Public Debt, Oil Rent, and Financial Development in MENA Countries: A Fractional Response Model Approach (FRM)

Abstract

The rapid accumulation of public debt raises global concern over its implications for financial markets. This study examines the effect of domestic public debt on financial development in Middle East and North Africa (MENA) countries, a region marked by sharp heterogeneity in institutions, debt dynamics, and oil dependence, using annual panel data for 16 countries over the period (2000–2020). Our analysis employs a fractional response model (FRM), which accounts for the bounded nature of the dependent variable, corrects for heteroskedasticity, and incorporates country fixed effects. The findings reveal a significant negative effect of domestic public debt on financial development, consistent with the lazy banks and crowding-out hypotheses. This adverse relationship persists across different income groups and debt percentiles, with modest attenuation at higher debt levels. Oil rents are also found to exert a robust negative effect, highlighting the structural vulnerabilities associated with oil dependence. These results emphasize the importance of debt management, fiscal frameworks that account for commodity cycles, and policies to reduce the sovereign–bank nexus in fostering sustainable financial development in the region.

1. Introduction

In recent years, government borrowing has surged to historic levels across advanced, emerging, and developing economies, driven by the fiscal demands of pandemic recovery, geopolitical tensions, energy price volatility, and ambitious public investment agendas. While public debt has played a critical role in stabilizing economies during crises, its rapid accumulation has reignited concerns about long-term financial sustainability and the unintended consequences for private sector credit and financial market development. Maintaining economic momentum, while preventing public debt from undermining the efficiency, depth, and access of financial systems, requires more attention from central banks and fiscal authorities. Against this backdrop, understanding the relationship between domestic public debt and financial development has become more urgent and policy relevant than ever.

The interaction between public debt and financial development has long been a focal point of scholarly and policy debate. While public borrowing may serve as an important tool for financing development and stabilizing the macroeconomy, its impact on financial sector performance remains contested. Theoretically, three dominant hypotheses frame this debate. The “lazy banks” hypothesis argues that high levels of public debt incentivize banks to shift lending toward government securities, thereby reducing private sector credit and impeding financial intermediation. The “crowding-out” hypothesis further suggests that government borrowing competes with the private sector for scarce financial resources, raising interest rates and diminishing productive investment. Conversely, the “safe assets” view maintains that government debt may enhance financial development by supplying stable and liquid instruments that support collateral frameworks and foster secondary market formation.

The main aim of this study is to examine the effect of domestic public debt on financial development across 161 MENA countries over the period 2000–2020, using a fractional response probit model (FRM).

The MENA region constitutes a particularly instructive case for studying the public debt–financial development nexus because of many reasons. Domestic banks are the primary holders of government securities, making the financial sector especially exposed to sovereign debt dynamics and thus an ideal ground for testing the “lazy banks” and “crowding-out” hypotheses. At the same time, fiscal revenues remain heavily tied to volatile oil rents, creating cycles of debt accumulation and repayment that few other regions experience with similar magnitude. This interaction between oil dependence, debt management, and financial development has received scarce empirical attention despite its immediate policy relevance, as governments across the MENA region pursue ambitious diversification agendas while expanding their domestic debt markets. Focusing on this region therefore provides sharper insights into how debt and resources jointly shape financial development, with lessons that extend to other emerging economies facing the dual challenge of resource dependence and debt sustainability.

This study focuses on domestic public debt because it directly influences the functioning of the domestic financial system, particularly through banks’ portfolio allocations and the crowding out of private sector credit. These are the key channels underlying the theoretical hypotheses examined here, such as the lazy banks and crowding-out effects. In contrast, external debt is shaped largely by global factors, such as international interest rates, investor risk appetite, and exchange rate regimes, that generate vulnerabilities like rollover risk, currency mismatches, and sudden stops. While these issues are important for financial stability, they operate through mechanisms distinct from the domestic intermediation processes that this paper seeks to analyze. Focusing on domestic debt therefore allows for a clearer and more precise test of the theoretical mechanisms under investigation.

The empirical literature reflects the theoretical divergence, with studies documenting both negative and positive effects of public debt on financial development, often conditional on institutional quality, macroeconomic structure, and the level of financial liberalization. In countries with weak institutions and shallow financial systems, public debt often displaces private credit and stifles innovation. In contrast, in economies with stronger governance and deep capital markets, government securities may function as safe assets that reinforce intermediation and monetary operations.

Despite these contributions, several important gaps remain. First, much of the research relies on narrow proxies such as domestic credit to the private sector, overlooking the multidimensional nature of financial development, while only a few studies employ broader measures like bank efficiency or composite indices. Second, most studies use traditional methods such as ARDL, NARDL, fixed effects, or GMM, which are limited in terms of addressing bounded dependent variables and heteroskedasticity. Third, empirical evidence on the MENA region is scarce, particularly regarding the role of oil rents and resource dependence. Finally, the intersection of oil rents, public debt, and financial development remains largely unexplored in hydrocarbon-dependent economies.

This study contributes to the literature by addressing these gaps by, first, utilizing the IMF’s multidimensional financial development index; second, applying a fractional response model (FRM) suitable for bounded outcomes; third, incorporating oil rents as a key control variable; and finally, examining heterogeneity in the debt–finance nexus across income levels and debt burdens.

A novel contribution of this study is the inclusion of oil rents as a control variable, capturing the influence of oil dependence on financial development, a factor particularly relevant for resource-rich economies. To the best of our knowledge, there is no existing study using oil rents in a model of financial development in this context.

The remainder of this paper is structured as follows. Section 2 introduces the concept of public debt and financial development, including its measurement, theoretical linkages, and the role of oil rents. Section 3 provides a review of the theoretical and empirical literature. Section 4 sets out the model specification, while Section 5 describes the data and econometric methods. Section 6 reports the empirical results, including robustness checks and marginal effects. Section 7 discusses the findings in light of existing theories and policy implications. Finally, Section 8 concludes and outlines directions for future research.

2. Public Debt and Financial Development

2.1. Definitions and Measurements

2.1.1. Financial Development

The financial sector comprises a set of institutions, instruments, markets, and the legal and regulatory frameworks that enable credit provision and the execution of transactions. Financial development is broadly understood as the process of reducing information, enforcement, and transaction costs, thereby facilitating the emergence of financial contracts, markets, and intermediaries that allocate resources more efficiently. A well-functioning financial system performs five essential functions: generating and processing information on investment opportunities; monitoring investments and enforcing corporate governance; enabling risk management, diversification, and trading; mobilizing and pooling savings; and easing the exchange of goods and services (For more, see (World Bank, 2019/2020)).

The importance of financial development extends beyond efficiency gains in financial intermediation. A substantial body of evidence demonstrates its positive causal effect on long-run economic growth through capital accumulation, technological progress, and improved resource allocation. Moreover, financial development reduces poverty and inequality by broadening access to finance for disadvantaged groups, supporting risk management, and fostering investment and productivity growth. It is also central to the growth of Small and Medium-Sized Enterprises (SMEs), which are typically more labor-intensive and contribute significantly to employment creation, particularly in emerging markets (For, more see (World Bank, 2019/2020)).

2.1.2. Measurement of Financial Development

Financial development is inherently difficult to measure, as it encompasses multiple dimensions of institutions and markets. Early empirical studies relied on single proxies such as private credit to GDP, stock market capitalization to GDP, or bank deposits and liquid liabilities relative to GDP. While informative, such measures capture only the size or depth of financial intermediation, neglecting broader issues of access, efficiency, and institutional quality. This limitation has generated increasing recognition of the need for composite, multidimensional indicators that can better reflect the complex role of financial systems in development.

In response, the IMF developed the financial development index (FDI). The index combines indicators of financial institutions and financial markets, thereby capturing the multidimensional character of financial development. On the institutional side, it incorporates measures of depth, such as credit to the private sector and the size of pension and mutual fund assets; access, reflected in the availability of banking services such as branches and Automated Teller Machines (ATMs); and efficiency, proxied by margins, spreads, costs, and returns. On the market side, it includes measures of depth, such as stock market capitalization and debt securities; access, reflected in the breadth of market participation and number of issuers; and efficiency, measured by trading activity and turnover ratios. Taking together, these components provide a comprehensive and comparable assessment of financial development across more than 180 countries and over several decades (For more, see IMF, 2023).

2.1.3. Domestic Public Debt

Consistent with prior studies (Ardagna et al., 2007); (Abbas et al., 2022); (Hauner, 2009), this study focuses on domestic public debt rather than total or foreign debt. The choice is theoretically motivated by the theoretical mechanisms under consideration: lazy banks, crowding out, and the safe assets mechanism all depend on how government securities would affect the accessibility, depth, and efficiency of the financial institutions; in other words, how the banks allocate their portfolios between government securities and private lending.

Domestic debt directly reflects banks’ portfolio choices, pricing, and liquidity management, whereas external debt is often determined by global interest rates, capital flows, and foreign investors’ risk appetite.

In contrast, total public debt combines domestic and external components that work through different channels, some of which may even offset one another. Consequently, using total debt would blur these mechanisms and make the estimated effect on financial development difficult to interpret.

Similarly, foreign debt primarily captures exposure to external vulnerabilities such as currency mismatches, rollover risks, and the so-called “original sin” of borrowing in foreign currency—factors that shape sovereign risk but do not directly influence the credit allocation decisions of domestic banks.

By focusing on domestic debt, we obtain a much cleaner identification of the bank–sovereign nexus, isolating the domestic intermediation channel from these confounding global-vulnerability factors and aligning our measure precisely with the theoretical hypotheses under investigation.

2.1.4. Domestic Public Debt Measurement

According to the World Bank’s statistical approach, domestic public debt is measured as the ratio of credit provided by domestic money banks to governments and state-owned enterprises relative to GDP.

2.2. Domestic Public Debt and Financial Development Linkage Hypothesis

The interaction between domestic public debt and financial development has been examined through three dominant hypotheses: the lazy banks hypothesis, the crowding-out hypothesis, and the safe assets hypothesis. Each perspective emphasizes distinct mechanisms by which sovereign borrowing shapes financial intermediation, market depth, and systemic efficiency.

2.2.1. The Lazy Banks Hypothesis

The lazy banks hypothesis suggests that higher levels of government borrowing alter the portfolio behavior of commercial banks. Because sovereign securities are liquid, relatively risk-free, and involve low monitoring and screening costs, banks may prefer to hold them rather than engage in private lending (Abdel-Halim & Ghazi, 2022; Hauner, 2009; Manove et al., 2001). This portfolio reallocation mechanism leads to a contraction of private sector credit, reduces the incentives for screening and monitoring, and suppresses financial innovation.

This mechanism resonates with classical concerns regarding the adverse effects of public borrowing. Classical economists, including Adam Smith and David Ricardo, argued that debt issuance diverts resources from productive activities and undermines the wealth-generating capacity of the economy (Tsoulfidis, 2007). The lazy banks hypothesis thus extends this classical view into modern financial systems, predicting a negative impact of public debt on financial development through weakened intermediation and a bias toward sovereign debt holdings.

H1 (Lazy Banks):

Higher domestic public debt reduces financial development.

2.2.2. The Crowding-Out Hypothesis

The crowding-out hypothesis highlights the competition between the public and private sectors for a limited pool of domestic savings. When governments issue debt, they absorb resources that could otherwise finance private investment. This process unfolds through several mechanisms.

First, the interest rate channel operates as increased public borrowing raises the demand for loanable funds, putting upward pressure on interest rates and restricting access to credit for households and firms (Diamond, 1965). Second, the intertemporal taxation channel suggests that the anticipation of future taxation to service public debt reduces disposable income and savings, discourages investment, and constrains intermediation (Barro, 1974; Tsoulfidis, 2007). Third, the investor confidence channel emphasizes that persistent increases in public debt can deteriorate sovereign credit ratings, raise risk premium, and undermine investor confidence. This erosion of credibility diminishes both foreign and domestic capital inflows, further suppressing financial sector development (Abbas et al., 2022; Akbar et al., 2021).

Together, these channels mirror neoclassical predictions that public debt competes directly with private capital accumulation, raises borrowing costs, and lowers long-run output. Translated into the financial sector, the crowding-out hypothesis predicts that debt accumulation hampers credit expansion, market depth, and efficiency.

H2 (Crowding Out):

Higher domestic public debt reduces financial development.

2.2.3. The Safe Assets Hypothesis

In contrast, the safe assets hypothesis emphasizes the potential for sovereign debt to strengthen financial development under conditions of fiscal sustainability and macroeconomic stability. Public debt can facilitate intermediation by supplying instruments that function as safe and liquid assets. These securities serve as high-quality collateral in interbank and repo markets, thereby reducing counterparty risk and supporting credit expansion (Hauner, 2009). They also underpin the development of benchmark yield curves, which improve risk pricing and hedging across financial markets (Weymuller, 2013). Moreover, active issuance and secondary trading of sovereign bonds deepen market liquidity, reduce transaction costs, and broaden participation (Agyapong & Bedjabeng, 2019; Benayed & Gabsi, 2020; Sağdıç et al., 2021).

However, these benefits are conditional. When sovereign debt remains within sustainable thresholds, it can enhance financial development by providing stability and collateral value. Yet excessive debt accumulation may undermine creditworthiness, diminish the safe-asset role of sovereign bonds, and erode the very mechanisms that initially supported market deepening (Benayed & Gabsi, 2020; Hauner, 2009). This perspective therefore suggests a nonlinear relationship between public debt and financial development, where moderate levels are beneficial but excessive levels become detrimental.

H3 (Safe Assets):

Under stable macro conditions, domestic public debt enhances financial development.

2.3. Oil Rent and Financial Development

The relationship between oil dependence and financial development (FD) is theoretically contested, reflecting the broader debate between the “resource curse” and “resource blessing” perspectives. The resource-curse hypothesis posits that abundant natural resource rents, rather than promoting development, often weaken financial systems through institutional fragility, rent-seeking, and macroeconomic distortions. Early contributions (Caselli & Cunningham, 2009; Rosser, 2006; Sachs & Warner, 1995; Van der Ploeg & Venables, 2009) argue that resource dependence fosters corruption, Dutch disease, and excessive volatility, all of which erode the foundations of FD.

(Beck, 2011) sharpened this argument with the Financial Resource Curse (FRC) hypothesis, suggesting that resources rents crowd out savings, dampen incentives for innovation, and distort capital allocation away from productive financial intermediation.

In contrast, the resource-blessing perspective emphasizes that resource wealth can foster financial development when mediated by strong institutions and prudent policies (Hussain, 2021; Yıldırım et al., 2020).

A further extension of this debate concerns the role of public debt in resource-dependent economies. This suggests that resource rents may interact with debt dynamics to magnify financial vulnerabilities under weak institutions, while in settings with robust governance frameworks, resource revenues and borrowing can be harnessed to strengthen financial intermediation and stability (Abdelaziz et al., 2025; Chengyonghui et al., 2023; Onifade et al., 2024).

3. Literature Review

Understanding the interaction between public debt and financial development has garnered significant attention in the economics literature. This section reviews key theoretical and empirical contributions that explore the nature of this relationship. Three dominant hypotheses underlie the theoretical debate: the “lazy banks”, the “crowding-out” hypothesis and the “safe assets” hypothesis. Both the “lazy banks” and the “crowding-out” hypothesis suggest a negative relationship between public debt and financial development. In contrast, the “safe assets” hypothesis emphasizes a positive effect of public debt on financial development.

3.1. Theoretical Review

3.1.1. Theoretical Studies on the “Lazy Banks” Hypothesis

The “lazy banks” hypothesis is rooted in the concept of moral hazard and profit-maximizing behavior among banks operating under risk-averse regulatory environments. Under such conditions, banks tend to prefer risk-free government securities over engaging in costly and uncertain screening of private sector projects. This preference leads to a reallocation of capital away from private sector lending, thereby constraining financial intermediation and stifling innovation (Ismihan & Ozkan, 2012; Manove et al., 2001).

(Manove et al., 2001) formalize the lazy bank hypothesis through a model where banks must choose between screening borrowers and relying on collateral. When legal systems strongly protect creditors’ rights, banks rationally prefer lending against collateral, resulting in inefficient credit allocation and favoring of government securities over private lending. Similarly, (Ismihan & Ozkan, 2012) formalize the crowding-out effect in economies with shallow financial systems, showing that public borrowing from domestic banks reduces credit availability for the private sector, especially in the presence of weak institutions and poor coordination between fiscal and monetary authorities.

3.1.2. Theoretical Studies on the Crowding-Out Hypothesis

The crowding-out hypothesis identifies three channels through which public debt can hinder financial development. First, increased public borrowing diverts national savings from productive private investment to unproductive public expenditure, reducing the pool of funds available for new private investment. Second, debt redemption may require higher taxation, further suppress private savings, and curb both investment and consumption. Third, excessive borrowing can exert upward pressure on interest rates, thereby reducing the private sector’s capacity to access credit (Tsoulfidis, 2007).

3.1.3. Theoretical Studies on the Safe Assets Hypothesis

Conversely, the “safe assets” view argues that government bonds provide a stable source of collateral and pricing benchmarks that can deepen financial markets and reduce credit risk. Government securities help develop a yield curve and encourage the creation of secondary markets, which in turn supports broader financial market development (Hauner, 2009; Weymuller, 2013).

(Weymuller, 2013) presents a framework in which banks act as “safety multipliers” by bundling risky investments with public debt to create private safe assets. When public safe assets are scarce, banks deleverage, reducing private credit supply and triggering credit crunches, highlighting how public debt may “crowd in” rather than “crowd out” investment. (Alupoaiei et al., 2024) develop a DSGE model incorporating external sectors, banking behavior, and disaggregated government spending, finding that government investment, but not consumption, can crowd in private investment and credit.

3.2. Empirical Review

Empirical studies offer mixed evidence. Many show negative effects of public debt on financial development, including (Abbas et al., 2022; Ahmed et al., 2024; Akbar et al., 2021; Ali & Ahmed, 2016; Bouis, 2019; Buljan et al., 2020; Chung-Yee et al., 2020; Emran & Farazi, 2009; İlgün, 2016; Janda & Kravtsov, 2017; Mun & Ismail, 2015). Others find positive effects, such as (Abdel-Halim & Ghazi, 2022; Alupoaiei et al., 2024; Kumhof & Tanner, 2005; Ozili, 2024). The relationship may depend on institutional quality, financial system depth, and the efficiency of governance, which might reverse the relationship at a certain point or threshold (Benayed & Gabsi, 2020; Hauner, 2009).

3.2.1. Evidence from MENA Countries

Despite the growing literature on public debt and financial development, empirical studies focusing specifically on the MENA region remain scarce.

Nevertheless, some recent studies provide valuable indirect insights. For instance, (Mohammed Daher et al., 2020) examine the debt-growth relationship in 20 MENA countries and identify a debt threshold of 58% of GDP beyond which the effect of public debt on growth turns negative, emphasizing the macro-financial risks of over-borrowing in the region. Similarly, (Alsamara et al., 2024) assess how energy endowments influence the debt-growth relationship and report that oil-rich MENA countries exhibit lower debt thresholds compared to their non-oil counterparts, suggesting that fiscal space is more constrained by volatility in resource revenues. Although these studies do not directly address financial development, they highlight key macroeconomic frictions, such as commodity dependence and debt sustainability, that are likely to shape the financial system’s capacity to intermediate.

In addition, (Abdelmonem Lotfy Mohamed et al., 2023) provide panel evidence for 16 MENA countries, showing that financial development positively affects economic growth but is hindered by limited private sector participation and crowding-out effects. While not focused solely on public debt, this study implies that excessive public sector dominance, including borrowing, may limit financial sector depth and efficiency in the region.

3.2.2. Oil Rents, Resource Dependence, and Financial Development

The literature on natural resources (NR) and financial development (FD) is divided between two contrasting perspectives: the “resource curse” and the “resource blessing.” (Çetin et al., 2023) found a negative NR–FD nexus in developing countries plagued by weak governance; (Zhang & Liang, 2023) confirmed similar effects in South Asia; and (Ding, 2023) observed resource-related financial fragility in G7 economies. (Ozili, 2023) further explained how windfalls often generate capital flight to offshore assets, eroding domestic finance, while (Han et al., 2022) and (Tang et al., 2022) demonstrated how resource dependence exacerbates Dutch disease, reduces competitiveness, and undermines macro-financial stability.

Conversely, (Hussain, 2021) showed that resource rents enhance FD in high-income countries by reinforcing human capital and governance structures, while (Yıldırım et al., 2020) found that oil and forest rents in developing economies expand credit creation and strengthen intermediation. Country-specific evidence lends further support: in Pakistan, (Asif et al., 2020) reported that resource rents stimulated FD through monetary expansion, while (Shahbaz et al., 2018) demonstrated that effective policy management of resource revenues increased labor absorption, investment, and demand for financial services. These studies collectively suggest that under conducive institutional and policy settings, (NR) can serve as a catalyst for financial deepening rather than a constraint.

In the MENA region, (Chengyonghui et al., 2023) find that both natural resources and public debt are negatively associated with FD, with debt amplifying the adverse effects of resource rents by heightening rent-seeking and weakening incentives for financial innovation. By contrast, (Onifade et al., 2024) provide evidence that natural resource abundance can strengthen FD—especially banking stability—when mediated by robust institutional frameworks. (Abdelaziz et al., 2025) further confirm that resource dependence undermines banking stability but demonstrate that improvements in regulatory quality, rule of law, and anti-corruption mechanisms mitigate these risks.

3.2.3. Rule of Institutions and Governance

Several studies document that weak institutions amplify the negative impact of public debt. (Ahmed et al., 2024) find that public debt harms the efficiency dimension of financial development in Pakistan, supporting the lazy bank hypothesis, though they report positive effects on depth and stability. (Abbas et al., 2022) find that debt has a negative effect in countries with weak institutions, but this reverses when governance indicators exceed certain thresholds. (Akbar et al., 2021) reach similar conclusions for 45 Asian countries, demonstrating the role of institutional performance in transforming the adverse effects of debt into favorable outcomes.

Public debt can undermine financial development by displacing private sector lending and distorting banks’ investment incentives, a pattern consistently observed across diverse empirical studies. (Buljan et al., 2020) examined the determinants of banks’ sovereign debt holdings in Central and Eastern European (CEE) countries and found a significant negative relationship between private credit demand and sovereign debt. Their findings suggest that public debt may crowd out financial development by incentivizing banks to shift resources away from private sector lending, especially in less developed economies with limited investment alternatives and strong home bias. (Chung-Yee et al., 2020) and (Janda & Kravtsov, 2017) similarly find a long- and short-run negative relationship between public debt and domestic credit to the private sector. In addition, they emphasize a symmetric relationship in both the long run as well as short run. (Bouis, 2019) shows that increased sovereign debt holdings are associated with slower private credit growth in emerging market and developing economies (EMDEs), due to portfolio rebalancing rather than coercive regulation. He also finds that such holdings improve banks’ after-tax returns mainly through lower provisioning rather than higher lending margins.

3.2.4. Linear Models

A considerable number of empirical studies have employed linear estimation techniques; some confirm the adverse impact of public debt on financial development, whereas others identify a positive effect. (İlgün, 2016) examines 18 emerging economies using cointegration techniques and second-generation panel data models that account for cross-sectional dependence and slope heterogeneity. His findings confirm a negative long-run association between bank credit to the government and a composite financial development index constructed via Principal Component Analysis (PCA). The inclusion of macroeconomic controls like trade openness, inflation, and GDP per capita further supports the hypothesis that excessive public debt is detrimental to financial sector growth. (Emran & Farazi, 2009) provide robust causal evidence on the crowding-out effect of government borrowing from domestic banks on private credit in developing countries, using panel data for 60 countries, employing an instrumental variable strategy based on political structure, parliamentary systems, polarization, and their interaction. At the country level, studies such as those by (Mun & Ismail, 2015) and (Ali & Ahmed, 2016) provide further support for the lazy banks and crowding-out hypotheses, emphasizing the role of inflation, savings behavior, and governance in shaping the debt–finance nexus.

In contrast, (Ozili, 2024) finds a positive correlation between public and private sector lending in 43 countries, especially before the Great Recession. (Abdel-Halim & Ghazi, 2022), using bank-level data in Jordan, find that government lending improves efficiency, although it does not significantly impact profitability. (Alupoaiei et al., 2024), using a VAR model for four Eastern European countries, shows that public investment supports private investment and GDP, though private investment plays a stronger role. (Kumhof & Tanner, 2005) emphasize that in countries with weak legal systems, public debt serves as safe collateral, supporting financial intermediation. They show how government securities help sustain banking operations in environments lacking enforceable contracts or property rights.

3.2.5. Non-Linear Models

An emerging strand of the literature highlights the non-linear nature of the relationship between public debt and financial development, where the direction and magnitude of the effect depend on structural conditions, institutional quality, and financial openness. (Benayed & Gabsi, 2020) detect an inverted-U relationship in Sub-Saharan Africa, where moderate debt supports development, but excessive borrowing undermines it. They identify a turning point at approximately 52% of GDP, beyond which public debt begins to crowd out private sector credit. (Bui, 2018) applies a Panel Smooth Transition Regression (PSTR) model to examine Asia-Pacific countries. His findings suggest that the adverse effects of public debt on financial development diminish in more liberalized and globally integrated financial systems. The study identifies a certain threshold value for financial freedom, above which the crowding-out effect dissipates. (Hauner, 2009), using country- and bank-level data for 73 middle-income countries, finds that moderate debt can provide collateral benefits, but excessive reliance on government securities reduces financial deepening and efficiency, especially in financially repressed settings.

3.2.6. Gaps

Despite this extensive body of work, several gaps remain. First, most studies rely on narrow proxies such as domestic credit to the private sector as a percentage of GDP (e.g., (Abbas et al., 2022; Ali & Ahmed, 2016; Benayed & Gabsi, 2020; Bui, 2018; Chung-Yee et al., 2020; İlgün, 2016; Janda & Kravtsov, 2017; Mun & Ismail, 2015)), which do not capture the full multidimensional nature of financial development. Only a few studies have used broader indicators such as bank efficiency or profitability (e.g., (Abdel-Halim & Ghazi, 2022; Hauner, 2009)) or constructed composite indices (e.g., (Chung-Yee et al., 2020; İlgün, 2016)).

Second, many studies employ traditional estimation techniques such as Autoregressive Distributed Lag (ARDL), such as (Ali & Ahmed, 2016); (Mun & Ismail, 2015), Nonlinear Autoregressive Distributed Lag (NARDL) (Chung-Yee et al., 2020), or panel fixed effects and Generalized Method of Moments (GMM); for example, (Abbas et al., 2022; Abdel-Halim & Ghazi, 2022; Benayed & Gabsi, 2020), which do not accommodate bounded dependent variables or account for heteroskedasticity. Third, empirical studies focusing on the MENA region are limited, and few incorporate the role of oil rents or resource dependence.

Third, Despite the breadth of the resource-curse debate, the specific intersection between oil rents, public debt, and financial development remains underexplored, particularly in the MENA region, which is a hydrocarbon-dependent region.

This study seeks to contribute to the literature by filling in these gaps by, first, using a multidimensional measure of financial development, the IMF’s financial development index; secondly, by applying a fractional response model (FRM) to a panel of MENA countries to address a limitation in the approaches used in the literature, which is related to the bounded nature of the dependent variable; third, by adding oil rents as a control variable to investigate the effect of oil rent on financial development in MENA countries. Additionally, the study explores whether the effect of public debt varies by income level and debt burden, two dimensions that have received limited empirical attention.

4. Model Specification

The empirical specification is motivated by the existing literature on the nexus between public debt and financial development, which has typically been explored through reduced-form relationships. By aligning with prior empirical studies and extending the specification to include oil rents, this study investigates whether domestic public debt has a significant effect on financial development, and determining whether this effect is negative, as predicted by the lazy bank and crowding-out hypotheses, or positive, as assumed in the safe asset hypothesis. The analysis is based on the following specifications:

$${FD}_{it}=f({GOVDEBT}_{it},x_{it})$$

(1)

where ${FD}_{it}$ denotes the International Monetary Fund (IMF) financial development index, as published. The index is bounded between 0 and 1, with higher values indicating more advanced financial development.

The key explanatory variable, ${GOVDEBT}_{it}$, represents the government domestic public debt as a percentage of GDP. The vector $x_{it}$ includes a set of control variables selected based on the existing literature: GDP per capita, the GDP deflator, oil rent, and the deposit interest rate.

5. Data and Methods

5.1. Econometric Methods

This study investigates the effect of domestic public debt on financial development in the Middle East and North Africa (MENA) region over the period 2002–2022. A fractional response probit model (FRM) is employed, to overcome limitations related to using standard linear regressions to estimate models that have dependent variables bounded between [0,1]. The model parameters are estimated utilizing Quasi-Maximum Likelihood Estimation (QMLE), which provides consistent estimates under correct mean specification without fully specifying the conditional distribution and is robust to the heteroskedasticity typical of fractional outcomes. Average partial effects (APEs) are then computed based on these parameter estimates, to translate coefficients into economically interpretable impacts. Diagnostic and robustness tests are conducted, including the cross-sectionally augmented IPS (CIPS) unit root test, heteroskedasticity tests, endogeneity test, and re-estimation of the model using a fixed effects model and a logit link function.

5.1.1. Unit Root Test

The fractional response model does not require strict stationarity of the regressors over time because it focuses on modelling the conditional mean of a bounded dependent variable, where outcomes are restricted to the unit interval [0,1] (Papke & Wooldridge, 2008). In such models, the focus lies on the relationship between the dependent and explanatory variables at the same period, and therefore the stationarity of the independent variables is not necessary for the consistency of the estimators. Nevertheless, given the macroeconomic nature of the variables used in this study, which may exhibit trending behavior, the cross-sectionally augmented IPS (CIPS) unit root test by (Pesaran, 2007) was conducted as a measure to ensure a deeper understanding of the underlying data characteristics.

The CIPS test accounts for cross-sectional dependence by estimating a cross-sectionally augmented Dickey–Fuller (CADF) regression for each cross-sectional unit, as follows:

$$∆y_{it}=a_i+b_i{y}_{i,t-1}+c_i{\overline{y}}_{t-1}+d_i∆{\overline{y}}_t+e_{it}$$

where ${\overline{y}}_t$ is the cross-sectional average at time t. The null hypothesis of a unit root $(H_0:b_i=0)$ is tested using individual t-statistics, and the overall CIPS statistic is obtained by taking the simple average across all cross-sectional units:

$$CIPS={\displaystyle \frac{1}{N}}\sum _{i=1}^N{t}_i$$

5.1.2. Fractional Response Model (FRM)

When the dependent variable is bound on [0,1], standard linear regression models are inappropriate for three main reasons. First, they can produce predicted values outside the range [0,1]. Second, they assume homoskedasticity, whereas bounded variables typically exhibit heteroskedasticity: variance is highest around the mid-range and lowest near the bounds. Third, they impose constant marginal effects, which ignores the potential non-linear relationship between the dependent and independent variables.

The FRM, introduced by (Papke & Wooldridge, 1996) and (Papke & Wooldridge, 2008), is an econometric approach that employs nonlinear link functions, logit or probit, to guarantee that all predicted values lie within the unit interval. This makes the FRM suitable for outcomes expressed as proportions, ratios, or indices bounded between zero and one, especially in panel-data settings where unobserved heterogeneity and heteroskedasticity are important considerations.

The FRM is specifically designed to (i) account for the fractional and bounded nature of the dependent variable; (ii) allow for flexible, non-linear functional forms; (iii) address heteroskedasticity, which naturally arises in fractional data where variance is typically highest around mid-range values and lowest near the bounds; and (iv) control for unobserved country-specific heterogeneity in panel data, such as persistent differences in governance, institutional quality, or other structural characteristics, through the inclusion of individual effects.

The FRM framework directly models the conditional mean of the fractional response variable, as follows:

$$E\left(y_i|x_i\right)=P\left(y_i|x_i\right)=G\left(x^{\prime }\beta \right)$$

(2)

where $E\left(y_i|x_i\right)$ denotes the conditional expectation of the dependent variable $y_i$ given the explanatory variables $x_i$, which is equivalent to the conditional probability $P\left(y_i|x_i\right)$. Here, $\mathrm{G}(.)$ is a known as a smooth function, satisfying $0<G\left(z\right)<1$ for all $z\in \mathbb{R}$. There are two choices for G, including the logistic cumulative distribution function (CDF), $G\left(z\right)={\displaystyle \frac{\mathrm{e}\mathrm{x}\mathrm{p}\left(z\right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left(z\right)}}$, and the standard normal CDF, probit function $\mathsf{\Phi }\left(z\right)=\Phi \left({\rm X}\beta \right)$. This approach ensures that the predicted values remain within the unit interval without requiring restrictive assumptions about the distributional form of the outcome variable.

In a subsequent extension, Papke and Wooldridge (2008) adapted this framework to a panel data setting by incorporating unobserved individual effects. They employed the probit function because it “leads to computationally simple estimators in the presence of unobserved heterogeneity or endogenous explanatory variables” (Papke & Wooldridge, 2008). The extended model is specified as follows:

$$E\left(y_{it}|x_{it},c_i\right)=\Phi \left(x_{it}\beta +c_i\right), t=1,\dots ..,T$$

(3)

where $c_i$ represents the unobserved individual effect. (Papke & Wooldridge, 2008) measured the importance of the observed individual effects $c_i$ by averaging the partial effects across the distribution of c, following (Chamberlain, 1980) approach to unobserved effects models. This methodology is particularly advantageous when the panel is characterized by a large cross-sectional and small period t. However, given that this study does not face such a panel structure, we instead account for cross-sectional fixed effects by including dummy variables to capture individual specific heterogeneity. In this paper, the fractional response function is as follows:

$$E\left(y_{it}|x_{it},c_i\right)=\Phi \left(x_{it}\beta +\sum _{j=1}^{N-1}{\delta }_j{D}_{ji}\right)$$

(4)

where $c_i=\sum _{j=1}^{N-1}{\delta }_j{D}_{ji}$, ${\delta }_j$ are estimated parameters for each dummy, $D_i$ is a vector of N − 1 dummy variables for individual countries $i$, and $D_{ji}$ is defined as follows:

$$D_{ji}=\left\{\begin{array}{c}1 if i=j\\ 0 otherwise\end{array}\right.$$

(5)

It is important to mention that the FRM retains the core properties of GLMs, relating the conditional mean of the outcome using a nonlinear link function, such as the logistic or normal CD, to a linear predictor ${\mathrm{x}}^{\prime }\beta$, and assumes a variance that is a function of the conditional mean of the dependent variable. In the generalized linear model (GLM) framework, the variance of $y_{it}$ conditional on $x_{it}$ is a known function of the mean, expressed as follows:

$$E\left(y_{it}|x_{it}\right)=\Phi \left(x_{it}^{\prime }\beta \right)$$

(6)

where ${\mu }_{it}=E\left(y_{it}|x_{it}\right)$. The fractional probit model specifies the conditional mean of the outcome $y_{it}ϵ\left[0.1\right]$, as follows:

$$Var\left(y_{it}|x_{it}\right)=v\left({\mu }_{it}\right)$$

(7)

Under the homoskedasticity assumption, the variance of $y_{it}$ conditional on $x_{it}$ is as follows:

$$Var\left(y_{it}|x_{it}\right)=\Phi \left(x_{it}^{\prime }\beta \right)[1-\Phi \left(x_{it}^{\prime }\beta \right)]$$

(8)

In the FRM model, where the dependent variable lies strictly within the (0,1) interval, the assumption of homoskedasticity is generally inappropriate. Homoskedasticity implies that the conditional variance of the dependent variable, given the explanatory variables, is constant across observations. However, due to the bounded nature of fractional outcomes, this assumption rarely holds. Specifically, the variance of the dependent variable tends to be lower when its value approaches the boundaries of zero or one, and higher when it is near the midpoint (e.g., around 0.5), reflecting the limited range of possible variation. As a result, the conditional variance is inherently heteroskedastic (Papke & Wooldridge, 2008). Although the structure in equation 8 allows variance to change with the mean, it still imposes a fixed functional form that may not capture other sources of heteroskedasticity across observations.

In panel data settings, unobserved heterogeneity or omitted variables may influence both the mean and the variance of the response. To allow for greater flexibility, the variance function can be extended as follows:

$$Var\left(y_{it}|x_{it},z_{it}\right)=\varnothing \left(z_{it}\right).v\left({\mu }_{it}\right)$$

(9)

Or

$$Var\left(y_{it}|x_{it}\right)=\Phi \left(x_{it}^{\prime }\beta \right)\left[1-\Phi \left(x_{it}^{\prime }\beta \right)\right].exp\left(z_{it}^{\prime }\gamma \right)$$

(10)

Here $\varnothing \left(z_{it}\right)$ is a positive function of variables $z_{it}$ associated with heteroskedasticity. A common specification is $\varnothing \left(z_{it}\right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(z_{it}^{\prime }\gamma \right)$, where γ controls how the heteroskedasticity (i.e., variance scaling) responds to changes in $z_{it}$, which ensures positivity and interpretable scaling. This extension accommodates individual-specific or group-specific dispersion and is compatible with robust quasi-likelihood estimation frameworks and generalized estimating equations (GEEs) (Zeger & Liang, 1986), (Wooldridge, 2010)

5.1.3. Heteroskedasticity Specification

To confirm the presence of heteroskedasticity, we follow the approach introduced by Papke and Wooldridge (2008) by testing the null and alternative hypotheses that the variance is constant, as follows:

$$\begin{gathered} H_0:\gamma =0, \\ {H}_0:\ne 0 \end{gathered}$$

(11)

This is equivalent to testing whether the covariates in the variance equation have any explanatory power. The test is implemented using a joint Wald test on the coefficients in the variance equation. Rejection of the null hypothesis provides evidence in favor of modelling heteroskedasticity explicitly.

5.1.4. Quasi-Maximum Likelihood Estimation (QMLE)

Following Papke and Wooldridge (1996), and using quasi-maximum likelihood estimation (QMLE), the QMLE approach specifies a conditional mean function of the form $E\left(y_{it}|x_{it},c_i\right)=\mathsf{\Phi }\left(x_{it}\beta +c_i\right)$. Estimation proceeds by maximizing the Bernoulli quasi-log-likelihood function:

$$Q_N\left(\beta ,\delta \right)=\sum _{i=1}^N\sum _{t=1}^T[y_{it}log\Phi \left(x_{it}^{\prime }\beta +D_i^{\prime }\delta \right)+\left(1-y_{it}\right)\mathit{log}\left(1-\Phi \left(x_{it}^{\prime }\beta +D_i^{\prime }\delta \right)\right)]$$

(12)

where $D_i$ is a vector of dummy variables for individual fixed effects and $\delta$ denotes their associated parameters. Importantly, this quasi-likelihood function does not assume that the dependent variable is Bernoulli-distributed; instead, consistency and asymptotic normality of the QMLE rely solely on correct specification of the conditional mean. This flexibility makes QMLE particularly appropriate for modelling fractional outcomes that do not arise from a known parametric distribution, while preserving the desirable properties of likelihood-based inference (Papke & Wooldridge, 1996); (Wooldridge, 2010). Therefore, the following model is estimated:

$$\begin{gathered} Q_N\left(\beta ,\delta \right)=\sum _{i=1}^N\sum _{t=1}^T[{FD}_{it}log\Phi \left(x_{it}^{\prime }\beta +D_i^{\prime }\delta \right) \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +\left(1-{FD}_{it}\right)\mathit{log}\left(1-\Phi \left(x_{it}^{\prime }\beta +D_i^{\prime }\delta \right)\right)] \end{gathered}$$

(13)

5.1.5. Average Partial Effect (APE)

In nonlinear models such as the fractional response model, marginal effects are not constant across observations because they depend on the specific values of the explanatory variables. The marginal effect of a covariate $x_{it}$ is given by the following:

$${\displaystyle \frac{\partial E\left(y_i|x_i\right)}{\partial x_{ij}}}=g\left(x_i^{\prime }\beta \right).{\beta }_j$$

(14)

Since this effect varies with $x_{it}$ (Papke & Wooldridge, 2008) using the average partial effect (APE), it is computed as follows:

$${APE}_j={\displaystyle \frac{1}{N}}\sum _{i=1}^{N}g\left(x_i^{\prime }\beta \right).{\beta }_j$$

(15)

which captures the average change in the predicted outcome associated with a one-unit increase in the covariate.

5.1.6. Endogeneity Test, Control Function Approach (CF)

The relationship between public debt and financial development is potentially bidirectional. A high level of financial development allows governments to issue more debt, while at the same time public debt could affect the financial market depth, efficiency, and accessibility in positive or negative ways. This simultaneity raises the probability of an existing endogeneity problem, the possibility of reverse causality, which, if unaddressed, would bias conventional estimates.

To examine the potential endogeneity of public debt, we apply the Control Function Approach (CFA) suggested by (Papke & Wooldridge, 2008). The procedure is implemented in two steps. In the first stage, the potentially endogenous variable is explained using suitable instruments to obtain residuals. In the second stage, these residuals are added to the main regression model, where their statistical significance provides a direct test of endogeneity and their inclusion corrects for any resulting bias.

In the first stage, current public debt is regressed on its own lagged values and other exogenous covariates, together with country and year fixed effects to obtain the predicted residuals from the debt equation.

$$\begin{gathered} {GOVDEBT}_{it}={\pi }_0+{\pi }_1{GOVDEBT}_{i,t-1}+{\pi }_2{GOVDEBT}_{i,t-2}+{\pi }_3{GOVDEBT}_{i,t-3} \\ +{\pi }_4{GOVDEBT}_{i,t-4}+{\pi }_5{GOVDEBT}_{i,t-5} \\ +\delta X_{it}+v_{it}+{\mu }_i \end{gathered}$$

(16)

where $X_{it}$ donate instrument variables, ${\mu }_i$ is the unobserved country effect, and ${\widehat{v}}_{it}$ is the error term.

Following the logic of instruments in dynamic panel models proposed by (Arellano & Bond, 1991), we rely on lagged values of public debt as instruments. Debt is highly persistent, making its past values strong predictors of current levels, which meets the relevance condition. At the same time, conditional controls, fixed effects, and deeper lags are unlikely to directly affect financial development. By including countries’ fixed effects and macroeconomic controls, we remove persistent cross-country differences and contemporaneous macro shocks that could jointly drive both debt and financial development. Under these conditions, deeper lags of debt affect current financial development only through their influence on current debt, thereby satisfying the exclusion restriction.

In the second stage, these residuals are included as an additional regressor in the main FRM, as follows:

$${FD}_{it}=\Phi (\alpha +\beta {GOVDEBT}_{it}+\delta X_{it}+\rho {\widehat{v}}_{it}+{\mu }_i)$$

(17)

where $\Phi (.)$ denotes the probit link function, ${\widehat{v}}_{it}$ is the first-stage residual, $X_{it}$ is the vector of control variables, and ${\mu }_i$ represent country fixed effects, respectively. The null and alternative hypotheses are as follows:

$$\begin{gathered} H_0:\rho =0, \\ {H}_1:\rho \ne 0 \end{gathered}$$

(18)

A statistically insignificant coefficient on the residual supports the exogeneity of debt, while a significant coefficient indicates the presence of endogeneity and validates the control function correction.

5.2. Data

This study utilizes annual panel data for 16 MENA countries over the period 2000–2022. The dependent variable is the financial development index (FD), obtained from the IMF. The main explanatory variable is government domestic public debt as % of GDP, sourced from the World Bank and national statistical agencies. Control variables include GDP per capita, the GDP deflator, the deposit interest rate, and oil rents as a percentage of GDP. Oil rents are introduced as a novel control to account for the macro-financial effects of oil dependence. Variable definitions and sources are provided in

Table 1. Descriptions of variables and data sources.

Code	Full Variable Name	Definition/Construction	Source(s)
FD	Financial Development Index	Composite index of financial institutions and markets covering three dimensions: depth (size and liquidity), access (availability of financial services), and efficiency (cost-effectiveness and sustainability).	International Monetary Fund (IMF)
GOVDEBT	Domestic Credit to Government and State-Owned Enterprises by Money Banks	Ratio between credit by domestic money banks to the government and state-owned enterprises and GDP.	World Bank
OILRENT	Oil Rents	The difference between the value of crude oil production at regional prices and total costs of production.	World Bank
DEF	Gross Domestic Product (GDP) Deflator	The GDP implicit deflator is the ratio of GDP in current local currency to GDP in constant local currency.	World Bank
R	Deposit Interest Rate	The rate is paid by commercial or similar banks for demand, time, or savings deposits.	World Bank and national central banks
GDP	GDP Per Capita (log)	Natural logarithm of GDP per capita is GDP divided by midyear population. GDP is the sum of the gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in current U.S. dollars.	World Bank

.

The dataset consists of a panel of 336 observations across 16 countries from the Middle East and North Africa (MENA) region.

Table 2. Descriptive statistics.

	Mean	Median	Std. Dev.	Min	Max	Skewness	Kurtosis
DEF	246.85	88.98	1294.51	7.52	20,260.56	12.51	177.15
GDP	15,902.43	6369.63	18,565.45	378.16	98,041.36	1.77	3.51
R	6.951	3.24	12.37	0.12	78.1	3.51	12.03
GOVDEBT	17.05	11.99	14.1	0.29	70.66	1.26	1.46
OILRENT	17.32	13.88	16.57	0	64.82	0.78	−0.33
FD	0.33	0.35	0.16	0.07	0.64	−0.10	−1.17
Obs.	336

displays descriptive statistics and highlights key distributional features of the data that carry important methodological implications. The financial development index, FD, which serves as the dependent variable, is fractional and bounded between 0 and 1. It exhibits a relatively symmetric distribution, with a skewness of minus 0.101 and a kurtosis of minus 1.172, indicating limited concentration at the tails. In contrast, several explanatory variables, such as the GDP deflator, DEF, and the interest rate, R, show substantial positive skewness of 12.515 and 3.506, respectively, along with high kurtosis values of 177.148 and 12.025. These patterns reflect the presence of extreme values and strong departures from normality. These characteristics, particularly the bounded nature of the dependent variable and the non-normality of key regressors, justify the use of the fractional response model, FRM. As emphasized by (Papke & Wooldridge, 2008), the FRM is well suited for modelling fractional outcomes and remains robust in the presence of skewed distributions and heavy-tailed explanatory variables.

6. Results

6.1. Unit Root Test—CIPS

The results of the cross-sectionally augmented IPS (CIPS) panel unit root tests, reported in

Table 3. CIPS unit root test.

	Level		First Difference
Variable	With Constant and Trend	With Constant	With Constant and Trend	With Constant
FD	−2.795 **	−2.159 *	−4.766 ***	−4.833 ***
DEF	−1.995	−0.298	−3.172 ***	−3.130 ***
GDP	−2.414	−1.689	−3.382 ***	−3.284 ***
R	−2.175	−1.839	−3.169 ***	−3.184 ***
GOVDEBT	−2.065	−1.561	−3.998 ***	−3.683 ***
OILRENT	−2.426	−2.270 **	−4.401 ***	−4.420 ***

* Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level.

, show that financial development rejects the null hypothesis of a unit root, indicating stationarity. In contrast, the explanatory variables, including domestic public debt, GDP per capita, deposit rate, and oil rents, appear non-stationary at levels under both specifications (constant only and constant with trend). However, after first differencing, all explanatory variables reject the null of a unit root at the 1% significance level, confirming that they are stationary in first differences and are therefore integrated with order one, I(1). This suggests that the model may capture long-run relationships between the independent variables and financial development. According to (Papke & Wooldridge, 2008), although the regressors are non-stationary at levels, their use in the fractional response framework remains valid because of the bounded nature of the dependent variable, which ensures the consistency of the estimators.

6.2. Heteroskedasticity Specification

To investigate the presence and sources of heteroskedasticity in financial development, we estimated a series of FRMs, allowing the conditional variance to depend on different candidates of explanatory variables. The results, presented in

Table 4. Heteroskedasticity tests.

Model	Log-Likelihood	${\mathit{R}}^2$	$\mathit{\gamma }$	p-Value
Country Dummy Variable	−194.304	0.08575	0.10769	<0.01
DEF	−194.38	0.08337	−4.00 × 10−5	<0.01
GOVDEBT	−194.369	0.08141	0.00466	0.018
R	−194.263	0.08219	0.01649	<0.01
OILRENT	−194.377	0.08497	−0.00317	0.072
log_GDP	−194.383	0.04857	−0.0286	0.524

, show that among all specifications, the model incorporating unobserved country-specific heterogeneity $c_i$, country dummy variables, demonstrated the best overall performance. It achieved the highest log-likelihood at −194.304 and the highest R² at 0.08575, and the estimated heteroskedastic component, $\gamma$ = 0.1077, was significant at 1%, suggesting substantial country-specific heterogeneity, such as institutional or structural differences, is an important source of heteroskedasticity. Although prices, DEF, the level of public debt, GODEBT, and interest rate, R, also have statistically significant variance parameters at 1%, their model was relatively weaker in terms of and not outperforming specification with $c_i$. They have less $R^2$ and a very small coefficient of variance component $\gamma$. Meanwhile, the difference between countries at the level of GDP and OILRENT was not statistically significant. These results indicate that different macroeconomic circumstances among countries do not affect the heteroskedasticity in the model, and unobserved heterogeneity $c_i$ is a primary source of heteroskedasticity, consistent with the theoretical predictions of this paper.

6.3. Estimation Results

Table 5. FRM estimation results.

Variable	FE Coef.	FE Std. Err.	FRM Coef.	FRM Std. Err.	APE (dy/dx)	APE Std. Err.
GOVDEBT	−0.00149	0.000367	−0.01048	0.001975	−0.00156	0.000313
log_GDP	0.023147	0.007804	0.098014	0.042968	0.014604	0.007853
R	−0.0019	0.00083	−0.02251	0.009711	−0.00335	0.001154
DEF	−1.99 × 10−7	2.30 × 10−6	−3.84 × 10−6	1.03 × 10−5	−5.73 × 10−7	1.55 × 10−6
OILRENT	−0.00088	0.000427	−0.00773	0.002448	−0.00115	0.000306
Joint Significance (test stat)	F-statistic 6.31 ***		Wald chi-square 2981.97 ***
Joint Significance (p-value)	0		0
Observations	336		336

*** Significant at 1% level.

presents the FRM estimation results for the fractional response probit model, which constitutes the main empirical specification, alongside a linear fixed effects (FE) model included solely to assess the robustness of the findings. The results indicate that government debt (GOVDEBT) is negatively and significantly associated with financial development across both models. In the FRM model, the coefficient estimate is −0.0105 (after rounding) with significance at the 1% level, and the corresponding average partial effect (APE) is −0.00156, implying that a one percentage point increase in the debt-to-GDP ratio is associated, on average, with a 0.16 percentage point reduction in the financial development index. The FE model, on the other hand, produces a qualitatively similar result, emphasizing the credibility of the main finding.

The estimated coefficient with respect to log GDP is positive and statistically significant in both models. The APE from the FRM is 0.0146, suggesting that higher levels of income per capita are associated with greater financial sector development, implying that an increase in income per capita is associated with a 1.46 percentage point rise in the predicted level of financial development, on average. The interest rate (R) also exhibits a negative and statistically significant effect on financial development in both specifications, with an APE of −0.00335, indicating that a one percentage point increase in the interest rate is associated with a 0.335 percentage point decrease in the predicted level of financial development, on average. While oil rents (OILRENT) are likewise negatively associated with the outcome, a one percentage point increase in oil rents as a share of GDP is associated with a 0.115 percentage point decrease in the predicted level of financial development, on average, consistent with the hypothesis that resource dependence may inhibit financial sector growth. In contrast, the GDP deflator (DEF) does not show a statistically significant effect in either model, which reflects that the opportunity cost does not affect financial development.

Joint significance tests confirm the overall explanatory strength of the models, with a Wald chi-square statistic of 2981.97, significant at the 1% level for the FRM, and an F-statistic of 6.31 with significance at the 1% level for the FE model. The consistency of estimated signs and statistical significance across specifications supports the robustness of the core results, while the FRM’s average partial effects offer a more interpretable and policy-relevant measure of the covariates’ marginal contributions.

6.4. Margins at Different Levels of Income

A growing body of empirical literature [such as (A. Othman et al., 2022; Abille & Kiliç, 2023; Alsamara et al., 2024; Elkhadrawi, 2025; Gomez-Puig et al., 2022; Makhoba et al., 2021; Mensah et al., 2020; Ndoricimpa, 2020; Reinhart & Rogoff, 2010)] has documented the existence of a nonlinear relationship between income levels and financial development, suggesting that the impact of GDP level on the financial sector may vary across different stages of development. This nonlinearity underscores the importance of exploring how different levels of income may influence the marginal effects of policy variables such as public debt.

Table 6. Summary statistics for log_GDP.

Statistic	Value	Statistic	Value
1st Percentile	6.408539	Minimum	5.935311
5th Percentile	6.923879	Maximum	11.49314
10th Percentile	7.243413	Mean	8.932087
25th Percentile	7.918963	Standard Deviation	1.310873
50th Percentile (Median)	8.759199	Variance	1.718388
75th Percentile	10.08181	Skewness	−0.02761
90th Percentile	10.66827	Kurtosis	1.886068
95th Percentile	10.86505	Observations	336
99th Percentile	11.44028

summarizes statistics for log-GDP in the sample, revealing a wide dispersion in income levels, with values ranging from 5.94 to 11.49 and a standard deviation of 1.31. The interquartile range spans from 7.92 (25th percentile) to 10.08 (75th percentile), indicating that 50% of the observations fall within this moderately wide income band. The mean of 8.93 is slightly higher than the median of 8.76, reflecting a mildly left-skewed but approximately symmetric distribution, as supported by the near-zero skewness of −0.03. A kurtosis value of 1.89 further suggests a relatively flat distribution compared to the normal. These distributional characteristics support the empirical strategy of estimating marginal effects at various points of the log-GDP distribution to investigate whether the influence of public debt on financial development varies systematically with income levels. The range and shape of the data provide a solid foundation for probing heterogeneity across low-, middle-, and high-income economies.

To explore the heterogeneity in the relationship between public debt and financial development across different income levels, we estimate the marginal effect of government debt at the 25th, 50th, and 75th percentiles of the log-GDP distribution. The results, reported in

Table 7. The marginal effect of public debt on financial development at different levels of GDP.

log-GDP Percentile	Countries	$\mathit{\partial }\mathit{F}\mathit{D}/\mathit{\partial }\mathit{G}\mathit{O}\mathit{V}\mathit{D}\mathit{E}\mathit{B}\mathit{T}$	Std. Err.	p-Value	95% CI
25th (7.92)	Egypt, Mauritania, Sudan	−0.00154	0.0003	0	[−0.00214, −0.00095]
50th (8.76)	Algeria, Iran, Jordan, Tunisia	−0.00157	0.00032	0	[−0.00220, −0.00095]
75th (10.08)	Bahrain, Libya, Morocco, Oman, Saudi Arabia, Israel, Kuwait, Qatar, United Arab Emirates	−0.00162	0.00034	0	[−0.00228, −0.00096]

, show that the negative effect of government debt persists across the income spectrum and intensifies slightly as income increases. Specifically, at the 25th percentile, log-GDP = 7.92, a one percentage point increase in the debt-to-GDP ratio is associated with a 0.15 percentage point decline in the financial development index. This marginal effect rises modestly to 0.16 percentage points at the 75th percentile, log-GDP = 10.08. All estimates are statistically significant at the 1% level. These findings suggest that the adverse effect of public debt on financial development is present across income groups but may be slightly more pronounced in higher-income economies.

6.5. Margins at Different Levels of Public Debt

Empirical research, such as (Benayed & Gabsi, 2020; Bui, 2018; Hauner, 2009), has shown that the relationship between public debt and financial development may be nonlinear, particularly as excessive levels of debt can have diminishing or even adverse effects on the functioning of the financial sector. Understanding the distribution of public debt across countries is therefore essential for investigating potential threshold effects. The summary in

Table 8. Summary of statistics for public debt.

Statistic	Value	Statistic	Value
1st Percentile	0.854473	Minimum	0.292208
5th Percentile	2.219363	Maximum	70.65548
10th Percentile	3.180247	Mean	17.05326
25th Percentile	5.830029	Standard Deviation	14.10485
50th Percentile (Median)	11.98536	Variance	198.9469
75th Percentile	24.34495	Skewness	1.256344
90th Percentile	36.22748	Kurtosis	4.463653
95th Percentile	44.53928	Observations	336
99th Percentile	63.64021

related to statistics for public debt reveals a highly dispersed and positively skewed distribution within the sample of 336 observations. The debt-to-GDP ratio ranges from a minimum of 0.29% to a maximum of 70.66%, with a mean of 17.05% and a median of 11.99%. The interquartile range spans from 5.83% (25th percentile) to 24.34% (75th percentile), suggesting that half of the countries in the sample have moderate debt levels. However, the relatively high standard deviation of 14.10 and the positive skewness coefficient of 1.26 indicate the presence of a long right tail, reflecting a group of countries with substantially higher debt burdens. Furthermore, the kurtosis value of 4.46 suggests a leptokurtic distribution, with fatter tails than the normal distribution. These characteristics provide empirical motivation for modelling nonlinear effects of debt on financial development, particularly for testing whether the marginal impact of debt intensifies beyond certain threshold levels.

To explore this relationship in our context, we estimate the marginal effect of public debt (GOVDEBT) on financial development (FD) at three distinct levels of public debt: 5.83%, 11.99%, and 24.34%, corresponding to the 25th, 50th, and 75th percentiles of the observed debt distribution. The results, presented in

Table 9. The marginal effect of public debt on financial development at different levels of public debt.

Public Debt Percentile		$\partial \mathit{F}\mathit{D}/\partial \mathit{G}\mathit{O}\mathit{V}\mathit{D}\mathit{E}\mathit{B}\mathit{T}$	Std. Err.	p-Value	95% CI
25th (5.83)	Iran, Mauritania, Sudan	−0.00162	0.000327	0	[−0.00226, −0.00097]
50th (11.99)	Israel, Oman	−0.0016	0.000321	0	[−0.00223, −0.00097]
75th (24.34)	Kuwait, Saudi Arabia, Tunisia, Algeria, Bahrain, Egypt, Jordan, Libya, Morocco, Qatar, United Arab Emirates	−0.00155	0.000306	0	[−0.00215, −0.00095]

, show that the marginal effect is consistently negative and statistically significant across all levels. At a debt level of 5.83%, a one percentage point increase in public debt is associated with a 0.162 percentage point reduction in the financial development index. This effect decreases slightly in magnitude to −0.00160 at 11.99% debt and to −0.00155 at 24.34% debt. While the differences are relatively modest, the pattern suggests a slightly diminishing marginal impact of government debt at higher levels. This may reflect better institutional capacity to manage debt in high-debt economies, nonlinear policy responses, or thresholds beyond which additional debt becomes less harmful. These findings reinforce the importance of modelling nonlinear effects in examining the public debt–financial development nexus.

6.6. Robustness Check

6.6.1. Endogeneity Test

To assess the robustness of our empirical findings and address potential endogeneity concerns, we implement the Control Function Approach (CFA), as we indicated in section (4.1.5). Our analysis proceeds in three steps. First, we estimate the first-stage debt equations using lagged values of debt and control variables as instruments (

Table 10. Debt estimation.

Variables	Lag2–Lag4 IV Set		Lag2–Lag5 IV Set
	Coef.	(SE)	Coef.	(SE)
${GOVDEBT}_{t-1}$	0.884	(0.053) ***	0.871	(0.055) ***
${GOVDEBT}_{t-2}$	−0.187	(0.147)	−0.182	(0.159)
${GOVDEBT}_{t-3}$	−0.086	(0.107)	−0.078	(0.118)
${GOVDEBT}_{t-4}$	0.172	(0.168)	0.164	(0.235)
${GOVDEBT}_{t-5}$	—	—	0.003	(0.117)
log_GDP	3.487	(1.531) **	3.675	(1.993) *
R	0.041	(0.045)	0.037	(0.047)
DEF	−0.000069	(0.000049)	−0.000052	(0.000074)
OILRENT	−0.252	(0.054) ***	−0.251	(0.056) ***
Constant	−23.026	(13.088) *	−24.595	(17.554)
Observations	272		256
R-squared	0.928		0.928
F (Excluded lags)	2.55		4.10
p-value (Excluded lags)	0.094		0.019

*, **, *** denote 10%, 5%, 1%.

). Second, we re-estimate the fractional probit models of financial development, augmenting the specification with the residual from the first stage to formally test for endogeneity of debt (

Table 11. FRM with endogeneity.

Variables	Baseline FRM (N = 336)		CF (L2–L4) (N = 272)		CF (L2–L5) (N = 256)
	Coef.	(SE)	Coef.	(SE)	Coef.	(SE)
GOVDEBT	−0.01048	(0.001) ***	−0.01372	(0.00463) ***	−0.01203	(0.00442) ***
Log_GDP	0.09801	(0.0429) **	−0.09014	(0.09691)	−0.18789	(0.10016) *
R	−0.02251	(0.0097) **	−0.01639	(0.01643)	−0.01083	(0.01427)
DEF	−0.00000384	(0.0000103)	0.00000391	(0.0000374)	−0.00000421	(0.0000225)
OILRENT	−0.00773	(0.00244) **	−0.01345	(0.00633) **	−0.01097	(0.00653) *
${\widehat{v}}_{it}$ (L2–L4)	—	—	0.00698	(0.00516)	—	—
${\widehat{v}}_{it}$ (L2–L5)	—	—	—	—	0.00654	(0.00463)
Log-likelihood	−194.304		−158.282		−149.140
Pseudo R2	0.0858		0.0886		0.0891

*, **, *** denote 10%, 5%, 1%.

). Finally, we compute the average partial effects (APEs) of debt and other covariates to provide interpretable magnitudes of their impact on financial development across all specifications (

Table 12. Average partial effects (APEs) with endogeneity.

Variable	Baseline FRM (N = 336)		CF (L2–L4) (N = 272)		CF (L2–L5) (N = 256)
	APE	(SE)	APE	(SE)	APE	(SE)
Domestic public debt	−0.00156	(0.000313)	−0.001689	(0.001003)	−0.001649	(0.000902)
log(GDP)	0.014604	(0.007853)	−0.011099	(0.010167)	−0.025752	(0.012775)
R	−0.00335	(0.001154)	−0.002018	(0.001252)	−0.001485	(0.001376)
DEF	−0.000000573	(0.00000155)	0.000000482	(0.00000439)	−0.000000578	(0.00000335)
Oil rents	−0.00115	(0.00306)	−0.001656	(0.000706)	−0.001503	(0.000644)

).

Table 10 reports the first-stage regression of domestic public debt on its lagged values and exogenous controls, with country fixed effects. The coefficient on ${GOVDEBT}_{t-1}$ is strongly significant and close to unity, reflecting persistence in debt dynamics. The joint F-statistic on the excluded lags ${(GOVDEBT}_{t-2} \mathrm{t}\mathrm{o} {GOVDEBT}_{t-4}$) is 2.55 with a probability of p = 0.094, suggesting weak relevance. However, extending the instrument set to ${GOVDEBT}_{t-5}$ increases the F-statistics to 4.10 with a probability of p = 0.019, which is reflects improving instrument strength. Among the other control variables, log_GDP and oil rents show significant associations with public debt, which is consistent with macroeconomic fundamentals influencing borrowing needs. These results confirm that lagged debt variables provide a reasonable and moderately strong instrument set for the (CFA).

Table 11 presents the FRM estimates of financial development, comparing the baseline specification with the control function models using ${GOVDEBT}_{t-2} \mathrm{t}\mathrm{o} {GOVDEBT}_{t-4}$ (L2–L4) and ${GOVDEBT}_{t-2} \mathrm{t}\mathrm{o} {GOVDEBT}_{t-5}$ (L2–L5) as instruments. Our results show that ${\widehat{v}}_{it}$ are insignificant across all specifications, which indicates that public debt can be treated as exogenous. Therefore, our empirical results suggest that endogeneity is not driving the main findings.

Nevertheless, across all specifications, domestic public debt exerts a negative and statistically significant effect on financial development. In the baseline model, the coefficient is –0.010 (p < 0.05), while the control function specifications yield slightly larger magnitudes (–0.012 to –0.014) with similar significance. The residual from the first stage (the control function term) is insignificant in both models (p = 0.176 and 0.158), indicating that we cannot reject the null hypothesis of the public debt exogeneity. Log_GDP switches sign under the CF (L2–L4) specification and is weak significant under the CF (L2–L5) specification. Oil rents consistently display a negative and significant effect overall, with the control function results demonstrating robustness of the debt–finance link, suggesting that endogeneity is not driving the main findings.

Table 12 reports the average partial effects (APEs) from the baseline and control function fractional probit models. Domestic public debt reduces financial development by 0.0016–0.0017 points on average. The magnitude of the effect is stable across baseline and CF models, reinforcing the robustness of the results. Among the controls, oil rents consistently show negative marginal effects, confirming the adverse impact of resource dependence on financial development. The effects of log_GDP, inflation, and interest rate remain economically modest relative to the debt effect. These results provide clear evidence that higher public debt is associated with lower levels of financial development in MENA countries, even after correcting for potential endogeneity.

In summary, the control function tests confirm that public debt is exogenous, as the residual terms from the first-stage regressions are statistically insignificant. Accordingly, there is no evidence of endogeneity in our specification. Nevertheless, across all models, public debt exerts a negative and robust effect on financial development, with stable average partial effects.

6.6.2. Estimates of the Model Using a Logit Link Function

To verify the robustness of our findings, we re-estimate the fractional response model using a logit link function in place of the probit specification.

Table 13. Robustness verification.

Variable	APE (Logit)	Logit Std. Err.	APE (Probit)	Probit Std. Err.
GOVDEBT	−0.00134	0.000391	−0.00156	0.000313
log_GDP	0.024223	0.007444	0.014604	0.007853
R	−0.00228	0.00136	−0.00335	0.001154
DEF	−2.22 × 10−7	9.97 × 10−7	−5.73 × 10−7	1.55 × 10−6
OILRENT	−0.00079	0.000393	−0.00115	0.000306
Joint significance (Wald ${Chi}^2$)	13,719.72 ***		2981.97 ***

*** Significant at 1% level.

reports the average partial effects (APEs) from both models, along with robust standard errors and joint significance statistics. Across both specifications, the estimated effects remain stable in sign, magnitude, and significance.

Specifically, government debt consistently shows a statistically significant negative association with financial development, with APEs of −0.00156 in the probit model and −0.00134 in the logit model. Log GDP maintains a positive and significant effect, while the interest rate R is negative in both cases, although marginally significant in the logit model. Oil rents exhibit a robust negative effect across both link functions. The GDP deflator (DEF) remains insignificant, further confirming the stability of core results.

Importantly, both models reject the null hypothesis of joint insignificance at the 1% level, as indicated by Wald χ² statistics of 2981.97 in the probit model and 13,719.72 in the logit. These results affirm the joint explanatory power of the model and demonstrate that the findings are not sensitive to the choice of link function, in line with the robustness properties of quasi-maximum likelihood estimation.

7. Discussion and Policy Implications

7.1. Discussion

The empirical findings of this study confirm a statistically significant and economically meaningful negative relationship between domestic public debt and financial development, consistent with the “lazy banks” and “crowding-out” hypotheses. The results, derived from a fractional response model and corroborated by a fixed effects specification, suggest that increased public debt leads to a measurable reduction in the financial development index. Specifically, the average partial effect implies that a one percentage point increase in the debt-to-GDP ratio results in a 0.16 percentage point decline in the financial development index, reinforcing the hypothesis that high public debt suppresses private credit intermediation.

This outcome is consistent with previous findings from studies such as (Abbas et al., 2022; Ahmed et al., 2024; Emran & Farazi, 2009), which showed that elevated public borrowing tends to crowd out private sector lending, particularly in economies with underdeveloped financial markets. It also aligns with the theoretical predictions of (Manove et al., 2001), where risk-averse banks prioritize safe government securities over private lending in the presence of public debt, reducing financial intermediation quality and innovation.

The analysis further indicates that the marginal negative effect of public debt on financial development is present across different income levels and persists at varying levels of debt. While the impact diminishes slightly at higher debt levels, possibly reflecting improved institutional capacities or policy frameworks, its consistent statistical significance highlights the nature of the public debt burden on financial markets. These results also contribute to the literature on nonlinear relationships, suggesting that even modest increases in debt can affect financial development adversely, particularly when institutional quality is weak or when financial markets are shallow.

The adverse association between oil rent and financial development is likely to operate through several separate channels. First, fiscal instability and procyclicality: price-driven swings in hydrocarbon revenues relax the government’s need to mobilize domestic savings in booms and compress credit in busts, weakening incentives to build broad intermediation capacity. Second, governance and soft budget constraints: large resource rents can dilute fiscal discipline and oversight, lowering the demand for efficient screening and monitoring by domestic financial institutions. Third, crowding-out and lazy-bank behavior: abundant sovereign securities and sizable government and National Oil Company (NOC) deposits provide low-risk, high-liquidity assets that shift banks’ portfolios away from private lending. Fourth, foreign-currency denomination: dollar-based receipts and offshore financing channels reduce the role of local-currency intermediation. Because oil revenues are largely denominated in U.S. dollars and accrue directly to governments, resource-rich countries often substitute international financial flows for domestic intermediation. Finally, Dutch disease and sectoral composition may shrink the tradable sector and the pipeline of bankable projects. These mechanisms are consistent with our empirical finding that higher oil rents coincide with lower financial development, including at higher income levels, and help rationalize why the effect persists across country groups.

Country-specific heterogeneity was identified as the most substantial source of heteroskedasticity in the model, emphasizing the critical role of unobserved structural and institutional factors in shaping the debt–finance nexus. This underscores that cross-country differences, such as legal protections, fiscal governance, and banking regulations, may mediate the extent to which public debt influences financial development.

7.2. Policy Implications

The results of this study underscore that rising domestic public debt is consistently associated with lower levels of financial development in MENA countries, even when controlling for oil rents and income heterogeneity. The evidence that the negative effect persists across both low- and high-income groups indicates that institutional strength alone does not fully shield economies from debt-related crowding-out or lazy-bank behavior. This has several implications for policy design.

First, fiscal policy should prioritize frameworks that explicitly account for the interaction between debt and financial development. The modest attenuation of adverse effects at higher debt levels in our results suggests that institutional capacity may mitigate, but not eliminate, debt-induced distortions. Fiscal rules that stabilize debt dynamics and build buffers during commodity booms can reduce procyclicality. However, these rules should be countercyclical in design; rigid targets risk forcing contraction during downturns, especially in resource-dependent economies.

Second, the strong negative association between oil rents and financial development highlights the importance of insulating domestic credit markets from commodity cycles. Sovereign wealth funds and stabilization mechanisms can help smooth expenditure paths, reduce reliance on debt during downturns, and maintain a more consistent demand for domestic financial intermediation.

Third, to mitigate the sovereign–bank nexus while safeguarding financial stability, policymakers should consider assigning a non-zero risk weight to government securities in banks’ balance sheets. Such a measure would enhance risk recognition, reduce incentives for “lazy banking,” and limit the crowding out of private sector credit, thereby strengthening the contribution of the financial sector to economic development. However, reforms must account for important trade-offs, in particular the potential for procyclical amplification during crises, the critical liquidity role of government bonds in interbank and central bank operations, and the risk of destabilizing fiscal financing in countries with weaker institutions. A balanced approach could involve a gradual and differentiated application, such as risk weights linked to sovereign credit ratings or debt sustainability indicators, the use of concentration limits to prevent excessive exposure, or the adoption of dynamic and countercyclical buffers that are tightened in good times but relaxed in downturns. Importantly, these measures should be implemented in parallel with credible fiscal rules and macroprudential frameworks, ensuring that the benefits of market discipline are realized without undermining financial stability or growth.

Finally, the heterogeneity observed across income levels and country contexts underscores the need for differentiated sequencing of reforms. In countries with limited institutional capacity, priority should be given to strengthening fiscal transparency, creditor-rights protection, and supervisory independence before adopting more advanced macroprudential tools. In higher-income economies, the focus should be on integrating debt management with broader macroprudential frameworks that align sovereign financing, collateral regulation, and liquidity management.

In conclusion, these implications emphasize that sustainable financial development in the MENA region requires more than controlling debt ratios; it demands coordinated reforms in fiscal governance, resource revenue management, and prudential regulation that explicitly account for the debt–finance–resource nexus identified in this study.

8. Conclusions

This study provides robust empirical evidence on the relationship between domestic public debt and financial development, using a fractional response model tailored for bounded dependent variables. The findings support the “lazy banks” and “crowding-out” hypotheses, showing that rising public debt is associated with a statistically and economically significant decline in financial development. Specifically, the average partial effect suggests that each one percentage point increase in the debt-to-GDP ratio reduces the financial development index by approximately 0.16 percentage points. These results are consistent across model specifications and remain robust with respect to alternative link functions and various levels of debt and income.

By integrating oil rents into the model, this paper also identifies a significant negative effect of resource dependence on financial development, further emphasizing the structural constraints faced by resource-rich economies. The analysis of heteroskedasticity underscores the importance of unobserved country-specific institutional and structural characteristics in shaping the debt–finance nexus. While higher income levels slightly attenuate the negative impact of debt, the effect persists, reinforcing the need for differentiated policy responses based on country context.

These results contribute to the ongoing debate on the dual role of public debt as both a potential stabilizing instrument and a source of financial repression. They suggest that while government borrowing can provide safe assets and liquidity in some contexts, excessive reliance on domestic debt may hinder financial intermediation, especially in environments characterized by institutional weaknesses and limited market depth.

Future research should explore the dynamic interactions between public debt, institutional quality, and financial development using structural macroeconomic models or panel VAR techniques. Further disaggregation of financial development dimensions—such as access, depth, and efficiency—could provide more granular insights into the channels through which public debt affects the financial system. Investigating nonlinear and asymmetric effects across different regions and economic structures also holds promise for refining policy prescriptions in diverse macro-financial environments.