Beyond Digital Transactions: Explaining the Compliance–Autonomy Gap in Microfinance Borrowers’ Adoption of Digital Financial Services in Jharkhand, India

1. Introduction

Digital financial services have moved from being a supplementary channel of financial service delivery to an important component of contemporary financial inclusion. Mobile applications, instant payment systems, agent networks, digital account management, and electronic repayment mechanisms have changed the way financial institutions interact with customers. For microfinance institutions (MFIs), digitalisation offers operational advantages such as faster processing, lower cash-handling requirements, improved record keeping, and greater transaction traceability. At the borrower level, however, the meaning of digitalisation is less straightforward. A transaction can be recorded as digital even when the borrower does not personally initiate, authenticate, or understand it.

This distinction matters particularly in low-income and financially constrained settings. The recent literature consistently shows that digital financial service adoption is shaped by a combination of usefulness, ease of use, trust, risk, literacy, infrastructure, and social context rather than by access to technology alone. Kajol et al. (2022) identify perceived usefulness, perceived ease of use, trust, security, and facilitating conditions among the most frequently examined determinants of digital financial transaction adoption. A later meta-analysis similarly confirms the importance of usefulness, ease of use, trust, and security-related dimensions (Neves et al., 2023). A systematic review focused specifically on low-income populations in low- and middle-income countries further argues that adoption is multidimensional, involving structural, knowledge, quality, and social-psychological conditions, with no single factor consistently dominating across contexts.

The Indian context is especially relevant because the country combines rapid digital-payment expansion with substantial variation in users’ capabilities and usage depth. Research on Indian consumers has shown that trust, facilitating conditions, performance expectancy, financial literacy, grievance redressal, and perceived risk can shape digital-payment adoption (Patil et al., 2020; Kumar et al., 2023; Jacob et al., 2024). Yet much of this literature examines bank customers, young consumers, urban users, or users who are already engaged with digital payment systems. Low-income microfinance borrowers constitute a different population. Their interaction with digital finance can be institutionally mediated, socially embedded, and linked to continued access to credit.

This paper therefore asks a more specific question: when a microfinance borrower performs a digital transaction, does that transaction represent genuine digital financial adoption or merely successful completion of an institutionally expected task? The question leads to the central concept developed in this study—the Compliance–Autonomy Gap. The concept distinguishes between digital activity that is driven by institutional requirements and digital behaviour that reflects independent financial agency. The distinction is important because repayment, fund transfer, savings, bill payment, balance checking, and new credit applications are not equivalent forms of digital engagement. A borrower may be fully capable of completing a repayment while remaining unable or unwilling to use digital channels for other financial decisions.

The study also develops a second conceptual distinction: Trust Asymmetry. The evidence suggests that borrowers may trust the MFI as an institution while remaining uncertain about the digital technology through which the institution delivers services. This is theoretically important because institutional trust should not be assumed to transfer automatically to a technological artefact. Recent Indian evidence on principal–agent trust similarly shows that low digital proficiency and trust deficits can make intermediaries central to digital financial service use (Yadav & Kalluru, 2024).

The paper makes four contributions. First, it reframes digital microfinance adoption from a single outcome into a continuum of increasingly autonomous behaviour. Second, it introduces the Compliance–Autonomy Gap as a borrower-centred lens for interpreting apparently high digital transaction rates. Third, it separates institutional trust from technological trust and uses the resulting Trust Asymmetry to explain why institutional confidence may coexist with digital hesitation. Fourth, it proposes the Trust-Enabled Phygital Adoption Framework (TEPAF), a conceptual model that places human assistance between institutional exposure and independent digital use. The framework is proposed from the present evidence; it is not claimed to be a statistically validated new theory.

2. Literature Review and Theoretical Background

2.1. Technology Acceptance Model and Digital Financial Behaviour

The Technology Acceptance Model (TAM) explains technology acceptance through perceived usefulness and perceived ease of use (Davis, 1989). Its continuing relevance in digital finance is evident from recent systematic reviews, which show that TAM and its extensions remain widely used in research on digital financial transactions and FinTech adoption (Kajol et al., 2022; Bhowal & Shil, 2026). Perceived usefulness captures the extent to which a user expects a technology to improve the effectiveness or convenience of a task, while perceived ease of use concerns the effort required to operate the system.

TAM provides a useful starting point for microfinance because borrowers may evaluate digital services in highly practical terms. If a digital repayment is faster, reduces travel, provides a clear record, or avoids carrying cash, it may be perceived as useful. Yet the same borrower may still hesitate to initiate a fund transfer if the process requires unfamiliar menus, authentication steps, or confidence in a digital interface. In this sense, usefulness can encourage acceptance without necessarily producing autonomous usage.

2.2. UTAUT, Social Context, and Facilitating Conditions

UTAUT broadens technology acceptance by incorporating performance expectancy, effort expectancy, social influence, and facilitating conditions (Venkatesh et al., 2003). This extension is valuable in financially constrained settings because digital behaviour is rarely determined by individual cognition alone. Family members, peers, field agents, community practices, network connectivity, device access, and service support can all shape actual opportunities to use technology.

Indian studies support the relevance of these factors. Patil et al. (2020) found that social influence and facilitating conditions were important in explaining mobile-payment intention and use behaviour. Nandru et al. (2023) similarly used an extended UTAUT2 model to examine Indian mobile-payment adoption, while Linge et al. (2023) found facilitating conditions to have a particularly strong role among Indian mobile-payment users. Among low-income Indian consumers, Jacob et al. (2024) reported positive effects of performance expectancy, effort expectancy, and facilitating conditions and a negative relationship between perceived risk and UPI intention. These findings are consistent with the argument that technology acceptance in low-income environments is partly a matter of whether the surrounding environment makes independent use feasible.

2.3. Trust as an Institutional and Technological Construct

Trust is central to financial technology because digital financial transactions involve uncertainty about system performance, privacy, security, error resolution, and the behaviour of service providers. Trust research distinguishes confidence in an institution or other social actor from trust in the technological system itself (McKnight et al., 2002). In digital finance, this distinction becomes especially important when the user already has a long-standing relationship with a bank, MFI, or field agent but is comparatively unfamiliar with the digital interface.

The distinction is supported by recent evidence. Kumar et al. (2023) found that perceived risk and trust shape mobile-banking adoption in India, while Yadav and Kalluru (2024) show that principal–agent trust and intermediary structures influence the use of digital financial services. More recent research continues to position trust as a central mechanism in FinTech adoption (Dahiya et al., 2026), and meta-analytic evidence indicates that trust and risk have systematic relationships with technological adoption across financial services (Ladeira et al., 2026).

The present study therefore treats trust as two related but distinct constructs: Trust in MFI (TMFI) and Trust in Technology (TTEC). The conceptual expectation is that institutional trust may facilitate exposure to digital finance, whereas technological trust is more directly relevant to independent execution. The empirical findings provide an opportunity to examine whether these forms of trust actually move together.

2.4. Perceived Risk and the Problem of Invisible Financial Loss

Perceived risk refers to the user’s expectation that a digital transaction may produce an undesirable outcome, such as monetary loss, failed transfer, privacy exposure, fraud, or inability to reverse an error. Digital financial services can therefore create a different psychological environment from cash-based transactions. Cash is tangible and its exchange is immediately visible; a digital balance or transaction status may be abstract, delayed, or difficult for a low-literacy user to interpret.

The literature consistently treats perceived risk as an inhibitor of digital financial adoption. Kajol et al. (2022) identify perceived risk as one of the most frequently cited barriers to digital financial transactions. Kumar et al. (2023) report a significant role for risk and trust in Indian mobile-banking adoption, while Jacob et al. (2024) report a negative association between perceived risk and UPI intention among low-income Indian consumers. The 2026 meta-analysis by Ladeira et al. further supports the general negative role of risk perception across technological financial services.

2.5. Digital Literacy and Capability

Digital literacy is broader than the possession of a smartphone. It includes the ability to navigate interfaces, interpret transaction messages, evaluate information, protect credentials, identify suspicious requests, and recover from errors. Evidence from digital-finance research increasingly treats digital literacy as a capability condition rather than a background demographic variable. Ullah et al. (2022), for example, connect financial skills and digital literacy with mobile-payment and mobile-banking intention. Sharma and Díaz Andrade (2023) identify technological skills and financial literacy as foundational dimensions of the human-development implications of digital financial services. Recent Indian research also continues to place digital capability near the centre of FinTech adoption.

2.6. From Transaction Adoption to Financial Autonomy

A limitation of conventional adoption studies is that adoption is often represented by one intention or usage variable. This is analytically convenient but can conceal differences in task complexity and user agency. Recent research has begun to examine the transition from intention to actual use and the role of contextual constraints. Chamboko (2024), for example, uses time-to-event analysis to examine digital financial service adoption, while Kumar et al. (2023) examine actual use alongside intention. Naeem et al. (2022) show how meanings, materials, and competencies shape the process of mobile-banking adoption. These studies suggest that adoption is better understood as a process than as a single binary event.

Building on this direction, the present study argues that microfinance digitalisation requires a distinction between institutional transaction completion and financial autonomy. The borrower-centred outcome is not simply whether a payment reached the MFI. It is whether the borrower can independently use digital finance for a broader range of legitimate financial purposes. This distinction is the conceptual basis of the Compliance–Autonomy Gap.

3. Conceptual Contribution: The Compliance–Autonomy Gap

The Compliance–Autonomy Gap refers to the difference between digitally completing institutionally required transactions and independently using digital financial services for self-directed financial management. In a microfinance setting, a borrower may be expected to repay digitally because the MFI has shifted from cash collection to electronic collection. The transaction is digital, but the borrower’s role may remain passive or assisted. By contrast, an autonomous user can initiate and complete a wider set of activities—such as transferring funds, checking balances, paying bills, managing savings, or seeking services—without depending on another person.

The concept does not imply that institutionally required transactions are unimportant. Digital repayment can reduce operational costs, increase traceability, and reduce cash-management burdens. The conceptual point is narrower: transaction volume should not be treated as a complete proxy for borrower empowerment. A system can be digitally efficient for the institution while remaining only partially empowering for the borrower.

Figure 1. Trust-Enabled Phygital Adoption Framework (TEPAF). Note. The framework is conceptual and derived from the theoretical literature and the empirical pattern observed in the present study. The intermediate stages are not directly estimated as latent variables in the current dataset and should therefore be treated as propositions for subsequent testing.

Figure 1. Trust-Enabled Phygital Adoption Framework (TEPAF). Note. The framework is conceptual and derived from the theoretical literature and the empirical pattern observed in the present study. The intermediate stages are not directly estimated as latent variables in the current dataset and should therefore be treated as propositions for subsequent testing.

Proposed framework for explaining the transition from institutional exposure to independent digital financial use.

4. Research Objectives and Hypotheses

The study pursues the following objectives:

• To assess the extent and form of digital financial use among active microfinance borrowers in Ranchi district.
• To examine the relationships between TAM/UTAUT-related constructs and behavioural intention toward digital financial services.
• To distinguish institutional trust from trust in digital technology.
• To interpret the gap between institutionally required digital transactions and broader autonomous financial use.
• To develop a borrower-centred conceptual framework for moving from assisted digital use toward financial autonomy.

Based on TAM, UTAUT, trust and risk literature, the following hypotheses are specified:

H1. Perceived usefulness is positively associated with borrowers’ behavioural intention to use digital financial services.

H2. Perceived ease of use is positively associated with borrowers’ behavioural intention to use digital financial services.

H3. Digital literacy is positively associated with borrowers’ behavioural intention to use digital financial services.

H4. Trust in the microfinance institution is positively associated with borrowers’ behavioural intention to use digital financial services.

H5. Trust in technology is positively associated with borrowers’ behavioural intention to use digital financial services.

H6. Perceived risk is negatively associated with borrowers’ behavioural intention to use digital financial services.

H7. Social influence is positively associated with borrowers’ behavioural intention to use digital financial services.

H8. Facilitating conditions are positively associated with borrowers’ behavioural intention to use digital financial services.

5. Methodology

5.1. Research Design and Setting

The study follows a quantitative, cross-sectional design. The empirical setting is Ranchi district of Jharkhand, India, and the target population comprises active borrowers of regulated Non-Banking Financial Company–Microfinance Institutions (NBFC-MFIs). The research focuses on borrowers who were already connected to the formal microfinance system, allowing the study to examine digital adoption within an existing institutional relationship rather than among the general population.

5.2. Sampling and Data Collection

A multi-stage cluster sampling strategy was used. Administrative areas including Kanke, Namkum, and Ratu were selected to provide variation between peri-urban and more rural settings. Active borrowers were approached at Joint Liability Group (JLG) centre meetings, and 663 usable responses were obtained. Data collection used Computer/Tablet-Assisted Personal Interviews (CAPI). Local enumerators administered the structured questionnaire in locally understandable languages and dialects, including Hindi, Nagpuri, and variations of Santhali. This approach was intended to reduce comprehension barriers associated with self-administered, text-heavy questionnaires.

5.3. Measures

The questionnaire included demographic items, digital-use questions, and five-point Likert-scale measures for Digital Literacy (DL), Perceived Ease of Use (PEOU), Perceived Usefulness (PU), Trust in MFI (TMFI), Trust in Technology (TTEC), Perceived Risk (PR), Social Influence (SI), Facilitating Conditions (FC), and Behavioural Intention (BI). The constructs were adapted to the microfinance context and administered through interviewer-assisted data collection.

5.4. Analytical Strategy

Descriptive statistics were used to profile respondents and compare digital tasks. Internal consistency was assessed with Cronbach’s alpha. Shapiro–Wilk tests were used to inspect distributional normality. Independent-samples t-tests were used to examine gender differences in behavioural intention and perceived risk. Multiple OLS regression was used to estimate the association between the eight explanatory constructs and behavioural intention.

Two diagnostic results require explicit qualification. First, the reported variance inflation factors are extremely high for several predictors, indicating severe multicollinearity. Second, the Breusch–Pagan test is statistically significant, indicating heteroscedasticity. Accordingly, the coefficient estimates should not be interpreted as clean estimates of unique causal effects. The reported OLS results are retained for transparency, but a submission-ready version should ideally rerun the model with heteroscedasticity-robust standard errors and consider a dimensionality-reduction, penalised-regression, or SEM/PLS-SEM strategy after checking the measurement model.

6. Results

6.1. Respondent Profile

Variable	Category	n	%
Gender	Female	398	60.0
	Male	265	40.0
Age	20–30 years	212	32.0
	31–40 years	225	34.0
	41–50 years	132	19.9
	>50 years	94	14.1
Education	Illiterate / Basic	198	29.9
	Secondary / High School	265	40.0
	Graduate / Higher	200	30.1
Occupation	Agriculture / Daily Labour	332	50.1
	Small Business / Trade	265	40.0
	Service / Other	66	9.9

The sample is predominantly female (60.0%), and 66.0% of respondents are between 20 and 40 years of age. Educational attainment is mixed: 29.9% report illiterate/basic education, 40.0% report secondary or high-school education, and 30.1% report graduate or higher education. Agriculture/daily labour is the largest occupational category (50.1%), followed by small business/trade (40.0%). These characteristics are important because digital adoption occurs within households and livelihoods where time, income, device access, and digital experience may vary considerably.

6.2. Evidence of the Compliance–Autonomy Gap

Digital activity	Reported use (%)	Interpretive position
Digital loan repayment	84.9	High institutional/transaction compliance
Fund transfer	12.1	Low evidence of broader autonomous financial use

The difference between repayment and fund transfer is the central descriptive finding. Digital loan repayment is reported by 84.9% of borrowers, whereas fund transfer is reported by only 12.1%. The gap is therefore 72.8 percentage points. This pattern suggests that aggregate digital transaction adoption may overstate the depth of borrower-centred digital financial engagement. The finding is consistent with the broader literature showing that adoption and actual use are context dependent and that low-income populations face structural, knowledge, quality, and social-psychological constraints (Neves et al., 2023; Yadav & Kalluru, 2024).

6.3. Descriptive Statistics

Construct	Code	Mean	SD
Digital Literacy	DL	3.15	1.29
Perceived Ease of Use	PEOU	3.71	1.09
Perceived Usefulness	PU	3.58	1.19
Trust in MFI	TMFI	4.37	0.50
Trust in Technology	TTEC	3.38	1.16
Perceived Risk	PR	2.79	1.25
Social Influence	SI	3.35	1.24
Facilitating Conditions	FC	3.85	1.16
Behavioural Intention	BI	3.55	1.39

The descriptive pattern provides direct evidence for Trust Asymmetry. Trust in the MFI is the highest-rated construct (M = 4.37, SD = 0.50), while Trust in Technology is lower (M = 3.38, SD = 1.16). Perceived Ease of Use (M = 3.71) and Facilitating Conditions (M = 3.85) are relatively favourable, while Digital Literacy is more moderate (M = 3.15). Behavioural Intention averages 3.55, suggesting a generally positive but heterogeneous orientation toward digital financial use.

6.4. Reliability and Distributional Diagnostics

Cronbach’s alpha values range from .702 for Trust in MFI to .983 for Behavioural Intention. The reported values indicate acceptable to very high internal consistency. However, the exceptionally high alpha for some constructs should not automatically be interpreted as superior measurement; very high values can also indicate item redundancy and should be examined alongside factor loadings and construct validity in a future CFA/SEM analysis.

Shapiro–Wilk tests were statistically significant for the principal constructs (p ≤ .001), indicating departures from normality. With N = 663, the OLS coefficient estimates may still be usable as large-sample approximations, but the more consequential concerns in this dataset are multicollinearity and heteroscedasticity.

6.5. Multicollinearity and Heteroscedasticity

Predictor	VIF
Digital Literacy	60.61
Perceived Ease of Use	44.61
Perceived Usefulness	35.02
Trust in MFI	1.82
Trust in Technology	79.17
Perceived Risk	89.08
Social Influence	36.29
Facilitating Conditions	54.89

The VIF results indicate severe multicollinearity. Values far above conventional thresholds suggest that several psychological and contextual constructs share substantial variance. In practical terms, this means that a regression coefficient should not be read as the isolated effect of one construct while all others are held constant in a substantively stable sense. The particularly high VIFs for perceived risk and trust in technology are important for interpreting their negative coefficients.

The Breusch–Pagan test produced LM = 26.3199, p = .0009, rejecting homoscedasticity. Therefore, the original standard errors are not ideal for final inferential reporting. The present manuscript reports the original estimates but explicitly labels them as unadjusted OLS results. Before journal submission, HC3 or another appropriate heteroscedasticity-robust estimator should be used and the hypothesis decisions should be updated if the robust p-values differ.

6.6. Regression Results

Predictor	β	SE	t	p	Decision at .05
Intercept	0.7301	0.781	0.934	.353	—
Digital Literacy	0.1395	0.106	1.321	.190	Not supported
Perceived Ease of Use	0.2086	0.108	1.931	.057	Not supported
Perceived Usefulness	0.2766	0.087	3.169	.002	Supported
Trust in MFI	0.0682	0.048	1.421	.159	Not supported
Trust in Technology	−0.2752	0.134	−2.050	.043	Not supported; opposite sign
Perceived Risk	−0.2735	0.133	−2.056	.043	Supported
Social Influence	0.2372	0.085	2.788	.006	Supported
Facilitating Conditions	0.3162	0.112	2.821	.006	Supported

The overall model reports R² = .985, adjusted R² = .984, and F = 768.6, p < .001. These values indicate a very high in-sample fit. However, given the severe VIF values and heteroscedasticity, the high R² should not be described as definitive evidence of exceptional explanatory power. It may partly reflect overlapping predictor information. The principal contribution of the regression is therefore pattern identification rather than causal attribution.

Perceived Usefulness is positively associated with behavioural intention (β = .2766, p = .002), supporting H1 in the reported unadjusted model. Perceived Ease of Use is positive but falls just above the conventional .05 threshold (β = .2086, p = .057), so H2 is not supported at the 5% level, although the result may be described as marginal under a pre-specified 10% criterion. Digital Literacy is positive but nonsignificant (β = .1395, p = .190), so H3 is not supported. Trust in MFI is also nonsignificant (β = .0682, p = .159), providing no direct support for H4.

Trust in Technology has a significant negative coefficient (β = −.2752, p = .043), which is opposite to the hypothesised positive relationship. H5 is therefore not supported. The unexpected sign should not be presented as evidence that trust itself discourages adoption. Given the VIF of 79.17 and the conceptual overlap between trust and risk, the coefficient may be unstable or may capture suppression effects. Perceived Risk is negatively associated with behavioural intention (β = −.2735, p = .043), supporting H6 in the unadjusted model. Social Influence (β = .2372, p = .006) supports H7, and Facilitating Conditions (β = .3162, p = .006) supports H8.

6.7. Gender Differences

The independent-samples t-test for behavioural intention produced t = −1.766, p = .0806. The result does not meet the conventional 5% significance criterion. The perceived-risk comparison produced t = 1.802, p = .0746 and is likewise nonsignificant at the 5% level. These results suggest that gender differences should not be presented as statistically established in the current sample.

7. Discussion

7.1. Digital Adoption Is not a Single Behaviour

The most important finding is the difference between digital repayment and fund-transfer use. An 84.9% repayment rate could easily be reported as evidence of successful digital adoption. Yet only 12.1% report fund-transfer use. The two measures represent different degrees of agency. Repayment is closely connected to the borrower’s institutional obligation, whereas fund transfer is more discretionary and requires the user to recognise a need, select an action, authenticate it, and evaluate the outcome.

This interpretation is consistent with research that distinguishes behavioural intention from actual use. Patil et al. (2020) explicitly examined use behaviour rather than intention alone, while Chamboko (2024) demonstrates that the timing and process of adoption can be studied as an event rather than as a static state. The present study extends that logic by distinguishing not merely intention from use, but institutionally required use from self-directed use.

The Compliance–Autonomy Gap therefore provides a useful interpretive correction. A digital repayment can demonstrate access to a digital channel, but it does not by itself demonstrate digital financial autonomy. For MFIs and policymakers, this distinction changes what should be measured. In addition to counting transactions, organisations should examine transaction diversity, independent initiation, user authentication, error-recovery capability, and the borrower’s confidence in using digital finance beyond the lending relationship.

7.2. Trust in the Institution Does Not Automatically Become Trust in the Technology

The descriptive evidence is particularly informative because Trust in MFI is substantially higher than Trust in Technology. The finding suggests that the borrower’s relationship with the institution and the borrower’s relationship with the digital channel are not identical. This is theoretically plausible. Institutional trust develops through repeated interpersonal interactions, predictable repayment procedures, group meetings, and familiarity with staff. Technology trust requires confidence that the system will execute the transaction correctly, protect credentials, display accurate information, and provide an understandable remedy when something goes wrong.

The finding complements Yadav and Kalluru’s (2024) evidence that intermediaries remain important in digital financial service adoption where digital proficiency is limited. It also aligns with Kumar et al. (2023), who show that risk and trust are central in Indian mobile-banking adoption. The present study’s contribution is to place these two trust objects side by side within a microfinance context.

7.3. Perceived Risk Matters, but Its Coefficient Requires Caution

The negative coefficient for perceived risk is substantively consistent with the wider literature. Users who fear financial loss, fraud, privacy exposure, or transaction failure are less likely to intend to use digital financial services. The result is also consistent with Jacob et al. (2024), who report a negative relationship between perceived risk and UPI intention among low-income Indian consumers.

At the same time, the present study has unusually high collinearity between perceived risk and trust in technology. This means that the two variables should not be interpreted as independent psychological forces with cleanly separable effects. The safer conclusion is that risk and technological confidence form a tightly connected part of the borrower’s evaluation environment. Future studies should model this relationship explicitly, rather than assuming that separate regression coefficients fully represent separate behavioural mechanisms.

7.4. Facilitating Conditions Are Central to the Local Adoption Environment

Facilitating Conditions is the strongest positive coefficient in the reported model (β = .3162, p = .006). This result is consistent with the broader UTAUT literature and with Indian evidence that infrastructure and support conditions matter for digital-payment use. For borrowers in rural or peri-urban areas, a technically good application is not sufficient if network availability, electricity, device access, account support, or transaction assistance is unreliable.

The result also supports a shift from an application-centred view of digital inclusion to an ecosystem-centred view. Adoption is partly produced by the environment surrounding the technology. If the user cannot reliably complete the transaction, interpret the confirmation message, or obtain help after an error, perceived usefulness may not translate into autonomous use.

7.5. The Negative Coefficient for Trust in Technology Should be Treated as a Diagnostic Finding

The negative TTEC coefficient is the most counterintuitive result. Because the hypothesis predicted a positive relationship, the finding is not theoretically interpreted as evidence that greater technological trust reduces adoption. Instead, it should be treated as a diagnostic signal. With VIF = 79.17, the coefficient is vulnerable to suppression and instability. It may also indicate that the operationalisation of trust in technology overlaps strongly with perceived risk or other predictors.

This is precisely why the revised manuscript avoids presenting the regression as a definitive causal model. A stronger next-stage analysis should examine correlations, tolerance statistics, condition indices, item-level factor structure, discriminant validity, and robust or penalised estimates. If the raw dataset is available, a PLS-SEM or CFA-based model could test whether trust and risk are empirically distinct constructs before estimating their structural relationships.

8. Trust-Enabled Phygital Adoption Framework (TEPAF)

The empirical and theoretical findings support a staged interpretation of digital microfinance adoption. TEPAF proposes that borrowers can move through five broad states: (1) institutional compliance, (2) assisted digital use, (3) familiarity and confidence, (4) independent digital use, and (5) financial autonomy. The stages are conceptual rather than statistically estimated in the present study.

Stage	Core borrower condition	Institutional implication
1. Institutional compliance	Uses digital channel mainly because the institution requires or strongly encourages the transaction.	Measure whether the transaction is mandatory and who initiates it.
2. Assisted digital use	Transaction is completed with help from staff, family, agents, or other intermediaries.	Treat assistance as a transition mechanism, not automatically as autonomy.
3. Familiarity and confidence	User becomes able to interpret screens, confirmations, risks, and routine procedures.	Use repeated demonstrations, vernacular guidance, and low-risk practice.
4. Independent digital use	Borrower can initiate and complete multiple financial tasks without routine assistance.	Measure transaction diversity and independent initiation.
5. Financial autonomy	Digital services are used for self-directed financial management beyond institutional obligations.	Evaluate digital inclusion through agency, choice, and capability.

TEPAF differs from a conventional TAM extension in one important respect. It does not treat all digital use as behaviourally equivalent. Instead, it connects acceptance constructs with the social organisation of use. Human assistance is not framed only as a barrier; it can also function as a bridge. The objective is to convert assisted use into independent capability rather than allowing assistance to become a permanent substitute for capability.

9. Managerial and Policy Implications

9.1. Measure Autonomy, Not Only Transaction Volume

MFIs should complement transaction-count metrics with borrower-level indicators of independent use. Useful indicators include the proportion of transactions initiated directly by the borrower, number of different digital services used, frequency of independent authentication, ability to check transaction status, and ability to resolve common errors. Such measures would distinguish institutional efficiency from borrower empowerment.

9.2. Convert Field Staff from Transaction Intermediaries into Digital Educators

Field officers and group facilitators can become digital trust intermediaries. Instead of merely completing transactions for borrowers, they can demonstrate how to perform the task, explain the confirmation message, show how to verify the amount, and teach safe authentication practices. This recommendation follows the logic of principal–agent trust and the evidence that intermediated use remains important among users with limited digital proficiency (Yadav & Kalluru, 2024).

9.3. Design for Vernacular and Low-Literacy Use

Digital interfaces serving microfinance borrowers should minimise text density, use familiar terminology, support local languages where feasible, and provide clear audio or visual confirmation. Digital literacy should be treated as a design constraint rather than merely as a user deficit.

9.4. Make Transaction Confirmation Immediate and Understandable

For users who are uncertain about invisible digital balances, immediate confirmation is central to trust formation. Confirmation should clearly display the amount, recipient, transaction status, and a reference number. Where literacy is limited, auditory or assisted confirmation can supplement conventional text receipts.

9.5. Strengthen Grievance Redressal

The literature shows that grievance redressal can influence mobile-payment use (Patil et al., 2020). In microfinance, failed transactions may have disproportionate psychological consequences because the borrower is managing limited financial resources. A fast, accessible, vernacular complaint and reversal mechanism can reduce the perceived consequences of failure and support trust formation.

9.6. Treat Cybersecurity as an Inclusion Issue

Cybersecurity education should be integrated with financial inclusion rather than treated as a specialist technical issue. Borrowers should be taught never to share PINs, passwords, or authentication codes and should be given practical demonstrations of how legitimate support differs from fraudulent requests. The goal is not simply to warn users but to make safe behaviour feasible and understandable.

10. Theoretical Implications

The study suggests that TAM and UTAUT are useful but incomplete when digital finance is introduced into a relationship-based microfinance ecosystem. The core models explain perceptions of usefulness, effort, social influence, and facilitating conditions, but they do not by themselves distinguish between institutionally induced use and autonomous use. The Compliance–Autonomy Gap therefore adds a contextual layer to technology-acceptance theory.

Second, the separation of institutional trust and technological trust contributes to trust-based adoption research. A borrower can have strong confidence in the institution while remaining uncertain about the technological channel. This suggests that trust transfer should be treated as an empirical question rather than an automatic assumption.

Third, the phygital perspective places human assistance inside the adoption process. Rather than treating physical and digital channels as mutually exclusive, TEPAF proposes that human intermediation can facilitate the transition to independent use. This is compatible with emerging practice-oriented research showing that technology adoption is shaped by the interaction of material features, competencies, and local practices (Naeem et al., 2022; Kumar et al., 2023).

11. Limitations and Future Research

The study has several limitations. First, the cross-sectional design limits causal interpretation. The relationships reported by OLS should therefore be described as associations. Second, the sample is confined to active microfinance borrowers in Ranchi district and should not be generalised automatically to all borrowers in Jharkhand or India. Third, the current analysis uses behavioural intention as the regression outcome while the descriptive usage measures provide only a limited picture of actual task-level behaviour. Fourth, the severe multicollinearity and significant heteroscedasticity require stronger inferential modelling before publication in a high-impact outlet.

Fifth, the Compliance–Autonomy Gap and TEPAF are conceptual contributions derived from the observed pattern; they have not yet been operationalised as validated scales. Future research should develop and validate measures of mandatory use, assisted use, independent initiation, transaction diversity, and financial autonomy. A Borrower Digital Autonomy Index could be developed and tested through exploratory and confirmatory factor analysis.

Sixth, future studies should collect data on who actually performs each transaction. A borrower may own the account but receive assistance from a family member, agent, or field worker. Distinguishing account ownership, transaction initiation, authentication, and decision-making would provide a stronger test of the Compliance–Autonomy Gap.

Finally, the next empirical stage should rerun the present model using robust standard errors and alternative specifications. If the raw data are available, a measurement model should first establish convergent and discriminant validity. Structural equation modelling could then test direct and indirect relationships. If moderation is theoretically desired—for example, whether perceived risk changes the effect of trust or usefulness—an explicit interaction term must be estimated. The current regression does not contain such an interaction term and therefore should not be described as a moderation model.

12. Conclusion

Digitalisation in microfinance should not be judged only by whether a transaction has become electronic. The evidence from 663 borrowers in Ranchi indicates a substantial difference between digital repayment and broader digital financial use. When 84.9% of borrowers report digital repayment but only 12.1% report fund-transfer use, the issue is not simply whether digital finance has arrived. The more important question is whether borrowers have acquired the capacity to use it independently.

The study’s conceptual answer is the Compliance–Autonomy Gap. Institutionally successful digitalisation can coexist with limited borrower autonomy. The descriptive trust pattern reinforces this argument: borrowers report very high trust in the MFI but substantially lower trust in technology. The reported regression further highlights the importance of usefulness, social influence, facilitating conditions, and perceived risk, while the unusually high multicollinearity cautions against overinterpreting individual coefficients.

The proposed Trust-Enabled Phygital Adoption Framework offers a practical way forward. Human assistance should be used as a bridge to competence rather than as a permanent substitute for competence. For digital microfinance to become genuinely inclusive, borrowers need not only access to a digital channel but also the confidence, literacy, security awareness, and infrastructural support required to use that channel independently. The ultimate outcome of digital financial inclusion is therefore not the number of digital transactions recorded by an institution, but the extent to which digital capability expands the borrower’s financial agency.