Artificial Intelligence as a Catalyst for Banker Knowledge Development: Econometric Evidence From South Asian Financial Markets And Economic Growth Dynamics

Artificial Intelligence as a Catalyst for Banker Knowledge Development: Econometric Evidence From South Asian Financial Markets And Economic Growth Dynamics | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial Intelligence as a Catalyst for Banker Knowledge Development: Econometric Evidence From South Asian Financial Markets And Economic Growth Dynamics Mohammad Abdullah-Al-Kafe This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9127571/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study constitutes the first systematic econometric investigation into whether the deployment of artificial intelligence for banker knowledge enhancement exerts a statistically significant, causal influence on credit intermediation efficiency, financial inclusion depth, and aggregate economic growth across the eight sovereign economies of South Asia. Drawing exclusively on publicly disclosed macroeconomic, financial sector, and technology adoption data spanning the period 2000 to 2023, a novel composite index of AI-augmented banking knowledge intensity (AABKI) is constructed and embedded within a panel data framework comprising fixed-effects, random-effects, and dynamic Generalised Method of Moments (GMM) estimators. Three foundational hypotheses are advanced and empirically evaluated: first, that AI-mediated knowledge diffusion within banking institutions constitutes an autonomous, hitherto unrecognised driver of credit-to-GDP ratio growth, independent of conventional financial development determinants; second, that the marginal effect of AI banking knowledge intensity on financial inclusion is heterogeneous across income strata and geographic regions within South Asia, thereby generating a novel distributional channel through which technology shapes economic convergence; and third, that the interaction between regulatory institutional quality and AI banking knowledge adoption generates nonlinear threshold effects on long-run economic growth that existing theoretical frameworks have not articulated. Empirical results confirm all three hypotheses at conventional significance levels. The AABKI index is positively and robustly associated with domestic credit depth (beta = 0.347, p < 0.01), financial inclusion breadth (beta = 0.289, p < 0.01), and real GDP per capita growth (beta = 0.214, p < 0.05), with threshold nonlinearities identified at a regulatory quality score of 0.62 on the World Governance Indicators scale. These findings carry profound implications for central bank policy, SAARC-level financial integration strategy, and the design of AI governance frameworks for emerging market banking systems. JEL Classification: G21, O33, O53, C23, G28, F36, O16 Artificial Intelligence and Machine Learning Banking Law Finance Other Economics Macroeconomics International Business Artificial intelligence banker knowledge financial inclusion panel data econometrics GMM estimation South Asia credit intermediation regulatory quality threshold effects SAARC 1. Introduction The architecture of economic development in South Asia has historically been shaped by the depth, reach, and efficiency of its banking systems. Across Bangladesh, India, Pakistan, Sri Lanka, Nepal, Bhutan, the Maldives, and Afghanistan, commercial banks remain the predominant intermediaries through which savings are mobilised, credit is allocated, and investment is facilitated. Yet the banking sectors of the region have for decades contended with persistent structural inefficiencies: entrenched information asymmetries between lenders and borrowers, high costs of credit assessment in geographically dispersed agricultural and informal economies, regulatory compliance burdens that consume disproportionate institutional resources, and a chronic deficit of expert analytical capacity at the level of the individual banker. The cumulative cost of these inefficiencies, measured in foregone credit allocation, elevated non-performing loan ratios, and suppressed financial inclusion, represents one of the most consequential yet underanalysed constraints on South Asian economic convergence. The emergence of artificial intelligence as a transformational technology in financial services opens a fundamentally new channel through which these structural constraints may be relaxed. Unlike prior waves of financial technology that automated transactional processing, the current generation of AI systems is distinguished by its capacity to augment, replicate, and disseminate sophisticated cognitive tasks: credit risk evaluation, regulatory interpretation, customer profiling, fraud detection, and macroeconomic scenario analysis. When such capabilities are embedded within banker training, decision support, and knowledge management systems, they do not merely improve operational efficiency; they alter the fundamental production function of banking knowledge itself. A banker equipped with AI-assisted knowledge tools is not the same productive agent as one operating without them. This distinction, obvious in principle, has been almost entirely absent from the empirical economic literature on South Asian financial development. This paper addresses that gap directly. The proposition advanced is that AI-mediated banking knowledge development represents a sui generis economic phenomenon: a technology-human knowledge interaction whose effects on credit intermediation, financial inclusion, and economic growth are measurable, substantial, and structurally distinct from those of conventional financial deepening or information technology adoption. No prior study has constructed a theoretically motivated, empirically operationalised measure of AI banking knowledge intensity at the country level for South Asia, nor has any prior study subjected such a construct to systematic panel econometric testing against growth and financial development outcomes across the full SAARC membership. The remainder of this paper is organised as follows. Section 2 reviews the relevant theoretical and empirical literature, identifying the precise gaps this study addresses. Section 3 develops the theoretical framework and formally states the three novel hypotheses. Section 4 describes data sources, variable construction, and the construction of the AABKI composite index. Section 5 presents the econometric methodology, encompassing pooled OLS, fixed-effects, random-effects, and system GMM estimations, along with panel cointegration and threshold regression analyses. Section 6 reports and interprets empirical results. Section 7 discusses policy implications. Section 8 concludes. 2. Literature Review 2.1 Financial Development and Economic Growth: The Established Paradigm The relationship between financial sector development and economic growth has been the subject of sustained theoretical and empirical inquiry since the foundational contributions of Goldsmith ( 1969 ), McKinnon ( 1973 ), and Shaw ( 1973 ). The central proposition that deeper, more efficient financial intermediation accelerates capital accumulation and total factor productivity growth has accumulated substantial econometric support. King and Levine ( 1993 ) demonstrated, using cross-country data, that financial depth, measured as the ratio of liquid liabilities to GDP, is a robust predictor of subsequent economic growth. Levine, Loayza, and Beck ( 2000 ) extended this finding using GMM estimation to address the endogeneity of financial development, confirming that the relationship is causal rather than spurious. Subsequent meta-analytic work by Valickova, Havranek, and Horvath ( 2015 ), synthesising 1,334 estimates from 67 studies, found a positive, statistically significant consensus effect of financial development on growth, though with considerable heterogeneity explained by methodological choices and country characteristics. Within the South Asian context, Ang ( 2008 ) found evidence of a finance-growth nexus in India using time-series cointegration methods, while Hassan, Sanchez, and Yu ( 2011 ) confirmed the relationship across a panel of developing economies including several South Asian nations. More recently, Bhattacharya and Sivasubramanian ( 2003 ) documented the role of financial sector reforms in accelerating Indian economic growth, and Raza, Shahzad, and Shahzad ( 2020 ) provided panel evidence that banking sector depth enhances growth across South Asian economies over the period 1980 to 2018. However, these studies conceptualise financial development as a structural macroeconomic attribute, leaving entirely unexamined the knowledge and capability characteristics of the human agents who operate within financial institutions. 2.2 Technology Adoption in Banking: An Emerging Literature A more recent literature has begun to examine the economic effects of technological innovation within banking. Framed initially around the adoption of automated teller machines, electronic payment systems, and core banking software, this literature has found broadly positive effects on banking efficiency and outreach. Beck, Senbet, and Simbanegavi ( 2015 ) documented that mobile banking adoption significantly expanded financial access in Sub-Saharan Africa, a finding partially applicable to South Asia given the rapid proliferation of mobile money platforms in Bangladesh and the transformative impact of the Unified Payments Interface in India. The economics of AI adoption in financial services remains at an earlier stage of empirical development, owing partly to data constraints and partly to the recency of meaningful deployment at scale. Fuster et al. ( 2022 ) examined AI-based mortgage lending in the United States and found that machine learning algorithms reduced processing time by 20 percent and expanded credit access to previously underserved demographics, though concerns about algorithmic bias were simultaneously documented. Jagtiani and Lemieux ( 2019 ) studied FinTech credit providers and found that alternative data utilisation, itself enabled by AI analytics, improved credit access for borrowers without conventional credit histories. Berg et al. ( 2020 ) demonstrated that digital footprint data processed by AI substantially predicted creditworthiness in a German FinTech context. For South Asia, the empirical literature on AI in banking remains sparse and largely descriptive. Studies by the International Finance Corporation ( 2019 ) and the Asian Development Bank ( 2021 ) documented the potential of AI for credit scoring in Indian and Bangladeshi microfinance, but neither offered econometric estimation of macroeconomic effects. The World Bank Global Findex Database, while invaluable for financial inclusion measurement, does not systematically capture the AI knowledge dimension of banking institutions. This study therefore occupies novel empirical territory. 2.3 Human Capital in Banking and Institutional Knowledge A third strand of literature relevant to this study concerns the role of human capital and institutional knowledge quality in banking performance. Barth, Caprio, and Levine ( 2004 ) established that banking supervisory capacity, itself a form of institutionalised knowledge, significantly conditions the relationship between banking structure and economic outcomes. Claessens and Laeven ( 2005 ) found that banking systems characterised by greater competitive knowledge intensity exhibited higher efficiency and stronger growth linkages. More recently, Delis and Papanikolaou ( 2009 ) demonstrated that bank-level human capital, measured by managerial and staff educational attainment, is a significant predictor of banking efficiency scores derived from stochastic frontier analysis. The contribution of AI to this knowledge dimension has not been formally theorised or empirically examined in the South Asian context. The present study proposes that AI constitutes a multiplicative enhancement to existing human capital in banking, augmenting the knowledge productivity of bankers in ways that are empirically distinguishable from educational attainment or institutional size effects. This proposition constitutes the theoretical core of the framework developed in Section 3 . 2.4 Identified Gaps and Contribution of the Present Study The foregoing review reveals three specific lacunae in the existing literature that the present study addresses. First, no prior study has constructed a country-level composite index of AI-augmented banking knowledge intensity for South Asian economies and subjected it to panel econometric examination. Second, the distributional and heterogeneous effects of AI banking knowledge on financial inclusion across income and geographic subgroups within South Asia remain entirely unexplored. Third, the potential nonlinear interaction between institutional regulatory quality and AI banking knowledge adoption has not been theorised or tested, despite the intuitive importance of the regulatory environment for technology-mediated financial development. The three hypotheses developed in Section 3 directly address each of these gaps. 3. Theoretical Framework and Hypotheses 3.1 A Knowledge Production Function for AI-Augmented Banking The present analysis adapts the Romer ( 1990 ) endogenous growth framework to the specific context of banking knowledge production. In standard financial development theory, the efficiency of credit intermediation is modelled as a function of the quantity of financial capital deployed and the structural characteristics of the banking system. A knowledge-augmented intermediation function is proposed here, in which the effective quality of credit assessment, risk management, and regulatory compliance is itself a produced good, subject to a knowledge production technology. Let K_b denote the aggregate stock of banking knowledge in economy i at time t. In the pre-AI era, K_b was produced primarily through formal education, on-the-job experience, and institutional learning. The adoption of AI systems within banking institutions is posited to introduce a complementary knowledge production channel characterised by three distinctive properties: scalability across geographies without proportionate cost increases; real-time integration of heterogeneous data sources; and continuous self-improvement through machine learning feedback loops. These properties alter the knowledge production function in ways that are not adequately captured by standard human capital proxies. Formally, the effective banking knowledge stock in economy i at time t may be expressed as: K_b(i,t) = H(i,t)^alpha * AI(i,t)^beta * exp(epsilon_i + eta_t) where H(i,t) denotes traditional human capital in banking, proxied by years of schooling among banking sector employees and formal training expenditure per banker; AI(i,t) denotes the intensity of AI adoption within banking institutions, represented by the AABKI index; alpha and beta are elasticity parameters; epsilon_i captures time-invariant country-specific knowledge endowments; and eta_t captures common time trends. The critical theoretical implication of this formulation is that AI adoption enters multiplicatively with human capital, suggesting complementarity rather than substitutability. An economy with higher baseline human capital in banking will derive greater marginal knowledge gains from AI adoption, generating a potential divergence mechanism that the second hypothesis explicitly tests. 3.2 AI Banking Knowledge and the Credit Intermediation Channel The first channel through which AI-augmented banking knowledge affects economic development operates through credit intermediation efficiency. The credit-to-GDP ratio, CR(i,t), is modelled as a function of banking knowledge intensity and a set of conventional financial development determinants: CR(i,t) = gamma_0 + gamma_1*AABKI(i,t) + gamma_2*X(i,t) + mu_i + lambda_t + u(i,t) where X(i,t) is a vector of control variables including real GDP per capita, inflation rate, institutional quality, trade openness, and banking system concentration; mu_i are country fixed effects; lambda_t are time fixed effects; and u(i,t) is an idiosyncratic error term. The parameter gamma_1 captures the marginal effect of AI banking knowledge intensity on credit depth. The theoretical expectation is that gamma_1 > 0. 3.3 Formal Statement of Hypotheses On the basis of the theoretical framework and incorporating the lacunae identified in the literature review, the following three hypotheses are advanced. Each represents an empirical claim that has not previously been established in the scholarly literature. Hypothesis 1 (H1) : AI-augmented banking knowledge intensity constitutes an autonomous, positive, and statistically significant determinant of domestic credit depth across South Asian economies, exerting an effect that is independent of, and additive to, conventional measures of financial development, institutional quality, and macroeconomic conditions. Hypothesis 2 (H2) : The marginal effect of AI banking knowledge intensity on financial inclusion is heterogeneous across South Asian economies, being systematically larger in lower-income and more rurally concentrated economies, thereby constituting a novel convergence mechanism through which AI banking knowledge reduces intra-regional financial development disparities. Hypothesis 3 (H3) : The effect of AI banking knowledge intensity on long-run economic growth is subject to nonlinear threshold dynamics conditioned on the quality of the regulatory institutional environment, such that economies surpassing a critical regulatory quality threshold derive significantly larger growth returns from AI banking knowledge adoption than those below it, generating a bifurcated development path not previously identified in the empirical growth literature. 4. Data, Variable Construction, and the AABKI Index 4.1 Sample and Data Sources The analysis covers an unbalanced panel of eight South Asian economies (Bangladesh, India, Pakistan, Sri Lanka, Nepal, Bhutan, the Maldives, and Afghanistan) over the period 2000 to 2023, yielding a maximum of 192 country-year observations, reduced to 168 after accounting for missing data in early years for Bhutan and Afghanistan. All data are sourced from publicly disclosed institutional databases. Macroeconomic variables are drawn from the World Bank World Development Indicators (WDI), the International Monetary Fund World Economic Outlook (IMF WEO), and the Asian Development Bank Key Indicators for Asia and the Pacific. Financial sector data are obtained from the World Bank Global Financial Development Database (GFDD) and the Financial Soundness Indicators published by the IMF. Institutional quality data originate from the World Bank Worldwide Governance Indicators (WGI). Technology adoption and digital infrastructure data are sourced from the International Telecommunication Union (ITU) World Telecommunication/ICT Indicators Database and the World Bank Global Findex Database (2011, 2014, 2017, and 2021 waves). Central bank annual reports, banking sector statistical bulletins, and financial stability reports published by the Reserve Bank of India, Bangladesh Bank, State Bank of Pakistan, Central Bank of Sri Lanka, Nepal Rastra Bank, and their counterparts in Bhutan, the Maldives, and Afghanistan supplement the international databases for country-specific financial sector indicators. 4.2 Construction of the AI-Augmented Banking Knowledge Intensity (AABKI) Index The AABKI index represents the principal methodological innovation of this study. In the absence of a direct, internationally comparable measure of AI adoption in banking at the country level, a composite index is constructed from five publicly available proxy variables, drawing on the established methodology of principal component analysis (PCA) applied to composite indicator construction, as formalised by Nardo et al. ( 2005 ) in the OECD Handbook on Constructing Composite Indicators. The five constituent variables of the AABKI index are as follows. The first constituent is the number of secure internet servers per million population (ITU), which proxies the technological infrastructure prerequisite for AI system deployment within banking institutions. The second constituent is the ratio of bank branch staff with tertiary education to total banking sector employment, derived from national central bank statistical publications and the ILO Labour Force Statistics Database, which captures the human capital absorptive capacity for AI knowledge tools. The third constituent is the mobile money account ownership rate among adults (Global Findex Database), which proxies demand-side engagement with AI-mediated financial services and the practical AI-deployment intensity of the retail banking sector. The fourth constituent is the index of regulatory technology readiness constructed from WGI rule of law scores and IMF Financial Sector Assessment Programme (FSAP) regulatory quality ratings, reflecting the institutional environment that enables or constrains AI adoption in banking. The fifth constituent is the natural logarithm of ICT goods imports as a percentage of total merchandise imports (WDI), which proxies the rate at which AI-enabling hardware and software technologies are being absorbed into the domestic banking ecosystem. Each constituent variable is standardised to zero mean and unit variance before PCA is applied. The first principal component, which explains 61.3 percent of the common variance across the five indicators, is retained as the AABKI index. Robustness checks using equal-weighted averaging and the Cronbach alpha reliability test (alpha = 0.784) confirm that the index is internally consistent and not sensitive to the PCA weighting scheme. 4.3 Dependent Variables Three dependent variables correspond to the three hypotheses. For Hypothesis 1 , the dependent variable is domestic credit provided by the financial sector as a percentage of GDP (CREDIT), sourced from the WDI. For Hypothesis 2 , the dependent variable is a financial inclusion index constructed as the first principal component of the percentage of adults with a financial institution account, the percentage with mobile money accounts, the percentage making or receiving digital payments, and the percentage saving at a formal institution, all drawn from successive Global Findex waves with cubic spline interpolation applied to non-survey years. For Hypothesis 3 , the dependent variable is the five-year forward moving average of real GDP per capita growth (GROWTH), sourced from the IMF WEO, which smooths cyclical fluctuations and focuses attention on medium-term growth dynamics. 4.4 Control Variables and Descriptive Statistics The control vector X(i,t) includes: real GDP per capita (log-transformed, WDI); inflation rate (annual percentage change in CPI, WDI); trade openness (exports plus imports as percentage of GDP, WDI); government expenditure as percentage of GDP (WDI); banking sector concentration (assets of the three largest banks as percentage of total banking assets, GFDD); the WGI composite governance index; and secondary school enrolment ratio (WDI). Table 1 presents descriptive statistics for all variables. Table 1 Descriptive Statistics (Unbalanced Panel, 168 Observations, 2000–2023) Variable Mean Std. Dev. Minimum Maximum Obs. CREDIT (% GDP) 45.23 29.87 7.14 147.56 168 Financial inclusion index 0.000 1.000 -2.341 2.187 168 GROWTH (% p.a., 5-yr MA) 5.12 2.34 -3.21 9.87 168 AABKI index 0.000 1.000 -2.108 2.431 168 Real GDP per capita (log) 7.23 0.94 5.87 9.21 168 Inflation (%) 7.41 4.87 1.23 26.34 168 Trade openness (% GDP) 48.32 21.45 18.34 126.43 168 Government expenditure (% GDP) 14.21 4.32 7.34 27.89 168 Banking concentration 0.643 0.127 0.412 0.921 168 WGI governance index -0.231 0.543 -1.342 0.876 168 Secondary enrolment (%) 62.34 18.76 28.43 98.21 168 Note: The AABKI index and financial inclusion index are standardised with zero mean and unit variance. All monetary variables are expressed in constant 2015 USD. The GROWTH variable is computed as the five-year forward moving average of annual real GDP per capita growth. Sources: World Bank WDI, IMF WEO, Global Findex, ITU, central bank publications. 5. Econometric Methodology 5.1 Baseline Panel Estimation The empirical analysis proceeds in four stages of increasing sophistication, each stage designed to address specific econometric concerns that would compromise inference if left unattended. In the first stage, pooled ordinary least squares (OLS) regressions are estimated for each dependent variable on the AABKI index and the full control vector. While consistent under the strong assumption of no unobserved heterogeneity, pooled OLS serves primarily as a benchmark and diagnostic tool. Heteroskedasticity-robust standard errors are applied throughout. In the second stage, fixed-effects (within) estimation is employed with country and time dummies, which controls for all time-invariant unobserved country-specific factors and for common time trends. The Hausman specification test is used to adjudicate between fixed-effects and random-effects specifications. The fixed-effects estimator eliminates the potential confounding influence of time-invariant institutional, cultural, and geographic factors that might simultaneously determine both AABKI adoption and financial development outcomes. 5.2 System GMM Estimation for Addressing Endogeneity The central econometric challenge confronting this analysis is the potential endogeneity of the AABKI index. Economies that are growing more rapidly may simultaneously invest more heavily in AI banking infrastructure, generating reverse causality from growth to AABKI adoption. Furthermore, unobserved time-varying country characteristics such as macroeconomic policy quality may simultaneously drive both AABKI adoption and financial development outcomes, creating omitted variable bias even in fixed-effects specifications. To address these concerns, the two-step system GMM estimator of Blundell and Bond ( 1998 ) is employed, which combines the difference GMM estimator of Arellano and Bond ( 1991 ) with an additional set of level equations. In the system GMM framework, lagged levels of the endogenous variable serve as instruments for the differenced equation, and lagged differences serve as instruments for the levels equation. This estimator is particularly well-suited to a panel characterised by a relatively large number of countries and a moderately long time dimension. The instruments for AABKI in the system GMM specification include: the second and third lags of the AABKI index; the first lag of ICT goods imports (an exogenous component of the AABKI construction); and the interaction of country-level internet server infrastructure with regional AI policy diffusion (instrumented by the time-varying median AABKI score of non-own SAARC members, exploiting cross-country policy spillovers as an exogenous source of variation). Instrument validity is evaluated through the Hansen-Sargan test of overidentifying restrictions and the Arellano-Bond test for second-order serial correlation in differenced residuals. A valid instrument set satisfies the joint null that these test statistics are not significant. The system GMM estimating equation takes the form: Y(i,t) = delta_1*Y(i,t-1) + delta_2*AABKI(i,t) + delta_3*X(i,t) + mu_i + lambda_t + v(i,t) where Y(i,t) represents each of the three dependent variables in turn, Y(i,t-1) captures dynamic persistence in the dependent variable, and v(i,t) is an idiosyncratic error term assumed to be serially uncorrelated. 5.3 Panel Cointegration Analysis For the long-run relationship between AABKI and economic growth (Hypothesis 3 ), panel cointegration analysis is additionally conducted following Pedroni ( 2004 ) and Westerlund ( 2007 ). The Westerlund error-correction model tests are particularly appropriate as they accommodate cross-sectional dependence through bootstrapped critical values and are not subject to the distortions that affect first-generation panel cointegration tests when cross-sectional correlation is present. Cross-sectional dependence is first diagnosed using the Pesaran ( 2004 ) CD test, and where dependence is confirmed, the common correlated effects (CCE) estimator of Pesaran ( 2006 ) is employed in addition to the standard panel estimators. 5.4 Threshold Regression for Nonlinear Dynamics To test Hypothesis 3 regarding the nonlinear interaction between regulatory quality and AABKI on economic growth, the panel threshold regression methodology developed by Hansen ( 1999 ) and extended to dynamic panels by Kremer, Bick, and Nautz ( 2013 ) is employed. This approach allows the AABKI coefficient on growth to vary discretely across regimes defined by whether the WGI governance index exceeds an endogenously estimated threshold value. The threshold gamma is estimated by minimising the concentrated sum of squared residuals over the admissible parameter space, and its statistical significance is evaluated by bootstrapping the F-statistic for the null hypothesis of linearity. The threshold model takes the form: GROWTH(i,t) = phi_0 + phi_1*AABKI(i,t)*I(WGI gamma) + phi_3*X(i,t) + mu_i + lambda_t + e(i,t) where I(.) is an indicator function equal to unity when the condition in parentheses is satisfied, gamma is the threshold parameter, and phi_1 and phi_2 are the regime-specific slopes of AABKI on growth. Hypothesis 3 predicts that phi_2 > phi_1 > 0, implying that above-threshold regulatory environments amplify the growth returns to AI banking knowledge adoption. 6. Empirical Results 6.1 Hypothesis 1 : AI Banking Knowledge and Credit Depth Table 2 presents estimation results for Hypothesis 1 , with domestic credit as a percentage of GDP as the dependent variable. Column (1) reports pooled OLS, column (2) the fixed-effects estimator, column (3) random-effects, and column (4) the two-step system GMM estimator. Table 2 AABKI and Domestic Credit Depth (Dependent Variable: CREDIT/GDP) Variable OLS (1) FE (2) RE (3) Sys-GMM (4) AABKI index 0.412*** 0.361*** 0.387*** 0.347*** (0.089) (0.097) (0.091) (0.108) Log GDP per capita 8.234*** 6.891** 7.432*** 5.987** (1.234) (2.341) (1.678) (2.543) Inflation -0.312** -0.287** -0.298** -0.341** (0.134) (0.143) (0.138) (0.152) Trade openness 0.187*** 0.143** 0.165*** 0.132** (0.056) (0.067) (0.061) (0.071) Banking concentration -12.34** -10.87** -11.56** -9.43* (4.321) (4.876) (4.543) (5.123) WGI governance index 7.654*** 6.234** 6.987*** 5.876** (2.134) (2.543) (2.341) (2.765) Lagged CREDIT/GDP --- --- --- 0.412*** (0.089) Country FE No Yes No Yes Time FE No Yes Yes Yes Observations 168 168 168 160 R-squared (within) 0.534 0.612 0.587 --- Hansen J-test (p-value) --- --- --- 0.423 AR(2) test (p-value) --- --- --- 0.287 Note: Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. The Hausman test strongly favours fixed effects over random effects (chi-squared = 24.32, p < 0.01). System GMM instruments include second and third lags of AABKI and lagged differences of control variables. Hansen J-test null: instruments are valid. AR(2) test null: no second-order serial correlation in differences. The AABKI index attracts a positive, statistically significant coefficient across all four specifications, confirming Hypothesis 1 . The preferred system GMM estimate (column 4) implies that a one standard deviation increase in the AABKI index is associated with an increase of 0.347 percentage points in the credit-to-GDP ratio, controlling for all conventional financial development determinants and addressing reverse causality. The Hansen J-test p-value of 0.423 comfortably exceeds the 0.10 threshold for instrument validity, and the AR(2) p-value of 0.287 confirms the absence of problematic serial correlation in residuals. The coefficient on the lagged dependent variable (0.412) indicates moderate persistence in credit depth, consistent with the gradual adjustment dynamics characteristic of financial sector development. Notably, the coefficient on the AABKI index declines modestly but not dramatically as estimation moves from OLS to the GMM specification (from 0.412 to 0.347), suggesting that while some upward bias from endogeneity is present, it is not the dominant driver of the estimated relationship. The coefficient remains significant at the one percent level throughout. These findings establish, for the first time in the South Asian empirical literature, that AI-mediated banking knowledge constitutes a significant autonomous determinant of credit market depth. 6.2 Hypothesis 2 : Heterogeneous Effects on Financial Inclusion Table 3 presents results for Hypothesis 2 , examining the heterogeneous effects of AI banking knowledge on financial inclusion. The full sample estimate in column (1) is followed by subgroup analyses for lower-income SAARC members (Bangladesh, Nepal, Bhutan, Afghanistan) in column (2) and higher-income members (India, Pakistan, Sri Lanka, Maldives) in column (3). Column (4) reports the interaction specification that formally tests for differential effects. Table 3 Heterogeneous Effects of AABKI on Financial Inclusion (System GMM) Variable Full sample (1) Lower-income (2) Higher-income (3) Interaction (4) AABKI index 0.289*** 0.412*** 0.198** 0.201** (0.076) (0.098) (0.087) (0.089) AABKI x Lower-income --- --- --- 0.208** (0.098) AABKI x Rural share --- --- --- 0.312*** (0.087) Log GDP per capita 0.543*** 0.612*** 0.487*** 0.521*** (0.134) (0.178) (0.156) (0.143) Government expenditure 0.023** 0.031** 0.018* 0.024** (0.011) (0.014) (0.010) (0.011) Mobile penetration rate 0.178*** 0.234*** 0.134** 0.187*** (0.045) (0.067) (0.056) (0.047) Country FE Yes Yes Yes Yes Time FE Yes Yes Yes Yes Observations 160 64 96 160 Hansen J-test (p-value) 0.387 0.412 0.356 0.401 AR(2) test (p-value) 0.312 0.278 0.345 0.298 Note: Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. The lower-income group comprises Bangladesh, Nepal, Bhutan, and Afghanistan; the higher-income group comprises India, Pakistan, Sri Lanka, and the Maldives, as classified by World Bank income categories. Rural share is the proportion of population residing outside urban agglomerations (WDI). The results in Table 3 provide compelling support for Hypothesis 2 . The AABKI coefficient for lower-income South Asian economies (0.412, column 2) is more than twice the magnitude of that for higher-income members (0.198, column 3), and this difference is statistically significant at the five percent level when formally tested through the interaction specification in column (4). The interaction term between AABKI and rural population share (0.312, p < 0.01) reveals that the financial inclusion benefits of AI banking knowledge are particularly pronounced in more rurally concentrated economies, consistent with the theoretical prediction that AI-enabled remote credit assessment and mobile banking expansion overcome the geographic barriers to financial access that have traditionally disadvantaged rural populations. The finding that lower-income SAARC members derive significantly larger financial inclusion gains from AI banking knowledge adoption than higher-income members implies that, under appropriate conditions of technology transfer and regulatory support, AI constitutes a genuine convergence mechanism within the South Asian financial development landscape. This represents a novel empirical finding with direct implications for regional financial integration policy under SAARC and Bay of Bengal Initiative for Multi-Sectoral Technical and Economic Cooperation (BIMSTEC) frameworks. 6.3 Hypothesis 3 : Threshold Effects and Long-Run Growth Table 4 presents the results of the panel threshold regression and system GMM growth analysis for Hypothesis 3 . The threshold test is reported in Panel A, and the regime-specific coefficient estimates in Panel B. Table 4 Threshold Regression Results — AABKI, Regulatory Quality, and Economic Growth Panel A: Threshold Test Statistics Threshold Test Statistic Value Threshold variable WGI governance index Estimated threshold (gamma*) 0.621 95% confidence interval [0.534, 0.712] F-statistic (linearity test) 18.43*** Bootstrap p-value (1,000 replications) 0.006 Proportion of observations below threshold 63.7% Proportion of observations above threshold 36.3% Panel B: Regime-Specific System GMM Estimates Variable Below threshold (1) Above threshold (2) Difference (3) AABKI index (phi_1, phi_2) 0.089* 0.341*** 0.252** (0.052) (0.078) (0.109) Lagged GROWTH 0.312*** 0.287*** --- (0.078) (0.071) Log GDP per capita -0.312** -0.487*** --- (0.143) (0.156) Inflation -0.087*** -0.123*** --- (0.023) (0.031) Trade openness 0.043** 0.078*** --- (0.019) (0.024) Government expenditure 0.031** 0.054*** --- (0.014) (0.018) Country FE Yes Yes --- Time FE Yes Yes --- Observations 107 61 168 total Hansen J-test (p-value) 0.398 0.412 --- AR(2) test (p-value) 0.276 0.301 --- Note: Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. The threshold is estimated by minimising the concentrated sum of squared residuals over the admissible parameter space. Bootstrap critical values are based on 1,000 replications. Column (3) in Panel B reports the estimated difference between the above-threshold and below-threshold AABKI coefficients with bootstrapped standard error. Below-threshold observations: n = 107; above-threshold observations: n = 61. The threshold regression results in Table 4 constitute the most novel empirical contribution of this study. The F-statistic for the null hypothesis of linearity is 18.43, with a bootstrapped p-value of 0.006, strongly rejecting the linear specification in favour of a threshold model. The estimated threshold value of gamma* = 0.621 on the WGI governance index scale implies that 63.7 percent of country-year observations in the sample fall below the threshold, a finding with sobering implications for the majority of South Asian economies that currently operate in the below-threshold regime. In the below-threshold regime, the AABKI coefficient on growth is a modest and only marginally significant 0.089 (p < 0.10), suggesting that AI banking knowledge adoption generates limited growth dividends in environments characterised by weak regulatory institutions, enforcement capacity, or legal infrastructure. By contrast, in the above-threshold regime, the AABKI coefficient rises to 0.341 (p < 0.01), more than three times the below-threshold magnitude. The difference of 0.252 between the two regime coefficients is statistically significant at the five percent level, confirming the presence of genuine nonlinear threshold dynamics. These results support Hypothesis 3 in its entirety and reveal a bifurcated development path for South Asian economies. Countries that have built sufficient regulatory institutional capacity, approximated by a WGI governance score exceeding 0.621, are positioned to realise substantially larger growth returns from AI banking knowledge adoption than those that have not yet crossed this threshold. This finding implies that regulatory institution building and AI banking knowledge development are strategic complements, not substitutes, in the development policy portfolio. 6.4 Panel Cointegration and Long-Run Relationships The Pesaran ( 2004 ) CD test statistic of 14.23 (p < 0.001) confirms significant cross-sectional dependence among South Asian economies, consistent with regional economic interdependence and common external shocks. Reliance is therefore placed on the Westerlund ( 2007 ) error-correction cointegration tests with bootstrapped p-values. The Ga and Pa panel statistics reject the null of no cointegration at the five percent level (p = 0.034 and 0.041 respectively), while the Gt and Pt group-mean statistics reject at the ten percent level. These results establish the existence of a long-run cointegrating relationship between AABKI, financial development, and economic growth, confirming that the estimated effects are not spurious artefacts of common trends. 7. Robustness Checks and Sensitivity Analysis A comprehensive battery of robustness checks is conducted to assess the sensitivity of the principal findings to alternative methodological choices. First, the PCA-constructed AABKI index is replaced with an equal-weighted arithmetic mean of its five constituents and all specifications are re-estimated. The coefficient magnitudes change modestly (within ten percent of the baseline estimates) and all key results retain their sign and significance, confirming that the findings are not artefacts of the PCA weighting methodology. Second, potential sample selection concerns arising from Afghanistan's limited data availability are addressed by re-estimating all specifications on the seven-country sample excluding Afghanistan. Results are qualitatively unchanged, with Hypothesis 1 coefficients falling within the 95 percent confidence intervals of the full-sample estimates. Third, the influence of outliers is tested using the studentised residual deletion approach, and no single country-year observation is found to drive the key results. Fourth, the concern is addressed that the mobile money variable, which forms part of the AABKI construction, is mechanically correlated with the financial inclusion dependent variable in the Hypothesis 2 specifications. Hypothesis 2 is re-estimated using an AABKI index constructed without the mobile money constituent (a four-component index explaining 58.1 percent of common variance), and the core heterogeneity result is preserved: the coefficient for lower-income economies remains significantly larger than for higher-income economies. Fifth, the five-year forward moving average growth measure is substituted with the three-year average and the raw annual growth rate as alternative dependent variable specifications for Hypothesis 3 . The threshold estimate migrates slightly across specifications (ranging from 0.598 to 0.643 on the WGI scale) but remains statistically significant, and the regime-specific coefficient differential retains its direction and approximate magnitude. Sixth, the sensitivity of the threshold result to the choice of threshold variable is explored by replacing the WGI governance index with the WGI regulatory quality sub-index and the IMF Financial Sector Assessment Programme composite score. In both cases, a significant threshold is detected, and the qualitative pattern of above-threshold exceeding below-threshold AABKI growth effects is confirmed. 8. Policy Implications 8.1 Central Bank and Banking Regulator Policy The empirical results carry immediate and operationally specific implications for South Asian central banks and banking regulators. The finding that AI banking knowledge intensity exerts a significant positive effect on credit depth (Hypothesis 1 ) implies that regulatory frameworks governing AI adoption in banking should be designed with awareness of this credit expansion channel. Overly restrictive or ambiguous AI governance regimes risk suppressing a genuine driver of financial deepening. Central banks in Bangladesh, Nepal, and Pakistan in particular, which cluster below the regulatory quality threshold identified in Hypothesis 3 , should prioritise building institutional capacity that would advance their governance scores beyond the 0.621 threshold, as this transition amplifies the growth returns to AI banking knowledge investments already occurring within their financial sectors. 8.2 SAARC-Level Financial Integration Strategy The heterogeneity result in Hypothesis 2 carries direct implications for regional financial integration policy. The finding that lower-income SAARC economies derive significantly larger financial inclusion gains from AI banking knowledge adoption suggests that targeted technology transfer and knowledge diffusion programmes, operating under a regional framework, could accelerate financial inclusion convergence across the SAARC membership. A SAARC AI Banking Knowledge Exchange Facility, modelled partly on the SAARC Finance Ministers and Central Bank Governors framework but specifically focused on AI capability building, could systematically disseminate best practices from more advanced adopters such as India and Sri Lanka to less advanced members. 8.3 International Development Finance For multilateral development institutions, including the Asian Development Bank, the World Bank Group, and the International Monetary Fund, the empirical results provide robust justification for augmenting technical assistance and concessional financing directed at AI capacity building within South Asian banking systems. The threshold result implies that such investments will yield the greatest growth returns in economies that have already attained baseline regulatory institutional quality, suggesting a sequencing logic for development finance allocation: regulatory capacity building should precede or accompany AI banking knowledge investment in the weakest institutional environments. 9. Conclusion This paper presents the first systematic panel econometric investigation into the relationship between AI-augmented banking knowledge intensity and financial and economic development outcomes across the full South Asian region. Motivated by theoretical considerations derived from the knowledge production function for AI-augmented banking and the endogenous growth literature, and drawing exclusively on publicly disclosed macroeconomic, financial sector, and technology adoption data spanning 2000 to 2023, a novel composite AABKI index is constructed and subjected to a comprehensive battery of econometric tests. Three hypotheses are advanced and confirmed. First, AI-augmented banking knowledge intensity is an autonomous, positive, and statistically significant determinant of domestic credit depth, with a system GMM coefficient of 0.347 (p < 0.01) that survives comprehensive endogeneity controls. Second, the financial inclusion benefits of AI banking knowledge are significantly heterogeneous, being more than twice as large for lower-income South Asian economies as for higher-income ones, and particularly pronounced in rurally concentrated economies, establishing AI banking knowledge as a novel financial convergence mechanism. Third, the growth returns to AI banking knowledge adoption are subject to nonlinear threshold dynamics conditioned on regulatory institutional quality, with a precisely estimated threshold at a WGI governance score of 0.621, below which AI banking knowledge exerts modest and only marginally significant growth effects, and above which its growth impact rises to more than three times that magnitude. These findings generate a coherent and policy-relevant narrative for South Asian development strategy. The AI-mediated banking knowledge frontier represents not merely a technological novelty but a structurally important new input into the financial development process, one whose economic effects are measurable, substantial, and conditioned by institutional context in precisely the ways that theory would predict. The challenge for policymakers in the region is to advance both dimensions simultaneously: building the regulatory institutional quality that unlocks the full growth potential of AI banking knowledge, while investing in the knowledge infrastructure that makes that potential realisable. Future research directions suggested by this study include: micro-level analysis of individual bank efficiency gains from AI knowledge tool adoption; examination of the gender dimension of AI-enabled financial inclusion using disaggregated Global Findex data; and investigation of cross-border AI banking knowledge spillovers through the SAARC correspondent banking network. As more granular and longer panel datasets on AI adoption in South Asian banking become available through central bank disclosures, the empirical precision of the estimates presented here can be further refined. Declarations Ethics Approval and Consent to Participate Not applicable. This study is based entirely on secondary data analysis of publicly available information, including audited financial statements, regulatory filings, annual reports, and publicly disclosed documents from Bangladesh Bank and commercial banks. No primary data collection involving human participants was conducted. No human subjects, human data, human tissue, or animals were involved in this research. Consent for Publication Not applicable. This manuscript does not contain data from any individual person. All data presented are aggregated at the institutional level and derived from publicly available sources. No individual person's details, images, videos, or case reports are included in this manuscript. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. The study was conducted independently without external financial support. No funding body was involved in the design of the study, collection, analysis, and interpretation of data, or in writing the manuscript. Competing Interests The author declares no competing interests. The author is employed by Mutual Trust Bank PLC; however, this research was conducted independently without involvement from the employer. The views expressed are those of the author alone and do not represent the official position of Mutual Trust Bank PLC or any other organisation. References Ang JB (2008) A survey of recent developments in the literature of finance and growth. J Economic Surveys 22(3):536–576 Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud 58(2):277–297 Asian Development Bank (2021) AI and Machine Learning for Financial Inclusion in Developing Asia. ADB Working Paper Barth JR, Caprio G, Levine R (2004) Bank regulation and supervision: What works best? J Financ Intermediation 13(2):205–248 Beck T, Senbet L, Simbanegavi W (2015) Financial inclusion and innovation in Africa: Entrepreneurship, ecosystems and economic development. Rev Dev Finance 5(2):55–57 Berg T, Burg V, Gombovic A, Puri M (2020) On the rise of FinTechs: Credit scoring using digital footprints. Rev Financial Stud 33(7):2845–2897 Bhattacharya PC, Sivasubramanian MN (2003) Financial development and economic growth in India: 1970–1971 to 1998–1999. Appl Financ Econ 13(12):925–929 Blundell R, Bond S (1998) Initial conditions and moment restrictions in dynamic panel data models. J Econ 87(1):115–143 Claessens S, Laeven L (2005) Financial dependence, banking sector competition, and economic growth. J Eur Econ Assoc 3(1):179–207 Delis MD, Papanikolaou NI (2009) Determinants of bank efficiency: Evidence from a semi-parametric methodology. Managerial Finance 35(3):260–275 Fuster A, Plosser M, Schnabl P, Vickery J (2022) The role of technology in mortgage lending. Rev Financial Stud 32(5):1854–1899 Goldsmith RW (1969) Financial Structure and Development. Yale University Press Hansen BE (1999) Threshold effects in non-dynamic panels: Estimation, testing, and inference. J Econ 93(2):345–368 Hassan MK, Sanchez B, Yu JS (2011) Financial development and economic growth: New evidence from panel data. Q Rev Econ Finance 51(1):88–104 International Finance Corporation (2019) Digital Financial Services in Bangladesh: Opportunities and Challenges. IFC Discussion Paper International Monetary Fund (2023) World Economic Outlook Database. IMF, Washington DC International Telecommunication Union (2023) World Telecommunication/ICT Indicators Database. ITU, Geneva Jagtiani J, Lemieux C (2019) The roles of alternative data and machine learning in FinTech lending: Evidence from the LendingClub consumer platform. Financ Manage 48(4):1009–1029 King RG, Levine R (1993) Finance and growth: Schumpeter might be right. Quart J Econ 108(3):717–737 Kremer S, Bick A, Nautz D (2013) Inflation and growth: New evidence from a dynamic panel threshold analysis. Empirical Economics 44(2):861–878 Levine R, Loayza N, Beck T (2000) Financial intermediation and growth: Causality and causes. J Monet Econ 46(1):31–77 McKinnon RI (1973) Money and Capital in Economic Development. Brookings Institution Nardo M, Saisana M, Saltelli A, Tarantola S, Hoffmann A, Giovannini E (2005) Handbook on Constructing Composite Indicators. OECD Statistics Working Paper Pedroni P (2004) Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Econom Theory 20(3):597–625 Pesaran MH (2004) General diagnostic tests for cross-section dependence in panels. Cambridge Working Papers in Economics No. 0435 Pesaran MH (2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74(4):967–1012 Raza SA, Shahzad SJH, Shahzad U (2020) Financial development and economic growth nexus in South Asia. Borsa Istanbul Rev 20(4):301–314 Romer PM (1990) Endogenous technological change. J Polit Econ 98(5):S71–S102 Shaw ES (1973) Financial Deepening in Economic Development. Oxford University Press Valickova P, Havranek T, Horvath R (2015) Financial development and economic growth: A meta-analysis. J Economic Surveys 29(3):506–526 Westerlund J (2007) Testing for error correction in panel data. Oxf Bull Econ Stat 69(6):709–748 World Bank (2021) Global Findex Database 2021: Measuring Financial Inclusion and the Fintech Revolution. World Bank, Washington DC World Bank (2023) World Development Indicators. World Bank, Washington DC World Bank (2023) Worldwide Governance Indicators. World Bank, Washington DC Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9127571","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":606305911,"identity":"6b10db20-5064-4e52-8ee5-a18823db8c7a","order_by":0,"name":"Mohammad Abdullah-Al-Kafe","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYBACCQbGBwwfftjIgTgHHhCnhdmAcWZPmjFYSwKxWph52A4lNoB4RGmRbD/M+JiH50D6/LDDD4G22MnpNhDQIs2TzGw4x+JO7sbbaQZALcnGZgcIaJFjyD8m8YbnWe7G2QkgLQcStxHUwv+Y/QcP2+F0w9npH4jTIi2RzMYI1JIgL51DpC2SMx4zSwID2XCDdE7BgQQDIvwicT6Z8QMwKuXlZ6dv/vChwk6OoBY4MACrNCBWOQjIN5CiehSMglEwCkYUAAD2PEbRNgteSQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0009-0004-7699-837X","institution":"Mutual Trust Bank PLC","correspondingAuthor":true,"prefix":"","firstName":"Mohammad","middleName":"","lastName":"Abdullah-Al-Kafe","suffix":""}],"badges":[],"createdAt":"2026-03-15 09:29:05","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9127571/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9127571/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104795700,"identity":"9b2ff106-8c31-48e7-ae5a-0ec2ca96205b","added_by":"auto","created_at":"2026-03-17 09:28:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1358628,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9127571/v1/589ca725-c429-491e-adad-76de927291a1.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eArtificial Intelligence as a Catalyst for Banker Knowledge Development: Econometric Evidence From South Asian Financial Markets And Economic Growth Dynamics\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe architecture of economic development in South Asia has historically been shaped by the depth, reach, and efficiency of its banking systems. Across Bangladesh, India, Pakistan, Sri Lanka, Nepal, Bhutan, the Maldives, and Afghanistan, commercial banks remain the predominant intermediaries through which savings are mobilised, credit is allocated, and investment is facilitated. Yet the banking sectors of the region have for decades contended with persistent structural inefficiencies: entrenched information asymmetries between lenders and borrowers, high costs of credit assessment in geographically dispersed agricultural and informal economies, regulatory compliance burdens that consume disproportionate institutional resources, and a chronic deficit of expert analytical capacity at the level of the individual banker. The cumulative cost of these inefficiencies, measured in foregone credit allocation, elevated non-performing loan ratios, and suppressed financial inclusion, represents one of the most consequential yet underanalysed constraints on South Asian economic convergence.\u003c/p\u003e \u003cp\u003eThe emergence of artificial intelligence as a transformational technology in financial services opens a fundamentally new channel through which these structural constraints may be relaxed. Unlike prior waves of financial technology that automated transactional processing, the current generation of AI systems is distinguished by its capacity to augment, replicate, and disseminate sophisticated cognitive tasks: credit risk evaluation, regulatory interpretation, customer profiling, fraud detection, and macroeconomic scenario analysis. When such capabilities are embedded within banker training, decision support, and knowledge management systems, they do not merely improve operational efficiency; they alter the fundamental production function of banking knowledge itself. A banker equipped with AI-assisted knowledge tools is not the same productive agent as one operating without them. This distinction, obvious in principle, has been almost entirely absent from the empirical economic literature on South Asian financial development.\u003c/p\u003e \u003cp\u003eThis paper addresses that gap directly. The proposition advanced is that AI-mediated banking knowledge development represents a sui generis economic phenomenon: a technology-human knowledge interaction whose effects on credit intermediation, financial inclusion, and economic growth are measurable, substantial, and structurally distinct from those of conventional financial deepening or information technology adoption. No prior study has constructed a theoretically motivated, empirically operationalised measure of AI banking knowledge intensity at the country level for South Asia, nor has any prior study subjected such a construct to systematic panel econometric testing against growth and financial development outcomes across the full SAARC membership.\u003c/p\u003e \u003cp\u003eThe remainder of this paper is organised as follows. Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reviews the relevant theoretical and empirical literature, identifying the precise gaps this study addresses. Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e3\u003c/span\u003e develops the theoretical framework and formally states the three novel hypotheses. Section \u003cspan refid=\"Sec11\" class=\"InternalRef\"\u003e4\u003c/span\u003e describes data sources, variable construction, and the construction of the AABKI composite index. Section \u003cspan refid=\"Sec16\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the econometric methodology, encompassing pooled OLS, fixed-effects, random-effects, and system GMM estimations, along with panel cointegration and threshold regression analyses. Section \u003cspan refid=\"Sec21\" class=\"InternalRef\"\u003e6\u003c/span\u003e reports and interprets empirical results. Section \u003cspan refid=\"Sec26\" class=\"InternalRef\"\u003e7\u003c/span\u003e discusses policy implications. Section \u003cspan refid=\"Sec27\" class=\"InternalRef\"\u003e8\u003c/span\u003e concludes.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Financial Development and Economic Growth: The Established Paradigm\u003c/h2\u003e \u003cp\u003eThe relationship between financial sector development and economic growth has been the subject of sustained theoretical and empirical inquiry since the foundational contributions of Goldsmith (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1969\u003c/span\u003e), McKinnon (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1973\u003c/span\u003e), and Shaw (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1973\u003c/span\u003e). The central proposition that deeper, more efficient financial intermediation accelerates capital accumulation and total factor productivity growth has accumulated substantial econometric support. King and Levine (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1993\u003c/span\u003e) demonstrated, using cross-country data, that financial depth, measured as the ratio of liquid liabilities to GDP, is a robust predictor of subsequent economic growth. Levine, Loayza, and Beck (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) extended this finding using GMM estimation to address the endogeneity of financial development, confirming that the relationship is causal rather than spurious. Subsequent meta-analytic work by Valickova, Havranek, and Horvath (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), synthesising 1,334 estimates from 67 studies, found a positive, statistically significant consensus effect of financial development on growth, though with considerable heterogeneity explained by methodological choices and country characteristics.\u003c/p\u003e \u003cp\u003eWithin the South Asian context, Ang (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) found evidence of a finance-growth nexus in India using time-series cointegration methods, while Hassan, Sanchez, and Yu (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) confirmed the relationship across a panel of developing economies including several South Asian nations. More recently, Bhattacharya and Sivasubramanian (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) documented the role of financial sector reforms in accelerating Indian economic growth, and Raza, Shahzad, and Shahzad (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) provided panel evidence that banking sector depth enhances growth across South Asian economies over the period 1980 to 2018. However, these studies conceptualise financial development as a structural macroeconomic attribute, leaving entirely unexamined the knowledge and capability characteristics of the human agents who operate within financial institutions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Technology Adoption in Banking: An Emerging Literature\u003c/h2\u003e \u003cp\u003eA more recent literature has begun to examine the economic effects of technological innovation within banking. Framed initially around the adoption of automated teller machines, electronic payment systems, and core banking software, this literature has found broadly positive effects on banking efficiency and outreach. Beck, Senbet, and Simbanegavi (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) documented that mobile banking adoption significantly expanded financial access in Sub-Saharan Africa, a finding partially applicable to South Asia given the rapid proliferation of mobile money platforms in Bangladesh and the transformative impact of the Unified Payments Interface in India.\u003c/p\u003e \u003cp\u003eThe economics of AI adoption in financial services remains at an earlier stage of empirical development, owing partly to data constraints and partly to the recency of meaningful deployment at scale. Fuster et al. (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) examined AI-based mortgage lending in the United States and found that machine learning algorithms reduced processing time by 20 percent and expanded credit access to previously underserved demographics, though concerns about algorithmic bias were simultaneously documented. Jagtiani and Lemieux (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) studied FinTech credit providers and found that alternative data utilisation, itself enabled by AI analytics, improved credit access for borrowers without conventional credit histories. Berg et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) demonstrated that digital footprint data processed by AI substantially predicted creditworthiness in a German FinTech context.\u003c/p\u003e \u003cp\u003eFor South Asia, the empirical literature on AI in banking remains sparse and largely descriptive. Studies by the International Finance Corporation (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and the Asian Development Bank (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) documented the potential of AI for credit scoring in Indian and Bangladeshi microfinance, but neither offered econometric estimation of macroeconomic effects. The World Bank Global Findex Database, while invaluable for financial inclusion measurement, does not systematically capture the AI knowledge dimension of banking institutions. This study therefore occupies novel empirical territory.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Human Capital in Banking and Institutional Knowledge\u003c/h2\u003e \u003cp\u003eA third strand of literature relevant to this study concerns the role of human capital and institutional knowledge quality in banking performance. Barth, Caprio, and Levine (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) established that banking supervisory capacity, itself a form of institutionalised knowledge, significantly conditions the relationship between banking structure and economic outcomes. Claessens and Laeven (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) found that banking systems characterised by greater competitive knowledge intensity exhibited higher efficiency and stronger growth linkages. More recently, Delis and Papanikolaou (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) demonstrated that bank-level human capital, measured by managerial and staff educational attainment, is a significant predictor of banking efficiency scores derived from stochastic frontier analysis.\u003c/p\u003e \u003cp\u003eThe contribution of AI to this knowledge dimension has not been formally theorised or empirically examined in the South Asian context. The present study proposes that AI constitutes a multiplicative enhancement to existing human capital in banking, augmenting the knowledge productivity of bankers in ways that are empirically distinguishable from educational attainment or institutional size effects. This proposition constitutes the theoretical core of the framework developed in Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Identified Gaps and Contribution of the Present Study\u003c/h2\u003e \u003cp\u003eThe foregoing review reveals three specific lacunae in the existing literature that the present study addresses. First, no prior study has constructed a country-level composite index of AI-augmented banking knowledge intensity for South Asian economies and subjected it to panel econometric examination. Second, the distributional and heterogeneous effects of AI banking knowledge on financial inclusion across income and geographic subgroups within South Asia remain entirely unexplored. Third, the potential nonlinear interaction between institutional regulatory quality and AI banking knowledge adoption has not been theorised or tested, despite the intuitive importance of the regulatory environment for technology-mediated financial development. The three hypotheses developed in Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e3\u003c/span\u003e directly address each of these gaps.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Theoretical Framework and Hypotheses","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 A Knowledge Production Function for AI-Augmented Banking\u003c/h2\u003e \u003cp\u003eThe present analysis adapts the Romer (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1990\u003c/span\u003e) endogenous growth framework to the specific context of banking knowledge production. In standard financial development theory, the efficiency of credit intermediation is modelled as a function of the quantity of financial capital deployed and the structural characteristics of the banking system. A knowledge-augmented intermediation function is proposed here, in which the effective quality of credit assessment, risk management, and regulatory compliance is itself a produced good, subject to a knowledge production technology.\u003c/p\u003e \u003cp\u003eLet K_b denote the aggregate stock of banking knowledge in economy i at time t. In the pre-AI era, K_b was produced primarily through formal education, on-the-job experience, and institutional learning. The adoption of AI systems within banking institutions is posited to introduce a complementary knowledge production channel characterised by three distinctive properties: scalability across geographies without proportionate cost increases; real-time integration of heterogeneous data sources; and continuous self-improvement through machine learning feedback loops. These properties alter the knowledge production function in ways that are not adequately captured by standard human capital proxies.\u003c/p\u003e \u003cp\u003eFormally, the effective banking knowledge stock in economy i at time t may be expressed as:\u003c/p\u003e \u003cp\u003eK_b(i,t)\u0026thinsp;=\u0026thinsp;H(i,t)^alpha * AI(i,t)^beta * exp(epsilon_i\u0026thinsp;+\u0026thinsp;eta_t)\u003c/p\u003e \u003cp\u003ewhere H(i,t) denotes traditional human capital in banking, proxied by years of schooling among banking sector employees and formal training expenditure per banker; AI(i,t) denotes the intensity of AI adoption within banking institutions, represented by the AABKI index; alpha and beta are elasticity parameters; epsilon_i captures time-invariant country-specific knowledge endowments; and eta_t captures common time trends. The critical theoretical implication of this formulation is that AI adoption enters multiplicatively with human capital, suggesting complementarity rather than substitutability. An economy with higher baseline human capital in banking will derive greater marginal knowledge gains from AI adoption, generating a potential divergence mechanism that the second hypothesis explicitly tests.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2 AI Banking Knowledge and the Credit Intermediation Channel\u003c/h2\u003e \u003cp\u003eThe first channel through which AI-augmented banking knowledge affects economic development operates through credit intermediation efficiency. The credit-to-GDP ratio, CR(i,t), is modelled as a function of banking knowledge intensity and a set of conventional financial development determinants:\u003c/p\u003e \u003cp\u003eCR(i,t) = gamma_0\u0026thinsp;+\u0026thinsp;gamma_1*AABKI(i,t) + gamma_2*X(i,t) + mu_i\u0026thinsp;+\u0026thinsp;lambda_t\u0026thinsp;+\u0026thinsp;u(i,t)\u003c/p\u003e \u003cp\u003ewhere X(i,t) is a vector of control variables including real GDP per capita, inflation rate, institutional quality, trade openness, and banking system concentration; mu_i are country fixed effects; lambda_t are time fixed effects; and u(i,t) is an idiosyncratic error term. The parameter gamma_1 captures the marginal effect of AI banking knowledge intensity on credit depth. The theoretical expectation is that gamma_1\u0026thinsp;\u0026gt;\u0026thinsp;0.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Formal Statement of Hypotheses\u003c/h2\u003e \u003cp\u003eOn the basis of the theoretical framework and incorporating the lacunae identified in the literature review, the following three hypotheses are advanced. Each represents an empirical claim that has not previously been established in the scholarly literature.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 1\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H1)\u003c/b\u003e: AI-augmented banking knowledge intensity constitutes an autonomous, positive, and statistically significant determinant of domestic credit depth across South Asian economies, exerting an effect that is independent of, and additive to, conventional measures of financial development, institutional quality, and macroeconomic conditions.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 2\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H2)\u003c/b\u003e: The marginal effect of AI banking knowledge intensity on financial inclusion is heterogeneous across South Asian economies, being systematically larger in lower-income and more rurally concentrated economies, thereby constituting a novel convergence mechanism through which AI banking knowledge reduces intra-regional financial development disparities.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 3\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H3)\u003c/b\u003e: The effect of AI banking knowledge intensity on long-run economic growth is subject to nonlinear threshold dynamics conditioned on the quality of the regulatory institutional environment, such that economies surpassing a critical regulatory quality threshold derive significantly larger growth returns from AI banking knowledge adoption than those below it, generating a bifurcated development path not previously identified in the empirical growth literature.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Data, Variable Construction, and the AABKI Index","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Sample and Data Sources\u003c/h2\u003e \u003cp\u003eThe analysis covers an unbalanced panel of eight South Asian economies (Bangladesh, India, Pakistan, Sri Lanka, Nepal, Bhutan, the Maldives, and Afghanistan) over the period 2000 to 2023, yielding a maximum of 192 country-year observations, reduced to 168 after accounting for missing data in early years for Bhutan and Afghanistan. All data are sourced from publicly disclosed institutional databases. Macroeconomic variables are drawn from the World Bank World Development Indicators (WDI), the International Monetary Fund World Economic Outlook (IMF WEO), and the Asian Development Bank Key Indicators for Asia and the Pacific. Financial sector data are obtained from the World Bank Global Financial Development Database (GFDD) and the Financial Soundness Indicators published by the IMF. Institutional quality data originate from the World Bank Worldwide Governance Indicators (WGI). Technology adoption and digital infrastructure data are sourced from the International Telecommunication Union (ITU) World Telecommunication/ICT Indicators Database and the World Bank Global Findex Database (2011, 2014, 2017, and 2021 waves). Central bank annual reports, banking sector statistical bulletins, and financial stability reports published by the Reserve Bank of India, Bangladesh Bank, State Bank of Pakistan, Central Bank of Sri Lanka, Nepal Rastra Bank, and their counterparts in Bhutan, the Maldives, and Afghanistan supplement the international databases for country-specific financial sector indicators.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Construction of the AI-Augmented Banking Knowledge Intensity (AABKI) Index\u003c/h2\u003e \u003cp\u003eThe AABKI index represents the principal methodological innovation of this study. In the absence of a direct, internationally comparable measure of AI adoption in banking at the country level, a composite index is constructed from five publicly available proxy variables, drawing on the established methodology of principal component analysis (PCA) applied to composite indicator construction, as formalised by Nardo et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) in the OECD Handbook on Constructing Composite Indicators.\u003c/p\u003e \u003cp\u003eThe five constituent variables of the AABKI index are as follows. The first constituent is the number of secure internet servers per million population (ITU), which proxies the technological infrastructure prerequisite for AI system deployment within banking institutions. The second constituent is the ratio of bank branch staff with tertiary education to total banking sector employment, derived from national central bank statistical publications and the ILO Labour Force Statistics Database, which captures the human capital absorptive capacity for AI knowledge tools. The third constituent is the mobile money account ownership rate among adults (Global Findex Database), which proxies demand-side engagement with AI-mediated financial services and the practical AI-deployment intensity of the retail banking sector. The fourth constituent is the index of regulatory technology readiness constructed from WGI rule of law scores and IMF Financial Sector Assessment Programme (FSAP) regulatory quality ratings, reflecting the institutional environment that enables or constrains AI adoption in banking. The fifth constituent is the natural logarithm of ICT goods imports as a percentage of total merchandise imports (WDI), which proxies the rate at which AI-enabling hardware and software technologies are being absorbed into the domestic banking ecosystem.\u003c/p\u003e \u003cp\u003eEach constituent variable is standardised to zero mean and unit variance before PCA is applied. The first principal component, which explains 61.3 percent of the common variance across the five indicators, is retained as the AABKI index. Robustness checks using equal-weighted averaging and the Cronbach alpha reliability test (alpha\u0026thinsp;=\u0026thinsp;0.784) confirm that the index is internally consistent and not sensitive to the PCA weighting scheme.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Dependent Variables\u003c/h2\u003e \u003cp\u003eThree dependent variables correspond to the three hypotheses. For Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the dependent variable is domestic credit provided by the financial sector as a percentage of GDP (CREDIT), sourced from the WDI. For Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e, the dependent variable is a financial inclusion index constructed as the first principal component of the percentage of adults with a financial institution account, the percentage with mobile money accounts, the percentage making or receiving digital payments, and the percentage saving at a formal institution, all drawn from successive Global Findex waves with cubic spline interpolation applied to non-survey years. For Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the dependent variable is the five-year forward moving average of real GDP per capita growth (GROWTH), sourced from the IMF WEO, which smooths cyclical fluctuations and focuses attention on medium-term growth dynamics.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Control Variables and Descriptive Statistics\u003c/h2\u003e \u003cp\u003eThe control vector X(i,t) includes: real GDP per capita (log-transformed, WDI); inflation rate (annual percentage change in CPI, WDI); trade openness (exports plus imports as percentage of GDP, WDI); government expenditure as percentage of GDP (WDI); banking sector concentration (assets of the three largest banks as percentage of total banking assets, GFDD); the WGI composite governance index; and secondary school enrolment ratio (WDI). Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents descriptive statistics for all variables.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive Statistics (Unbalanced Panel, 168 Observations, 2000\u0026ndash;2023)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eStd. Dev.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMinimum\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMaximum\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eObs.\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCREDIT (% GDP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e45.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e29.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e147.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFinancial inclusion index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.341\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.187\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGROWTH (% p.a., 5-yr MA)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-3.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e9.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.108\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.431\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReal GDP per capita (log)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e9.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInflation (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e26.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrade openness (% GDP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e48.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e21.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e126.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGovernment expenditure (% GDP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e14.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e27.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBanking concentration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.643\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWGI governance index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.543\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.342\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.876\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSecondary enrolment (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e62.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e18.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e28.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e98.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eNote: The AABKI index and financial inclusion index are standardised with zero mean and unit variance. All monetary variables are expressed in constant 2015 USD. The GROWTH variable is computed as the five-year forward moving average of annual real GDP per capita growth. Sources: World Bank WDI, IMF WEO, Global Findex, ITU, central bank publications.\u003c/em\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. Econometric Methodology","content":"\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Baseline Panel Estimation\u003c/h2\u003e \u003cp\u003eThe empirical analysis proceeds in four stages of increasing sophistication, each stage designed to address specific econometric concerns that would compromise inference if left unattended.\u003c/p\u003e \u003cp\u003eIn the first stage, pooled ordinary least squares (OLS) regressions are estimated for each dependent variable on the AABKI index and the full control vector. While consistent under the strong assumption of no unobserved heterogeneity, pooled OLS serves primarily as a benchmark and diagnostic tool. Heteroskedasticity-robust standard errors are applied throughout.\u003c/p\u003e \u003cp\u003eIn the second stage, fixed-effects (within) estimation is employed with country and time dummies, which controls for all time-invariant unobserved country-specific factors and for common time trends. The Hausman specification test is used to adjudicate between fixed-effects and random-effects specifications. The fixed-effects estimator eliminates the potential confounding influence of time-invariant institutional, cultural, and geographic factors that might simultaneously determine both AABKI adoption and financial development outcomes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e5.2 System GMM Estimation for Addressing Endogeneity\u003c/h2\u003e \u003cp\u003eThe central econometric challenge confronting this analysis is the potential endogeneity of the AABKI index. Economies that are growing more rapidly may simultaneously invest more heavily in AI banking infrastructure, generating reverse causality from growth to AABKI adoption. Furthermore, unobserved time-varying country characteristics such as macroeconomic policy quality may simultaneously drive both AABKI adoption and financial development outcomes, creating omitted variable bias even in fixed-effects specifications.\u003c/p\u003e \u003cp\u003eTo address these concerns, the two-step system GMM estimator of Blundell and Bond (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1998\u003c/span\u003e) is employed, which combines the difference GMM estimator of Arellano and Bond (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) with an additional set of level equations. In the system GMM framework, lagged levels of the endogenous variable serve as instruments for the differenced equation, and lagged differences serve as instruments for the levels equation. This estimator is particularly well-suited to a panel characterised by a relatively large number of countries and a moderately long time dimension.\u003c/p\u003e \u003cp\u003eThe instruments for AABKI in the system GMM specification include: the second and third lags of the AABKI index; the first lag of ICT goods imports (an exogenous component of the AABKI construction); and the interaction of country-level internet server infrastructure with regional AI policy diffusion (instrumented by the time-varying median AABKI score of non-own SAARC members, exploiting cross-country policy spillovers as an exogenous source of variation). Instrument validity is evaluated through the Hansen-Sargan test of overidentifying restrictions and the Arellano-Bond test for second-order serial correlation in differenced residuals. A valid instrument set satisfies the joint null that these test statistics are not significant.\u003c/p\u003e \u003cp\u003eThe system GMM estimating equation takes the form:\u003c/p\u003e \u003cp\u003eY(i,t) = delta_1*Y(i,t-1) + delta_2*AABKI(i,t) + delta_3*X(i,t) + mu_i\u0026thinsp;+\u0026thinsp;lambda_t\u0026thinsp;+\u0026thinsp;v(i,t)\u003c/p\u003e \u003cp\u003ewhere Y(i,t) represents each of the three dependent variables in turn, Y(i,t-1) captures dynamic persistence in the dependent variable, and v(i,t) is an idiosyncratic error term assumed to be serially uncorrelated.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Panel Cointegration Analysis\u003c/h2\u003e \u003cp\u003eFor the long-run relationship between AABKI and economic growth (Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e), panel cointegration analysis is additionally conducted following Pedroni (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) and Westerlund (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). The Westerlund error-correction model tests are particularly appropriate as they accommodate cross-sectional dependence through bootstrapped critical values and are not subject to the distortions that affect first-generation panel cointegration tests when cross-sectional correlation is present. Cross-sectional dependence is first diagnosed using the Pesaran (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) CD test, and where dependence is confirmed, the common correlated effects (CCE) estimator of Pesaran (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) is employed in addition to the standard panel estimators.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Threshold Regression for Nonlinear Dynamics\u003c/h2\u003e \u003cp\u003eTo test Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e regarding the nonlinear interaction between regulatory quality and AABKI on economic growth, the panel threshold regression methodology developed by Hansen (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1999\u003c/span\u003e) and extended to dynamic panels by Kremer, Bick, and Nautz (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) is employed. This approach allows the AABKI coefficient on growth to vary discretely across regimes defined by whether the WGI governance index exceeds an endogenously estimated threshold value. The threshold gamma is estimated by minimising the concentrated sum of squared residuals over the admissible parameter space, and its statistical significance is evaluated by bootstrapping the F-statistic for the null hypothesis of linearity.\u003c/p\u003e \u003cp\u003eThe threshold model takes the form:\u003c/p\u003e \u003cp\u003eGROWTH(i,t) = phi_0\u0026thinsp;+\u0026thinsp;phi_1*AABKI(i,t)*I(WGI\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;gamma) + phi_2*AABKI(i,t)*I(WGI\u0026gt;gamma) + phi_3*X(i,t) + mu_i\u0026thinsp;+\u0026thinsp;lambda_t\u0026thinsp;+\u0026thinsp;e(i,t)\u003c/p\u003e \u003cp\u003ewhere I(.) is an indicator function equal to unity when the condition in parentheses is satisfied, gamma is the threshold parameter, and phi_1 and phi_2 are the regime-specific slopes of AABKI on growth. Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e predicts that phi_2\u0026thinsp;\u0026gt;\u0026thinsp;phi_1\u0026thinsp;\u0026gt;\u0026thinsp;0, implying that above-threshold regulatory environments amplify the growth returns to AI banking knowledge adoption.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Empirical Results","content":"\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e6.1 Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e: AI Banking Knowledge and Credit Depth\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents estimation results for Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e, with domestic credit as a percentage of GDP as the dependent variable. Column (1) reports pooled OLS, column (2) the fixed-effects estimator, column (3) random-effects, and column (4) the two-step system GMM estimator.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAABKI and Domestic Credit Depth (Dependent Variable: CREDIT/GDP)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOLS (1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFE (2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRE (3)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSys-GMM (4)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.412***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.361***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.387***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.347***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.089)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.097)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.091)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.108)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLog GDP per capita\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8.234***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.891**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.432***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5.987**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1.234)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2.341)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.678)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(2.543)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInflation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.312**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.287**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.298**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.341**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.134)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.143)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.138)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.152)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrade openness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.187***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.143**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.165***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.132**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.056)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.067)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.061)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.071)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBanking concentration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-12.34**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-10.87**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-11.56**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-9.43*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(4.321)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(4.876)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(4.543)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(5.123)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWGI governance index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.654***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.234**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.987***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5.876**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2.134)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2.543)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(2.341)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(2.765)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLagged CREDIT/GDP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.412***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.089)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared (within)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.534\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.612\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.587\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHansen J-test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.423\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAR(2) test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.287\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eNote: Robust standard errors in parentheses. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. The Hausman test strongly favours fixed effects over random effects (chi-squared\u0026thinsp;=\u0026thinsp;24.32, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). System GMM instruments include second and third lags of AABKI and lagged differences of control variables. Hansen J-test null: instruments are valid. AR(2) test null: no second-order serial correlation in differences.\u003c/em\u003e \u003c/p\u003e \u003cp\u003eThe AABKI index attracts a positive, statistically significant coefficient across all four specifications, confirming Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The preferred system GMM estimate (column 4) implies that a one standard deviation increase in the AABKI index is associated with an increase of 0.347 percentage points in the credit-to-GDP ratio, controlling for all conventional financial development determinants and addressing reverse causality. The Hansen J-test p-value of 0.423 comfortably exceeds the 0.10 threshold for instrument validity, and the AR(2) p-value of 0.287 confirms the absence of problematic serial correlation in residuals. The coefficient on the lagged dependent variable (0.412) indicates moderate persistence in credit depth, consistent with the gradual adjustment dynamics characteristic of financial sector development.\u003c/p\u003e \u003cp\u003eNotably, the coefficient on the AABKI index declines modestly but not dramatically as estimation moves from OLS to the GMM specification (from 0.412 to 0.347), suggesting that while some upward bias from endogeneity is present, it is not the dominant driver of the estimated relationship. The coefficient remains significant at the one percent level throughout. These findings establish, for the first time in the South Asian empirical literature, that AI-mediated banking knowledge constitutes a significant autonomous determinant of credit market depth.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e6.2 Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e: Heterogeneous Effects on Financial Inclusion\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents results for Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e, examining the heterogeneous effects of AI banking knowledge on financial inclusion. The full sample estimate in column (1) is followed by subgroup analyses for lower-income SAARC members (Bangladesh, Nepal, Bhutan, Afghanistan) in column (2) and higher-income members (India, Pakistan, Sri Lanka, Maldives) in column (3). Column (4) reports the interaction specification that formally tests for differential effects.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eHeterogeneous Effects of AABKI on Financial Inclusion (System GMM)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFull sample (1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLower-income (2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHigher-income (3)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eInteraction (4)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.289***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.412***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.198**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.201**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.076)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.098)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.087)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.089)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI x Lower-income\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.208**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.098)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI x Rural share\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.312***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.087)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLog GDP per capita\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.543***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.612***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.487***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.521***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.134)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.178)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.156)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.143)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGovernment expenditure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.023**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.031**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.018*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.024**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.011)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.014)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.010)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.011)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMobile penetration rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.178***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.234***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.134**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.187***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.045)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.067)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.056)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.047)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHansen J-test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.387\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.356\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.401\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAR(2) test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.312\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.278\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.345\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.298\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eNote: Robust standard errors in parentheses. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. The lower-income group comprises Bangladesh, Nepal, Bhutan, and Afghanistan; the higher-income group comprises India, Pakistan, Sri Lanka, and the Maldives, as classified by World Bank income categories. Rural share is the proportion of population residing outside urban agglomerations (WDI).\u003c/em\u003e \u003c/p\u003e \u003cp\u003eThe results in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provide compelling support for Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The AABKI coefficient for lower-income South Asian economies (0.412, column 2) is more than twice the magnitude of that for higher-income members (0.198, column 3), and this difference is statistically significant at the five percent level when formally tested through the interaction specification in column (4). The interaction term between AABKI and rural population share (0.312, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) reveals that the financial inclusion benefits of AI banking knowledge are particularly pronounced in more rurally concentrated economies, consistent with the theoretical prediction that AI-enabled remote credit assessment and mobile banking expansion overcome the geographic barriers to financial access that have traditionally disadvantaged rural populations.\u003c/p\u003e \u003cp\u003eThe finding that lower-income SAARC members derive significantly larger financial inclusion gains from AI banking knowledge adoption than higher-income members implies that, under appropriate conditions of technology transfer and regulatory support, AI constitutes a genuine convergence mechanism within the South Asian financial development landscape. This represents a novel empirical finding with direct implications for regional financial integration policy under SAARC and Bay of Bengal Initiative for Multi-Sectoral Technical and Economic Cooperation (BIMSTEC) frameworks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e6.3 Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e: Threshold Effects and Long-Run Growth\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the results of the panel threshold regression and system GMM growth analysis for Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The threshold test is reported in Panel A, and the regime-specific coefficient estimates in Panel B.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThreshold Regression Results \u0026mdash; AABKI, Regulatory Quality, and Economic Growth Panel A: Threshold Test Statistics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThreshold Test Statistic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValue\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThreshold variable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWGI governance index\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEstimated threshold (gamma*)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.621\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e95% confidence interval\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e[0.534, 0.712]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF-statistic (linearity test)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18.43***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBootstrap p-value (1,000 replications)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.006\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProportion of observations below threshold\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63.7%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProportion of observations above threshold\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36.3%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003ePanel B: Regime-Specific System GMM Estimates\u003c/b\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBelow threshold (1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAbove threshold (2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDifference (3)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAABKI index (phi_1, phi_2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.089*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.341***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.252**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.052)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.078)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.109)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLagged GROWTH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.312***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.287***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.078)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.071)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLog GDP per capita\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.312**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.487***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.143)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.156)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInflation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.087***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.123***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.023)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.031)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrade openness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.043**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.078***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.019)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.024)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGovernment expenditure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.031**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.054***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.014)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.018)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e107\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e168 total\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHansen J-test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAR(2) test (p-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.276\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.301\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eNote: Robust standard errors in parentheses. *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.10. The threshold is estimated by minimising the concentrated sum of squared residuals over the admissible parameter space. Bootstrap critical values are based on 1,000 replications. Column (3) in Panel B reports the estimated difference between the above-threshold and below-threshold AABKI coefficients with bootstrapped standard error. Below-threshold observations: n\u0026thinsp;=\u0026thinsp;107; above-threshold observations: n\u0026thinsp;=\u0026thinsp;61.\u003c/em\u003e \u003c/p\u003e \u003cp\u003eThe threshold regression results in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e constitute the most novel empirical contribution of this study. The F-statistic for the null hypothesis of linearity is 18.43, with a bootstrapped p-value of 0.006, strongly rejecting the linear specification in favour of a threshold model. The estimated threshold value of gamma* = 0.621 on the WGI governance index scale implies that 63.7 percent of country-year observations in the sample fall below the threshold, a finding with sobering implications for the majority of South Asian economies that currently operate in the below-threshold regime.\u003c/p\u003e \u003cp\u003eIn the below-threshold regime, the AABKI coefficient on growth is a modest and only marginally significant 0.089 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.10), suggesting that AI banking knowledge adoption generates limited growth dividends in environments characterised by weak regulatory institutions, enforcement capacity, or legal infrastructure. By contrast, in the above-threshold regime, the AABKI coefficient rises to 0.341 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), more than three times the below-threshold magnitude. The difference of 0.252 between the two regime coefficients is statistically significant at the five percent level, confirming the presence of genuine nonlinear threshold dynamics.\u003c/p\u003e \u003cp\u003eThese results support Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e in its entirety and reveal a bifurcated development path for South Asian economies. Countries that have built sufficient regulatory institutional capacity, approximated by a WGI governance score exceeding 0.621, are positioned to realise substantially larger growth returns from AI banking knowledge adoption than those that have not yet crossed this threshold. This finding implies that regulatory institution building and AI banking knowledge development are strategic complements, not substitutes, in the development policy portfolio.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e6.4 Panel Cointegration and Long-Run Relationships\u003c/h2\u003e \u003cp\u003eThe Pesaran (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2004\u003c/span\u003e) CD test statistic of 14.23 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) confirms significant cross-sectional dependence among South Asian economies, consistent with regional economic interdependence and common external shocks. Reliance is therefore placed on the Westerlund (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) error-correction cointegration tests with bootstrapped p-values. The Ga and Pa panel statistics reject the null of no cointegration at the five percent level (p\u0026thinsp;=\u0026thinsp;0.034 and 0.041 respectively), while the Gt and Pt group-mean statistics reject at the ten percent level. These results establish the existence of a long-run cointegrating relationship between AABKI, financial development, and economic growth, confirming that the estimated effects are not spurious artefacts of common trends.\u003c/p\u003e \u003c/div\u003e"},{"header":"7. Robustness Checks and Sensitivity Analysis","content":"\u003cp\u003eA comprehensive battery of robustness checks is conducted to assess the sensitivity of the principal findings to alternative methodological choices. First, the PCA-constructed AABKI index is replaced with an equal-weighted arithmetic mean of its five constituents and all specifications are re-estimated. The coefficient magnitudes change modestly (within ten percent of the baseline estimates) and all key results retain their sign and significance, confirming that the findings are not artefacts of the PCA weighting methodology.\u003c/p\u003e \u003cp\u003eSecond, potential sample selection concerns arising from Afghanistan's limited data availability are addressed by re-estimating all specifications on the seven-country sample excluding Afghanistan. Results are qualitatively unchanged, with Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e coefficients falling within the 95 percent confidence intervals of the full-sample estimates. Third, the influence of outliers is tested using the studentised residual deletion approach, and no single country-year observation is found to drive the key results.\u003c/p\u003e \u003cp\u003eFourth, the concern is addressed that the mobile money variable, which forms part of the AABKI construction, is mechanically correlated with the financial inclusion dependent variable in the Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e specifications. Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e is re-estimated using an AABKI index constructed without the mobile money constituent (a four-component index explaining 58.1 percent of common variance), and the core heterogeneity result is preserved: the coefficient for lower-income economies remains significantly larger than for higher-income economies.\u003c/p\u003e \u003cp\u003eFifth, the five-year forward moving average growth measure is substituted with the three-year average and the raw annual growth rate as alternative dependent variable specifications for Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The threshold estimate migrates slightly across specifications (ranging from 0.598 to 0.643 on the WGI scale) but remains statistically significant, and the regime-specific coefficient differential retains its direction and approximate magnitude. Sixth, the sensitivity of the threshold result to the choice of threshold variable is explored by replacing the WGI governance index with the WGI regulatory quality sub-index and the IMF Financial Sector Assessment Programme composite score. In both cases, a significant threshold is detected, and the qualitative pattern of above-threshold exceeding below-threshold AABKI growth effects is confirmed.\u003c/p\u003e"},{"header":"8. Policy Implications","content":"\u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e8.1 Central Bank and Banking Regulator Policy\u003c/h2\u003e \u003cp\u003eThe empirical results carry immediate and operationally specific implications for South Asian central banks and banking regulators. The finding that AI banking knowledge intensity exerts a significant positive effect on credit depth (Hypothesis \u003cspan refid=\"FPar4\" class=\"InternalRef\"\u003e1\u003c/span\u003e) implies that regulatory frameworks governing AI adoption in banking should be designed with awareness of this credit expansion channel. Overly restrictive or ambiguous AI governance regimes risk suppressing a genuine driver of financial deepening. Central banks in Bangladesh, Nepal, and Pakistan in particular, which cluster below the regulatory quality threshold identified in Hypothesis \u003cspan refid=\"FPar6\" class=\"InternalRef\"\u003e3\u003c/span\u003e, should prioritise building institutional capacity that would advance their governance scores beyond the 0.621 threshold, as this transition amplifies the growth returns to AI banking knowledge investments already occurring within their financial sectors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e8.2 SAARC-Level Financial Integration Strategy\u003c/h2\u003e \u003cp\u003eThe heterogeneity result in Hypothesis \u003cspan refid=\"FPar5\" class=\"InternalRef\"\u003e2\u003c/span\u003e carries direct implications for regional financial integration policy. The finding that lower-income SAARC economies derive significantly larger financial inclusion gains from AI banking knowledge adoption suggests that targeted technology transfer and knowledge diffusion programmes, operating under a regional framework, could accelerate financial inclusion convergence across the SAARC membership. A SAARC AI Banking Knowledge Exchange Facility, modelled partly on the SAARC Finance Ministers and Central Bank Governors framework but specifically focused on AI capability building, could systematically disseminate best practices from more advanced adopters such as India and Sri Lanka to less advanced members.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e8.3 International Development Finance\u003c/h2\u003e \u003cp\u003eFor multilateral development institutions, including the Asian Development Bank, the World Bank Group, and the International Monetary Fund, the empirical results provide robust justification for augmenting technical assistance and concessional financing directed at AI capacity building within South Asian banking systems. The threshold result implies that such investments will yield the greatest growth returns in economies that have already attained baseline regulatory institutional quality, suggesting a sequencing logic for development finance allocation: regulatory capacity building should precede or accompany AI banking knowledge investment in the weakest institutional environments.\u003c/p\u003e \u003c/div\u003e"},{"header":"9. Conclusion","content":"\u003cp\u003eThis paper presents the first systematic panel econometric investigation into the relationship between AI-augmented banking knowledge intensity and financial and economic development outcomes across the full South Asian region. Motivated by theoretical considerations derived from the knowledge production function for AI-augmented banking and the endogenous growth literature, and drawing exclusively on publicly disclosed macroeconomic, financial sector, and technology adoption data spanning 2000 to 2023, a novel composite AABKI index is constructed and subjected to a comprehensive battery of econometric tests.\u003c/p\u003e \u003cp\u003eThree hypotheses are advanced and confirmed. First, AI-augmented banking knowledge intensity is an autonomous, positive, and statistically significant determinant of domestic credit depth, with a system GMM coefficient of 0.347 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) that survives comprehensive endogeneity controls. Second, the financial inclusion benefits of AI banking knowledge are significantly heterogeneous, being more than twice as large for lower-income South Asian economies as for higher-income ones, and particularly pronounced in rurally concentrated economies, establishing AI banking knowledge as a novel financial convergence mechanism. Third, the growth returns to AI banking knowledge adoption are subject to nonlinear threshold dynamics conditioned on regulatory institutional quality, with a precisely estimated threshold at a WGI governance score of 0.621, below which AI banking knowledge exerts modest and only marginally significant growth effects, and above which its growth impact rises to more than three times that magnitude.\u003c/p\u003e \u003cp\u003eThese findings generate a coherent and policy-relevant narrative for South Asian development strategy. The AI-mediated banking knowledge frontier represents not merely a technological novelty but a structurally important new input into the financial development process, one whose economic effects are measurable, substantial, and conditioned by institutional context in precisely the ways that theory would predict. The challenge for policymakers in the region is to advance both dimensions simultaneously: building the regulatory institutional quality that unlocks the full growth potential of AI banking knowledge, while investing in the knowledge infrastructure that makes that potential realisable.\u003c/p\u003e \u003cp\u003eFuture research directions suggested by this study include: micro-level analysis of individual bank efficiency gains from AI knowledge tool adoption; examination of the gender dimension of AI-enabled financial inclusion using disaggregated Global Findex data; and investigation of cross-border AI banking knowledge spillovers through the SAARC correspondent banking network. As more granular and longer panel datasets on AI adoption in South Asian banking become available through central bank disclosures, the empirical precision of the estimates presented here can be further refined.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eEthics Approval and Consent to Participate\u003c/h2\u003e \u003cp\u003eNot applicable. This study is based entirely on secondary data analysis of publicly available information, including audited financial statements, regulatory filings, annual reports, and publicly disclosed documents from Bangladesh Bank and commercial banks. No primary data collection involving human participants was conducted. No human subjects, human data, human tissue, or animals were involved in this research.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for Publication\u003c/strong\u003e \u003cp\u003eNot applicable. This manuscript does not contain data from any individual person. All data presented are aggregated at the institutional level and derived from publicly available sources. No individual person's details, images, videos, or case reports are included in this manuscript.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. The study was conducted independently without external financial support. No funding body was involved in the design of the study, collection, analysis, and interpretation of data, or in writing the manuscript.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCompeting Interests\u003c/strong\u003e \u003cp\u003eThe author declares no competing interests. The author is employed by Mutual Trust Bank PLC; however, this research was conducted independently without involvement from the employer. The views expressed are those of the author alone and do not represent the official position of Mutual Trust Bank PLC or any other organisation.\u003c/p\u003e \u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAng JB (2008) A survey of recent developments in the literature of finance and growth. J Economic Surveys 22(3):536\u0026ndash;576\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud 58(2):277\u0026ndash;297\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsian Development Bank (2021) AI and Machine Learning for Financial Inclusion in Developing Asia. ADB Working Paper\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarth JR, Caprio G, Levine R (2004) Bank regulation and supervision: What works best? J Financ Intermediation 13(2):205\u0026ndash;248\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeck T, Senbet L, Simbanegavi W (2015) Financial inclusion and innovation in Africa: Entrepreneurship, ecosystems and economic development. Rev Dev Finance 5(2):55\u0026ndash;57\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerg T, Burg V, Gombovic A, Puri M (2020) On the rise of FinTechs: Credit scoring using digital footprints. Rev Financial Stud 33(7):2845\u0026ndash;2897\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhattacharya PC, Sivasubramanian MN (2003) Financial development and economic growth in India: 1970\u0026ndash;1971 to 1998\u0026ndash;1999. Appl Financ Econ 13(12):925\u0026ndash;929\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlundell R, Bond S (1998) Initial conditions and moment restrictions in dynamic panel data models. J Econ 87(1):115\u0026ndash;143\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClaessens S, Laeven L (2005) Financial dependence, banking sector competition, and economic growth. J Eur Econ Assoc 3(1):179\u0026ndash;207\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDelis MD, Papanikolaou NI (2009) Determinants of bank efficiency: Evidence from a semi-parametric methodology. Managerial Finance 35(3):260\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFuster A, Plosser M, Schnabl P, Vickery J (2022) The role of technology in mortgage lending. Rev Financial Stud 32(5):1854\u0026ndash;1899\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoldsmith RW (1969) Financial Structure and Development. Yale University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHansen BE (1999) Threshold effects in non-dynamic panels: Estimation, testing, and inference. J Econ 93(2):345\u0026ndash;368\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHassan MK, Sanchez B, Yu JS (2011) Financial development and economic growth: New evidence from panel data. Q Rev Econ Finance 51(1):88\u0026ndash;104\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInternational Finance Corporation (2019) Digital Financial Services in Bangladesh: Opportunities and Challenges. IFC Discussion Paper\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInternational Monetary Fund (2023) World Economic Outlook Database. IMF, Washington DC\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInternational Telecommunication Union (2023) World Telecommunication/ICT Indicators Database. ITU, Geneva\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJagtiani J, Lemieux C (2019) The roles of alternative data and machine learning in FinTech lending: Evidence from the LendingClub consumer platform. Financ Manage 48(4):1009\u0026ndash;1029\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKing RG, Levine R (1993) Finance and growth: Schumpeter might be right. Quart J Econ 108(3):717\u0026ndash;737\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKremer S, Bick A, Nautz D (2013) Inflation and growth: New evidence from a dynamic panel threshold analysis. Empirical Economics 44(2):861\u0026ndash;878\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLevine R, Loayza N, Beck T (2000) Financial intermediation and growth: Causality and causes. J Monet Econ 46(1):31\u0026ndash;77\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcKinnon RI (1973) Money and Capital in Economic Development. Brookings Institution\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNardo M, Saisana M, Saltelli A, Tarantola S, Hoffmann A, Giovannini E (2005) Handbook on Constructing Composite Indicators. OECD Statistics Working Paper\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedroni P (2004) Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Econom Theory 20(3):597\u0026ndash;625\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePesaran MH (2004) General diagnostic tests for cross-section dependence in panels. Cambridge Working Papers in Economics No. 0435\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePesaran MH (2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74(4):967\u0026ndash;1012\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaza SA, Shahzad SJH, Shahzad U (2020) Financial development and economic growth nexus in South Asia. Borsa Istanbul Rev 20(4):301\u0026ndash;314\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRomer PM (1990) Endogenous technological change. J Polit Econ 98(5):S71\u0026ndash;S102\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShaw ES (1973) Financial Deepening in Economic Development. Oxford University Press\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eValickova P, Havranek T, Horvath R (2015) Financial development and economic growth: A meta-analysis. J Economic Surveys 29(3):506\u0026ndash;526\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWesterlund J (2007) Testing for error correction in panel data. Oxf Bull Econ Stat 69(6):709\u0026ndash;748\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank (2021) Global Findex Database 2021: Measuring Financial Inclusion and the Fintech Revolution. World Bank, Washington DC\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank (2023) World Development Indicators. World Bank, Washington DC\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank (2023) Worldwide Governance Indicators. World Bank, Washington DC\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Artificial intelligence, banker knowledge, financial inclusion, panel data econometrics, GMM estimation, South Asia, credit intermediation, regulatory quality, threshold effects, SAARC","lastPublishedDoi":"10.21203/rs.3.rs-9127571/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9127571/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study constitutes the first systematic econometric investigation into whether the deployment of artificial intelligence for banker knowledge enhancement exerts a statistically significant, causal influence on credit intermediation efficiency, financial inclusion depth, and aggregate economic growth across the eight sovereign economies of South Asia. Drawing exclusively on publicly disclosed macroeconomic, financial sector, and technology adoption data spanning the period 2000 to 2023, a novel composite index of AI-augmented banking knowledge intensity (AABKI) is constructed and embedded within a panel data framework comprising fixed-effects, random-effects, and dynamic Generalised Method of Moments (GMM) estimators. Three foundational hypotheses are advanced and empirically evaluated: first, that AI-mediated knowledge diffusion within banking institutions constitutes an autonomous, hitherto unrecognised driver of credit-to-GDP ratio growth, independent of conventional financial development determinants; second, that the marginal effect of AI banking knowledge intensity on financial inclusion is heterogeneous across income strata and geographic regions within South Asia, thereby generating a novel distributional channel through which technology shapes economic convergence; and third, that the interaction between regulatory institutional quality and AI banking knowledge adoption generates nonlinear threshold effects on long-run economic growth that existing theoretical frameworks have not articulated. Empirical results confirm all three hypotheses at conventional significance levels. The AABKI index is positively and robustly associated with domestic credit depth (beta = 0.347, p \u0026lt; 0.01), financial inclusion breadth (beta = 0.289, p \u0026lt; 0.01), and real GDP per capita growth (beta = 0.214, p \u0026lt; 0.05), with threshold nonlinearities identified at a regulatory quality score of 0.62 on the World Governance Indicators scale. These findings carry profound implications for central bank policy, SAARC-level financial integration strategy, and the design of AI governance frameworks for emerging market banking systems.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJEL Classification: \u003c/strong\u003eG21, O33, O53, C23, G28, F36, O16\u003c/p\u003e","manuscriptTitle":"Artificial Intelligence as a Catalyst for Banker Knowledge Development: Econometric Evidence From South Asian Financial Markets And Economic Growth Dynamics","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-17 09:27:24","doi":"10.21203/rs.3.rs-9127571/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9a86c76e-3f3d-4958-8f5e-d4c73e573bb3","owner":[],"postedDate":"March 17th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64638942,"name":"Artificial Intelligence and Machine Learning"},{"id":64638943,"name":"Banking Law"},{"id":64638944,"name":"Finance"},{"id":64638945,"name":"Other Economics"},{"id":64638946,"name":"Macroeconomics"},{"id":64638947,"name":"International Business"}],"tags":[],"updatedAt":"2026-03-17T09:27:24+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-17 09:27:24","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9127571","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9127571","identity":"rs-9127571","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}