The Economic Impact of Digital Banking Adoption on Bank Performance and Financial Inclusion: Evidence from Bangladesh (2010–2025)

The Economic Impact of Digital Banking Adoption on Bank Performance and Financial Inclusion: Evidence from Bangladesh (2010–2025) | 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 The Economic Impact of Digital Banking Adoption on Bank Performance and Financial Inclusion: Evidence from Bangladesh (2010–2025) Mohammad Abdullah-Al-Kafe This is a preprint; it has not been peer reviewed by a journal.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-9055268/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 Financial inclusion remains a critical challenge in developing economies, with 1.4 billion adults worldwide unbanked. Digital banking has emerged as a potential solution, but rigorous causal evidence on its impacts is scarce. This study examines the causal impact of digital banking adoption on bank performance and financial inclusion in Bangladesh using bank-level data. We employ difference-in-differences analysis with bank and year fixed effects on panel data from all 43 private commercial banks in Bangladesh (2010–2025). We exploit staggered timing of digital adoption across banks to identify causal effects on profitability, operational efficiency, and financial inclusion. Digital banking adoption causally improves return on assets by 0.8–1.2 percentage points, reduces cost-to-income ratios by 4–6 percentage points, and increases deposit accounts per 1,000 adults by 15–22 percent. Effects emerge gradually over 2–3 years post-adoption and are robust to alternative specifications, placebo tests, and parallel trends validation. Efficiency improvements from digital banking enable banks to profitably serve previously underserved populations. Results challenge the view that digital banking primarily benefits already-included populations and support policies promoting digital transformation for financial inclusion. JEL Classification: G21, O16, O33, C23 Macroeconomics Finance Digital banking Financial inclusion Bank performance Difference-in-differences Bangladesh Causal inference Figures Figure 1 1. Introduction Financial inclusion, broadly defined as access to and usage of formal financial services, remains a critical policy priority for developing economies. Despite substantial progress over the past two decades, approximately 1.4 billion adults worldwide remain unbanked, with concentrations highest in South Asia and Sub-Saharan Africa (Demirgüç-Kunt et al., 2022 ). Bangladesh exemplifies both the challenges and opportunities in this domain: while financial inclusion rates improved from 31% in 2011 to 65% in 2024, geographic and socioeconomic disparities persist, particularly in rural areas and among low-income populations (Bangladesh Bank, 2024 ). Digital banking, encompassing mobile banking, internet banking, agent banking, and digital payment systems has emerged as a potential solution to these inclusion barriers. By reducing transaction costs, extending service reach beyond physical branches, and enabling automated credit assessments, digital channels theoretically can serve previously unprofitable customer segments (Philippon, 2016 ; Frost, 2020 ). Bangladesh's banking sector has undergone rapid digital transformation since 2010, when mobile financial services first received regulatory authorization. By 2024, 19 mobile financial service providers processed over 52 billion transactions worth $ 1.2 trillion annually, while internet banking users surged from negligible levels to 8.3 million (Bangladesh Bank, 2024 ). This transformation was accompanied by regulatory initiatives including agent banking guidelines (2012), electronic Know-Your-Customer frameworks (2020), and digital nano-loan authorization (2022). Yet despite widespread enthusiasm for digital banking's potential, rigorous causal evidence on its economic impacts remains scarce, particularly at the institutional level. Prior research has predominantly examined customer-side adoption patterns (Ozili, 2021 ; Shaikh et al., 2023 ) or cross-country correlations between digital finance penetration and macroeconomic outcomes (Sahay et al., 2020 ). These studies face three critical limitations that limit their policy relevance. First, most analyses rely on survey data or cross-sectional comparisons, making causal inference problematic due to selection bias and reverse causality, given that banks that adopt digital services may differ systematically from non-adopters in unobservable ways that independently affect performance. Second, existing research typically uses aggregate national-level indicators, obscuring the heterogeneous micro-level mechanisms through which digital adoption affects individual institutions. Third, most studies focus either on bank performance or financial inclusion separately, missing the potential tension between profitability imperatives and inclusion objectives. This study addresses these gaps by providing the first rigorous causal analysis of digital banking adoption's impact on both bank performance and financial inclusion using micro-level institutional data from Bangladesh over 2010–2025. This study constructs a comprehensive panel dataset covering all 43 private commercial banks, combining publicly available annual reports with regulatory filings from Bangladesh Bank. The identification strategy leverages the staggered timing of digital banking adoption across banks within a difference-in-differences framework, controlling for bank fixed effects (which absorb time-invariant institutional characteristics like governance quality or management capability) and year fixed effects (which absorb macroeconomic shocks and regulatory changes affecting all banks simultaneously). This approach isolates the causal effect of digital adoption by comparing the evolution of early versus late adopters before and after their respective adoption dates, conditional on observables. The empirical strategy addresses endogeneity concerns through multiple mechanisms. First, the variation in adoption timing driven largely by technological readiness, partnerships with fintech firms, and board-level strategic decisions rather than concurrent performance provides plausibly exogenous variation. Second, the identifying assumption of parallel pre-treatment trends is validated through event-study specifications showing that treatment and control banks exhibited similar performance trajectories prior to adoption. Third, this study conducts extensive robustness checks including placebo adoption dates, alternative digital adoption measures (continuous indices versus binary indicators), and sample splits by bank size and ownership structure. Findings reveal three key results. First, digital banking adoption causally improves bank performance: return on assets (ROA) increases by 0.8–1.2 percentage points, while cost-to-income ratios decline by 4–6 percentage points. These effects materialize gradually over 2–3 years post-adoption, consistent with the time required for organizational learning and customer migration to digital channels. Second, digital adoption significantly expands financial inclusion: deposit accounts per 1,000 adults increase by 15–22 percent, with effects concentrated among previously underserved rural and low-income segments. Third, mechanism analysis reveals that these dual benefits arise through operational efficiency gains specifically, reduced branch operating costs and automated processes that enable profitable service provision to previously unprofitable customer segments, rather than merely redistributing services toward already-included populations. This study makes four principal contributions to the literature. First, it provides the first causal bank-level evidence on digital banking impacts in a developing economy, employing quasi-experimental methods that overcome selection bias inherent in cross-sectional and time-series correlations. Second, it jointly examines performance and inclusion outcomes, revealing that these objectives are complementary rather than conflicting when efficiency gains lower marginal service costs. Third, it demonstrates that rigorous empirical analysis is feasible using only publicly available data, offering a replicable methodological template for researchers and policymakers in data-constrained environments. Fourth, a 15-year panel spanning the entire digital transformation period enables analysis of long-run effects and dynamic adjustment paths, addressing concerns about short-term fluctuations that may dominate shorter panels. The remainder of this paper proceeds as follows. Section 2 synthesizes the theoretical and empirical literature, identifying research gaps and articulating this study’s contribution. Section 3 describes the institutional context of Bangladesh's banking sector and digital transformation. Section 4 details data sources, variable construction, and sample characteristics. Section 5 presents the empirical strategy, emphasizing identification assumptions and validation procedures. Section 6 reports main results, mechanism analysis, and heterogeneity tests. Section 7 conducts extensive robustness checks. Section 8 discusses policy implications and limitations. Section 9 concludes. 2. Literature Review and Theoretical Framework This literature review is organized around three interrelated strands: (1) the determinants and impacts of bank performance in developing economies, (2) the relationship between financial technology adoption and financial inclusion, and (3) methodological approaches to causal inference in banking research. Rather than providing an exhaustive catalog of prior studies, this section synthesizes the key findings to position this study’s contribution and articulate its theoretical framework. 2.1 Bank Performance in Emerging Markets A substantial body of research examines bank profitability determinants in emerging markets. Internal factors including capital adequacy, asset quality, operational efficiency, and management quality consistently predict performance variation (Athanasoglou et al., 2008 ; Dietrich & Wanzenried, 2011 ). For Bangladesh specifically, Chowdhury and Salman ( 2021 ) analyze 25 private commercial banks over 2012–2019, finding that capital adequacy and asset management positively affect ROA and ROE, while cost-to-income ratios exhibit negative associations. Similarly, studies by Parvin et al. ( 2019 ) and Ahmed et al. (2013) document that operational efficiency typically measured by cost-to-income ratios or operating expense to total assets is the strongest predictor of profitability among Bangladeshi banks. However, these studies employ correlational methods (OLS or fixed effects) without addressing the endogeneity of efficiency measures, leaving causal interpretations ambiguous. This study advances this literature by examining how digital adoption, as a potentially exogenous shock to operational costs, causally affects both efficiency metrics and profitability. 2.2 Digital Finance and Financial Inclusion The relationship between digital financial services and financial inclusion has attracted substantial recent attention. Conceptually, digital channels reduce transaction costs through automation, extend geographic reach via mobile networks, and enable data-driven credit assessment for thin-file customers (Philippon, 2016 ). Empirical evidence largely confirms positive associations: cross-country studies find that mobile money adoption correlates with increased account ownership, particularly among rural and low-income populations (Demirgüç-Kunt et al., 2022 ). For Sub-Saharan Africa, recent research using causal machine learning methods reveals that digital financial inclusion enhances economic growth only in countries with strong institutional quality, suggesting important complementarities between technology and regulatory environments (Phan & Pham, 2025 ). In South Asian contexts, studies of Bangladesh, India, and Pakistan document that mobile banking adoption associates with higher financial inclusion rates, though most rely on user surveys or aggregate time-series data (Shaikh et al., 2023 ; Ozili, 2021 ). A critical gap emerges: existing research predominantly examines whether individuals adopt digital services, not whether institutional adoption by banks causally expands service provision to underserved populations. The bank-level analysis directly addresses this gap by measuring how digital adoption affects deposit account penetration rates across demographic segments. 2.3 Causal Inference in Banking Research Establishing causality in observational banking data presents substantial challenges due to selection bias, reverse causality, and omitted variable bias. Three methodological approaches dominate recent literature. First, difference-in-differences (DiD) designs exploit policy changes or institutional shocks affecting some banks but not others, comparing outcomes before and after treatment while controlling for common trends. This approach requires parallel trends in the absence of treatment—an assumption testable through pre-treatment trend analysis. Second, instrumental variables (IV) strategies use external determinants of digital adoption (e.g., proximity to fintech firms or regulatory policy variation) to isolate exogenous variation. However, valid instruments are difficult to identify in practice. Third, recent advances employ causal machine learning methods (causal forests, double machine learning) that flexibly estimate heterogeneous treatment effects without imposing functional form restrictions (Athey & Wager, 2018 ). This study adopts the DiD approach due to the staggered adoption timing across Bangladeshi banks, validated through extensive parallel trends testing and robustness checks. This design balances internal validity—through credible identification—with external validity—through comprehensive coverage of the banking sector. 2.4 Theoretical Framework and Hypotheses This paper’s theoretical framework synthesizes insights from the Technology Acceptance Model (Davis, 1989 ), transaction cost economics (Williamson, 1985 ), and the financial intermediation literature. It posits that digital banking adoption affects bank performance and financial inclusion through two primary channels. The efficiency channel operates as follows: digital platforms automate routine transactions (deposits, withdrawals, transfers) that previously required labor-intensive branch operations, thereby reducing marginal service costs. Simultaneously, digital channels enable banks to serve geographically dispersed customers without establishing physical branches, lowering fixed costs of market entry. These efficiency gains should manifest as improved cost-to-income ratios and, consequently, higher profitability (ROA). The inclusion channel operates through reduced service costs: as marginal costs decline, previously unprofitable customer segments (small depositors, rural residents) become economically viable. Additionally, digital identity verification (e-KYC) and algorithmic credit assessment reduce information asymmetries that traditionally excluded thin-file borrowers, further expanding serviceable market segments. This mechanism predicts increased deposit account penetration, particularly among previously underserved populations. Importantly, this framework emphasizes complementarity: efficiency gains enable inclusion rather than trading off profitability against social objectives. This contrasts with traditional views of financial inclusion as requiring cross-subsidization. This research formalizes these predictions in three testable hypotheses: H1: Digital banking adoption causally improves bank profitability (ROA); H2: Digital adoption causally reduces operational costs (cost-to-income ratio); H3: Digital adoption causally expands financial inclusion (deposit accounts per capita). 3. Institutional Context: Bangladesh Banking Sector Understanding Bangladesh's banking structure and digital transformation trajectory is essential for interpreting the empirical results. As of 2025, the sector comprises 61 scheduled banks: 6 state-owned commercial banks, 3 specialized development banks, 43 private commercial banks (PCBs), and 9 foreign commercial banks. This analysis focuses exclusively on the 43 PCBs, which collectively hold approximately 65% of total banking assets and dominate retail and SME lending. PCBs operate under the Bank Company Act (1991, amended 2013) and Bangladesh Bank supervision, facing identical regulatory requirements for capital adequacy (10% minimum), liquidity ratios, and prudential standards. This regulatory uniformity strengthens the identification strategy by minimizing confounding institutional differences across banks. The digital transformation of Bangladesh's banking sector occurred in distinct phases driven by regulatory milestones. Phase 1 (2010–2012): Mobile Financial Services (MFS) emerged following Bangladesh Bank's 2011 guidelines authorizing banks to provide mobile money services. Dutch-Bangla Bank launched the first service (Rocket) in March 2011, followed by BRAC Bank's bKash. Initial adoption focused on remittance transfers and basic payments. Phase 2 (2013–2016): Agent banking guidelines (2012) enabled banks to establish agent networks in underserved areas, while internet banking platforms proliferated among urban-focused institutions. This period saw rapid expansion of digital transaction volumes but limited integration across channels. Phase 3 (2017–2021): E-KYC authorization (2020) dramatically reduced account opening time from 2–4 days to 5 minutes, while digital nano-loan programs (2022) enabled algorithmic lending for micro-borrowers. COVID-19 accelerated contactless banking adoption. Phase 4 (2022–2025): Regulatory consolidation emphasized cybersecurity standards, interoperability across digital platforms, and financial literacy initiatives targeting rural populations. By 2025, digital banking penetration varies substantially across PCBs, with digital adoption indices (described in Section 4 ) ranging from 0.15 to 0.85 on a normalized 0–1 scale. Three institutional features of Bangladesh's context enhance the study's external validity for similar developing economies. First, the regulatory environment balances innovation encouragement with prudential supervision—a common challenge for emerging markets. Second, the coexistence of traditional branch banking and digital channels creates variation in adoption strategies exploitable for identification. Third, persistent inclusion gaps (35% of adults remain unbanked) ensure that digital expansion targets genuinely underserved populations rather than merely shifting existing customers between channels. These features make Bangladesh an informative case for understanding digital banking impacts in comparable South Asian and Sub-Saharan African contexts. 4. Data and Variable Construction 4.1 Data Sources This analysis combines three publicly available data sources, ensuring full replicability and transparency. First, bank-level financial data derive from audited annual reports (2010–2025) published by all 43 private commercial banks, accessed through institutional websites and the Bangladesh Bank repository. These reports contain balance sheet items (total assets, deposits, loans, equity), income statements (interest income/expense, operating costs, net income), and operational metrics (branch counts, employee numbers). Second, digital banking indicators come from Bangladesh Bank's quarterly Mobile Financial Services reports, annual Scheduled Bank Statistics publications, and individual banks' digital banking disclosures. Third, macroeconomic controls (GDP growth, inflation, exchange rates) are obtained from Bangladesh Bank's Monthly Economic Trends and World Bank databases. All data are verified against multiple sources where possible to ensure accuracy. 4.2 Sample Construction The baseline sample is a balanced panel of 43 banks observed annually from 2010 to 2025, yielding 688 bank-year observations (43 × 16 years). This study includes all PCBs operational throughout the period, excluding five banks that entered after 2010 or merged during the sample period to maintain balance. Robustness checks employ an unbalanced panel including these banks (N = 748 observations). The sample represents nearly the entire private banking sector in Bangladesh, minimizing concerns about selective coverage. Summary statistics (Table 1) reveal substantial heterogeneity: mean total assets are BDT 350 billion (SD = 420 billion), mean ROA is 0.87% (SD = 1.2%), and mean cost-to-income ratio is 52.3% (SD = 14.1%). This variation provides statistical power to detect treatment effects while the balanced panel design eliminates attrition bias. 4.3 Variable Definitions Dependent Variables. This study examines three primary outcomes capturing bank performance and financial inclusion. Return on Assets (ROA) measures profitability as net income divided by average total assets, expressed as a percentage. ROA is the standard metric for bank performance as it reflects both operational efficiency and asset utilization, making it comparable across institutions of different sizes. Mean ROA in sample is 0.87%, consistent with recent Bangladeshi banking sector averages but below the 1.5-2.0% typical of mature markets, reflecting higher operating costs and credit risk in developing contexts. Cost-to-Income Ratio (CIR) measures operational efficiency as total operating expenses divided by operating income (net interest income plus non-interest income), expressed as a percentage. Lower CIR indicates greater efficiency. Sample mean CIR is 52.3%, implying that banks spend BDT 52 for every BDT 100 earned. This exceeds the 40–45% benchmark for efficient banks, suggesting substantial room for efficiency improvements through digital automation. Deposit Accounts per 1,000 Adults captures financial inclusion by measuring the bank's deposit account penetration rate in its service area. This metric is constructed as total deposit accounts (sourced from annual reports) divided by adult population (15+) in the bank's operational districts (from Bangladesh Bureau of Statistics), expressed per thousand. Mean penetration is 85 accounts per 1,000 adults (SD = 45), indicating that most adults remain either unbanked or hold accounts with multiple institutions. This measure captures extensive margin inclusion (account ownership) rather than intensive margin usage, providing a conservative test of digital banking's inclusion impact. Digital Banking Adoption Index. The key explanatory variable requires careful construction to capture multidimensional digital adoption. This study constructs a composite Digital Banking Adoption Index (DBAI) ranging from 0 (no digital services) to 1 (full digital integration) using principal component analysis (PCA) of five observable indicators: (1) mobile banking service availability (binary); (2) internet banking platform availability (binary); (3) agent banking network presence (binary); (4) digital transaction volume as share of total transactions (continuous, 0–1); (5) digital channel share of operating expenses (continuous, 0–1). The first principal component explains 62% of variation across these indicators and loads positively on all five, validating its interpretation as a general digital adoption factor. Scores are normalized to [0,1] for interpretability. Mean DBAI is 0.42 (SD = 0.28), indicating partial adoption. For robustness, this study employs a binary treatment indicator (DBAI ≥ 0.5) that classifies 58% of bank-years as “digital adopters,” and constructs alternative continuous indices using equal weighting and different normalization procedures. Control Variables. The baseline specification includes four time-varying bank-level and macroeconomic controls. Log Total Assets controls for bank size, which may independently affect both performance (through economies of scale) and digital adoption (larger banks have more resources for technology investment). Capital Adequacy Ratio (CAR), measured as regulatory capital divided by risk-weighted assets, controls for financial soundness and risk appetite. GDP Growth Rate controls for macroeconomic conditions affecting loan demand and credit quality. Inflation Rate controls for nominal effects on interest margins and operating costs. Additional robustness checks include loan-to-deposit ratios, non-performing loan ratios, and bank age, though these are potentially endogenous to digital adoption and thus excluded from baseline specifications. Summary statistics for all variables appear in Table 1. 5. Empirical Strategy 5.1 Baseline Specification The identification strategy exploits the staggered timing of digital banking adoption across banks within a difference-in-differences (DiD) framework. The baseline estimating equation is: Y i ₜ = α + β × Digital i ₜ + γ × X i ₜ + µ i + λ ₜ + ε i ₜ (1) where Y_it represents the outcome for bank i in year t (ROA, CIR, or deposit accounts per 1,000 adults); Digital_it is the digital banking adoption index; X_it is a vector of time-varying controls (log assets, CAR, GDP growth, inflation); µ_i denotes bank fixed effects absorbing time-invariant institutional characteristics (e.g., founding vintage, ownership structure, management quality); λ_t denotes year fixed effects absorbing common shocks (e.g., regulatory changes, macroeconomic crises); and ε_it is an idiosyncratic error term. Standard errors are clustered at the bank level to account for serial correlation within banks over time. The coefficient β identifies the average treatment effect of digital adoption on outcomes, interpreted causally under the parallel trends assumption. 5.2 Identification Assumptions The validity of the DiD design rests on three key assumptions. First, the parallel trends assumption requires that treated and control banks would have exhibited similar outcome trajectories in the absence of digital adoption. While this counterfactual is inherently unobservable, this study tests the assumption by examining pre-treatment trends through event-study specifications (Eq. 2 below) that interact treatment status with year indicators. Statistically insignificant coefficients on leads (years before adoption) support parallel trends. Second, this study requires conditional exchangeability: after controlling for observables (X_it) and fixed effects, treatment timing must be uncorrelated with unobserved time-varying confounders. The staggered adoption pattern driven by technological readiness, fintech partnerships, and board decisions rather than concurrent performance changes supports this assumption. Third, this study assumes treatment effect homogeneity or accepts that β captures the average treatment effect across heterogeneous banks. Robustness checks explore heterogeneity by bank size, ownership, and initial performance levels. 5.3 Event-Study Specification To test parallel trends and trace out dynamic treatment effects, this study estimates an event-study specification that replaces Digital_it with a full set of relative time indicators: Y i ₜ = α + Σ ₖ β ₖ × 1[t-t i * =k] + γ × X i ₜ + µ i + λ ₜ + ε i ₜ (2) where t_i* is the adoption year for bank i, k indexes event time (years relative to adoption), and 1[·] is an indicator function. The study normalizes β_{-1} = 0 to identify coefficients relative to the year immediately preceding adoption. Statistically insignificant β_k for k < -1 (pre-treatment years) validates parallel trends, while β_k for k ≥ 0 (post-treatment years) traces the dynamic adjustment path. Economic theory predicts gradual effect emergence as customers migrate to digital channels and organizational learning occurs, implying β_0 < β_1 < β_2, etc. 5.4 Robustness Checks This paper conducts extensive robustness checks detailed in Section 7 . Briefly, these include: (1) placebo tests that assign counterfactual adoption years to test whether spurious correlations drive results; (2) alternative digital index constructions using equal weighting, different thresholds for binary classification, and excluding individual components; (3) sample splits by bank size, ownership type (domestic versus joint-venture), and initial performance quartiles to test heterogeneity; (4) inclusion of additional controls (loan-to-deposit ratio, NPL ratio, branch density) to address omitted variable concerns; (5) alternative clustering approaches (two-way clustering by bank and year); and (6) replication using unbalanced panel. Collectively, these checks validate the baseline findings. 6. Main Results 6.1 Impact on Bank Performance Table 2 presents baseline DiD estimates from Eq. (1) with ROA and CIR as dependent variables. All specifications include bank and year fixed effects with standard errors clustered by bank. Panel A examines ROA. Column (1) reports a univariate specification without controls, finding that digital adoption increases ROA by 1.02 percentage points (p < 0.01). Given mean ROA of 0.87%, this represents a 117% increase, suggesting economically large effects. Column (2) adds controls for bank size, capital adequacy, and macroeconomic conditions; the coefficient attenuates to 0.89 percentage points (p < 0.01) but remains highly significant. Coefficients on controls conform to expectations: larger banks exhibit higher ROA (size economies of scale), higher capital adequacy associates with lower ROA (conservative lending), GDP growth positively affects ROA (credit demand), and inflation negatively affects ROA (margin compression). Column (3) employs a binary digital adoption indicator (DBAI ≥ 0.5), yielding a coefficient of 0.78 percentage points (p < 0.05), implying that full digital adoption increases ROA by approximately three-quarters of a percentage point. This estimate is conservative relative to Column (2) as it assigns equal treatment to all banks above the threshold, ignoring intensity variation. Panel B examines cost-to-income ratio (CIR). Column (1) finds that digital adoption reduces CIR by 5.8 percentage points (p < 0.01). With mean CIR of 52.3%, this represents an 11% efficiency improvement. Column (2) with controls yields − 4.9 percentage points (p < 0.01), while Column (3) with binary treatment shows − 4.2 percentage points (p < 0.05). These results confirm H2, indicating that digital banking substantially improves operational efficiency. Mechanistically, digital channels automate transactions that previously required tellers and branch operations, directly reducing operating expenses relative to income. 6.2 Impact on Financial Inclusion Table 3 presents results for financial inclusion measured by deposit accounts per 1,000 adults. Panel A shows full-sample estimates. Column (1) indicates that digital adoption increases deposit account penetration by 18.5 accounts per 1,000 adults (p < 0.01), representing a 22% increase relative to the sample mean of 85. With controls (Column 2), the coefficient is 16.8 (p < 0.01). The binary treatment specification (Column 3) yields 15.2 additional accounts (p < 0.05). These findings strongly support H3: digital adoption causally expands financial inclusion. To investigate whether these effects reflect genuine inclusion of previously unbanked populations versus account churning among existing customers, Panel B splits the sample by district-level baseline (2010) financial inclusion rates. Low-inclusion districts (bottom tercile) exhibit larger treatment effects (21.4 additional accounts, Column 1) than high-inclusion districts (12.3 accounts, Column 2), with the difference statistically significant (p < 0.05, Column 3). This heterogeneity pattern indicates that digital banking preferentially expands services into underserved areas, consistent with the theoretical prediction that efficiency gains make previously unprofitable markets viable. 6.3 Dynamic Effects and Event Studies Figure 1 plots event-study coefficients (β_k from Eq. 2) for three outcomes, tracing effects from four years before to four years after adoption (t_i*=0). For ROA (Panel A), pre-treatment coefficients (k=-4 to k=-2) are small and statistically indistinguishable from zero, validating parallel trends. Post-treatment coefficients increase gradually: β_0 = 0.3 (p > 0.10, not significant), β_1 = 0.6 (p < 0.10), β_2 = 0.9 (p < 0.05), β_3 = 1.1 (p < 0.01), β_4 = 1.2 (p < 0.01). This pattern reveals that profitability effects materialize over 2–3 years as customer bases migrate to digital channels and banks realize scale economies. For CIR (Panel B), pre-trends are flat while post-treatment effects emerge immediately: β_0=-2.8 (p < 0.05), stabilizing at approximately − 5.0 by year 3. The immediate effect likely reflects automation of routine transactions, while gradual deepening reflects broader organizational process re-engineering. For deposit accounts (Panel C), pre-trends are again flat with post-treatment effects building over time: β_0 = 8.2 (p < 0.10), β_2 = 16.5 (p < 0.05), β_4 = 20.1 (p < 0.01). The delayed inclusion impact may reflect time required for customer awareness, trust-building, and overcoming behavioral barriers to digital adoption. Collectively, these event studies support causal interpretation: effects absent before treatment, emerge after treatment, and exhibit economically sensible dynamic patterns. 6.4 Mechanism Analysis To elucidate the mechanisms underlying the main results, this study examines intermediate outcomes. Table 4 presents regressions with operating expenses per branch, transaction automation rate, customer acquisition cost, and branch expansion as dependent variables. Digital adoption reduces operating expenses per branch by 18% (Column 1, p < 0.01), increases transaction automation from 35% to 52% (Column 2, p < 0.01), and reduces customer acquisition costs by 32% (Column 3, p 0.10), suggesting that digital channels complement rather than substitute for physical presence. Together, these findings support the efficiency channel hypothesis: automation lowers marginal service costs, enabling profitable service provision to previously excluded segments without requiring extensive branch networks. 7. Robustness Checks This section subjects the baseline findings to extensive robustness checks addressing potential threats to identification. 7.1 Placebo Tests This analysis assigns counterfactual adoption years randomly to banks and re-estimate Eq. (1). If the results reflect true causal effects rather than spurious correlation or pre-existing trends, placebo treatments should yield null effects. Table 5 reports results from 1,000 random reassignments. For ROA, the median placebo coefficient is 0.05 with 95% percentile range [-0.3, 0.4], while actual estimate (0.89) falls far outside this range (p < 0.01). Similarly, placebo CIR and inclusion coefficients are tightly centered on zero, with the actual estimates clearly outside null distributions. These placebo tests strongly support causal interpretation. 7.2 Alternative Digital Adoption Measures Table 6 examines sensitivity to alternative digital adoption index constructions. Row 1 replicates baseline PCA-based index results. Row 2 uses equal-weighted averaging of five components; coefficients on ROA (0.82), CIR (-4.5), and inclusion (15.9) are nearly identical to baseline. Row 3 constructs the index using only transaction volume shares, excluding binary indicators; results remain similar (ROA: 0.76, CIR: -4.1, inclusion: 14.8). Row 4 employs factor analysis instead of PCA; results unchanged. Row 5 uses a higher binary threshold (DBAI ≥ 0.7); estimated effects are larger (ROA: 1.15), consistent with more intensive treatment. Collectively, these tests indicate that results are not artifacts of index construction choices. 7.3 Sample Splits and Heterogeneity Table 7 explores treatment effect heterogeneity across bank characteristics. Panel A splits by bank size (above versus below median assets). Large banks exhibit slightly larger ROA effects (1.05 versus 0.72, difference not significant) but similar CIR and inclusion effects, suggesting efficiency gains and inclusion are not size-dependent. Panel B splits by ownership (domestic versus joint-venture with foreign partners). Joint-venture banks show marginally larger effects, potentially reflecting technology spillovers from international partners, though differences are not statistically significant. Panel C splits by initial (2010) performance quartiles. Interestingly, initially weak performers (bottom quartile ROA) exhibit the largest treatment effects (1.38 percentage point ROA increase), suggesting digital adoption offers catch-up opportunities for lagging institutions. This pattern mitigates concerns that only high-performing banks successfully adopt digital services. 7.4 Additional Controls and Alternative Specifications Table 8 examines sensitivity to additional controls that may confound the digital adoption-performance relationship. Column 1 adds loan-to-deposit ratio (LDR) to capture liquidity management; baseline coefficients remain stable. Column 2 adds non-performing loan (NPL) ratio to control for asset quality; again, coefficients unchanged. Column 3 includes both LDR and NPL; results robust. Column 4 adds bank age (years since establishment) to control for experience; no material change. Column 5 controls for branch density (branches per square km) to capture geographic strategy; coefficients stable. Column 6 simultaneously includes all additional controls; baseline findings persist. These tests indicate that omitted variable bias does not drive the results. 7.5 Alternative Clustering and Inference Baseline standard errors are clustered at the bank level to account for within-bank serial correlation. Table 9 explores alternative inference approaches. Row 1 replicates baseline one-way clustering. Row 2 employs two-way clustering by bank and year to allow for arbitrary correlation within banks over time and across banks within years; standard errors increase modestly but significance levels unchanged. Row 3 uses wild cluster bootstrap with 1,000 replications; p-values similar to baseline. Row 4 reports Conley (1999) spatial standard errors accounting for correlation across geographically proximate banks; again, results robust. These tests confirm that baseline inference is conservative. 8. Discussion 8.1 Policy Implications These findings yield several policy implications for financial regulators and development practitioners. First, the positive causal effects on both bank performance and financial inclusion suggest that digital banking adoption is a rare "win-win" intervention, obviating traditional trade-offs between profitability and social objectives. Policymakers should therefore prioritize removing barriers to digital adoption including regulatory uncertainty, cybersecurity concerns, and infrastructure gaps rather than viewing digital finance as requiring subsidies or mandates. Second, the 2–3 year lag between adoption and full effect realization implies that patience is warranted; short-term performance declines during digital transitions should not trigger premature regulatory intervention. Third, the larger inclusion effects in initially underserved areas suggest that targeted incentives for digital expansion into rural and low-income markets could amplify inclusion gains. Such incentives might include regulatory flexibility for agent banking in remote areas or public investment in digital infrastructure (internet connectivity, electricity reliability). Fourth, the complementarity between digital channels and physical branches reflected in the mechanism analysis indicates that digital banking should be viewed as augmenting rather than replacing traditional infrastructure, with implications for branch licensing policies. Finally, the use of publicly available data demonstrates that rigorous impact evaluation is feasible without costly proprietary surveys, potentially informing regulatory data collection priorities. 8.2 Limitations and Future Research Several limitations warrant acknowledgment. First, while the DiD design identifies average treatment effects, it does not fully illuminate distributional impacts, specifically whether inclusion gains benefit the poorest quintile or merely expand services to the “almost included.” Future research combining bank-level administrative data with household surveys could decompose inclusion effects across income deciles. Second, the measure of digital adoption (DBAI) captures supply-side availability but not demand-side usage; banks may offer digital services that customers underutilize. Incorporating transaction-level data on actual digital channel usage would refine measurement. Third, this study cannot definitively rule out all confounding unobservables despite extensive controls and robustness checks; future research could exploit quasi-random variation such as staggered regulatory policy changes or natural experiments. Fourth, the focus on private commercial banks excludes state-owned and foreign banks, which may respond differently to digital opportunities given distinct objectives and constraints. Fifth, this study examines a single country; cross-country replication is needed to assess generalizability, particularly to contexts with different regulatory regimes, infrastructure quality, or financial literacy levels. Sixth, the outcome measures (ROA, CIR, deposit account penetration) capture immediate effects but miss longer-run welfare impacts such as poverty reduction or entrepreneurship promotion, which require longitudinal household tracking. Finally, this study does not examine financial stability implications, including whether digital banking increases systemic risk through operational dependencies or interconnectedness, which remains an important avenue for future research. 8.3 Broader Implications for Development Finance Beyond Bangladesh, these findings contribute to ongoing debates about digital finance’s role in economic development. The efficiency channel documented here suggests that technology-driven cost reductions can overcome financial inclusion’s “last mile problem,” which refers to the challenge of profitably serving geographically dispersed, low-balance customers. This mechanism differs from microfinance’s group lending approach (which addressed information asymmetries through social collateral) and government-mandated priority sector lending (which cross-subsidized unprofitable segments), instead relying on automation to align commercial incentives with social objectives. Whether this model scales to the most excluded populations, particularly those lacking mobile phones, digital literacy, or reliable internet, remains an open question requiring further research. Additionally, the finding that joint-venture banks with foreign partnerships exhibit marginally stronger effects hints at technology transfer benefits from financial sector openness, though this study lacks statistical power to firmly establish this channel. Finally, the relatively modest absolute magnitude of inclusion effects (15–22 additional accounts per 1,000 adults) underscores that digital banking is not a panacea; complementary interventions addressing financial literacy, consumer protection, and infrastructure gaps remain essential. 9. Conclusion This study provides the first rigorous causal evidence on digital banking adoption's impact on bank performance and financial inclusion using comprehensive institutional data from Bangladesh over 2010–2025. Exploiting staggered adoption timing within a difference-in-differences framework, this study finds that digital banking adoption causally improves bank profitability (ROA increases by 0.8–1.2 percentage points), enhances operational efficiency (cost-to-income ratios decline by 4–6 percentage points), and expands financial inclusion (deposit accounts per 1,000 adults increase by 15–22 percent). These effects materialize gradually over 2–3 years post-adoption, are robust to extensive specification checks and alternative measurement approaches, and are largest in initially underserved markets. Mechanistically, this study shows that efficiency gains from transaction automation and reduced branch operating costs enable profitable service provision to previously unprofitable customer segments, reconciling performance and inclusion objectives. This contrasts with traditional views of financial inclusion as requiring cross-subsidization or regulatory mandates, instead highlighting technology's potential to align commercial incentives with social goals. These findings suggest that removing barriers to digital adoption—through regulatory clarity, infrastructure investment, and cybersecurity capacity-building—may be more effective than supply-side mandates in promoting inclusive finance. Methodologically, this analysis demonstrates that rigorous causal inference is achievable using only publicly available regulatory data, offering a replicable template for researchers and policymakers in data-constrained environments. The DiD design with extensive validation (parallel trends tests, placebo tests, robustness checks) provides confidence in causal interpretation while acknowledging limitations inherent in observational data. Future research should extend this approach to other developing economies, incorporate distributional welfare analysis, and examine longer-run impacts on poverty reduction and economic mobility. As digital transformation continues reshaping financial services globally, understanding its causal impacts on institutional performance and societal outcomes becomes increasingly critical. The findings from Bangladesh a country representative of South Asian and broader developing-world challenges and opportunities provide evidence-based guidance for policymakers navigating the opportunities and risks of digital finance. 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 (bank) 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 by the author 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 and do not represent the official position of Mutual Trust Bank PLC or any other organization. Data Availability Statement All data supporting the findings of this study are publicly available: Bank Financial Data (2010–2025): Audited annual reports from all 43 private commercial banks are publicly available through institutional websites and the Bangladesh Bank repository (https://www.bb.org.bd). Regulatory Data: Bangladesh Bank publications including Annual Reports, Financial Stability Reports, and Mobile Financial Services Reports are publicly accessible at https://www.bb.org.bd/pub/. Financial Inclusion Data: World Bank Global Findex Database (https://globalfindex.worldbank.org/) and FinScope Bangladesh Survey data. Replication Materials: The author commits to providing detailed data extraction procedures, variable construction code, and analysis scripts upon reasonable request. Aggregated datasets (with bank identifiers anonymized if required) can be shared for verification purposes, subject to Bangladesh Bank data sharing guidelines. No proprietary or confidential data were used. All analyses can be independently replicated using the publicly available sources listed above. References Athey, S., & Wager, S. (2018). Estimation and inference of heterogeneous treatment effects using random forests. Journal of the American Statistical Association, 113(523), 1228-1242. Athanasoglou, P. P., Brissimis, S. N., & Delis, M. D. (2008). Bank-specific, industry-specific and macroeconomic determinants of bank profitability. Journal of International Financial Markets, Institutions and Money, 18(2), 121-136. Bangladesh Bank. (2024). Annual Report 2023-2024. Bangladesh Bank, Dhaka. Chowdhury, M. A., & Salman, M. A. G. (2021). Bank-specific & macroeconomic determinants of profitability: Empirical evidence from Bangladeshi private commercial banks. American Journal of Theoretical and Applied Business, 7(4), 72-80. Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319-340. Demirgüç-Kunt, A., Klapper, L., Singer, D., & Ansar, S. (2022). The Global Findex Database 2021: Financial Inclusion, Digital Payments, and Resilience in the Age of COVID-19. World Bank, Washington, DC. Dietrich, A., & Wanzenried, G. (2011). Determinants of bank profitability before and during the crisis: Evidence from Switzerland. Journal of International Financial Markets, Institutions and Money, 21(3), 307-327. Frost, J. (2020). The economic forces driving fintech adoption across countries. BIS Working Papers No. 838, Bank for International Settlements. Ozili, P. K. (2021). Financial inclusion research around the world: A review. Forum for Social Economics, 50(4), 457-479. Parvin, S., Chowdhury, A. N. M. M. H., Siddiqua, A., & Ferdous, J. (2019). Effect of liquidity and bank size on the profitability of commercial banks in Bangladesh. Asian Business Review, 9(1), 7-10. Phan, D. V., & Pham, T. T. (2025). Causal and nonlinear effects of digital financial inclusion on bank stability: Evidence from emerging and advanced economies. Banks and Bank Systems, 20(4), 153-171. Philippon, T. (2016). The fintech opportunity. BIS Working Papers No. 655, Bank for International Settlements. Sahay, R., von Allmen, U. E., Lahreche, A., Khera, P., Ogawa, S., Bazarbash, M., & Beaton, K. (2020). The promise of fintech: Financial inclusion in the post COVID-19 era. IMF Departmental Paper No. 2020/009, International Monetary Fund. Shaikh, A. A., Alamoudi, H., Alharthi, M., & Glavee-Geo, R. (2023). Advances in mobile financial services: A review of the literature and future research directions. International Journal of Bank Marketing, 41(1), 1-33. Williamson, O. E. (1985). The Economic Institutions of Capitalism. Free Press, New York. Additional Declarations The authors declare no competing interests. Supplementary Files APPENDICES.docx 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-9055268","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":602184019,"identity":"91e2461d-d2e3-471b-9b2f-2dc0fb9a31e5","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":"2026-03-07 04:29:06","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-9055268/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9055268/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104366251,"identity":"7baeed93-6732-4ca4-953c-c2f14762ee25","added_by":"auto","created_at":"2026-03-11 03:17:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":6752,"visible":true,"origin":"","legend":"\u003cp\u003eThis image is not available with this version.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9055268/v1/b55839e6f14c4bb2ff4bf4fb.png"},{"id":104409597,"identity":"fcc1311c-a346-4abc-b42f-d8be6bc65bbc","added_by":"auto","created_at":"2026-03-11 12:46:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":882198,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9055268/v1/f8baa969-5cf5-47da-a316-45eb323d85d8.pdf"},{"id":104406325,"identity":"08f5c3c7-68f3-4c34-a35a-6bed7e563d66","added_by":"auto","created_at":"2026-03-11 12:25:20","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":31635,"visible":true,"origin":"","legend":"","description":"","filename":"APPENDICES.docx","url":"https://assets-eu.researchsquare.com/files/rs-9055268/v1/dd750a11214a5eaae3a619f8.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eThe Economic Impact of Digital Banking Adoption on Bank Performance and Financial Inclusion: Evidence from Bangladesh (2010–2025)\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eFinancial inclusion, broadly defined as access to and usage of formal financial services, remains a critical policy priority for developing economies. Despite substantial progress over the past two decades, approximately 1.4\u0026nbsp;billion adults worldwide remain unbanked, with concentrations highest in South Asia and Sub-Saharan Africa (Demirg\u0026uuml;\u0026ccedil;-Kunt et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Bangladesh exemplifies both the challenges and opportunities in this domain: while financial inclusion rates improved from 31% in 2011 to 65% in 2024, geographic and socioeconomic disparities persist, particularly in rural areas and among low-income populations (Bangladesh Bank, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDigital banking, encompassing mobile banking, internet banking, agent banking, and digital payment systems has emerged as a potential solution to these inclusion barriers. By reducing transaction costs, extending service reach beyond physical branches, and enabling automated credit assessments, digital channels theoretically can serve previously unprofitable customer segments (Philippon, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Frost, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Bangladesh's banking sector has undergone rapid digital transformation since 2010, when mobile financial services first received regulatory authorization. By 2024, 19 mobile financial service providers processed over 52\u0026nbsp;billion transactions worth \u003cspan\u003e$\u003c/span\u003e1.2 trillion annually, while internet banking users surged from negligible levels to 8.3\u0026nbsp;million (Bangladesh Bank, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This transformation was accompanied by regulatory initiatives including agent banking guidelines (2012), electronic Know-Your-Customer frameworks (2020), and digital nano-loan authorization (2022).\u003c/p\u003e \u003cp\u003eYet despite widespread enthusiasm for digital banking's potential, rigorous causal evidence on its economic impacts remains scarce, particularly at the institutional level. Prior research has predominantly examined customer-side adoption patterns (Ozili, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Shaikh et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) or cross-country correlations between digital finance penetration and macroeconomic outcomes (Sahay et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). These studies face three critical limitations that limit their policy relevance. First, most analyses rely on survey data or cross-sectional comparisons, making causal inference problematic due to selection bias and reverse causality, given that banks that adopt digital services may differ systematically from non-adopters in unobservable ways that independently affect performance. Second, existing research typically uses aggregate national-level indicators, obscuring the heterogeneous micro-level mechanisms through which digital adoption affects individual institutions. Third, most studies focus either on bank performance or financial inclusion separately, missing the potential tension between profitability imperatives and inclusion objectives.\u003c/p\u003e \u003cp\u003eThis study addresses these gaps by providing the first rigorous causal analysis of digital banking adoption's impact on both bank performance and financial inclusion using micro-level institutional data from Bangladesh over 2010\u0026ndash;2025. This study constructs a comprehensive panel dataset covering all 43 private commercial banks, combining publicly available annual reports with regulatory filings from Bangladesh Bank. The identification strategy leverages the staggered timing of digital banking adoption across banks within a difference-in-differences framework, controlling for bank fixed effects (which absorb time-invariant institutional characteristics like governance quality or management capability) and year fixed effects (which absorb macroeconomic shocks and regulatory changes affecting all banks simultaneously). This approach isolates the causal effect of digital adoption by comparing the evolution of early versus late adopters before and after their respective adoption dates, conditional on observables.\u003c/p\u003e \u003cp\u003eThe empirical strategy addresses endogeneity concerns through multiple mechanisms. First, the variation in adoption timing driven largely by technological readiness, partnerships with fintech firms, and board-level strategic decisions rather than concurrent performance provides plausibly exogenous variation. Second, the identifying assumption of parallel pre-treatment trends is validated through event-study specifications showing that treatment and control banks exhibited similar performance trajectories prior to adoption. Third, this study conducts extensive robustness checks including placebo adoption dates, alternative digital adoption measures (continuous indices versus binary indicators), and sample splits by bank size and ownership structure.\u003c/p\u003e \u003cp\u003eFindings reveal three key results. First, digital banking adoption causally improves bank performance: return on assets (ROA) increases by 0.8\u0026ndash;1.2 percentage points, while cost-to-income ratios decline by 4\u0026ndash;6 percentage points. These effects materialize gradually over 2\u0026ndash;3 years post-adoption, consistent with the time required for organizational learning and customer migration to digital channels. Second, digital adoption significantly expands financial inclusion: deposit accounts per 1,000 adults increase by 15\u0026ndash;22 percent, with effects concentrated among previously underserved rural and low-income segments. Third, mechanism analysis reveals that these dual benefits arise through operational efficiency gains specifically, reduced branch operating costs and automated processes that enable profitable service provision to previously unprofitable customer segments, rather than merely redistributing services toward already-included populations.\u003c/p\u003e \u003cp\u003eThis study makes four principal contributions to the literature. First, it provides the first causal bank-level evidence on digital banking impacts in a developing economy, employing quasi-experimental methods that overcome selection bias inherent in cross-sectional and time-series correlations. Second, it jointly examines performance and inclusion outcomes, revealing that these objectives are complementary rather than conflicting when efficiency gains lower marginal service costs. Third, it demonstrates that rigorous empirical analysis is feasible using only publicly available data, offering a replicable methodological template for researchers and policymakers in data-constrained environments. Fourth, a 15-year panel spanning the entire digital transformation period enables analysis of long-run effects and dynamic adjustment paths, addressing concerns about short-term fluctuations that may dominate shorter panels.\u003c/p\u003e \u003cp\u003eThe remainder of this paper proceeds as follows. Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e synthesizes the theoretical and empirical literature, identifying research gaps and articulating this study\u0026rsquo;s contribution. Section \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e3\u003c/span\u003e describes the institutional context of Bangladesh's banking sector and digital transformation. Section \u003cspan refid=\"Sec8\" class=\"InternalRef\"\u003e4\u003c/span\u003e details data sources, variable construction, and sample characteristics. Section \u003cspan refid=\"Sec12\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the empirical strategy, emphasizing identification assumptions and validation procedures. Section \u003cspan refid=\"Sec17\" class=\"InternalRef\"\u003e6\u003c/span\u003e reports main results, mechanism analysis, and heterogeneity tests. Section \u003cspan refid=\"Sec22\" class=\"InternalRef\"\u003e7\u003c/span\u003e conducts extensive robustness checks. Section \u003cspan refid=\"Sec28\" class=\"InternalRef\"\u003e8\u003c/span\u003e discusses policy implications and limitations. Section \u003cspan refid=\"Sec32\" class=\"InternalRef\"\u003e9\u003c/span\u003e concludes.\u003c/p\u003e"},{"header":"2. Literature Review and Theoretical Framework","content":"\u003cp\u003eThis literature review is organized around three interrelated strands: (1) the determinants and impacts of bank performance in developing economies, (2) the relationship between financial technology adoption and financial inclusion, and (3) methodological approaches to causal inference in banking research. Rather than providing an exhaustive catalog of prior studies, this section synthesizes the key findings to position this study\u0026rsquo;s contribution and articulate its theoretical framework.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Bank Performance in Emerging Markets\u003c/h2\u003e \u003cp\u003eA substantial body of research examines bank profitability determinants in emerging markets. Internal factors including capital adequacy, asset quality, operational efficiency, and management quality consistently predict performance variation (Athanasoglou et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Dietrich \u0026amp; Wanzenried, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). For Bangladesh specifically, Chowdhury and Salman (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) analyze 25 private commercial banks over 2012\u0026ndash;2019, finding that capital adequacy and asset management positively affect ROA and ROE, while cost-to-income ratios exhibit negative associations. Similarly, studies by Parvin et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and Ahmed et al. (2013) document that operational efficiency typically measured by cost-to-income ratios or operating expense to total assets is the strongest predictor of profitability among Bangladeshi banks. However, these studies employ correlational methods (OLS or fixed effects) without addressing the endogeneity of efficiency measures, leaving causal interpretations ambiguous. This study advances this literature by examining how digital adoption, as a potentially exogenous shock to operational costs, causally affects both efficiency metrics and profitability.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Digital Finance and Financial Inclusion\u003c/h2\u003e \u003cp\u003eThe relationship between digital financial services and financial inclusion has attracted substantial recent attention. Conceptually, digital channels reduce transaction costs through automation, extend geographic reach via mobile networks, and enable data-driven credit assessment for thin-file customers (Philippon, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Empirical evidence largely confirms positive associations: cross-country studies find that mobile money adoption correlates with increased account ownership, particularly among rural and low-income populations (Demirg\u0026uuml;\u0026ccedil;-Kunt et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). For Sub-Saharan Africa, recent research using causal machine learning methods reveals that digital financial inclusion enhances economic growth only in countries with strong institutional quality, suggesting important complementarities between technology and regulatory environments (Phan \u0026amp; Pham, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In South Asian contexts, studies of Bangladesh, India, and Pakistan document that mobile banking adoption associates with higher financial inclusion rates, though most rely on user surveys or aggregate time-series data (Shaikh et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Ozili, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). A critical gap emerges: existing research predominantly examines whether individuals adopt digital services, not whether institutional adoption by banks causally expands service provision to underserved populations. The bank-level analysis directly addresses this gap by measuring how digital adoption affects deposit account penetration rates across demographic segments.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Causal Inference in Banking Research\u003c/h2\u003e \u003cp\u003eEstablishing causality in observational banking data presents substantial challenges due to selection bias, reverse causality, and omitted variable bias. Three methodological approaches dominate recent literature. First, difference-in-differences (DiD) designs exploit policy changes or institutional shocks affecting some banks but not others, comparing outcomes before and after treatment while controlling for common trends. This approach requires parallel trends in the absence of treatment\u0026mdash;an assumption testable through pre-treatment trend analysis. Second, instrumental variables (IV) strategies use external determinants of digital adoption (e.g., proximity to fintech firms or regulatory policy variation) to isolate exogenous variation. However, valid instruments are difficult to identify in practice. Third, recent advances employ causal machine learning methods (causal forests, double machine learning) that flexibly estimate heterogeneous treatment effects without imposing functional form restrictions (Athey \u0026amp; Wager, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This study adopts the DiD approach due to the staggered adoption timing across Bangladeshi banks, validated through extensive parallel trends testing and robustness checks. This design balances internal validity\u0026mdash;through credible identification\u0026mdash;with external validity\u0026mdash;through comprehensive coverage of the banking sector.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Theoretical Framework and Hypotheses\u003c/h2\u003e \u003cp\u003eThis paper\u0026rsquo;s theoretical framework synthesizes insights from the Technology Acceptance Model (Davis, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1989\u003c/span\u003e), transaction cost economics (Williamson, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1985\u003c/span\u003e), and the financial intermediation literature. It posits that digital banking adoption affects bank performance and financial inclusion through two primary channels. The efficiency channel operates as follows: digital platforms automate routine transactions (deposits, withdrawals, transfers) that previously required labor-intensive branch operations, thereby reducing marginal service costs. Simultaneously, digital channels enable banks to serve geographically dispersed customers without establishing physical branches, lowering fixed costs of market entry. These efficiency gains should manifest as improved cost-to-income ratios and, consequently, higher profitability (ROA). The inclusion channel operates through reduced service costs: as marginal costs decline, previously unprofitable customer segments (small depositors, rural residents) become economically viable. Additionally, digital identity verification (e-KYC) and algorithmic credit assessment reduce information asymmetries that traditionally excluded thin-file borrowers, further expanding serviceable market segments. This mechanism predicts increased deposit account penetration, particularly among previously underserved populations. Importantly, this framework emphasizes complementarity: efficiency gains enable inclusion rather than trading off profitability against social objectives. This contrasts with traditional views of financial inclusion as requiring cross-subsidization. This research formalizes these predictions in three testable hypotheses: H1: Digital banking adoption causally improves bank profitability (ROA); H2: Digital adoption causally reduces operational costs (cost-to-income ratio); H3: Digital adoption causally expands financial inclusion (deposit accounts per capita).\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Institutional Context: Bangladesh Banking Sector","content":"\u003cp\u003eUnderstanding Bangladesh's banking structure and digital transformation trajectory is essential for interpreting the empirical results. As of 2025, the sector comprises 61 scheduled banks: 6 state-owned commercial banks, 3 specialized development banks, 43 private commercial banks (PCBs), and 9 foreign commercial banks. This analysis focuses exclusively on the 43 PCBs, which collectively hold approximately 65% of total banking assets and dominate retail and SME lending. PCBs operate under the Bank Company Act (1991, amended 2013) and Bangladesh Bank supervision, facing identical regulatory requirements for capital adequacy (10% minimum), liquidity ratios, and prudential standards. This regulatory uniformity strengthens the identification strategy by minimizing confounding institutional differences across banks.\u003c/p\u003e \u003cp\u003eThe digital transformation of Bangladesh's banking sector occurred in distinct phases driven by regulatory milestones. Phase 1 (2010\u0026ndash;2012): Mobile Financial Services (MFS) emerged following Bangladesh Bank's 2011 guidelines authorizing banks to provide mobile money services. Dutch-Bangla Bank launched the first service (Rocket) in March 2011, followed by BRAC Bank's bKash. Initial adoption focused on remittance transfers and basic payments. Phase 2 (2013\u0026ndash;2016): Agent banking guidelines (2012) enabled banks to establish agent networks in underserved areas, while internet banking platforms proliferated among urban-focused institutions. This period saw rapid expansion of digital transaction volumes but limited integration across channels. Phase 3 (2017\u0026ndash;2021): E-KYC authorization (2020) dramatically reduced account opening time from 2\u0026ndash;4 days to 5 minutes, while digital nano-loan programs (2022) enabled algorithmic lending for micro-borrowers. COVID-19 accelerated contactless banking adoption. Phase 4 (2022\u0026ndash;2025): Regulatory consolidation emphasized cybersecurity standards, interoperability across digital platforms, and financial literacy initiatives targeting rural populations. By 2025, digital banking penetration varies substantially across PCBs, with digital adoption indices (described in Section \u003cspan refid=\"Sec8\" class=\"InternalRef\"\u003e4\u003c/span\u003e) ranging from 0.15 to 0.85 on a normalized 0\u0026ndash;1 scale.\u003c/p\u003e \u003cp\u003eThree institutional features of Bangladesh's context enhance the study's external validity for similar developing economies. First, the regulatory environment balances innovation encouragement with prudential supervision\u0026mdash;a common challenge for emerging markets. Second, the coexistence of traditional branch banking and digital channels creates variation in adoption strategies exploitable for identification. Third, persistent inclusion gaps (35% of adults remain unbanked) ensure that digital expansion targets genuinely underserved populations rather than merely shifting existing customers between channels. These features make Bangladesh an informative case for understanding digital banking impacts in comparable South Asian and Sub-Saharan African contexts.\u003c/p\u003e"},{"header":"4. Data and Variable Construction","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Data Sources\u003c/h2\u003e \u003cp\u003eThis analysis combines three publicly available data sources, ensuring full replicability and transparency. First, bank-level financial data derive from audited annual reports (2010\u0026ndash;2025) published by all 43 private commercial banks, accessed through institutional websites and the Bangladesh Bank repository. These reports contain balance sheet items (total assets, deposits, loans, equity), income statements (interest income/expense, operating costs, net income), and operational metrics (branch counts, employee numbers). Second, digital banking indicators come from Bangladesh Bank's quarterly Mobile Financial Services reports, annual Scheduled Bank Statistics publications, and individual banks' digital banking disclosures. Third, macroeconomic controls (GDP growth, inflation, exchange rates) are obtained from Bangladesh Bank's Monthly Economic Trends and World Bank databases. All data are verified against multiple sources where possible to ensure accuracy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Sample Construction\u003c/h2\u003e \u003cp\u003eThe baseline sample is a balanced panel of 43 banks observed annually from 2010 to 2025, yielding 688 bank-year observations (43 \u0026times; 16 years). This study includes all PCBs operational throughout the period, excluding five banks that entered after 2010 or merged during the sample period to maintain balance. Robustness checks employ an unbalanced panel including these banks (N\u0026thinsp;=\u0026thinsp;748 observations). The sample represents nearly the entire private banking sector in Bangladesh, minimizing concerns about selective coverage. Summary statistics (Table\u0026nbsp;1) reveal substantial heterogeneity: mean total assets are BDT 350\u0026nbsp;billion (SD\u0026thinsp;=\u0026thinsp;420\u0026nbsp;billion), mean ROA is 0.87% (SD\u0026thinsp;=\u0026thinsp;1.2%), and mean cost-to-income ratio is 52.3% (SD\u0026thinsp;=\u0026thinsp;14.1%). This variation provides statistical power to detect treatment effects while the balanced panel design eliminates attrition bias.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Variable Definitions\u003c/h2\u003e \u003cp\u003e \u003cb\u003eDependent Variables.\u003c/b\u003e This study examines three primary outcomes capturing bank performance and financial inclusion.\u003c/p\u003e \u003cp\u003eReturn on Assets (ROA) measures profitability as net income divided by average total assets, expressed as a percentage. ROA is the standard metric for bank performance as it reflects both operational efficiency and asset utilization, making it comparable across institutions of different sizes. Mean ROA in sample is 0.87%, consistent with recent Bangladeshi banking sector averages but below the 1.5-2.0% typical of mature markets, reflecting higher operating costs and credit risk in developing contexts.\u003c/p\u003e \u003cp\u003eCost-to-Income Ratio (CIR) measures operational efficiency as total operating expenses divided by operating income (net interest income plus non-interest income), expressed as a percentage. Lower CIR indicates greater efficiency. Sample mean CIR is 52.3%, implying that banks spend BDT 52 for every BDT 100 earned. This exceeds the 40\u0026ndash;45% benchmark for efficient banks, suggesting substantial room for efficiency improvements through digital automation.\u003c/p\u003e \u003cp\u003eDeposit Accounts per 1,000 Adults captures financial inclusion by measuring the bank's deposit account penetration rate in its service area. This metric is constructed as total deposit accounts (sourced from annual reports) divided by adult population (15+) in the bank's operational districts (from Bangladesh Bureau of Statistics), expressed per thousand. Mean penetration is 85 accounts per 1,000 adults (SD\u0026thinsp;=\u0026thinsp;45), indicating that most adults remain either unbanked or hold accounts with multiple institutions. This measure captures extensive margin inclusion (account ownership) rather than intensive margin usage, providing a conservative test of digital banking's inclusion impact.\u003c/p\u003e \u003cp\u003e \u003cb\u003eDigital Banking Adoption Index.\u003c/b\u003e The key explanatory variable requires careful construction to capture multidimensional digital adoption.\u003c/p\u003e \u003cp\u003eThis study constructs a composite Digital Banking Adoption Index (DBAI) ranging from 0 (no digital services) to 1 (full digital integration) using principal component analysis (PCA) of five observable indicators: (1) mobile banking service availability (binary); (2) internet banking platform availability (binary); (3) agent banking network presence (binary); (4) digital transaction volume as share of total transactions (continuous, 0\u0026ndash;1); (5) digital channel share of operating expenses (continuous, 0\u0026ndash;1). The first principal component explains 62% of variation across these indicators and loads positively on all five, validating its interpretation as a general digital adoption factor. Scores are normalized to [0,1] for interpretability. Mean DBAI is 0.42 (SD\u0026thinsp;=\u0026thinsp;0.28), indicating partial adoption. For robustness, this study employs a binary treatment indicator (DBAI\u0026thinsp;\u0026ge;\u0026thinsp;0.5) that classifies 58% of bank-years as \u0026ldquo;digital adopters,\u0026rdquo; and constructs alternative continuous indices using equal weighting and different normalization procedures.\u003c/p\u003e \u003cp\u003e \u003cb\u003eControl Variables.\u003c/b\u003e The baseline specification includes four time-varying bank-level and macroeconomic controls.\u003c/p\u003e \u003cp\u003eLog Total Assets controls for bank size, which may independently affect both performance (through economies of scale) and digital adoption (larger banks have more resources for technology investment). Capital Adequacy Ratio (CAR), measured as regulatory capital divided by risk-weighted assets, controls for financial soundness and risk appetite. GDP Growth Rate controls for macroeconomic conditions affecting loan demand and credit quality. Inflation Rate controls for nominal effects on interest margins and operating costs. Additional robustness checks include loan-to-deposit ratios, non-performing loan ratios, and bank age, though these are potentially endogenous to digital adoption and thus excluded from baseline specifications. Summary statistics for all variables appear in Table\u0026nbsp;1.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Empirical Strategy","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Baseline Specification\u003c/h2\u003e \u003cp\u003eThe identification strategy exploits the staggered timing of digital banking adoption across banks within a difference-in-differences (DiD) framework. The baseline estimating equation is:\u003c/p\u003e \u003cp\u003e \u003cem\u003eY\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e = α\u0026thinsp;+\u0026thinsp;β\u0026thinsp;\u0026times;\u0026thinsp;Digital\u003cem\u003e\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e + γ\u0026thinsp;\u0026times;\u0026thinsp;X\u003cem\u003e\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e + \u0026micro;\u003cem\u003e\u003csub\u003ei\u003c/sub\u003e\u003c/em\u003e + λ\u003cem\u003eₜ\u003c/em\u003e + ε\u003cem\u003e\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e (1)\u003c/p\u003e \u003cp\u003ewhere Y_it represents the outcome for bank i in year t (ROA, CIR, or deposit accounts per 1,000 adults); Digital_it is the digital banking adoption index; X_it is a vector of time-varying controls (log assets, CAR, GDP growth, inflation); \u0026micro;_i denotes bank fixed effects absorbing time-invariant institutional characteristics (e.g., founding vintage, ownership structure, management quality); λ_t denotes year fixed effects absorbing common shocks (e.g., regulatory changes, macroeconomic crises); and ε_it is an idiosyncratic error term. Standard errors are clustered at the bank level to account for serial correlation within banks over time. The coefficient β identifies the average treatment effect of digital adoption on outcomes, interpreted causally under the parallel trends assumption.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Identification Assumptions\u003c/h2\u003e \u003cp\u003eThe validity of the DiD design rests on three key assumptions. First, the parallel trends assumption requires that treated and control banks would have exhibited similar outcome trajectories in the absence of digital adoption. While this counterfactual is inherently unobservable, this study tests the assumption by examining pre-treatment trends through event-study specifications (Eq.\u0026nbsp;2 below) that interact treatment status with year indicators. Statistically insignificant coefficients on leads (years before adoption) support parallel trends. Second, this study requires conditional exchangeability: after controlling for observables (X_it) and fixed effects, treatment timing must be uncorrelated with unobserved time-varying confounders. The staggered adoption pattern driven by technological readiness, fintech partnerships, and board decisions rather than concurrent performance changes supports this assumption. Third, this study assumes treatment effect homogeneity or accepts that β captures the average treatment effect across heterogeneous banks. Robustness checks explore heterogeneity by bank size, ownership, and initial performance levels.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Event-Study Specification\u003c/h2\u003e \u003cp\u003eTo test parallel trends and trace out dynamic treatment effects, this study estimates an event-study specification that replaces Digital_it with a full set of relative time indicators:\u003c/p\u003e \u003cp\u003e \u003cem\u003eY\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e = α\u0026thinsp;+\u0026thinsp;Σ\u003cem\u003eₖ\u003c/em\u003e β\u003cem\u003eₖ\u003c/em\u003e \u0026times; 1[t-t\u003cem\u003e\u003csub\u003ei\u003c/sub\u003e*\u003c/em\u003e=k] + γ\u0026thinsp;\u0026times;\u0026thinsp;X\u003cem\u003e\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e + \u0026micro;\u003cem\u003e\u003csub\u003ei\u003c/sub\u003e\u003c/em\u003e + λ\u003cem\u003eₜ\u003c/em\u003e + ε\u003cem\u003e\u003csub\u003ei\u003c/sub\u003eₜ\u003c/em\u003e (2)\u003c/p\u003e \u003cp\u003ewhere t_i* is the adoption year for bank i, k indexes event time (years relative to adoption), and 1[\u0026middot;] is an indicator function. The study normalizes β_{-1} = 0 to identify coefficients relative to the year immediately preceding adoption. Statistically insignificant β_k for k \u0026lt; -1 (pre-treatment years) validates parallel trends, while β_k for k\u0026thinsp;\u0026ge;\u0026thinsp;0 (post-treatment years) traces the dynamic adjustment path. Economic theory predicts gradual effect emergence as customers migrate to digital channels and organizational learning occurs, implying β_0\u0026thinsp;\u0026lt;\u0026thinsp;β_1\u0026thinsp;\u0026lt;\u0026thinsp;β_2, etc.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Robustness Checks\u003c/h2\u003e \u003cp\u003eThis paper conducts extensive robustness checks detailed in Section \u003cspan refid=\"Sec22\" class=\"InternalRef\"\u003e7\u003c/span\u003e. Briefly, these include: (1) placebo tests that assign counterfactual adoption years to test whether spurious correlations drive results; (2) alternative digital index constructions using equal weighting, different thresholds for binary classification, and excluding individual components; (3) sample splits by bank size, ownership type (domestic versus joint-venture), and initial performance quartiles to test heterogeneity; (4) inclusion of additional controls (loan-to-deposit ratio, NPL ratio, branch density) to address omitted variable concerns; (5) alternative clustering approaches (two-way clustering by bank and year); and (6) replication using unbalanced panel. Collectively, these checks validate the baseline findings.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Main Results","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e6.1 Impact on Bank Performance\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;2 presents baseline DiD estimates from Eq.\u0026nbsp;(1) with ROA and CIR as dependent variables. All specifications include bank and year fixed effects with standard errors clustered by bank.\u003c/p\u003e \u003cp\u003ePanel A examines ROA. Column (1) reports a univariate specification without controls, finding that digital adoption increases ROA by 1.02 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Given mean ROA of 0.87%, this represents a 117% increase, suggesting economically large effects. Column (2) adds controls for bank size, capital adequacy, and macroeconomic conditions; the coefficient attenuates to 0.89 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) but remains highly significant. Coefficients on controls conform to expectations: larger banks exhibit higher ROA (size economies of scale), higher capital adequacy associates with lower ROA (conservative lending), GDP growth positively affects ROA (credit demand), and inflation negatively affects ROA (margin compression). Column (3) employs a binary digital adoption indicator (DBAI\u0026thinsp;\u0026ge;\u0026thinsp;0.5), yielding a coefficient of 0.78 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), implying that full digital adoption increases ROA by approximately three-quarters of a percentage point. This estimate is conservative relative to Column (2) as it assigns equal treatment to all banks above the threshold, ignoring intensity variation.\u003c/p\u003e \u003cp\u003ePanel B examines cost-to-income ratio (CIR). Column (1) finds that digital adoption reduces CIR by 5.8 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). With mean CIR of 52.3%, this represents an 11% efficiency improvement. Column (2) with controls yields\u0026thinsp;\u0026minus;\u0026thinsp;4.9 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), while Column (3) with binary treatment shows\u0026thinsp;\u0026minus;\u0026thinsp;4.2 percentage points (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). These results confirm H2, indicating that digital banking substantially improves operational efficiency. Mechanistically, digital channels automate transactions that previously required tellers and branch operations, directly reducing operating expenses relative to income.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e6.2 Impact on Financial Inclusion\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;3 presents results for financial inclusion measured by deposit accounts per 1,000 adults. Panel A shows full-sample estimates.\u003c/p\u003e \u003cp\u003eColumn (1) indicates that digital adoption increases deposit account penetration by 18.5 accounts per 1,000 adults (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), representing a 22% increase relative to the sample mean of 85. With controls (Column 2), the coefficient is 16.8 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). The binary treatment specification (Column 3) yields 15.2 additional accounts (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). These findings strongly support H3: digital adoption causally expands financial inclusion. To investigate whether these effects reflect genuine inclusion of previously unbanked populations versus account churning among existing customers, Panel B splits the sample by district-level baseline (2010) financial inclusion rates. Low-inclusion districts (bottom tercile) exhibit larger treatment effects (21.4 additional accounts, Column 1) than high-inclusion districts (12.3 accounts, Column 2), with the difference statistically significant (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Column 3). This heterogeneity pattern indicates that digital banking preferentially expands services into underserved areas, consistent with the theoretical prediction that efficiency gains make previously unprofitable markets viable.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e6.3 Dynamic Effects and Event Studies\u003c/h2\u003e \u003cp\u003eFigure 1 plots event-study coefficients (β_k from Eq.\u0026nbsp;2) for three outcomes, tracing effects from four years before to four years after adoption (t_i*=0). For ROA (Panel A), pre-treatment coefficients (k=-4 to k=-2) are small and statistically indistinguishable from zero, validating parallel trends. Post-treatment coefficients increase gradually: β_0\u0026thinsp;=\u0026thinsp;0.3 (p\u0026thinsp;\u0026gt;\u0026thinsp;0.10, not significant), β_1\u0026thinsp;=\u0026thinsp;0.6 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.10), β_2\u0026thinsp;=\u0026thinsp;0.9 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), β_3\u0026thinsp;=\u0026thinsp;1.1 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), β_4\u0026thinsp;=\u0026thinsp;1.2 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). This pattern reveals that profitability effects materialize over 2\u0026ndash;3 years as customer bases migrate to digital channels and banks realize scale economies. For CIR (Panel B), pre-trends are flat while post-treatment effects emerge immediately: β_0=-2.8 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), stabilizing at approximately\u0026thinsp;\u0026minus;\u0026thinsp;5.0 by year 3. The immediate effect likely reflects automation of routine transactions, while gradual deepening reflects broader organizational process re-engineering. For deposit accounts (Panel C), pre-trends are again flat with post-treatment effects building over time: β_0\u0026thinsp;=\u0026thinsp;8.2 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.10), β_2\u0026thinsp;=\u0026thinsp;16.5 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), β_4\u0026thinsp;=\u0026thinsp;20.1 (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). The delayed inclusion impact may reflect time required for customer awareness, trust-building, and overcoming behavioral barriers to digital adoption. Collectively, these event studies support causal interpretation: effects absent before treatment, emerge after treatment, and exhibit economically sensible dynamic patterns.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e6.4 Mechanism Analysis\u003c/h2\u003e \u003cp\u003eTo elucidate the mechanisms underlying the main results, this study examines intermediate outcomes. Table\u0026nbsp;4 presents regressions with operating expenses per branch, transaction automation rate, customer acquisition cost, and branch expansion as dependent variables. Digital adoption reduces operating expenses per branch by 18% (Column 1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), increases transaction automation from 35% to 52% (Column 2, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), and reduces customer acquisition costs by 32% (Column 3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Interestingly, digital adoption does not significantly affect branch counts (Column 4, p\u0026thinsp;\u0026gt;\u0026thinsp;0.10), suggesting that digital channels complement rather than substitute for physical presence. Together, these findings support the efficiency channel hypothesis: automation lowers marginal service costs, enabling profitable service provision to previously excluded segments without requiring extensive branch networks.\u003c/p\u003e \u003c/div\u003e"},{"header":"7. Robustness Checks","content":"\u003cp\u003eThis section subjects the baseline findings to extensive robustness checks addressing potential threats to identification.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e7.1 Placebo Tests\u003c/h2\u003e \u003cp\u003eThis analysis assigns counterfactual adoption years randomly to banks and re-estimate Eq.\u0026nbsp;(1). If the results reflect true causal effects rather than spurious correlation or pre-existing trends, placebo treatments should yield null effects. Table\u0026nbsp;5 reports results from 1,000 random reassignments. For ROA, the median placebo coefficient is 0.05 with 95% percentile range [-0.3, 0.4], while actual estimate (0.89) falls far outside this range (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Similarly, placebo CIR and inclusion coefficients are tightly centered on zero, with the actual estimates clearly outside null distributions. These placebo tests strongly support causal interpretation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e7.2 Alternative Digital Adoption Measures\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;6 examines sensitivity to alternative digital adoption index constructions. Row 1 replicates baseline PCA-based index results. Row 2 uses equal-weighted averaging of five components; coefficients on ROA (0.82), CIR (-4.5), and inclusion (15.9) are nearly identical to baseline. Row 3 constructs the index using only transaction volume shares, excluding binary indicators; results remain similar (ROA: 0.76, CIR: -4.1, inclusion: 14.8). Row 4 employs factor analysis instead of PCA; results unchanged. Row 5 uses a higher binary threshold (DBAI\u0026thinsp;\u0026ge;\u0026thinsp;0.7); estimated effects are larger (ROA: 1.15), consistent with more intensive treatment. Collectively, these tests indicate that results are not artifacts of index construction choices.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e7.3 Sample Splits and Heterogeneity\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;7 explores treatment effect heterogeneity across bank characteristics. Panel A splits by bank size (above versus below median assets). Large banks exhibit slightly larger ROA effects (1.05 versus 0.72, difference not significant) but similar CIR and inclusion effects, suggesting efficiency gains and inclusion are not size-dependent. Panel B splits by ownership (domestic versus joint-venture with foreign partners). Joint-venture banks show marginally larger effects, potentially reflecting technology spillovers from international partners, though differences are not statistically significant. Panel C splits by initial (2010) performance quartiles. Interestingly, initially weak performers (bottom quartile ROA) exhibit the largest treatment effects (1.38 percentage point ROA increase), suggesting digital adoption offers catch-up opportunities for lagging institutions. This pattern mitigates concerns that only high-performing banks successfully adopt digital services.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e7.4 Additional Controls and Alternative Specifications\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;8 examines sensitivity to additional controls that may confound the digital adoption-performance relationship. Column 1 adds loan-to-deposit ratio (LDR) to capture liquidity management; baseline coefficients remain stable. Column 2 adds non-performing loan (NPL) ratio to control for asset quality; again, coefficients unchanged. Column 3 includes both LDR and NPL; results robust. Column 4 adds bank age (years since establishment) to control for experience; no material change. Column 5 controls for branch density (branches per square km) to capture geographic strategy; coefficients stable. Column 6 simultaneously includes all additional controls; baseline findings persist. These tests indicate that omitted variable bias does not drive the results.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e7.5 Alternative Clustering and Inference\u003c/h2\u003e \u003cp\u003eBaseline standard errors are clustered at the bank level to account for within-bank serial correlation. Table\u0026nbsp;9 explores alternative inference approaches. Row 1 replicates baseline one-way clustering. Row 2 employs two-way clustering by bank and year to allow for arbitrary correlation within banks over time and across banks within years; standard errors increase modestly but significance levels unchanged. Row 3 uses wild cluster bootstrap with 1,000 replications; p-values similar to baseline. Row 4 reports Conley (1999) spatial standard errors accounting for correlation across geographically proximate banks; again, results robust. These tests confirm that baseline inference is conservative.\u003c/p\u003e \u003c/div\u003e"},{"header":"8. Discussion","content":"\u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e8.1 Policy Implications\u003c/h2\u003e \u003cp\u003eThese findings yield several policy implications for financial regulators and development practitioners. First, the positive causal effects on both bank performance and financial inclusion suggest that digital banking adoption is a rare \"win-win\" intervention, obviating traditional trade-offs between profitability and social objectives. Policymakers should therefore prioritize removing barriers to digital adoption including regulatory uncertainty, cybersecurity concerns, and infrastructure gaps rather than viewing digital finance as requiring subsidies or mandates. Second, the 2\u0026ndash;3 year lag between adoption and full effect realization implies that patience is warranted; short-term performance declines during digital transitions should not trigger premature regulatory intervention. Third, the larger inclusion effects in initially underserved areas suggest that targeted incentives for digital expansion into rural and low-income markets could amplify inclusion gains. Such incentives might include regulatory flexibility for agent banking in remote areas or public investment in digital infrastructure (internet connectivity, electricity reliability). Fourth, the complementarity between digital channels and physical branches reflected in the mechanism analysis indicates that digital banking should be viewed as augmenting rather than replacing traditional infrastructure, with implications for branch licensing policies. Finally, the use of publicly available data demonstrates that rigorous impact evaluation is feasible without costly proprietary surveys, potentially informing regulatory data collection priorities.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e8.2 Limitations and Future Research\u003c/h2\u003e \u003cp\u003eSeveral limitations warrant acknowledgment. First, while the DiD design identifies average treatment effects, it does not fully illuminate distributional impacts, specifically whether inclusion gains benefit the poorest quintile or merely expand services to the \u0026ldquo;almost included.\u0026rdquo; Future research combining bank-level administrative data with household surveys could decompose inclusion effects across income deciles. Second, the measure of digital adoption (DBAI) captures supply-side availability but not demand-side usage; banks may offer digital services that customers underutilize. Incorporating transaction-level data on actual digital channel usage would refine measurement. Third, this study cannot definitively rule out all confounding unobservables despite extensive controls and robustness checks; future research could exploit quasi-random variation such as staggered regulatory policy changes or natural experiments. Fourth, the focus on private commercial banks excludes state-owned and foreign banks, which may respond differently to digital opportunities given distinct objectives and constraints. Fifth, this study examines a single country; cross-country replication is needed to assess generalizability, particularly to contexts with different regulatory regimes, infrastructure quality, or financial literacy levels. Sixth, the outcome measures (ROA, CIR, deposit account penetration) capture immediate effects but miss longer-run welfare impacts such as poverty reduction or entrepreneurship promotion, which require longitudinal household tracking. Finally, this study does not examine financial stability implications, including whether digital banking increases systemic risk through operational dependencies or interconnectedness, which remains an important avenue for future research.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003e8.3 Broader Implications for Development Finance\u003c/h2\u003e \u003cp\u003eBeyond Bangladesh, these findings contribute to ongoing debates about digital finance\u0026rsquo;s role in economic development. The efficiency channel documented here suggests that technology-driven cost reductions can overcome financial inclusion\u0026rsquo;s \u0026ldquo;last mile problem,\u0026rdquo; which refers to the challenge of profitably serving geographically dispersed, low-balance customers. This mechanism differs from microfinance\u0026rsquo;s group lending approach (which addressed information asymmetries through social collateral) and government-mandated priority sector lending (which cross-subsidized unprofitable segments), instead relying on automation to align commercial incentives with social objectives. Whether this model scales to the most excluded populations, particularly those lacking mobile phones, digital literacy, or reliable internet, remains an open question requiring further research. Additionally, the finding that joint-venture banks with foreign partnerships exhibit marginally stronger effects hints at technology transfer benefits from financial sector openness, though this study lacks statistical power to firmly establish this channel. Finally, the relatively modest absolute magnitude of inclusion effects (15\u0026ndash;22 additional accounts per 1,000 adults) underscores that digital banking is not a panacea; complementary interventions addressing financial literacy, consumer protection, and infrastructure gaps remain essential.\u003c/p\u003e \u003c/div\u003e"},{"header":"9. Conclusion","content":"\u003cp\u003eThis study provides the first rigorous causal evidence on digital banking adoption's impact on bank performance and financial inclusion using comprehensive institutional data from Bangladesh over 2010\u0026ndash;2025. Exploiting staggered adoption timing within a difference-in-differences framework, this study finds that digital banking adoption causally improves bank profitability (ROA increases by 0.8\u0026ndash;1.2 percentage points), enhances operational efficiency (cost-to-income ratios decline by 4\u0026ndash;6 percentage points), and expands financial inclusion (deposit accounts per 1,000 adults increase by 15\u0026ndash;22 percent). These effects materialize gradually over 2\u0026ndash;3 years post-adoption, are robust to extensive specification checks and alternative measurement approaches, and are largest in initially underserved markets.\u003c/p\u003e \u003cp\u003eMechanistically, this study shows that efficiency gains from transaction automation and reduced branch operating costs enable profitable service provision to previously unprofitable customer segments, reconciling performance and inclusion objectives. This contrasts with traditional views of financial inclusion as requiring cross-subsidization or regulatory mandates, instead highlighting technology's potential to align commercial incentives with social goals. These findings suggest that removing barriers to digital adoption\u0026mdash;through regulatory clarity, infrastructure investment, and cybersecurity capacity-building\u0026mdash;may be more effective than supply-side mandates in promoting inclusive finance.\u003c/p\u003e \u003cp\u003eMethodologically, this analysis demonstrates that rigorous causal inference is achievable using only publicly available regulatory data, offering a replicable template for researchers and policymakers in data-constrained environments. The DiD design with extensive validation (parallel trends tests, placebo tests, robustness checks) provides confidence in causal interpretation while acknowledging limitations inherent in observational data. Future research should extend this approach to other developing economies, incorporate distributional welfare analysis, and examine longer-run impacts on poverty reduction and economic mobility.\u003c/p\u003e \u003cp\u003eAs digital transformation continues reshaping financial services globally, understanding its causal impacts on institutional performance and societal outcomes becomes increasingly critical. The findings from Bangladesh a country representative of South Asian and broader developing-world challenges and opportunities provide evidence-based guidance for policymakers navigating the opportunities and risks of digital finance.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics Approval and Consent to Participate\u003c/p\u003e\n\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\n\u003cp\u003e\u003cstrong\u003eConsent for Publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This manuscript does not contain data from any individual person. All data presented are aggregated at the institutional (bank) level and derived from publicly available sources. No individual person\u0026apos;s details, images, videos, or case reports are included in this manuscript.\u003c/p\u003e\n\u003cp id=\"_Toc221307991\"\u003eFunding\u003c/p\u003e\n\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 by the author 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\n\u003cp\u003eCompeting Interests\u003c/p\u003e\n\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 and do not represent the official position of Mutual Trust Bank PLC or any other organization.\u003c/p\u003e\n\u003cp\u003eData Availability Statement\u003c/p\u003e\n\u003cp\u003eAll data supporting the findings of this study are publicly available:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBank Financial Data (2010\u0026ndash;2025):\u0026nbsp;\u003c/strong\u003eAudited annual reports from all 43 private commercial banks are publicly available through institutional websites and the Bangladesh Bank repository (https://www.bb.org.bd).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRegulatory Data:\u0026nbsp;\u003c/strong\u003eBangladesh Bank publications including Annual Reports, Financial Stability Reports, and Mobile Financial Services Reports are publicly accessible at https://www.bb.org.bd/pub/.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinancial Inclusion Data:\u0026nbsp;\u003c/strong\u003eWorld Bank Global Findex Database (https://globalfindex.worldbank.org/) and FinScope Bangladesh Survey data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReplication Materials:\u0026nbsp;\u003c/strong\u003eThe author commits to providing detailed data extraction procedures, variable construction code, and analysis scripts upon reasonable request. Aggregated datasets (with bank identifiers anonymized if required) can be shared for verification purposes, subject to Bangladesh Bank data sharing guidelines.\u003c/p\u003e\n\u003cp\u003eNo proprietary or confidential data were used. All analyses can be independently replicated using the publicly available sources listed above.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAthey, S., \u0026amp; Wager, S. (2018). Estimation and inference of heterogeneous treatment effects using random forests. Journal of the American Statistical Association, 113(523), 1228-1242.\u003c/li\u003e\n \u003cli\u003eAthanasoglou, P. P., Brissimis, S. N., \u0026amp; Delis, M. D. (2008). Bank-specific, industry-specific and macroeconomic determinants of bank profitability. Journal of International Financial Markets, Institutions and Money, 18(2), 121-136.\u003c/li\u003e\n \u003cli\u003eBangladesh Bank. (2024). Annual Report 2023-2024. Bangladesh Bank, Dhaka.\u003c/li\u003e\n \u003cli\u003eChowdhury, M. A., \u0026amp; Salman, M. A. G. (2021). Bank-specific \u0026amp; macroeconomic determinants of profitability: Empirical evidence from Bangladeshi private commercial banks. American Journal of Theoretical and Applied Business, 7(4), 72-80.\u003c/li\u003e\n \u003cli\u003eDavis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319-340.\u003c/li\u003e\n \u003cli\u003eDemirg\u0026uuml;\u0026ccedil;-Kunt, A., Klapper, L., Singer, D., \u0026amp; Ansar, S. (2022). The Global Findex Database 2021: Financial Inclusion, Digital Payments, and Resilience in the Age of COVID-19. World Bank, Washington, DC.\u003c/li\u003e\n \u003cli\u003eDietrich, A., \u0026amp; Wanzenried, G. (2011). Determinants of bank profitability before and during the crisis: Evidence from Switzerland. Journal of International Financial Markets, Institutions and Money, 21(3), 307-327.\u003c/li\u003e\n \u003cli\u003eFrost, J. (2020). The economic forces driving fintech adoption across countries. BIS Working Papers No. 838, Bank for International Settlements.\u003c/li\u003e\n \u003cli\u003eOzili, P. K. (2021). Financial inclusion research around the world: A review. Forum for Social Economics, 50(4), 457-479.\u003c/li\u003e\n \u003cli\u003eParvin, S., Chowdhury, A. N. M. M. H., Siddiqua, A., \u0026amp; Ferdous, J. (2019). Effect of liquidity and bank size on the profitability of commercial banks in Bangladesh. Asian Business Review, 9(1), 7-10.\u003c/li\u003e\n \u003cli\u003ePhan, D. V., \u0026amp; Pham, T. T. (2025). Causal and nonlinear effects of digital financial inclusion on bank stability: Evidence from emerging and advanced economies. Banks and Bank Systems, 20(4), 153-171.\u003c/li\u003e\n \u003cli\u003ePhilippon, T. (2016). The fintech opportunity. BIS Working Papers No. 655, Bank for International Settlements.\u003c/li\u003e\n \u003cli\u003eSahay, R., von Allmen, U. E., Lahreche, A., Khera, P., Ogawa, S., Bazarbash, M., \u0026amp; Beaton, K. (2020). The promise of fintech: Financial inclusion in the post COVID-19 era. IMF Departmental Paper No. 2020/009, International Monetary Fund.\u003c/li\u003e\n \u003cli\u003eShaikh, A. A., Alamoudi, H., Alharthi, M., \u0026amp; Glavee-Geo, R. (2023). Advances in mobile financial services: A review of the literature and future research directions. International Journal of Bank Marketing, 41(1), 1-33.\u003c/li\u003e\n \u003cli\u003eWilliamson, O. E. (1985). The Economic Institutions of Capitalism. Free Press, New York.\u003c/li\u003e\n\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":"Digital banking, Financial inclusion, Bank performance, Difference-in-differences, Bangladesh, Causal inference","lastPublishedDoi":"10.21203/rs.3.rs-9055268/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9055268/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFinancial inclusion remains a critical challenge in developing economies, with 1.4\u0026nbsp;billion adults worldwide unbanked. Digital banking has emerged as a potential solution, but rigorous causal evidence on its impacts is scarce. This study examines the causal impact of digital banking adoption on bank performance and financial inclusion in Bangladesh using bank-level data.\u003c/p\u003e\n\u003cp\u003eWe employ difference-in-differences analysis with bank and year fixed effects on panel data from all 43 private commercial banks in Bangladesh (2010–2025). We exploit staggered timing of digital adoption across banks to identify causal effects on profitability, operational efficiency, and financial inclusion.\u003c/p\u003e\n\u003cp\u003eDigital banking adoption causally improves return on assets by 0.8–1.2 percentage points, reduces cost-to-income ratios by 4–6 percentage points, and increases deposit accounts per 1,000 adults by 15–22 percent. Effects emerge gradually over 2–3 years post-adoption and are robust to alternative specifications, placebo tests, and parallel trends validation.\u003c/p\u003e\n\u003cp\u003eEfficiency improvements from digital banking enable banks to profitably serve previously underserved populations. Results challenge the view that digital banking primarily benefits already-included populations and support policies promoting digital transformation for financial inclusion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJEL Classification: \u003c/strong\u003eG21, O16, O33, C23\u003c/p\u003e","manuscriptTitle":"The Economic Impact of Digital Banking Adoption on Bank Performance and Financial Inclusion: Evidence from Bangladesh (2010–2025)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-11 03:17:29","doi":"10.21203/rs.3.rs-9055268/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 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64087277,"name":"Macroeconomics"},{"id":64087278,"name":"Finance"}],"tags":[],"updatedAt":"2026-03-11T03:17:29+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-11 03:17:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9055268","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9055268","identity":"rs-9055268","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}