Diversification and nonlinear credit cyclicality in dual-banking systems Open Access

The cyclicality of bank lending, reflected in its expansion during booms and contraction in downturns, is a well-documented yet persistent driver of macro-financial instability. This amplification effect exacerbates macroeconomic volatility and systemic risk (Andrieș and Sprincean, 2021; Micco and Panizza, 2006). In response, regulators have implemented macro-prudential tools such as countercyclical capital buffers to dampen these cycles (Auer et al., 2022). While various bank-specific characteristics have been found to moderate procyclicality, including size, capitalization, asset quality and ownership structure (Allen et al., 2017; Bertay et al., 2015; Dursun-de Neef and Schandlbauer, 2020; Iwanicz-Drozdowska and Witkowski, 2016), the role of diversification in mitigating cyclicality remains insufficiently explored. This study addresses that gap by examining whether diversification stabilizes lending behavior across business cycles, focusing on dual-banking systems comprising both Islamic and conventional banks.

Islamic banks are theoretically expected to behave differently over the business cycle, owing to their reliance on Islamic financing contracts that tie financing to real-sector performance (Mirakhor, 2018). Yet empirical findings are mixed. Some studies find Islamic banks as procyclical as conventional banks or even more so (Albaity et al., 2022; Ascarya et al., 2016; Aysan and Ozturk, 2018), while others report more countercyclical or less procyclical patterns, especially during downturns (Ibrahim, 2016; Saadaoui and Hamza, 2020; Šeho et al., 2024a; Zulkhibri and Sakti, 2018). These inconsistencies may reflect differences in bank size, contract structures and jurisdictional contexts.

From a theoretical standpoint, diversification may enable banks to smooth income and reduce exposure to sector-specific shocks, thereby sustaining lending when macroeconomic conditions deteriorate. However, excessive diversification may dilute monitoring efficiency (Acharya et al., 2006; Tabak et al., 2011) or lead banks into riskier segments, raising the question of whether diversification is always stabilizing or whether its effects depend on the bank’s business model, regulatory environment and cycle stage.

Parallel to this debate, an emerging literature explores the stabilizing potential of diversification. While sectoral or geographic diversification has been associated with more resilient credit supply during crises (Doerr and Schaz, 2021; Gelman et al., 2022), its effect on lending cyclicality over the full business cycle, especially in dual-banking systems, remains under-investigated. Notably, Šeho et al. (2024b) show that diversification can amplify procyclicality in Malaysian banks, particularly among Islamic institutions, suggesting that diversification’s impact is nonlinear and context-dependent.

Despite these advances, several limitations remain. First, much of the credit cyclicality literature continues to model the relationship between economic activity and lending as linear, potentially overlooking nonlinear and phase-dependent dynamics. Even studies that incorporate quadratic terms typically infer significance from coefficient estimates alone, without examining whether cyclical effects vary across different points of the business cycle. Second, while diversification has been linked to bank stability and crisis resilience, its role in shaping the sensitivity of credit supply to macroeconomic fluctuations has not been systematically examined. Third, existing studies rarely integrate loan portfolio and income diversification within a unified dynamic framework that allows for heterogeneous effects across banking models. These gaps limit our understanding of whether diversification merely affects the level of credit growth or fundamentally alters its cyclical responsiveness.

To address these limitations, this study makes three distinct contributions to the literature on bank lending cyclicality and financial stability. First, it advances the procyclicality literature by modeling the relationship between economic activity and credit growth as nonlinear and by explicitly evaluating the marginal effects across different phases of the business cycle. Rather than inferring nonlinear dynamics solely from the significance of the quadratic coefficient, the analysis identifies the specific ranges of economic activity over which credit responses are statistically significant. It further compares these phase-dependent dynamics across Islamic and conventional banks. Second, it moves beyond the traditional focus on diversification and bank performance by examining whether loan portfolio and income diversification systematically moderate the responsiveness of credit supply to macroeconomic fluctuations. In doing so, it distinguishes between the direct and nonlinear effects of diversification and its role in shaping cyclical sensitivity. Third, it integrates nonlinear cyclicality and diversification-based moderation within a unified dynamic framework, allowing the moderating effects of diversification to vary across banking models and thereby offering a structural perspective on how portfolio composition influences macro-financial transmission in dual-banking systems.

The Gulf Cooperation Council (GCC) provides an ideal empirical setting for this analysis for several reasons. First, GCC countries host dual banking systems where Islamic and conventional banks operate under comparable regulatory and macro-financial conditions. Second, since 2010, these economies have undertaken substantial diversification efforts away from hydrocarbons, broadening the base of real economic activity and creating new lending opportunities (Esam, 2021). Third, the rapid expansion of domestic credit to the private sector in the GCC, from 50.4% of GDP in 2008–107.3% in 2020 ^[1], raises concerns about potential credit bubbles and systemic instability (Alessi and Detken, 2018). Studying this region helps isolate banking effects from broader institutional differences that often confound cross-country comparisons.

Understanding whether internal strategies such as income and loan diversification can serve as buffers during downturns has significant implications for macroprudential design. If diversification reduces cyclicality, it could complement regulatory tools by building resilience from within bank balance sheets. Moreover, evidence from dual-banking systems provides a useful testbed for identifying bank-type heterogeneity in response to economic shocks.

In preview, our findings reveal three central results. First, both bank types exhibit nonlinear procyclicality in lending but with distinct patterns: conventional banks respond positively across the cycle, while Islamic banks are procyclical only during downturns. Second, income and loan diversification influence credit behavior asymmetrically. Income diversification supports credit growth but amplifies procyclicality in highly diversified conventional banks. In contrast, sectoral loan diversification reduces procyclicality in both bank types, particularly at moderate levels. Third, these effects are sensitive to bank ownership. Private banks exhibit clear cyclical sensitivity, while state-owned banks show no significant credit response to macroeconomic fluctuations. These results are robust to alternative credit measures and highlight the conditions under which internal bank strategies can complement external regulatory tools in stabilizing credit over the cycle.

The remainder of the paper is structured as follows. Section 2 reviews the relevant literature on lending procyclicality and diversification and develops the conceptual foundation of the study. Section 3 describes the data, variables and empirical methodology. Section 4 presents the empirical findings, including the main results and robustness checks. Section 5 discusses the implications of the findings and concludes.

The cyclicality of bank lending has long been a subject of scrutiny in banking and macroeconomics due to its destabilizing effects on financial systems. A defining feature of modern banking systems is the procyclical nature of credit expanding during economic upswings and contracting during downturns (Aiyar et al., 2014; Jiménez et al., 2017). This behavior is fueled by both supply- and demand-side dynamics. On the supply side, banks tend to ease lending standards in boom periods due to rising asset values and improved borrower performance, thus encouraging greater risk-taking (Bernanke and Gertler, 1995; Dell’Ariccia and Marquez, 2006). Conversely, downturns trigger deleveraging, tighter credit standards and greater caution, amplifying macroeconomic contractions (Gambacorta and Shin, 2018). This cyclical lending behavior creates an intermediation paradox: credit is most abundant when least needed and most constrained when demand is highest.

The economic consequences of such procyclicality are particularly acute for small- and medium-sized enterprises (SMEs), which often lack access to alternative sources of financing. During downturns, elevated credit risk and rising nonperforming loans push banks to retrench, leaving vulnerable firms credit-constrained when they need liquidity to survive (Kashyap and Stein, 2004). The 2008 Global Financial Crisis exemplified this dynamic, as excessive precrisis credit fueled asset price bubbles, while postcrisis credit withdrawal intensified the downturn (Brunnermeier, 2009). Similar patterns were observed during the 1997 Asian Financial Crisis, where aggressive precrisis lending was followed by sharp contractions, exacerbating capital outflows and financial distress (Kaminsky and Reinhart, 1999).

While most studies examine conventional banks, Islamic banks, operating under distinct contractual and ethical frameworks, offer an important comparative lens on credit cyclicality. Rooted in Shariah principles, these institutions rely on contracts that tie financing to real-sector performance and discourage speculative lending (Beck et al., 2013; Hasan and Dridi, 2011; Seho et al., 2016; Šeho et al., 2020). In theory, this alignment with real economic activity should reduce procyclicality. However, empirical evidence presents a more complex picture. Some studies report countercyclical or less procyclical behavior in Islamic banks, particularly during downturns (Ibrahim, 2016; Saadaoui and Hamza, 2020; Šeho et al., 2024a; Zulkhibri and Sakti, 2018). Others find Islamic banks as procyclical, or even more so, than their conventional counterparts (Albaity et al., 2022; Ascarya et al., 2016; Aysan and Ozturk, 2018).

These mixed findings may reflect differences in institutional context, regulatory environments and contract structures. For instance, the predominance of Murabaha, a debt-like contract, in many Islamic banks may weaken the theoretical promise of risk-sharing (Šeho et al., 2020). Bank size also plays a role: smaller Islamic banks may emulate conventional peers due to competitive pressures or governance limitations (Soedarmono and Yusgiantoro, 2023). Recent work by Šeho et al. (2024b) suggests that Islamic banks in Malaysia exhibit nonlinear credit behavior – procyclical during expansions but countercyclical during contractions – highlighting the importance of the business cycle phase.

Beyond the behavior of Islamic banks, procyclicality has broader implications for financial stability. Lending booms inflate asset prices and fuel leverage, increasing systemic fragility (Adrian and Shin, 2010). Credit retrenchment during recessions deepens macroeconomic distress, prompting spillover effects across the real economy. Recognizing these risks, Basel III introduced the countercyclical capital buffer (CCyB) to force banks to accumulate capital during expansions and draw it down during contractions (Supervision, 2011). While conceptually sound, the effectiveness of CCyBs remains debated. Some studies show they help smooth credit cycles (Drehmann et al., 2018), while others argue that banks remain risk-averse even with capital buffers in place, particularly when uncertainty is high (Jiménez et al., 2017).

A bank’s internal characteristics further influence lending cyclicality. Well-capitalized banks tend to be more resilient and maintain credit lines during stress (Berger and Bouwman, 2013), while poorly capitalized banks cut lending more sharply. Loan quality also matters: banks with weaker portfolios face greater contraction pressures during downturns (Foos et al., 2010). Ownership plays a stabilizing role; state-owned banks, less driven by profit maximization, are often deployed as countercyclical tools (Micco and Panizza, 2006; Šeho et al., 2024a). Bank size, market share and risk governance structures likewise shape responsiveness to macro shocks (Gambacorta and Shin, 2018).

Finally, recent literature has challenged the assumption that procyclicality is linear. Instead, banks may respond asymmetrically across business cycle phases (Bouvatier et al., 2014; Šeho et al., 2024b). Lending increases tend to be smooth during expansions, but retrenchments can be sudden and steep during contractions. For Islamic banks, nonlinear dynamics may be exacerbated by Shariah-compliant constraints that restrict restructuring options during distress.

Despite growing regulatory attention, credit procyclicality remains a persistent feature of banking systems, with mixed evidence on the effectiveness of institutional countermeasures. Moreover, much of the existing literature models lending cyclicality as linear, potentially overlooking nonlinear dynamics and asymmetric responses across different phases of the business cycle. These limitations raise important questions about whether credit sensitivity to economic activity follows a nonlinear pattern and whether such dynamics differ across banking models operating within dual systems.

Among bank-level strategies for managing risk and ensuring credit continuity, few have received as much sustained attention as diversification. At its core, diversification is seen as a means of reducing dependence on specific sectors or revenue streams, thereby improving resilience to shocks. Yet, the extent to which diversification enables banks to lend more, particularly during periods of macroeconomic stress, remains an open empirical question. This issue lies at the intersection of risk management and macro-financial stability, and its resolution holds substantial implications for both scholars and policymakers. While practitioners and regulators have long viewed diversification as a stabilizing force, academic findings have presented a more nuanced and often contradictory picture.

The theoretical literature is divided between two main schools of thought. The first draws from Modern Portfolio Theory (Markowitz, 1952), arguing that diversification enhances the risk-return tradeoff by reducing exposure to idiosyncratic shocks, thereby expanding the bank’s capacity to lend consistently. In contrast, the “focus hypothesis” contends that specialization allows banks to develop deeper expertise, enhance monitoring and reduce agency costs, thus improving lending quality (Acharya et al., 2006; Šeho et al., 2021; Tabak et al., 2011). Under this view, diversification can dilute institutional knowledge, reduce screening efficiency and increase organizational complexity, discouraging credit provision, especially under stress.

Empirical studies tend to support both views, depending on the context. For example, Šeho et al. (2024a) report that sectoral loan diversification is positively associated with credit growth in both Islamic and conventional banks in Malaysia. Likewise, Gelman et al. (2022) find that banks with higher geographic diversification sustained significantly more small business lending during the 2008 financial crisis. These banks proactively maintained credit exposure even when market conditions deteriorated, suggesting supply-side resilience. These findings suggest that diversification may enhance banks’ ability to smooth earnings, reduce capital volatility and maintain regulatory capital ratios, thereby preserving credit capacity even when traditional income sources decline. In regions with a strong presence of such banks, small business lending was 4.2% higher, with measurable local spillover effects (Gelman et al., 2022).

Diversification into nonlending activities can also enhance lending resilience. Gelman et al. (2022) show that banks with insurance subsidiaries extended approximately 3% more loans prior to the crisis and maintained 38% higher levels of small business lending during the downturn. These results imply that stable, alternative income sources, such as fees, insurance premiums or trading revenue, can ease the pressure on loan-dependent earnings and help sustain credit in adverse conditions.

However, the benefits of diversification are not universal. Several studies emphasize its context-dependent nature. For example, Acharya et al. (2006) find that loan diversification in Italy does not improve performance, while Berger et al. (2010) report increased risk and lower profitability among diversified Russian banks. Similarly, Chen et al. (2014) document profit reductions among Chinese banks pursuing diversification, and Tabak et al. (2011) associate diversification with elevated default risk in Brazil. Although these studies focus on risk and return rather than credit growth per se, they suggest that diversification can be destabilizing when it compromises monitoring quality or organizational coherence.

More recent research has introduced additional nuance, particularly in dual-banking systems. Al-kayed and Aliani (2020) show that loan diversification has adverse effects on profitability and risk for Islamic banks in the GCC. Šeho et al. (2021) find that diversification worsens both risk and returns in dual-banking environments. Moreover, Šeho et al. (2023) demonstrate that the relationship between diversification and stability is nonlinear and differs systematically between Islamic and conventional banks. While conventional banks appear to benefit from greater diversification, Islamic banks show heightened sensitivity to moderate levels, likely due to differences in contract structures, governance mechanisms and regulatory frameworks.

These findings imply that the impact of diversification on credit behavior cannot be generalized across institutions. Bank size, regulatory oversight and operating model shape both the feasibility and effectiveness of diversification strategies. Large banks, for example, may benefit from economies of scale and superior risk management infrastructure, while smaller institutions may struggle with the complexity and coordination costs introduced by diversification. For Islamic banks, diversification is further constrained by Shariah principles, limiting participation in certain sectors and instruments. Nonetheless, they can pursue sectoral or loan-side diversification within Shariah-compliant boundaries.

It is important to clarify that the diversification measures employed in this study refer to sectoral allocation of financing and the composition of income streams within permissible balance-sheet activities. While Islamic banks are restricted from financing certain non-Shariah-compliant sectors and from engaging in interest-based or derivative-intensive activities, lending in the GCC is predominantly concentrated in real-economy sectors such as construction, trade, manufacturing and energy, in which both banking models actively participate. Diversification in this context, therefore, reflects relative allocation choices across available sectors and income sources rather than access to capital market securitization or complex financial engineering. The analysis does not assume identical opportunity sets but evaluates how diversification within institutional boundaries influences credit behavior across banking models.

The nature of diversification also matters. Much of the literature has focused on income diversification, particularly expansion into noninterest income activities. Moderate engagement in these areas has sometimes improved financial performance (Sanya and Wolfe, 2011), but excessive reliance on noninterest income is often linked to heightened earnings volatility and reduced credit stability (Nguyen et al., 2023). In contrast, asset-side or loan diversification appears to have a more direct relationship with credit continuity. For instance, AlKhouri (2019) and Sahul Hamid and Ibrahim (2021) show that sectoral loan diversification improves both stability and lending outcomes in dual-banking systems.

Geographic diversification, especially through cross-border operations, is also associated with a more stable credit supply. Doerr and Schaz (2021) report that banks involved in international syndicated lending continued extending credit to foreign borrowers even during domestic crises. This suggests that geographic diversification can act as a shock absorber, mitigating the impact of local financial disruptions on overall lending capacity.

In sum, while diversification does not uniformly enhance credit growth, it can strengthen a bank’s lending capacity when implemented strategically. Its effects depend critically on execution, institutional context and the type of diversification pursued. Loan portfolio diversification and income diversification may operate through different channels and, given structural and contractual differences, may influence Islamic and conventional banks in distinct ways. These considerations suggest that diversification is unlikely to have a uniform effect on credit dynamics across banking models.

Beyond their direct impact on credit growth, diversification strategies may also shape how banks respond to macroeconomic fluctuations. We turn to this stabilizing role in the next section.

While diversification has been widely examined for its impact on bank risk, return and credit growth, its role as a stabilizing force over the business cycle remains relatively underexplored. A critical question for both researchers and policymakers is whether diversification moderates the sensitivity of bank credit supply to macroeconomic conditions, attenuating procyclicality or supporting countercyclical behavior. This question is especially pertinent in dual-banking systems, where banks operate under heterogeneous business models and constraints.

Emerging evidence suggests that diversification may enhance credit continuity during adverse conditions. For instance, Gelman et al. (2022) show that diversified banks, especially those with insurance subsidiaries, sustained more lending during the 2008 crisis. Doerr and Schaz (2021) similarly find that geographic diversification helped banks maintain cross-border lending amid domestic turmoil. While informative, these studies focus on crisis-specific dynamics and do not address whether diversification systematically moderates credit behavior across the full business cycle.

One of the few studies that directly engages this question is Šeho et al. (2024a), who explore the role of loan diversification in moderating lending cyclicality within Malaysia’s dual-banking system. They find that both Islamic and conventional banks exhibit nonlinear credit responses, procyclical during expansions and, for Islamic banks, countercyclical during downturns. Strikingly, their results suggest that diversification intensifies procyclicality in Islamic banks, while loan concentration appears to dampen it. This effect persists across ownership structures: diversification is associated with procyclical behavior in public and local banks and countercyclical in foreign banks. These findings challenge the notion that diversification is inherently stabilizing and underscore its context-dependent nature.

The implications are particularly significant for Islamic banks, where diversification effects may be shaped by the use of risk-sharing contracts, Shariah governance and unique balance sheet structures. These features can influence both the capacity and the incentive of Islamic banks to maintain lending during different phases of the cycle. What emerges is a more complex and conditional view of diversification, one that depends not only on bank type but also on the prevailing macroeconomic environment.

As discussed earlier, diversification outcomes depend heavily on institutional characteristics. This research extends that insight to examine whether such contextual factors also influence the stabilizing potential of diversification. In particular, it is important to recognize that Malaysian Islamic banking differs substantially from that of the GCC. Malaysia’s model is often characterized by a centralized regulatory structure and a pragmatic, sometimes criticized, reliance on debt-like instruments such as Tawarruq (El-Gamal, 2006; Šeho et al., 2024b). In contrast, GCC Islamic banks tend to adopt more classical contract forms, such as Mudarabah and Musharakah, and operate within more diverse jurisprudential interpretations. These doctrinal and structural differences affect how banks approach risk, governance and portfolio construction. These differences, together with the unique institutional arrangements found in the GCC, including more concentrated markets, varying degrees of state involvement and diverse Shariah compliance practices, make it essential to test whether earlier findings from Malaysia hold in this distinct setting.

This study aims to fill that gap by investigating whether diversification moderates the responsiveness of credit to business cycle fluctuations among Islamic and conventional banks in the GCC. It also examines how this moderating effect varies across bank ownership types. In doing so, the paper contributes to broader debates on financial stability and macroprudential policy by asking whether internal strategies, such as loan and income diversification, can complement regulatory tools in smoothing credit cycles.

Although limited in scope, existing evidence suggests that diversification may act as a financial buffer by spreading exposures and stabilizing income streams. However, this perspective has not been systematically integrated into empirical models of credit cyclicality, particularly in dual-banking environments.

Ultimately, the relevant question is not whether diversification is beneficial in the abstract, but under what conditions and for which types of banks it moderates cyclical lending dynamics. Given differences in balance sheet composition, risk-sharing mechanisms and regulatory constraints, the stabilizing role of diversification may vary between Islamic and conventional banks.

Our sample is an unbalanced panel of 44 conventional and 21 Islamic banks from the GCC countries covering 2008–2021. We include only banks that have at least three consecutive years of observations. ^[2] We winsorize all continuous bank-specific variables at the 1st and 99th percentiles to eliminate the potential extreme influence of outliers on the regression results.

Since most economic behaviors are dynamic (Nerlove, 2000), we specify a dynamic model that includes a lagged dependent variable. The dynamic specification is consistent with prior studies of bank lending cyclicality that employ system generalized method of moments (GMM) to address persistence in credit growth and potential endogeneity [e.g. Ibrahim (2016), Bouvatier et al. (2014), Šeho et al. (2024a)]. The inclusion of a lagged dependent variable renders fixed-effects estimation biased and inconsistent (Nickell, 1981), motivating the use of the system GMM estimator. Given the panel structure, which features a relatively large cross-section and limited time dimension, system GMM is well-suited to this empirical setting. Taking lead from these studies, we adopt the following baseline model:

where Δ is the first difference operator; Credit_bt is the natural logarithm of gross loans of bank b at time t; GDP_t is the CPI-adjusted gross domestic product, where 2010 is used as the base year, in time t; Loan Diversification_bt-1 is calculated using the inverse Herfindahl-Hirschman Index (HHI) based on sectoral loan exposures across 13 economic sectors of bank b at time t − 1; Income Diversification_bt-1 is income-based diversification measure of bank b at time t − 1, calculated using the inverse HHI of interest and noninterest income shares; B_bt-1 is a vector of bank-specific variables at time t – 1; INF_t is the annual inflation rate at time t; τ_t is a time-specific effect; ν_b is a bank-specific time-invariant effect and ε_bt is the common error term. We include the lagged dependent variable to allow for potential dynamic adjustments. Definitions and sources of the variables used in our analysis are provided in Table 1.

To examine the potential nonlinear relationship between economic shocks and bank lending and if the relationships are different for Islamic and conventional banks, we modify Eq.(1) by introducing the square term of ΔGDP_t, that is, ΔGDP_t², and interact it with the Islamic bank dummy (IB).

Knowing that diversification may have a nonlinear relationship with certain bank-specific performance measure variables, such as risk and returns (Šeho et al., 2021) and z-score (Šeho et al., 2024a; Šeho et al., 2023), we modify equation (1) by introducing the square term of Loan Diversification_bt-1 and Income Diversification_bt-1, that is, Loan Diversification²_bt-1 and Income Diversification²_bt-1, respectively, and interact it with the Islamic bank dummy (IB), to examine the potential nonlinear relationship between our two diversification measures and bank credit, and if the relationships are different for Islamic and conventional banks:

Finally, to investigate the potential credit stabilizing role of diversification and whether it works differently for conventional and Islamic banks, as found in the case of Malaysia (Šeho et al., 2024b), we extend equation (1) by introducing an interaction term of our two diversification measures, that is, Loan Diversification_bt-1 with ΔGDP_t and IB:

and Income Diversification_bt-1 with ΔGDP_t and IB:

The empirical specifications above are designed to test the hypotheses developed in Section 2 through the model’s interaction structure. H1 is evaluated through the coefficients on GDP growth and its squared term in equation (2), which capture nonlinear lending dynamics, as well as their interaction with the Islamic bank dummy to assess differences in cyclical sensitivity across the two banking models. H2 is examined through the coefficients on loan and income diversification and their squared terms in equation (3), which capture potential nonlinear effects on credit growth. Interactions between diversification and the Islamic bank dummy allow these effects to differ across banking models. H3 is tested through interactions between GDP growth and the diversification measures, including triple interactions with the Islamic bank dummy in equation (4) and equation (5), to determine whether diversification moderates credit cyclicality and whether this moderating effect varies across bank types.

Because several interaction terms involve continuous variables, the marginal effects of GDP growth and diversification cannot be inferred solely from individual coefficients. Following Brambor et al. (2006), we compute and graphically present marginal effects to properly evaluate the conditional relationships implied by the triple interaction terms. This approach ensures an accurate interpretation of how diversification modifies the sensitivity of credit growth to economic activity across the two banking models.

Given the dynamic specification of the model, which includes a lagged dependent variable, traditional fixed-effects estimators would yield biased and inconsistent estimates due to the correlation between individual effects and the lagged dependent variable (Nickell, 1981). We choose the two-step system GMM proposed by (Blundell and Bond, 1998) and (Arellano and Bover, 1995) over the difference GMM due to its superiority in handling the potential precision and bias issues associated with the difference GMM (Blundell and Bond, 1998). To make the two-step standard errors more robust, we apply the (Windmeijer, 2005) correction. We use the Hansen test of over-identification and the Arellano-Bond tests of first- and second-order autocorrelation in the differenced residuals to assess the validity of our estimations.

We begin by presenting descriptive statistics in Table 2 and plotting annual averages of key variables in Figure 1. Islamic banks, on average, appear less diversified across both loan and income streams than their conventional counterparts, a pattern consistent with Shariah-based prohibitions on sectors such as gambling, alcohol and defense. This pattern is consistent with evidence that Islamic banks face institutional and contractual constraints that shape balance sheet composition and sectoral participation (Beck et al., 2013; Hasan and Dridi, 2011; Mirakhor, 2018; Šeho et al., 2021). At the same time, higher average credit growth among Islamic banks is not inconsistent with prior work showing that lending dynamics can differ by banking model and by phase of the cycle in dual systems (Ibrahim, 2016; Saadaoui and Hamza, 2020; Šeho et al., 2024a). Interestingly, Islamic banks registered higher average credit growth over the sample period, with the exception of 2008. Despite these structural differences, Figure 1 shows that credit growth across both bank types broadly tracks macroeconomic fluctuations in the same direction.

Table 3 reports sectoral correlations of loan exposures. The results show that sectoral returns are not perfectly correlated and, in some cases, are negatively correlated, suggesting that diversification across sectors could generate stabilizing effects through imperfect co-movement. This underscores the potential value of portfolio diversification in reducing risk exposure and maintaining lending capacity.

We next turn to the regression results. Table 4 presents estimates for equations (1) and (2), while Table 5 reports estimates for equations (3), (4) and (5). All explanatory variables are treated as weakly exogenous. Given that bank-specific variables are included in the model with lags, it is unlikely that shocks to individual bank credit contemporaneously influence aggregate GDP growth, in line with the assumptions of Ibrahim (2016) and Šeho et al. (2024b). Accordingly, only the lagged dependent variable is treated as endogenous and instrumented using the GMM-style option, while all other regressors enter under IV-style, as recommended by Roodman (2009). Following Roodman (2009)’s rule of thumb, lagged instruments are used starting with the second lag to address potential endogeneity while avoiding overfitting. To preserve the reliability of the Hansen test and minimize the risk of instrument proliferation, we ensure that the number of instruments does not exceed the number of cross-sectional groups. Postestimation diagnostics support the validity of the specification. The Hansen test of overidentifying restrictions fails to reject the null hypothesis of instrument validity, and the Arellano–Bond tests indicate the presence of first-order but not second-order serial correlation, as expected in a correctly specified dynamic panel model. These results confirm the consistency of the system GMM estimator in our setting.

Next, we turn to the regression results in Table 4. Regression (1), estimated on the full sample using a linear model, finds no evidence of procyclical credit behavior. The absence of an average linear effect in the pooled specification is consistent with the view that cyclicality can be heterogeneous across bank types and asymmetric across cycle phases, which linear models may mask (Bouvatier et al., 2014; Ibrahim, 2016; Šeho et al., 2024a). However, once we distinguish between Islamic and conventional banks in regression (2), we find that only conventional banks exhibit statistically significant procyclicality. In regressions (3) and (4), we incorporate a quadratic term for ΔGDP to capture potential nonlinearities. The squared term is negative and significant, indicating an inverted U-shaped relationship between GDP growth and credit growth. This nonlinear pattern aligns with the growing evidence that bank credit responds asymmetrically across the cycle, rather than proportionally at all GDP growth rates (Bouvatier et al., 2014; Šeho et al., 2024b). Our marginal effects analysis below extends this approach by identifying where along the curve the response is statistically meaningful. The estimated turning point occurs at a GDP growth rate of 0.379 for conventional banks and 0.137 for Islamic banks, suggesting that Islamic banks reach their peak lending responsiveness earlier in the cycle.

While this parametric test indicates nonlinearities, it does not fully capture how lending behavior evolves across the distribution of GDP growth rates. To address this, we calculate average marginal effects based on regression (4) and plot them in Figure 2. These estimates confirm a nonlinear pattern and reveal important differences between bank types.

For conventional banks, Figure 2 shows strong procyclicality during economic downturns and somewhat milder procyclicality in early expansions, with effects tapering off at higher levels of GDP growth. For instance, at a ΔGDP of −0.29, a one-percentage-point decrease in GDP growth reduces credit growth by 0.703 percentage points. At ΔGDP = 0.10 ^[3], the marginal effect is 0.293, indicating weaker responsiveness in upturns.

Islamic banks exhibit a markedly different pattern. As shown in Figure 2, they are significantly procyclical only during downturns. For example, at ΔGDP = −0.29, the marginal effect is 0.668, similar in magnitude to that of conventional banks. However, no statistically significant procyclicality is observed during economic expansions, and the effects become insignificant when ΔGDP exceeds −0.07. This asymmetric response suggests that Islamic banks contract their financing less during adverse conditions, while exhibiting no statistically distinct behavior during upturns. These results align with Saadaoui and Hamza (2020) and extend previous findings by uncovering a threshold effect in the procyclicality of Islamic bank credit. This phase-dependent finding also helps reconcile mixed evidence in the Islamic bank cyclicality literature by showing that differences may arise from where observations fall on the business-cycle distribution rather than from the banking model alone (Šeho et al., 2024a).

We now turn to the role of diversification. We distinguish between the direct association of diversification with credit growth, which has been widely studied in the diversification-performance literature, and the moderating role of diversification over the business cycle, which remains comparatively underexplored. In the linear specifications reported in Table 4, income diversification is positively and significantly associated with credit growth, whereas loan diversification is not statistically significant. The positive association between income diversification and credit growth is consistent with the argument that non-lending income can relax capacity constraints and support lending (Sanya and Wolfe, 2011), although the literature also cautions that extensive reliance on such income may increase volatility (Nguyen et al., 2023). However, the nonlinear specifications presented in Table 5, particularly regressions (1) and (2), reveal more complex dynamics. These regressions incorporate interaction terms that allow the effects of diversification to vary across different levels of sectoral exposure or income. Figures 3 and 4 illustrate the marginal effects of loan and income diversification, respectively, highlighting their conditional influence on credit growth.

For conventional banks, loan diversification does not significantly affect credit growth at most levels of diversification. Income diversification, however, has a positive effect, especially at moderate to high levels of diversification. Its effect diminishes as income diversification approaches the theoretical upper bound (0.5), suggesting decreasing marginal effects of income diversification on credit growth. For Islamic banks, the results differ. Loan diversification is positively associated with credit growth at moderate levels, reflecting the benefits of a broader distribution of financing across sectors. Similarly, income diversification enhances credit growth at moderate to high levels. The nonlinear and bank-type-specific patterns are consistent with evidence that diversification effects depend on institutional setting and business model, and can be nonlinear in dual systems (Šeho et al., 2021; Šeho et al., 2023).

To assess whether diversification moderates credit cyclicality, we interact ΔGDP with loan and income diversification, as specified in regressions (3) and (4) of Table 5. Since both interaction variables are continuous, we compute marginal effects, as suggested by Brambor et al. (2006), and present them in Figures 5 and 6. Figure 5 shows that loan diversification attenuates procyclicality: banks with more diversified loan portfolios exhibit a weaker response to GDP fluctuations. While diversification does not render lending behavior countercyclical, it clearly dampens the procyclicality of credit growth across both Islamic and conventional banks. These findings are in line with the broader notion that spreading exposures can support credit continuity when shocks are sector-specific (Gelman et al., 2022) and that diversification can act as a shock absorber in stress episodes (Doerr and Schaz, 2021). At the same time, evidence directly linking diversification to credit cyclicality over the full business cycle remains limited, particularly in cross-country dual-banking settings. Our results extend the small set of studies that examine diversification as a moderator of cyclicality by jointly considering loan and income diversification within a unified dynamic framework.

Figure 6 examines the moderating effect of income diversification on bank credit cyclicality. For conventional banks, high levels of income diversification appear to amplify procyclicality, especially at the upper end of the diversification range. This suggests that reliance on diverse income sources may make banks more sensitive to macroeconomic conditions when those income streams are cyclical. For Islamic banks, income diversification does not significantly affect the cyclical pattern of credit growth, indicating that its stabilizing potential may be limited under Shariah-compliant income structures. The amplification effect for highly diversified conventional banks is consistent with findings that increased reliance on noninterest income can raise earnings volatility and weaken stability, especially under heightened uncertainty (Nguyen et al., 2023).

Finally, we examine the role of bank-level controls. Across specifications, credit growth is negatively associated with noninterest income, impaired credit quality and inflation. Liquidity is positively related to lending, while bank size, capital, profitability and deposit levels are generally insignificant.

To assess the robustness of our main findings, we conduct two sets of supplementary analyses. First, we test the sensitivity of our results to an alternative measure of bank credit. Following Ibrahim (2016) and Šeho et al. (2024b), we replace gross loans with net loans to capture a more conservative measure of credit exposure. The regression results using net loans, presented alongside those for gross loans in Tables 4 and 5 and Figures 2 through 6, yield qualitatively similar conclusions. While the magnitudes of the coefficients vary slightly, the direction and statistical significance of the main effects remain largely intact.

Second, given the well-documented influence of bank ownership on credit behavior over the business cycle (Bertay et al., 2015; Micco and Panizza, 2006; Panizza, 2023), we examine whether the procyclicality and moderating role of diversification differ across ownership types. In particular, we distinguish between state-owned (public) and non-state-owned (private) banks. As reported in Table 6 and illustrated in Figure 7, private banks exhibit procyclical lending behavior during economic downturns, whereas state-owned banks do not show any statistically significant cyclical pattern. Moreover, loan diversification plays a stabilizing role for private banks but is ineffective for public banks (Figure 8). Income diversification, by contrast, amplifies procyclicality only among highly diversified private banks and has no measurable effect on credit cyclicality among state-owned institutions (Figure 9). This pattern is consistent with the view that state-owned banks can dampen credit cyclicality due to policy objectives and softer profit constraints, while private banks transmit macro shocks more strongly (Bertay et al., 2015; Micco and Panizza, 2006; Panizza, 2023).

While a growing body of literature documents the procyclical behavior of conventional banks, the evidence for Islamic banks remains mixed. Moreover, the role of diversification – across income sources or sectoral loan exposures – in moderating credit cyclicality is still not well understood, particularly in dual-banking environments. This study addresses these gaps by examining three core questions in the GCC context: (i) do Islamic and conventional banks exhibit procyclical credit behavior, and is this relationship nonlinear? (ii) does diversification influence credit growth? and (iii) does diversification moderate credit sensitivity to GDP fluctuations?

While a growing body of literature documents the procyclical behavior of conventional banks (Bertay et al., 2015; Bouvatier et al., 2014), the evidence for Islamic banks remains mixed (Aysan and Ozturk, 2018; Ibrahim, 2016; Saadaoui and Hamza, 2020). Moreover, although diversification has been extensively studied in relation to bank risk and performance (Acharya et al., 2006; Sanya and Wolfe, 2011; Šeho et al., 2021; Šeho et al., 2023; Tabak et al., 2011), its role in moderating credit cyclicality remains comparatively underexplored, particularly in dual-banking environments. This study addresses these gaps by examining three core questions in the GCC context: (i) do Islamic and conventional banks exhibit procyclical credit behavior, and is this relationship nonlinear? (ii) does diversification influence credit growth? and (iii) does diversification moderate credit sensitivity to GDP fluctuations?

Using an unbalanced panel of 44 conventional and 21 Islamic banks from the GCC over the period 2008–2021, we estimate dynamic panel models using the system GMM estimator. We find that both conventional and Islamic banks exhibit nonlinear procyclicality, with an inverted U-shaped relationship between GDP growth and credit. However, Islamic banks reach their turning point at a significantly lower level of GDP growth and do not respond significantly during economic expansions. This suggests that Islamic banks are procyclical only during downturns, while conventional banks respond more strongly across the entire cycle. By explicitly evaluating marginal effects across different levels of GDP growth, our analysis shows that cyclical sensitivity is phase-dependent rather than uniform across the nonlinear curve. This helps reconcile prior mixed findings in the Islamic banking literature, which may have reflected differences in the distribution of observations across business-cycle phases.

Diversification plays a nuanced role. Income diversification contributes positively to credit growth across both bank types, consistent with the argument that alternative income streams can relax lending constraints. However, the marginal gains diminish at higher levels in conventional banks, echoing concerns that excessive reliance on noninterest income may increase volatility. Sectoral loan diversification appears particularly important for Islamic banks, where moderate levels are associated with increased credit expansion. More critically, loan diversification reduces procyclicality in both conventional and Islamic banks, while income diversification amplifies it, but only in highly diversified conventional banks. These findings reveal that diversification is not unconditionally stabilizing; its effects depend on bank type, the form of diversification and the business cycle phase.

Importantly, this study extends the literature by jointly modeling nonlinear credit cyclicality and the moderating role of both loan and income diversification within a unified dynamic framework. While prior work has examined diversification and performance, and a small number of studies have considered nonlinear effects in dual-banking systems, cross-country evidence on diversification as a moderator of lending cyclicality remains limited. By distinguishing between direct diversification effects and its interaction with macroeconomic fluctuations, and by allowing these relationships to vary across banking models, this study provides a structural perspective on how portfolio composition shapes macro-financial transmission in dual-banking systems.

While we cannot fully isolate the underlying mechanisms, our results are consistent with theoretical expectations. In conventional banks, income diversification may increase exposure to cyclical fee-based or trading activities, thereby amplifying sensitivity to macro conditions. In Islamic banks, Shariah-compliant constraints limit such exposures, potentially muting the channel through which income diversification could affect credit cyclicality. This institutional distinction helps explain why similar diversification strategies produce heterogeneous cyclical outcomes across banking models.

These findings remain robust when credit is measured using net loans and when bank ownership is considered. Private banks exhibit procyclical behavior during downturns, while state-owned banks show no significant cyclical patterns, consistent with the view that public ownership may dampen credit cyclicality (Bertay et al., 2015; Micco and Panizza, 2006; Panizza, 2023). Loan diversification stabilizes credit growth only in private banks, and income diversification amplifies procyclicality only among highly diversified private institutions, reinforcing the importance of institutional context in shaping macro-financial transmission.

These results offer at least two policy implications. First, the different responses across bank types and diversification channels suggest that macroprudential regulation should not adopt a one-size-fits-all approach. Specifically, sectoral exposure limits and countercyclical capital buffers could be calibrated differently for Islamic and conventional banks, reflecting their asymmetric credit behavior across the cycle. Second, regulatory frameworks should recognize that income diversification, particularly when it relies on market-based or cyclical revenue streams, may elevate risk in periods of economic expansion, particularly for conventional banks. Promoting prudent sectoral loan diversification, especially among Islamic institutions, may help enhance financial resilience without triggering excessive credit growth.

Beyond regulatory design, these findings also carry broader societal implications. Credit supply stability is closely linked to investment continuity, small business survival and employment dynamics, particularly in bank-based financial systems such as those in the GCC. If diversification strategies can dampen excessive credit contraction during downturns, they may indirectly support household income stability and business resilience. In dual-banking environments, understanding the asymmetric responses of Islamic and conventional banks can also help inform public discourse on banking resilience and consumer protection. Rather than viewing Islamic and conventional models as inherently more or less stable, our findings suggest that stability depends on portfolio composition and diversification strategy. This nuanced understanding may contribute to more balanced public perceptions of financial system resilience.

Our analysis is subject to several limitations. The results are drawn from GCC countries and may not generalize to jurisdictions with different legal, regulatory or economic structures. Moreover, we do not directly observe the underlying composition of noninterest income or the contractual characteristics of sectoral exposures. Future research could explore the interaction between contractual diversification among Islamic banks and other stabilizing tools, such as countercyclical capital buffers or stress testing frameworks. Another promising direction is to examine how fintech-driven changes to bank income structures, particularly in dual-banking systems, may alter the relationship between diversification and credit cyclicality.

The authors gratefully acknowledge the valuable feedback received from participants at the 26th Malaysian Finance Association International Conference 2024, the 14th Financial Markets and Corporate Governance Conference 2024, and the Values for Impact Conference 2025.

The authors acknowledge that they have used ChatGPT to improve the structure and clarity of the language and grammar of their original existing work.