highlights differences between within-firm and between-firm associations. In Model 2, asset tangibility does not exhibit a statistically significant within association, indicating that short-term variations in tangible assets within a given firm are not systematically related to short-term bank borrowing. However, the between coefficient is negative (${\mathsf{\beta }}_b$ = −0.458) and statistically significant at the 1% level, suggesting that firms with structurally higher average tangibility tend to display lower reliance on short-term bank debt. This pattern indicates that collateral capacity is primarily associated with persistent cross-sectional differences rather than short-term adjustments.

Taken together, the comparison between RE and CRE estimations indicates that the main findings are generally robust in terms of both coefficient signs and orders of magnitude, thereby strengthening the credibility of the empirical conclusions.

4.3.3. Additional Robustness Checks

To further evaluate the stability of the empirical findings, additional model specifications were estimated.

First, the baseline CRE models were augmented by including year dummy variables to capture time-specific fixed effects. This specification accounts for macroeconomic and sector-wide shocks affecting all firms within a given year, including the COVID-19 pandemic, global economic volatility, and recurrent drought episodes. The inclusion of these time controls does not materially alter the estimated relationships. For instance, liquidity remains negatively associated with non-interest-bearing liabilities (β = −0.112, p < 0.05), consistent with the baseline specification. Similarly, asset tangibility continues to display a strong negative association with NIBCL (β = −0.519, p < 0.01), indicating that firms with greater collateral capacity rely less on informal financing. In the short-term debt model, sales growth retains a positive and statistically significant relationship with short-term bank leverage (β = 0.0075, p < 0.05). Overall, the direction and magnitude of the key coefficients remain broadly stable after controlling for time-specific shocks.

Second, to address potential reverse causality between profitability and leverage, the models were re-estimated using lagged profitability instead of contemporaneous profitability. Since current financing decisions cannot influence past performance, this specification helps mitigate simultaneity concerns. The results remain broadly consistent with the baseline estimates. In Model 1 (NIBCL), liquidity continues to exhibit a negative and statistically significant association with non-interest-bearing liabilities (β = −0.099, p < 0.01), indicating that firms with stronger liquidity positions rely less on informal short-term financing. Lagged profitability retains the expected negative sign but remains statistically insignificant (β = −0.020), consistent with the baseline specification. Overall, the stability of the main coefficients suggests that the core relationships are not driven by contemporaneous simultaneity between profitability and financing decisions.

Third, to assess whether extreme observations influence the empirical results, additional estimations were conducted using winsorized versions of the sales growth (SGR) and non-interest-bearing liabilities (NIBCL) variables at the 1st and 99th percentiles. The CRE estimations using the winsorized variables yield qualitatively similar results in terms of coefficient signs and statistical significance. In particular, liquidity, asset tangibility, and firm size remain negatively associated with non-interest-bearing liabilities, while the positive association between sales growth and short-term bank debt persists. These findings suggest that the main results are not driven by extreme observations.

For brevity, the detailed regression tables corresponding to these alternative specifications are not reported in the manuscript but are available upon request. Overall, the consistency of the estimated relationships across specifications supports the internal stability of the empirical results within the panel framework.

5. Empirical Analysis and Discussion

This section discusses the associations revealed by the CRE estimations in light of the institutional environment of Moroccan agricultural SMEs, characterized by limited access to bank credit, high exposure to climatic risks, and persistent segmentation of financial markets. Given the observational nature of the panel data, the results should be interpreted as associational relationships rather than causal effects. In this context, distinguishing between interest-bearing debt and non-interest-bearing liabilities is particularly informative for understanding the observed financial structure.

The results concerning non-interest-bearing current liabilities (NIBCL) reveal significant negative associations with liquidity, asset tangibility, and firm size. This pattern suggests that reliance on trade credit and shareholders’ current accounts tends to occur more frequently among firms facing more limited access to bank financing rather than reflecting an optimal capital structure choice in the traditional sense.

The negative association with liquidity is consistent with the Pecking Order Theory (Myers & Majluf, 1984), according to which firms prioritize internal funds before resorting to external financing. When internal liquidity is insufficient, agricultural SMEs turn to financing mechanisms embedded in commercial relationships. This interpretation aligns with Petersen and Rajan (1997), who document that trade credit is often used as a substitute for bank credit in financially constrained environments.

Similarly, the negative association between asset tangibility and NIBCL suggests that firms with greater collateral capacity tend to rely less on informal financing sources. This pattern is consistent with the view that non-interest-bearing liabilities constitute a distinct financing instrument rather than merely a residual component of total debt (García-Teruel & Martínez-Solano, 2010). This interpretation is also consistent with the theoretical framework of Burkart and Ellingsen (2004), which highlights that trade credit and bank debt operate through different commitment and monitoring mechanisms.

The negative association between firm size and NIBCL further suggests that smaller agricultural enterprises tend to rely more heavily on non-interest-bearing liabilities. In economies characterized by relatively underdeveloped financial markets, trade credit often plays an important role as an alternative financing channel (Fisman & Love, 2003). In this respect, the results are consistent with the interpretation that, in the Moroccan context, NIBCL is associated with structural financing constraints rather than with a tax-based leverage trade-off as emphasized by the Trade-Off Theory.

Regarding short-term bank debt, the negative association with profitability is consistent with hierarchical financing behavior predicted by the Pecking Order Theory. This finding aligns with a broad empirical consensus documenting a negative profitability–debt relationship among SMEs in diverse institutional contexts, including Ecuador, Ghana, Algeria, Portugal, Malaysia, and Sweden (Agyei et al., 2020; Gutiérrez-Ponce, 2024; Hacini et al., 2022; Öhman & Yazdanfar, 2017; Saarani & Shahadan, 2013; Serrasqueiro et al., 2016). It is also consistent with recent evidence from Morocco (Boumlik et al., 2025), where a robust negative association between profitability and leverage is identified among unlisted family SMEs. Taken together, these findings suggest that hierarchical financing patterns are widely observed within the Moroccan SME environment.

The positive association between sales growth and short-term bank debt is consistent with increased working capital requirements and seasonal financing needs, which are particularly pronounced in agriculture due to volatile and cyclical cash flows (FAO, 2016; OECD, 2020). Similar associations have been reported in empirical studies conducted in Ghana and India, where growth opportunities are positively related to short-term borrowing (Agyei et al., 2020; Rao et al., 2019). In this sense, short-term debt appears to function primarily as an operational liquidity management instrument rather than as a strategic mechanism related to tax optimization.

The largely insignificant results in the long-term bank debt model—except for profitability under the CRE specification—constitute a particularly revealing finding. While the negative association between profitability and long-term debt is consistent with evidence reported for Portugal and France (Adair & Adaskou, 2015; Serrasqueiro et al., 2016), the limited explanatory power of firm-level variables contrasts with studies documenting a positive role of asset tangibility in long-term borrowing, such as in Ghana, India, and GCC countries (Abdalla et al., 2025; Agyei et al., 2020; Rao et al., 2019). This divergence is consistent with the interpretation that, in the Moroccan agricultural context, long-term indebtedness may be more closely associated with supply-side constraints—such as stringent collateral requirements, financial institutions’ risk aversion, and maturity mismatches—than with firm-level balance-sheet characteristics alone. This interpretation is consistent with evidence of structural credit rationing in emerging economies (Beck & Demirguc-Kunt, 2006; Boutfssi & Quamar, 2024; Stein et al., 2013).

Finally, total leverage is primarily characterized by a robust negative association with profitability, which is consistent with hierarchical financing patterns widely documented in SME studies across Croatia, United Kingdom, and Sweden (Öhman & Yazdanfar, 2017; Poutziouris et al., 2022; Šarlija & Harc, 2016). However, the absence or instability of size and tangibility associations in the aggregate leverage model contrasts with findings reported for Germany and Ecuador, where firm size is often positively related to debt capacity (Abdalla et al., 2025; Agyei et al., 2020; Gutiérrez-Ponce, 2024; Shugliashvili et al., 2023).

Importantly, the absence of strong firm-level associations in the long-term debt model—except for profitability—suggests that traditional firm-level capital structure predictors may play a limited role in explaining long-term borrowing within this sample. This pattern is consistent with the interpretation that long-term indebtedness among Moroccan agricultural SMEs appears to be associated more closely with structural credit supply constraints than with balance-sheet characteristics alone.

Taken together, the differentiated behavior observed across debt components highlights the importance of reassessing the relative explanatory power of traditional capital structure theories in this context. The empirical patterns identified in this study do not support the Pecking Order Theory and the Trade-Off Theory symmetrically.

Predictions derived from the Pecking Order framework receive relatively consistent empirical support. Profitability displays a robust negative association across interest-bearing debt measures, and liquidity is negatively associated with external financing, which is consistent with hierarchical financing behavior. By contrast, core predictions derived from the Trade-Off Theory receive more limited empirical support. Asset tangibility does not appear to be systematically associated with long-term leverage, and firm size does not display a robust association with aggregate leverage. Moreover, the limited explanatory power of firm-level variables in the long-term debt model suggests that observed leverage patterns may be more closely related to credit supply conditions than to tax–distress cost optimization mechanisms.

These conclusions resonate with prior evidence on Moroccan SMEs more broadly. In particular, Boumlik et al. (2025) report that Moroccan family-owned SMEs also display financing patterns dominated by internal funds and only limited support for Trade-Off predictions. The convergence of findings across family and agricultural SMEs suggests that hierarchical financing behavior may constitute a structural characteristic of the Moroccan SME environment rather than a sector-specific phenomenon. The differentiated validation of hypotheses across debt components further underscores the analytical importance of distinguishing between bank financing and non-interest-bearing liabilities when examining capital structure in financially constrained agricultural settings.

Taken together, the empirical findings allow for a structured assessment of the proposed hypotheses. Hypothesis 1 receives strong support, as profitability is negatively associated with interest-bearing debt and total leverage, consistent with the hierarchical financing behavior emphasized by the Pecking Order Theory. Hypothesis 2 receives partial support: asset tangibility is significantly associated with lower reliance on non-interest-bearing liabilities, while its association with long-term bank debt remains weak. Hypothesis 3 also receives partial support, as higher liquidity is associated with lower reliance on non-interest-bearing liabilities, although no robust relationship with interest-bearing debt is observed. Hypothesis 4 is not supported, since sales growth is positively associated with short-term bank borrowing, contrary to the predicted negative relationship with interest-bearing debt. Hypothesis 5 receives partial support, as firm size is associated with lower reliance on non-interest-bearing liabilities, while its relationship with interest-bearing debt remains statistically weak. Finally, Hypothesis 6 is not supported, as firm age does not exhibit a consistent association with leverage patterns in the estimations. Overall, the differentiated validation of the hypotheses underscores the importance of debt composition for understanding financing patterns in segmented and risk-intensive agricultural environments.

6. Conclusions, Implications, and Future Recommendations

The empirical results show that profitability and liquidity are negatively associated with leverage, a pattern consistent with the hierarchical financing behavior emphasized by the Pecking Order Theory. In contrast, asset tangibility and firm size are negatively associated with reliance on non-interest-bearing liabilities, suggesting that firms with greater collateral capacity and organizational scale tend to rely less on informal financing channels. Meanwhile, the limited explanatory power of firm-level characteristics for long-term debt suggests that this component may be more closely associated with credit supply conditions than with balance-sheet optimization mechanisms typically emphasized by the Trade-Off Theory.

From a theoretical perspective, these findings suggest that debt composition can be interpreted as an outcome of heterogeneous financing mechanisms rather than as a homogeneous leverage decision. In financially segmented environments, interest-bearing bank debt and non-interest-bearing liabilities appear to follow distinct economic logics. Interest-bearing bank debt reflects firms’ integration into formal credit markets governed by collateral requirements and screening mechanisms, whereas non-interest-bearing liabilities are frequently observed in contexts where firms face credit rationing. Consequently, when credit supply constraints are sufficiently binding, aggregate leverage measures may conceal offsetting patterns across debt components. Firm characteristics such as profitability, asset tangibility, and firm size may therefore be associated with formal and informal financing in different ways. Models relying solely on total leverage may thus attenuate or misrepresent the underlying financing structure. The contribution of this study lies not only in the disaggregation of leverage measures but also in highlighting that the empirical relevance of capital structure theories appears to depend on the degree of financial segmentation. When institutional constraints dominate firm-level optimization incentives, debt composition itself becomes analytically informative.

From a policy perspective, the findings suggest that more targeted interventions may be more effective than broad credit expansion policies. First, the strong association between asset tangibility and lower reliance on non-interest-bearing liabilities suggests that collateral capacity is closely related to firms’ integration into formal financial channels. Policies aimed at improving the formal registration and valuation of movable agricultural assets—such as machinery, irrigation equipment, and livestock—could expand the pool of pledgeable collateral and potentially facilitate SMEs’ access to bank lending. Second, the weak association between firm characteristics and long-term debt is consistent with the presence of supply-side rigidities that may limit maturity extension. Strengthening agricultural credit guarantee schemes specifically designed to support long-term investment projects, rather than short-term working capital financing, may help reduce perceived lending risks and alleviate maturity mismatches. Finally, the persistent reliance on non-interest-bearing liabilities among smaller and less liquid firms suggests that financial literacy programs and incentives for financial formalization could help reduce dependence on relationship-based financing and support gradual integration into formal credit markets.

Several limitations should be acknowledged. First, the empirical analysis relies on a relatively small balanced panel of formally registered agricultural SMEs operating within a single national context. Although this structure ensures internal consistency and comparability over the 2017–2022 period, it inevitably limits the external validity of the findings. The documented relationships should therefore be interpreted as context-specific rather than universally generalizable, as financing patterns may differ across institutional and financial systems characterized by alternative credit allocation mechanisms. Second, while the correlated random effects framework accounts for time-invariant unobserved heterogeneity and year fixed effects capture common macroeconomic shocks, residual endogeneity inherent to observational panel data cannot be entirely ruled out. Unobserved time-varying firm characteristics or evolving credit market conditions may influence financing decisions beyond the associations identified in this study. Third, the empirical specification focuses primarily on firm-level determinants, and the absence of explicit macro-financial or credit supply variables limits the ability to fully disentangle firm-level financing behavior from broader systemic constraints.

Future research could extend this analysis by examining similar dynamics in other institutional contexts and sectoral environments. Comparative studies across countries or industries would help determine whether the debt composition patterns identified here reflect structural characteristics of agricultural SMEs or more general features of SME financing under conditions of financial constraint.