Articles
Authors: Muhammad Khan (University of Salford) , Shahrukh Khan (Independent Researcher GIZ)
Digital Public Infrastructure (DPI) encompassing digital identity, interoperable payment systems, and consent-based data exchange has become a central organising concept in development finance and financial inclusion policy. Proponents argue that these foundational digital layers can accelerate financial access, reduce transaction costs, and catalyse broad-based economic development across the Global South. However, the relationship between DPI deployment and financial inclusion outcomes is neither automatic nor linear, and the conditions under which DPI delivers equitable, sustainable inclusion have remained underspecified in both theory and policy. This paper provides a structured comparative synthesis of twenty-two empirical and conceptual studies spanning India, Sub-Saharan Africa, Brazil, Ukraine, and Indonesia, supplemented by thirty additional academic references, to argue that DPI delivers equitable and sustainable financial inclusion only when three conditions hold simultaneously: the infrastructure must be interoperable and genuinely public in its governance; institutional quality must cross context-specific thresholds; and human capability development must be pursued as a co-investment rather than an afterthought. We term this the Conditional DPI-Inclusion Thesis and specify the empirical conditions under which it would be disconfirmed. Critically, we situate this synthesis within the development finance literature, addressing how DPI shapes allocation decisions, sequencing imperatives, DFI and MDB conditionality design, and aid effectiveness. We identify five mechanisms through which DPI affects financial inclusion, transaction cost reduction, information asymmetry mitigation, network externality generation, formalisation spillovers, and trust anchoring and assess the evidence quality for each. We demonstrate that where governance falls below context-specific thresholds, DPI risks entrenching inequalities rather than dismantling them. The paper concludes with a prioritised policy agenda for development finance institutions and a forward-looking research programme.
Keywords: Digital Public Infrastructure, Financial Inclusion, Development Finance, Digital Identity, Mobile Money, Governance, Aid Effectiveness
How to Cite: Khan, M. & Khan, S. (2026) “Digital Public Infrastructure, Financial Inclusion, and the Governance Imperative in Development Finance”, Ledger: The Salford Journal of Accounting and Finance. 1(1). doi: https://doi.org//ledger.416
Around 2019, the global development finance conversation shifted in a way that had been building for years before anyone named it properly. What changed was not the existence of mobile money or digital payments, those had been growing since M-Pesa rewired Kenya’s financial landscape in 2007, but rather the conceptual leap from thinking about digital financial services as products to thinking about digital infrastructure as a public good. The emergence of the Digital Public Infrastructure (DPI) paradigm, crystallised in India’s JAM (Jan Dhan accounts, Aadhaar, and Mobile numbers) trinity and later formalised across G20 working groups, the UNDP, and the World Bank, marks a genuine inflection point in how development finance practitioners and scholars conceptualise the architecture of financial inclusion.
The stakes for development finance are significant. If DPI functions as a multiplier of inclusion outcomes when combined with appropriate institutional investments, then development finance institutions (DFIs) and multilateral development banks (MDBs) face a fundamental reallocation question: should they finance digital infrastructure as a standalone technical intervention or as a component of bundled ecosystem investments that include governance strengthening, consumer protection, and digital literacy? The answer to this question determines not only project design but also conditionality frameworks, sequencing priorities, and the allocation logic of billions of dollars in concessional and blended finance annually.
And yet the gap between DPI’s conceptual ambitions and its empirical track record remains substantial. For every India Stack success story, 470 million new bank accounts opened between 2011 and 2017, a gender gap in account ownership that narrowed from 17% to 6%, KYC costs that collapsed from fifteen dollars to seven cents (D’Silva et al., 2019; Balasundaram et al., 2026), there is a Liberia where the national biometric ID procurement was cancelled in August 2025, leaving digital financial services geographically concentrated in Monrovia and structurally disconnected from the rural majority (Kanu, 2025). For every Brazil where Pix reached 93% of adults and transformed payment system competition within three years (Mazzucato et al., 2024), there is a Ukraine where a collapse in bank branch density failed to generate compensatory digital adoption because trust deficits and infrastructure weakness outweighed technology availability (Naumenkova et al., 2019).
This paper takes these contradictions seriously. Its central claim, the Conditional DPI-Inclusion Thesis, is that DPI delivers equitable and sustainable financial inclusion only when three conditions hold simultaneously: the infrastructure must be interoperable and genuinely public in its governance; institutional quality must cross context-specific thresholds; and human capability development must be pursued as a co-investment. Critically, the paper frames this claim not merely as a contribution to the DPI or fintech literature, but as a contribution to development finance scholarship: it addresses how DPI changes the allocation logic of development finance, how DFIs should sequence investments, what this implies for conditionality and risk-sharing, and what political economy constraints shape what donors can realistically fund.
This is not a claim against DPI, the evidence reviewed here consistently supports DPI’s potential. It is a claim against techno-optimism, and against the policy habit of mistaking a well-designed payments rail for a development outcome.
The term Digital Public Infrastructure suffers from a definitional ambiguity that has both accelerated its adoption and weakened its analytical purchase. In its most precise formulation, following D’Silva et al. (2019), Arner et al. (2018), and the UNDP’s 2023 DPI framework, DPI comprises three interlocking layers: (1) digital identity systems that enable secure, low-cost verification and authentication at scale; (2) interoperable digital payment rails that allow real-time, frictionless value transfer across institutions and channels; and (3) consent-based data exchange frameworks that enable individuals and organisations to share and act on financial and personal data without sacrificing privacy or control.
This three-pillar architecture matters because each layer enables specific inclusion mechanisms that the others cannot replicate. Digital identity addresses the documentation and trust barriers that have historically excluded the poorest from formal financial systems, the e-KYC process enabled by Aadhaar reduced onboarding costs by 99.5% for Indian banks (D’Silva et al., 2019). Payment rails address transaction cost and distance barriers, UPI’s zero-marginal-cost interoperability enabled 12 billion transactions monthly by late 2023 (Gupta, 2024). Data exchange addresses information asymmetry barriers that prevent credit markets from functioning for the poor, account aggregators in India’s system enable cash-flow-based lending for individuals and MSMEs with no conventional credit history (D’Silva et al., 2019).
What distinguishes DPI from ordinary digital government services or commercial fintech is its explicitly public character, a character that Mazzucato, Eaves, and Vasconcellos (2024) argue cannot be assumed but must be deliberately constructed through governance. Their framework distinguishes between DPI that is merely technically interoperable (attribute-based publicness), and DPI designed with explicit normative purpose, participatory governance, and accountability mechanisms (common-good DPI). This distinction has direct empirical consequences: Aadhaar’s exclusion incidents, in which biometric failures denied welfare payments to eligible beneficiaries, arose precisely because technical infrastructure was rolled out without adequate consumer protection, grievance redressal, or legal safeguards.
The development finance literature has engaged with digital technology primarily through the lenses of ICT4D and e-government, with relatively limited engagement with DPI as a distinct category of public investment that changes the allocation logic of the broader development finance portfolio (Gabor and Brooks, 2017). This paper argues for a more direct engagement.
First, DPI investment affects the allocation logic of development finance by generating positive complementarities with other portfolio investments. A functioning digital identity layer reduces the per-beneficiary cost of social protection delivery, health system access, and voter registration, effectively increasing the return on each of these separate investments. A payment rail that enables digital G2P transfers reduces leakage, which improves the effective disbursement rate of social protection programmes funded by DFIs and bilateral donors. The BIS estimates that India’s Aadhaar-enabled direct benefit transfers reduced leakage by US$12.7 billion in subsidies and transfers between 2014 and 2019 (D’Silva et al., 2019). From a portfolio perspective, this means that DPI investment, when well-governed, can shift the cost-effectiveness frontier of the entire development finance portfolio, not merely the digital finance component.
Second, DPI has direct implications for the sequencing of infrastructure, governance, and capability investments, a long-standing challenge in development finance. The evidence reviewed here consistently suggests that the failure mode of DPI investment is not technical but institutional: payment rails deployed ahead of consumer protection legislation, biometric identity systems rolled out without grievance redressal mechanisms, and digital lending products introduced before regulatory capacity can assess algorithmic fairness. The correct development finance response is not to delay DPI investment but to bundle it with institutional co-investments as a condition of financing rather than a separate, later intervention.
Third, DPI changes what DFI and MDB conditionality can and should encompass. Traditional financial sector conditionality has focused on macroeconomic stability, banking sector capitalisation, and regulatory frameworks for credit markets. DPI-era conditionality needs to extend to: interoperability mandates that prevent digital monopolisation; data protection legislation with enforcement capacity; fee oversight that prevents regressive pricing of digital payment services; and consumer literacy standards that ensure populations can meaningfully use, rather than merely access, digital financial services. Several MDBs, including the World Bank and the Asian Development Bank. have begun incorporating these elements into digital finance operations, but they remain inconsistently applied and rarely bundled as jointly financed packages.
Fourth, DPI has implications for aid effectiveness and public investment design. The evidence suggests that DPI investments have particularly high returns when they are state-led and interoperability-first, because private-sector-led DPI architectures tend to lock in proprietary standards and concentrate market power in ways that undermine the public-good properties that justify development finance engagement in the first place. This has a direct implication for public investment design: donors funding DPI in fragile and transitional states should condition support on open-standard architectures, mandatory interoperability, and non-exclusive licensing terms, not because of ideological commitment to openness, but because proprietary DPI consistently underperforms on inclusion metrics relative to open-architecture alternatives.
Fifth, political economy constraints shape what donors can actually fund. In many developing countries, the institutional actors who control the procurement and governance of digital identity and payment systems include telecoms incumbents, domestic banks, and government agencies with conflicting interests in the outcomes. A biometric identity system that enables effective tax collection may face resistance from informal economy actors with political influence. A payment interoperability mandate may face resistance from a dominant mobile money operator that benefits from current lock-in. Development finance institutions that treat these political economy constraints as externalities to the technical design process will consistently be surprised when technically sound DPI investments underperform. A more realistic development finance approach treats political economy feasibility assessment as a front-end component of DPI investment appraisal.
The argument developed in this paper is explicitly a development finance argument, not a fintech or ICT4D argument. The distinction matters. The fintech literature is primarily concerned with product innovation, market competition, and the disruption of incumbents by digital challengers, questions that are relevant to DPI design but not to development finance allocation decisions. The ICT4D literature is primarily concerned with the adoption and impact of information and communication technologies in development contexts, questions that overlap with DPI but do not engage the institutional, political economy, and sequencing dimensions that are central to development finance practice.
A development finance framing asks different and more practically relevant questions: What should a DFI’s conditionality framework look like when financing a national digital identity system? How should an MDB apportion concessional finance between infrastructure hardware, regulatory capacity, and consumer protection when the three are complements? What political economy conditions must be in place before a donor can reasonably expect a DPI investment to deliver the projected inclusion returns? How should risk be shared between public development finance, private sector co-investors, and the host government when DPI has significant public-good externalities but also commercial revenue potential?
These are questions that the DPI literature has not adequately addressed, and that this paper directly engages. The Conditional DPI-Inclusion Thesis is developed in this paper as a contribution to development finance scholarship, not merely as a descriptive account of when DPI works for financial inclusion.
The available evidence suggests that DPI can deliver significant financial inclusion gains, but the strength of that evidence varies considerably by context and methodological approach. The most compelling causal evidence comes from India. Balasundaram et al.’s (2026) difference-in-differences study, tracking 200 rural villages across four Indian states from 2019 to 2023, finds that high UPI adoption villages experienced a 27% rise in business registrations, a 58.3% increase in digital payment use, a 34% growth in account ownership, and a 42.1% increase in female-owned enterprises. Informal borrowing fell 53.1% and cash transactions declined 42.8%. The study deploys event-study pre-trend diagnostics, synthetic control counterfactuals, geographic buffer exclusions, and Oster bounds for unobservable selection, a methodological rigour that makes these findings among the most credible in the DPI literature.
The causal mechanisms are quantified through mediation decomposition: transaction cost reduction (42.2%), improved financial information access (42.4%), and network effects (39.1%) explain the bulk of UPI’s inclusion dividend. Gupta’s (2024) qualitative study, drawing on 150 low-income consumer and retailer interviews in Delhi-NCR, adds important texture: adoption of UPI among marginalised communities is fundamentally anchored in institutional trust generated by government ownership, multi-layered security authentication, and transparent grievance redressal, elements that commercial platforms structurally cannot replicate.
In Sub-Saharan Africa, the evidence base is extensive but more heterogeneous, and caution is warranted in generalising from any single study. Kouladoum, Wirajing and Nchofoung (2022) analyse 43 SSA countries from 2004 to 2019 using IV-Tobit and System GMM, finding that major ICT indicators have positive, statistically significant, and robust effects on multiple dimensions of financial inclusion. Notably, fixed broadband produces larger effects than mobile subscriptions, partly challenging the dominant mobile-first narrative, though this finding may reflect the study’s panel structure, which captures long-run infrastructure effects rather than short-run adoption dynamics.
Khera et al. (2022), constructing a multidimensional digital financial inclusion index across 52 emerging market and developing economies, document that digital financial inclusion increased in all countries between 2014 and 2017, including in eight where traditional inclusion declined, indicating substitution effects as users migrated to mobile channels. Zimbabwe’s mobile money transaction values reaching 140% of GDP is an extraordinary statistic, but it should be read in context: Zimbabwe’s mobile money dominance reflects the near collapse of formal banking rather than merely the success of digital infrastructure. This important contextual nuance is underplayed in much of the citation of this finding.
Van and Quoc (2025) provide cross-country evidence across 117 countries from 2004 to 2022 using panel threshold regression, revealing a critical non-linearity: digital financial inclusion positively affects sustainable development in countries with low and medium financial development, but exerts a statistically significant negative effect once financial development exceeds a threshold of 0.7176. This finding, suggesting that in highly financialised economies, expanded digital access may accelerate credit-financed consumption of energy-intensive goods, is among the more uncomfortable in the DFI literature and merits substantially more attention than it has received.
No finding in the DPI-financial inclusion literature is more consistently replicated than the governance moderation effect, though the mechanism through which governance conditions DPI’s effectiveness varies across studies. Chinoda and Kapingura (2024), analysing 25 SSA countries from 2014 to 2020 using System GMM with interaction terms, find that while digital financial inclusion has a direct positive effect on GDP per capita growth, this effect is substantially amplified when interacted with governance quality. Without adequate governance, the economic dividends of DFI are substantially dissipated.
Akpa, Asongu and Batuo (2025) compute governance thresholds below which internet diffusion fails to translate into financial inclusion gains in SSA: voice and accountability at approximately 0.300, government effectiveness at approximately 0.300, and rule of law at approximately 0.250 on standard WGI scales. These are empirically derived turning points from SGMM interaction terms, and they should be treated as approximate, model-specific signals rather than universal policy cutoffs, a critical qualification that the policy discussion in this paper returns to explicitly in Section 8. The majority of SSA’s fragile and conflict-affected states fall below these thresholds, which means, if the threshold estimates have any generalisability, that connectivity investments in these contexts may not deliver projected inclusion returns without simultaneous institutional investment.
Suhrab, Chen and Ullah’s (2024) study of BRICS economies adds that technological innovation and infrastructure development both strengthen DFI’s inequality-reducing effect. Without these complements, DFI’s standalone effect on inequality can be occasionally positive, meaning it can worsen rather than improve distributional outcomes in economies with insufficient innovation capacity or physical infrastructure. This finding directly challenges generic DFI-for-development narratives.
Gallego-Losada et al.’s (2023) bibliometric analysis of 387 DFI papers reveals a field whose intellectual structure is technically sophisticated but conceptually narrow. The keyword co-occurrence network shows financial inclusion dominating at 200 occurrences with 485 link-strength connections, clustered with fintech, mobile money, ICT, and economic development. The research-area concentration in Business and Economics reflects a field built primarily by economists and information systems researchers, with limited contribution from political scientists, sociologists, and legal scholars who could provide the governance and equity analysis the field structurally needs.
The measurement problem is epistemological as well as bibliometric. As Ozili (2018) argues, financial data inclusion is routinely conflated with financial inclusion. Individuals who hold dormant accounts registered through G2P programmes appear as included in standard metrics without ever meaningfully participating in the formal financial system. Khera et al.’s (2022) finding that access and usage can diverge sharply, and Malladi, Soni and Srinivasan’s (2021) documentation that PMJDY’s 355 million accounts coexist with persistent low transactional usage in rural India, both point to the same measurement failure: development finance metrics have been counting architecture and calling it inclusion.
Mazzucato, Eaves and Vasconcellos’s (2024) challenge to the assumption that DPI is inherently public by virtue of its technical properties is the sharpest conceptual contribution in the literature reviewed here. Their typology, distinguishing attribute-based publicness (interoperability, open standards) from functional publicness (enabling essential capabilities) and governance-based publicness (explicit values, co-creation, accountability), exposes the insufficiency of much DPI advocacy. Interoperability enables competition but does not guarantee equity. Open source reduces excludability but does not ensure accessibility.
Brazil’s Pix involved 130+ stakeholder actors in co-design, a Pix Forum that shaped both technical architecture and user trust in ways that Jamaica’s digital identity system, designed without participatory governance and characterised by institutional trust deficits, simply could not achieve (Mazzucato et al., 2024). Zuckerman (2020) extends this political economy logic: the architectures societies build for digital public goods reflect political choices, not technological determinism. Applied to DPI in development finance, the implication is that the same identity, payment, and data exchange infrastructure that in India produced expanded financial inclusion could, under different governance arrangements, produce financial surveillance, authoritarian control, or corporate data capture. Infrastructure is not neutral.
What the reviewed literature does not adequately address, and what a more complete political economy of DPI would require, is a sustained account of how incumbent financial interests shape DPI design, how donor incentives affect governance choices in recipient countries, and how elite capture operates in DPI procurement processes.
The literature consistently exposes what we term the equity paradox of DPI: the populations most in need of financial inclusion are frequently the hardest to reach with digital infrastructure, and the most vulnerable to its failures when it does reach them. Djatmiko, Sinaga and Pawirosumarto (2025) document that algorithmic bias in AI-enabled public services disproportionately disadvantages those lacking digital proficiency, which is precisely the population that DPI inclusion programmes most need to reach. Infrastructure investment that does not reach the last mile, rural areas, elderly populations, women in low-income settings, and persons with disabilities risks amplifying existing inequalities under the cover of technical progress.
Mpofu and Mhlanga’s (2022) analysis of DFS taxation in Africa adds a distributional dimension routinely absent from DPI discussions: Uganda’s 2018 1% mobile money levy produced P2P transaction usage declines exceeding 50%, with the tax burden falling disproportionately on low-income users in low-competition markets. Rwanda’s fee waivers increased adoption and Uganda’s levy reversed it within the same regional context, demonstrating both the price elasticity of DFS demand and the regressive potential of poorly designed fiscal policy layered onto inclusion infrastructure. However, it is important to note that this finding comes from a single African context and should not be extrapolated as a universal rule about DFS taxation without attention to the significant variation in market structure, fiscal capacity, and alternative revenue bases across the continent.
Gender represents the most acute dimension of the equity paradox. Despite female-owned enterprise growth of 42.1% in high-UPI Indian villages (Balasundaram et al., 2026), Tay, Tai and Tan (2022) document that 56% of the world’s 1.7 billion unbanked adults are women. DFI’s inequality-reducing effects are strongest when infrastructure and innovation complements are present, which are precisely the conditions most likely to be absent in the low-income, predominantly female agricultural communities where the gender inclusion gap is deepest.
Our theoretical framework synthesises four traditions in a conditional causal architecture. Transaction Cost Theory (North, 1990; as applied in Balasundaram et al., 2026) establishes the baseline mechanism: DPI reduces the friction, search, verification, and settlement costs that otherwise price the poor out of formal financial markets. Network Effects Theory (Rochet and Tirole, 2003) explains the self-reinforcing dynamics of payment platform adoption: once merchant acceptance and consumer use cross critical thresholds, network externalities produce non-linear inclusion gains, visible in UPI’s explosive merchant adoption trajectory and in community influencer effects documented in the rural India literature.
Institutional Complementarity Theory (North, 1991; Acemoglu and Robinson, 2012) frames the governance moderation finding: DPI does not operate in an institutional vacuum. Its effectiveness is conditioned by the quality of property rights, rule of law, regulatory capacity, and anti-corruption mechanisms that determine whether digital financial transactions are trusted, enforced, and accessible across the income distribution. The Common Good Framework (Mazzucato et al., 2024) provides the normative architecture: DPI governance must be deliberate, participatory, and oriented toward explicitly defined public values, not merely technically efficient.
These four traditions combine into the Conditional DPI-Inclusion Framework, modelled in figure 1, which depicts financial inclusion as the joint outcome of DPI architecture quality, governance capacity, and human capability development, mediated by five mechanisms: transaction cost reduction, information asymmetry mitigation, network externality generation, formalisation spillovers, and trust anchoring. It is important to note that while transaction cost reduction and network effects are doing the primary explanatory work in the empirical literature, the Common Good Framework functions more as a normative architecture than as an empirically tested mechanism. The paper presents all four traditions as conceptually relevant while acknowledging this hierarchy of empirical support.
Figure 1: Conditional DPI-Inclusion Framework. The three DPI pillars (identity, payments, data exchange) interact with governance quality and human capability development to produce financial inclusion outcomes via five mediating mechanisms. Below governance thresholds, the pathway is blocked or reversed. Dashed pathways indicate weaker or primarily qualitative evidence.
The Conditional DPI-Inclusion Thesis states: DPI delivers equitable and sustainable financial inclusion if and only if three conditions are co-present above critical thresholds: (C1) architecture quality, the DPI is interoperable, open-standard, and governed as a common good rather than a proprietary system; (C2) governance capacity, institutional quality, measured across dimensions of rule of law, regulatory effectiveness, and accountability, meets context-specific minimum levels; and (C3) human capability development, targeted investment in digital and financial literacy accompanies infrastructure deployment.
The logical structure of the thesis requires clarification. The three conditions are jointly necessary but individually insufficient, meaning that strong performance on any one or two conditions does not compensate for failure on the third. Ukraine’s case illustrates C1 and C3 partial failure: despite reasonable internet infrastructure, the absence of institutional trust (C2) and limited digital financial capability (C3) produces persistent exclusion. Liberia illustrates C1 and C2 partial failure: a functioning mobile payment rail exists, but the absence of a national digital identity system (C1 incomplete) and weak regulatory enforcement (C2 insufficient) leaves the payment rail structurally unable to onboard the unconnected majority.
The conditions do not operate as binary switches. They function as gradients with threshold effects: below certain minimum levels, inclusion returns to DPI investment approach zero or turn negative; above the threshold, returns increase, but not necessarily linearly. The governance thresholds estimated by Akpa et al. (2025) should be understood as indicative turning points in the observed data, not as stable, universal breakpoints that apply equally in all institutional contexts. Practitioner use of these thresholds should be as diagnostic signals that trigger additional governance assessment, not as hard financing rules.
A theoretical contribution that cannot specify what would disconfirm it is not falsifiable in the scientifically useful sense. We therefore specify the evidence that would count against the Conditional DPI-Inclusion Thesis:
First, the thesis would be disconfirmed if a well-designed study documented large, sustained, and equitably distributed financial inclusion gains from DPI deployment in a context where institutional quality is demonstrably below the governance thresholds identified in the literature and where no significant governance co-investment occurred. Such a finding would suggest that governance is not a necessary condition for DPI’s inclusion effects, which is the strongest specific claim the thesis makes.
Second, the thesis would be weakened but not disconfirmed if studies systematically showed that any single condition alone, strong architecture, or strong governance alone, or strong capability investment alone, was sufficient to produce inclusion gains. This would shift the thesis from ‘jointly necessary’ to ‘weighted and partially substitutable’, which is a more nuanced but less tractable claim.
Third, the thesis would be disconfirmed for the governance threshold specification specifically if future research with larger samples and more precise governance measurement found that the threshold relationship is not significant once country-level unobservable are more completely controlled. Threshold panel regression findings are sensitive to measurement choices, and the current evidence base, while suggestive, does not establish threshold governance effects as a stable law.
Fourth, the formalisation spillover and trust anchoring mechanisms, which currently rest on moderate or qualitative evidence, would be disconfirmed as genuine DPI mechanisms if future causal studies found that these effects are fully explained by confounders, such as general economic development, rather than by DPI-specific pathways.
Specifying these falsification conditions is not merely a methodological gesture: it helps practitioners and future researchers identify the most productive empirical tests of the thesis and prevents it from collapsing into the unhelpfully broad claim that ‘context matters’, which is true but not actionable.
For the thesis to be practically useful to development finance practitioners, each condition must be assessable before a financing decision is made. We propose the following ex ante assessment approach:
For C1 (Architecture Quality): Assess whether the proposed DPI architecture adopts open and interoperable standards, whether private operators can participate on equal terms with incumbents, whether a governance body with independent oversight has been established, and whether the architecture includes analogue (‘phygital’) channels for populations unable to use digital interfaces. These can be assessed from procurement documentation, regulatory frameworks, and design specifications at appraisal stage.
For C2 (Governance Capacity): Use World Governance Indicators, particularly Rule of Law, Government Effectiveness, and Voice and Accountability, as approximate indicators, with the caveat that scores in the range of 0.25-0.30 on these scales signal elevated governance risk rather than a hard financing threshold. Supplement with country-specific diagnostic assessments of consumer data protection legislation, payment market regulation, and anti-corruption enforcement capacity.
For C3 (Human Capability): Assess current digital literacy rates among the target population for the DPI investment, the presence and capacity of existing digital and financial literacy programmes, and the degree to which capability-building is integrated into the DPI project design rather than treated as a separate line item.
| DPI Pillar | Core Mechanism | Key Governance Prerequisite | Risk if Governance Absent |
|---|---|---|---|
| Digital Identity | Eliminates documentation barriers; enables e-KYC and G2P targeting | Legal data protection framework; anti-exclusion safeguards; biometric appeals process | Surveillance; biometric exclusion; denial of welfare to eligible beneficiaries |
| Payment Rails | Reduces transaction costs; enables real-time settlement and G2P disbursement | Interoperability mandates; consumer protection regulation; fee oversight authority | Monopolisation; regressive fee structures; over-indebtedness among low-income users |
| Data Exchange | Mitigates information asymmetry; enables credit scoring for thin-file borrowers | Consent architecture; data fiduciary regulation; algorithmic accountability standards | Predatory lending; discriminatory credit scoring; mass commercial data capture |
This paper employs a Structured Comparative Literature Synthesis (SCLS) methodology. The research question, under what conditions does DPI deliver equitable financial inclusion? is essentially a question about causal heterogeneity across contexts. It cannot be answered by a single primary dataset, since no single country captures the relevant variation in DPI architecture, governance quality, and human capability development simultaneously.
SCLS is not presented here as superior to all alternatives, but as appropriate for this specific question and evidence base. A standard systematic review with meta-analysis would be unsuitable because the outcome variables are heterogeneous across studies (GDP growth, GINI coefficients, account ownership rates, digital payment use, business registrations) and are not combinable into a common effect size without strong and questionable comparability assumptions. A realist review would be a methodological alternative with genuine merit; it is also designed for conditional causal questions and explicitly incorporates mechanisms and contexts. We chose SCLS over realist review because the evidence base, while theoretically motivated, does not yet have the depth of qualitative mechanism evidence that realist synthesis requires for the iterative ‘theory testing and refining’ cycles that realist review depends on. A meta-ethnographic approach would be valuable for the qualitative subset of the evidence but would lose the quantitative panel and experimental studies that contain the highest-quality causal evidence in the corpus. SCLS allows theoretically structured integration across quantitative, qualitative, and conceptual studies while remaining explicitly transparent about evidence quality differences.
We conducted a systematic search across the following databases: Web of Science, Scopus, Google Scholar, SSRN, World Bank Open Knowledge Repository, IMF e-Library, and NBER Working Papers. We also searched the working paper series of the BIS, UCL Institute for Innovation and Public Purpose, and the Digital Frontiers Institute.
The following keyword combinations were used: ‘digital public infrastructure’ AND (‘financial inclusion’ OR ‘financial access’); ‘digital financial inclusion’ AND ‘governance’; ‘DPI’ AND ‘development finance’; ‘mobile money’ AND ‘institutions’; ‘UPI’ AND (‘rural’ OR ‘inclusion’); ‘digital identity’ AND ‘financial access’; ‘payment systems’ AND ‘inclusion’ AND ‘developing countries’; ‘fintech’ AND ‘governance threshold’; and ‘digital finance’ AND ‘SDG’. The search covered publications from January 2017 to February 2026, with seminal pre-2017 works (D’Silva et al., 2019; Arner et al., 2018; Jack and Suri, 2014) included on the basis of their foundational theoretical relevance.
Figure 2: PRISMA-Style Search Flow Diagram. Records identified across all databases: n = 847 | After duplicate removal: n = 623 | Title/abstract screened: n = 623 | Excluded at title/abstract: n = 481 | Full texts assessed for eligibility: n = 142 | Excluded at full-text stage: n = 120 | Final retained for primary synthesis: n = 22 | Additional references (theoretical framework and contextual literature): n = 30
| Dimension | Inclusion Criteria | Exclusion Criteria |
|---|---|---|
| Publication type | Peer-reviewed journal articles; high-quality working papers from reputable institutions (BIS, World Bank, IMF, UNDP, UCL IIPP); dissertations with rigorous methodology | Opinion pieces and editorials without primary or secondary evidence; pure technical/engineering papers without development focus |
| Substantive focus | Papers addressing DPI/DFI and financial inclusion outcomes; papers addressing governance moderation of DFI effects; papers addressing equity dimensions of digital finance | Papers focused exclusively on fintech product innovation without institutional analysis; smart city infrastructure papers without financial inclusion connection |
| Geography | Developing and emerging market economies; cross-country studies including significant LMIC share | Studies exclusively covering high-income OECD countries without relevance to developing economy DPI design |
| Methodology | Quantitative (panel, experimental, quasi-experimental), qualitative, mixed-methods, conceptual with empirical grounding | Purely descriptive or journalistic accounts; papers with no identifiable methodology section |
| Date range | 2017–2026 (primary); pre-2017 foundational works included where explicitly theoretically relevant | Pre-2017 papers without direct theoretical or empirical relevance to DPI specifically (as distinct from broader mobile money/fintech literature) |
| Language | English-language publications | Non-English publications (acknowledged as a limitation of the search) |
Each retained study was scored on a ten-point instrument adapted from the Mixed Methods Appraisal Tool (MMAT), modified to reflect the specific evidentiary requirements of a governance-conditioned DPI synthesis. The ten criteria and their maximum scores are: (1) causal identification strategy (2 points, highest weight, reflecting the centrality of causal inference to the thesis); (2) data quality and representativeness (1 point); (3) construct validity of financial inclusion measure (1 point); (4) governance/institutional variable quality (1 point); (5) methodological transparency and replicability (1 point); (6) generalisability, explicit acknowledgement of scope conditions (1 point); (7) equity and distributional analysis (1 point); (8) mechanism specification (1 point); (9) policy relevance (0.5 points); (10) literature engagement (0.5 points). Studies scoring 6 or above were retained for primary synthesis; studies scoring 4-5 were included in the supplementary literature. The full item-by-item appraisal table for all 22 retained studies is presented in Appendix C.
The retained sample has the following quality distribution: six studies scored 8–10 (strong causal identification, typically panel data with instrument or quasi-experimental design); nine studies scored 6–7.5 (moderate causal identification, typically System GMM or rigorous cross-country analysis); and seven studies scored 6 (primarily qualitative, conceptual, or bibliometric, retained for mechanism and governance analysis rather than causal estimation). The synthesis reflects this quality distribution by differentiating between strong causal evidence, moderate associational evidence, and qualitative or conceptual evidence throughout Sections 6 and 7.
Several limitations warrant explicit acknowledgement. First, the retained sample is geographically concentrated in India and Sub-Saharan Africa, reflecting the distribution of published high-quality research rather than a deliberate selection decision. This means the synthesis is better positioned to speak to DPI outcomes in these contexts than to Latin America, Southeast Asia, or Eastern Europe. Second, the search strategy was conducted in English only, potentially excluding high-quality research published in Portuguese, French, Swahili, or other languages relevant to major DPI contexts. Third, the evidence base contains a plausible positive-findings bias: the incentive structure of academic publishing, combined with the policy interest in documenting DPI success, may mean that null or negative findings are underrepresented. We have attempted to mitigate this by actively including studies documenting failure cases (Liberia, Ukraine) and negative conditional effects (Van and Quoc, 2025), but readers should interpret the synthesis with this limitation in mind.
The reviewed studies suggest, with reasonable but not absolute consistency, that DPI’s inclusion effects are stronger when the full three-pillar architecture is deployed with adequate governance. India’s JAM trinity, the clearest empirical case for which high-quality causal evidence exists, produced the largest documented increase in account ownership across 121 countries: account ownership rose from 35% to 80% between 2011 and 2017, the gender gap narrowed from 17% to 6%, and the income gap from 14% to 5% (D’Silva et al., 2019). These gains appear to be genuinely complementary across pillars rather than driven by any single component.
In contrast, Liberia, which has mobile money penetration reaching 50% of the population and US$959 million in e-money transactions in 2024, exhibits persistent geographic concentration of financial access in Monrovia and a 12% gap between richest and poorest household access rates (Kanu, 2025). Without a functioning national digital ID system (only 720,000 registered as of 2025) or an interoperable data exchange framework, the payment rail operates in institutional isolation — useful for those already connected but structurally unable to onboard the unconnected.
It is important not to overstate the pillar-completeness claim. The reviewed evidence does not include controlled comparisons of single-pillar versus full-architecture DPI deployments with otherwise similar conditions. The claim that all three pillars are necessary is theoretically motivated and consistent with the available evidence, but it has not been directly tested with a design capable of isolating pillar-specific effects from country-level confounders.
Multiple reviewed studies find that governance does not improve DPI’s inclusion effects continuously, it appears to operate through thresholds, below which the inclusion dividend is effectively absent. Akpa et al.’s (2025) SGMM analysis identifies approximate threshold values: voice and accountability at 0.300, government effectiveness at 0.300, and rule of law at 0.250. Van and Quoc’s (2025) panel threshold regression finds structural breaks in the DFI-sustainable development relationship at financial development levels of 0.2895 and 0.7176.
These findings have policy relevance for development finance: a DFI financing broadband rollout in a country with rule of law scores below 0.25 may be financing infrastructure that will not deliver projected inclusion returns, because the institutional conditions for translating connectivity into financial access are absent. The counterfactual investment is not no broadband but broadband plus institutional capacity building, a bundled design that MDBs have historically been slow to implement.
However, threshold findings from panel threshold regression are model-specific turning points, not universal governance laws. The specific threshold values depend on the governance indicators used, the sample of countries, the time period, and the functional form assumed for the relationship. These numbers should be treated as approximate, contextually grounded signals, useful for identifying countries where governance assessment warrants careful attention before DPI financing, but not as hard cutoffs that mechanically determine funding decisions.
The comparative synthesis identifies five mechanisms through which DPI affects financial inclusion. The evidence quality for each mechanism varies substantially, and the policy and theoretical weight assigned to each should reflect this variation.
| Mechanism | Evidence Quality | Key Finding | Supporting Studies | Nature of Evidence |
|---|---|---|---|---|
| Transaction Cost Reduction | Strong (causal) | 42.2% of UPI inclusion effect attributable to transaction cost reduction; KYC cost reduced 99.5% under Aadhaar e-KYC | Balasundaram et al. (2026); D’Silva et al. (2019) | Causal, DiD mediation decomposition; programme evaluation |
| Information Asymmetry Mitigation | Moderate (associational) | Cash-flow-based lending enabled for thin-file borrowers; mechanism is theoretically grounded, but direct causal identification is limited in available studies | Arner et al. (2018); D’Silva et al. (2019) | Associational, mechanism inferred from programme logic and partial observational evidence |
| Network Externality Generation | Strong (causal) | 39.1% of UPI inclusion effect; self-reinforcing merchant adoption dynamics; community influencer effects documented qualitatively | Balasundaram et al. (2026); Gupta (2024) | Causal (DiD) supplemented by qualitative mechanism documentation |
| Formalisation Spillovers | Moderate (panel) | 27% business registration increase in high-UPI villages; Uganda added 350,000 new taxpayers following mobile money rollout. Mechanism is plausible but evidence reflects association rather than direct causal test | Balasundaram et al. (2026); Kanu (2025) | Panel associational; plausible causal story but confounders not fully ruled out |
| Trust Anchoring | Moderate (qualitative) | Government ownership of DPI is central driver of adoption in LMIC contexts; trust deficits block DFI adoption in Ukraine; commercial platforms cannot replicate state-derived trust | Gupta (2024); Naumenkova et al. (2019) | Qualitative, interviews and case analysis; not causally identified in the strict sense |
Cross-regional comparison reveals patterns that are not reducible to income level differences, though the analysis should be read with awareness that the regional cases selected are not a systematic comparative sample. Sub-Saharan Africa’s digital-led inclusion trajectory, where digital indicators drive improvements even as traditional bank-based inclusion stalls in some countries (Khera et al., 2022), appears to reflect a genuine leapfrogging dynamic, the absence of incumbent banking infrastructure becoming an advantage rather than an obstacle for mobile money penetration. The most digitally included SSA countries (Ghana, Kenya, Senegal, Uganda, Rwanda) are not the continent’s richest economies; they are economies that invested early in mobile money regulatory frameworks, consistent with the thesis’s emphasis on governance sequencing.
South and East Asia’s trajectory, exemplified by India, reflects a different model: state-led DPI investment as the foundation for market competition. This architecture’s defining feature is that the rail is free and public while applications are competitive and private, a design choice that prevented the monopolisation characterising early mobile money markets elsewhere. Brazil’s Pix represents a central bank-built interoperable payment system with compulsory bank participation, a third model that combines state mandate with private application development (Mazzucato et al., 2024).
Eastern Europe presents the cautionary case. Ukraine’s digital inclusion profile, only 10.8% of adults using digital channels for financial services against a high-income country average of 46.7%, despite 62.8% internet penetration, demonstrates that digital readiness without institutional trust and regulatory quality does not produce digital financial inclusion (Naumenkova et al., 2019). This finding should be interpreted with appropriate caution: Ukraine in the period studied also experienced significant political instability, which may be a confounding factor rather than merely an illustration of the governance moderation effect.
These regional cases were selected for their illustrative diversity rather than through a systematic comparative case selection protocol, and readers should be cautious about treating the regional patterns as representative of all countries within each region. Bangladesh, Indonesia, Mexico, Nigeria, and the Philippines, all of which have significant and varied DPI experiences, are notable absences from the primary synthesis that reflect the concentration of the reviewed literature rather than a deliberate exclusion decision.
| Country / Region | DPI Maturity | Governance Context | Inclusion Outcome | Main Constraint | Policy Lesson |
|---|---|---|---|---|---|
| India (JAM) | High, full three-pillar architecture | Moderate-strong; state administrative capacity to implement at scale; some consumer protection gaps | Account ownership 35%→80% (2011-17); gender gap 17%→6%; KYC cost -99.5% | Biometric exclusion incidents: rural usage gap persists | Full architecture + state capacity critical; consumer protection must be co-designed not added post-hoc |
| Sub-Saharan Africa (Kenya, Ghana, Rwanda) | Medium, strong payment rails; identity and data exchange less developed | Heterogeneous; leading economies invested early in mobile money regulation | Mobile money leaders: transaction values 140% of GDP (Zimbabwe); digital inclusion rising even as bank-based inclusion stalls | Governance quality heterogeneous; DFS taxation risk; last-mile connectivity gaps | Governance-first sequencing in leading economies; DFS tax design critical for inclusion protection |
| Brazil (Pix) | High, central bank-mandated interoperable payment system | Strong central bank independence; compulsory bank participation mandate; participatory design process | 93% adult reach within 3 years; payment system competition transformed | Compulsory mandate required strong regulator; lesson may not transfer to weaker regulatory contexts | Mandatory interoperability requires strong regulatory authority; participatory design (Pix Forum) built user trust |
| Ukraine | Low, digital readiness without DPI architecture | Weak institutional trust; bank branch collapse not offset by digital adoption | Only 10.8% use digital channels despite 62.8% internet penetration | Trust deficit; infrastructure without institutional confidence | Digital readiness ≠ digital inclusion; trust and regulatory quality preconditions must be met |
| Liberia | Partial, payment rail without identity or data exchange | Weak governance; biometric ID procurement cancelled 2025 | 50% mobile money penetration but geographic concentration in Monrovia; 12% richest-poorest gap | Missing identity and data exchange pillars; governance fragility | Pillar incompleteness limits inclusion reach; governance fragility creates procurement sustainability risk |
Perhaps the most politically significant finding of this synthesis is the consistent evidence that DPI, absent deliberate equity design, can reproduce and sometimes amplify existing inequalities. Several mechanisms drive this paradox. Digital infrastructure investment typically reaches urban and wealthier populations first, widening urban-rural and income access gaps before they narrow. Digital financial services have historically benefited the near-bankable, those slightly below the formal threshold, more than the ‘deeply excluded,’ who face language barriers, disability, extreme poverty, or geographic remoteness.
Ozili (2018) documents that ‘forced inclusion’ policies mandating digital channels without addressing underlying barriers create compliance without access: individuals are registered as digitally included while remaining functionally excluded. The equity paradox cannot be resolved by better DPI architecture alone, it requires explicit pro-poor design choices in product standards, fee regulation, language accessibility, and consumer redress mechanisms, combined with genuine capability investment in the populations most at risk of exclusion.
Gender represents the most acute dimension of the equity paradox. Despite female-owned enterprise growth of 42.1% in high-UPI Indian villages (Balasundaram et al., 2026), the largest documented gender dividend in the DPI literature, Tay, Tai and Tan (2022) document that 56% of the world’s 1.7 billion unbanked adults are women. The mechanisms through which DPI produces gender dividends, and the conditions under which it fails to do so, remain inadequately researched.
The findings presented above allow us to address an apparent paradox in the DPI literature: why do some studies find large, positive, and well-identified effects of DPI on financial inclusion (Balasundaram et al., 2026; D’Silva et al., 2019) while others find weak, mixed, or even negative effects (Naumenkova et al., 2019; Van and Quoc, 2025)? The synthesis suggests that all of these studies are correctly characterising different regions of the conditional distribution. When DPI architecture is complete, governance is above threshold, and human capability conditions are met, the effects appear large and positive. When any of the three conditions fails, and in the majority of developing economies, at least one does, the effects diminish, disappear, or invert.
This unified explanation has important implications for how the development finance community should interpret DPI success stories. India’s UPI revolution required decades of prior institutional investment: Aadhaar from 2010, regulatory ecosystem development through the 2000s, and a state with sufficient administrative capacity to implement a layered architecture at billion-person scale. Brazil’s Pix required the Banco Central do Brasil to mandate participation by all financial institutions with over 500,000 clients, a regulatory authority that most developing country central banks lack the political independence and legal capacity to exercise. The structural risk in development finance is the tendency to replicate architectures without replicating the institutional conditions that make them work.
The most practically significant implication of this synthesis for development finance practitioners is what we term the bundled ecosystem imperative. The evidence consistently supports financing DPI not as a standalone technical project but as an integrated ecosystem in which infrastructure, governance strengthening, consumer protection, digital literacy, grievance redressal, and monitoring and evaluation are co-financed and sequenced together.
This is not merely a design preference; it follows from the evidence on complementarity. DPI infrastructure deployed ahead of consumer protection legislation creates the conditions for the Aadhaar exclusion incidents documented by Mazzucato et al. (2024): technically functional infrastructure that fails specific, vulnerable populations because the institutional safeguards were treated as someone else’s responsibility. DPI deployed without digital literacy investment creates the conditions documented by Malladi et al. (2021): 355 million registered accounts in India’s PMJDY programme coexisting with persistent low transactional usage among rural and elderly populations who were never adequately equipped to use them.
A bundled ecosystem approach is more operationally realistic than threshold-based financing rules. Current governance threshold estimates are model-specific and context-dependent, and a rigid rule that no DFI should finance DPI where WGI Rule of Law scores fall below 0.25 would deny financing to fragile states where DPI investment may, under the right design conditions and with appropriate governance co-investment, still deliver meaningful inclusion gains. The better approach, and the one better supported by the evidence, is staged financing conditional on measurable governance improvement plans, consumer protection milestones, and usage outcomes monitored post-deployment.
The bibliometric analysis by Gallego-Losada et al. (2023) confirms that the dominant research paradigm in DFI is supply-side and access-focused: account ownership, ATM density, bank branches, mobile money registrations. These are measures of DPI architecture, not measures of whether the architecture is producing the welfare outcomes that justify public investment. Ozili’s (2018) distinction between financial data inclusion and financial inclusion, the former being mere registration, the latter being active participation, remains methodologically underexplored.
What the field needs, and what development finance institutions should require in their results frameworks, are outcome-based measurement frameworks that track: (a) whether registered users actively use accounts for transactions, savings, and credit; (b) whether digital credit products are improving consumption smoothing and investment outcomes for low-income households; (c) whether DPI-enabled G2P transfers are reaching intended beneficiaries without leakage; (d) whether digital financial access is reducing dependence on predatory informal credit; and (e) whether the quality of digital financial services, fee levels, security, grievance resolution rates, product suitability, is improving for the most vulnerable. The Balasundaram et al. (2026) study, which tracks business registrations, savings balances, transaction frequency, and informal borrowing rates simultaneously, represents the methodological frontier. It is not representative of standard evaluation practice.
The current trajectory of global DPI development contains at least three distinct political economy models with different inclusion implications. The state-led model (India, Estonia, Brazil) positions government as the infrastructure provider and regulator, with private sector innovation layered atop public rails. The market-led model (Kenya M-Pesa in its early phase, much of West Africa’s mobile money ecosystem) positions a dominant private operator as the de facto infrastructure provider, with regulation following rather than preceding market development. The donor-assisted model positions development finance institutions and bilateral donors as capacity builders and sometimes infrastructure investors, with attendant complications around ownership, sustainability, and alignment with national interests.
Each model has documented failure modes. State-led DPI requires state capacity that many developing countries lack. Market-led DPI produces exclusion incidents when profit incentives are not constrained by consumer protection. Donor-assisted DPI creates dependency and sustainability gaps when funding ends, as the cancelled Liberia biometric procurement illustrates. The question of which model is appropriate in which context is not primarily technical but political economy, and development finance institutions must assess it explicitly at appraisal stage.
Tay, Tai and Tan’s (2022) PRISMA-based synthesis maps DFI contributions to 13 of 17 SDGs, and Van and Quoc’s (2025) empirical work confirms significant effects on composite sustainable development scores in low- and middle-income countries. The SDG integration argument for DPI investment is compelling in principle: the same identity infrastructure that enables financial account opening also enables health service delivery and civic participation; the same payment rails that enable mobile money also to enable digitised G2P transfers.
However, Van and Quoc’s (2025) finding that DFI undermines sustainable development in highly financially developed economies through consumption-financed energy-intensive spending introduces an environmental dimension that the SDG-DPI integration literature has not adequately confronted. This finding should be treated with appropriate caution, it comes from one cross-country threshold study, and its extrapolation to DPI design principles requires additional evidence, but it does raise a genuine design question: without green finance principles embedded in the data exchange and credit scoring layers of DPI, there is no mechanism that preferentially channels DFI-enabled credit toward climate-positive investments. This is an architectural challenge for future DPI design, not merely a post-hoc regulatory adjustment.
The evidence base reviewed here supports a prioritised and sequenced policy agenda organised around the three pillars of the Conditional DPI-Inclusion Framework. Table 4 links each recommendation to its evidence base, indicates evidence strength, identifies the primary implementation barrier, and designates the lead actor.
| Recommendation | Supporting Evidence | Evidence Strength | Primary Implementation Barrier | Lead Actor | Urgency |
|---|---|---|---|---|---|
| Finance DPI as a bundled ecosystem (infrastructure + governance + consumer protection + digital literacy + grievance redress + monitoring) rather than standalone infrastructure | Balasundaram et al. (2026); Mazzucato et al. (2024); Malladi et al. (2021); Kanu (2025) | Strong, consistent across multiple study types | Project finance structures are ill-suited to multi-sector bundled investments; governance components are harder to measure and slower to deliver | MDBs, DFIs, portfolio design and conditionality | High |
| Require interoperability mandates and open standards as a condition of DFI/MDB financing for payment infrastructure | Mazzucato et al. (2024); D’Silva et al. (2019); Chinoda & Kapingura (2024) | Strong, both comparative evidence and theory support | Incumbent operator resistance; weak regulatory authority in many developing country central banks | MDBs (conditionality); central banks (mandate design) | High |
| Use governance thresholds as diagnostic signals for staged financing design, not as hard cutoffs | Akpa et al. (2025); Chinoda & Kapingura (2024); Van & Quoc (2025) | Moderate, threshold estimates are model-specific; the directional finding is robust | DFI appraisal processes are not designed for staged, milestone-conditioned digital infrastructure financing | DFIs, bilateral donors, appraisal frameworks | High |
| Treat digital and financial literacy as infrastructure investment, co-financed alongside DPI hardware and software | Malladi et al. (2021); Djatmiko et al. (2025); Balasundaram et al. (2026) | Moderate, associational evidence strong; direct causal evidence limited | Capability programmes are classified as ‘soft’ costs and deprioritised in infrastructure budgets | National governments; development banks; MDBs | High |
| Design DFS tax policy to protect inclusion: single-point, low-rate, exemptions for low-value transactions, regular review against usage data | Mpofu & Mhlanga (2022), Uganda and Rwanda comparison | Moderate, robust within single regional context; extrapolation to other regions requires care | Revenue pressures in low-income governments; limited alternative tax bases | Finance ministries; regulators; IMF advisory | Medium |
| Embed outcome-based measurement in DPI financing: active usage, welfare outcomes, grievance resolution rates, not just enrolment and access metrics | Ozili (2018); Khera et al. (2022); Gallego-Losada et al. (2023) | Strong conceptual foundation: measurement design is a technical gap | MDB results frameworks historically focused on infrastructure outputs, not usage and welfare outcomes | MDBs; evaluation units; national statistics offices | Medium |
| Commission political economy feasibility assessments as front-end components of DPI investment appraisal, addressing incumbent interests, donor incentives, and elite capture risks | Mazzucato et al. (2024); Zuckerman (2020); Kanu (2025) | Qualitative, case evidence; not causally identified | Political economy assessment is methodologically contested and often avoided for diplomatic reasons in DFI operations | DFIs; MDB country teams | Medium |
Development finance institutions and national governments must resist defining DPI success in purely technical terms. Every DPI architecture should be evaluated against Mazzucato et al.’s (2024) five common-good governance pillars: explicit societal purpose; co-creation with affected communities (including participatory forums like Brazil’s Pix Forum); collective learning mechanisms; universal access pathways including analogue (‘phygital’) channels; and transparency and accountability mechanisms through which citizens can audit data use and contest service failures. DPI procurement processes should address each pillar explicitly as a design specification, not as an aspirational goal.
Given the evidence on governance thresholds and the practical limitations of hard funding cutoffs, the recommended approach for DFIs and MDBs is staged financing: an initial tranche that supports DPI design and governance capacity-building; a second tranche conditional on measurable governance milestones (consumer protection legislation enacted, regulatory enforcement capacity demonstrated, grievance redressal systems operational); and a third tranche for scale deployment conditional on usage outcome evidence from the initial rollout.
Specific governance investments that the evidence supports include consumer data protection legislation with meaningful enforcement capacity; independent competition regulation for payment markets; risk-based, proportional KYC/AML frameworks enabling financial access for the poor; and grievance redressal systems with measurable resolution rates and enforceable consumer rights.
The consistent finding that digital literacy mediates DPI’s inclusion effects, and that low-literacy populations are the most likely to remain excluded, fall victim to fraud, and abandon digital financial products, implies that digital and financial literacy programmes must be treated as infrastructure investment rather than discretionary add-ons. Malladi et al.’s (2021) finding that 22% of India’s Business Correspondent agents faced fraud in 2017, up from 2% in 2015, points to a critical principle: the populations most targeted by DPI inclusion programmes are also the most vulnerable to its failure modes. DFI financing that does not include capability co-investment is architecturally incomplete.
This synthesis identifies five high-priority research gaps. First, the literature on DPI’s long-run effects is essentially non-existent. The best causal study in the corpus (Balasundaram et al., 2026) covers 2019–2023, sufficient to identify adoption effects but insufficient to observe wealth accumulation, firm survival, or the compounding effects of sustained financial inclusion. Long-run panel studies tracking cohorts from DPI onboarding through to durable income and wealth outcomes are a priority.
Second, gender mechanisms within DPI-enabled financial inclusion demand more rigorous decomposition. The 42.1% growth in female-owned enterprises in high-UPI Indian villages is the largest documented gender dividend in the DPI literature, but the mechanisms are only partially identified. Are effects driven primarily by credit history generation, income autonomy, time savings, or social norm change? These mechanisms have very different policy implications, and only mixed-methods research combining digital transaction data with household surveys and qualitative fieldwork can distinguish them.
Third, DPI and climate finance is an emerging priority. Van and Quoc’s (2025) finding that DFI undermines sustainable development in highly financially developed economies through consumption-intensive credit expansion opens a research agenda on green DPI design: how can payment rails, credit scoring systems, and data exchange frameworks be designed to preferentially channel digital financial access toward climate-positive investments?
Fourth, the political economy of DPI, who benefits from different governance models, how incumbent financial interests shape DPI design, and how to build political coalitions for common-good DPI governance, remains dramatically under-researched. Economic and technical studies dominate the corpus; political science contributes at the margins.
Fifth, DPI in conflict-affected and fragile states is almost entirely absent from the literature. The conditions facing DPI development in active conflict zones and post-conflict reconstruction require specialised analytical frameworks that existing work does not provide.
Digital Public Infrastructure is real, and its potential to transform financial inclusion in the Global South is real. The evidence reviewed here, spanning India’s UPI rollout, Sub-Saharan Africa’s mobile money expansion, Brazil’s Pix experiment, and dozens of country-level studies, consistently demonstrates that when DPI is correctly designed, properly governed, and accompanied by investments in human capability, it delivers measurable and meaningful inclusion gains at population scale. The quantitative evidence from India alone, 470 million new accounts, a 99.5% reduction in KYC costs, a 53% reduction in informal borrowing in high-UPI rural villages, and a 42% increase in female-owned enterprises, represents genuine development achievement.
But the evidence also demonstrates, with comparable consistency, that DPI is not a development solution in itself. It is an enabler, and an enabler that can as easily facilitate exploitation, surveillance, and entrenched inequality as it can facilitate inclusion and empowerment. The difference between these outcomes is not technical. It lies in whether DPI is governed as a common good, with explicit normative purpose, participatory design, accountability mechanisms, and deliberate equity protections, or whether it becomes what too many infrastructure investments in development history have become: a modernisation project that reaches the relatively advantaged while leaving the most marginalised further from the frontier.
The Conditional DPI-Inclusion Thesis advanced here is both a positive claim and a prescription. Descriptively, the reviewed evidence suggests that DPI delivers equitable financial inclusion only when architecture quality, governance capacity, and human capability development co-occur above critical thresholds, and we have specified the empirical conditions under which this claim would be disconfirmed. Prescriptively, development finance institutions must stop financing DPI architecture and start financing DPI ecosystems, integrated programmes that build institutional and human complements to digital infrastructure as co-investments, not afterthoughts.
For development finance practice, the central implication is straightforward even if the implementation is difficult: DPI must be financed as a bundle. The infrastructure, the governance capacity to regulate and enforce it, the consumer protections to prevent its misuse, the digital literacy to enable its equitable adoption, and the monitoring systems to track whether it is working, these are not separable investments to be sequenced across different project cycles and different funders. They are a single investment in a digital public good, and they must be funded as one.
The choice facing the multilateral development finance system is between financing cathedrals, technically impressive infrastructure that serves those already privileged enough to reach its doors, and financing systems that are designed from the outset to serve those for whom the doors have always been too far away. The architecture is not enough. The governance makes it work. The capability investment makes it equitable. Only when all three are present does DPI fulfil the promise that has made it one of the most discussed and most financed concepts in contemporary development practice.
Acemoglu, D. and Robinson, J.A. (2012) Why Nations Fail: The Origins of Power, Prosperity, and Poverty. Crown Publishers, New York.
Akpa, A.F., Asongu, S.A. and Batuo, M.E. (2025) ‘The role of governance in the effect of the internet on financial inclusion in Sub-Saharan Africa’, Information Technology for Development, 31(3), pp. 664–682.
Arner, D.W., Buckley, R.P. and Zetzsche, D.A. (2018) FinTech for Financial Inclusion: A Framework for Digital Financial Transformation. Alliance for Financial Inclusion.
Balasundaram, E., Anandavel, V., Navalgund, N., Aranganathan, P. and Hanji, S. (2026) ‘The impact of digital public infrastructure on financial inclusion and economic growth in rural India: Evidence from the UPI revolution’, Indian Growth and Development Review, 19(1), pp. 109–126.
Beck, T., Demirgüç-Kunt, A. and Levine, R. (2007) ‘Finance, inequality and the poor’, Journal of Economic Growth, 12(1), pp. 27–49.
Chinoda, T. and Kapingura, F.M. (2024) ‘Digital financial inclusion and economic growth in Sub-Saharan Africa: The role of institutions and governance’, African Journal of Economic and Management Studies, 15(1), pp. 15–30.
Demirgüç-Kunt, A., Klapper, L., Singer, D., Ansar, S. and Hess, J. (2018) The Global Findex Database 2017: Measuring Financial Inclusion and the Fintech Revolution. World Bank, Washington DC.
Demirgüç-Kunt, A., Klapper, L., Singer, D. and Ansar, S. (2022) The Global Findex Database 2021: Financial Inclusion, Digital Payments, and Resilience in the Age of COVID-19. World Bank, Washington DC.
Djatmiko, G.H., Sinaga, O. and Pawirosumarto, S. (2025) ‘Digital transformation and social inclusion in public services: A qualitative analysis of e-government adoption for marginalized communities in sustainable governance’, Sustainability, 17(7), 2908.
D’Silva, D., Filková, Z., Packer, F. and Tiwari, S. (2019) The Design of Digital Financial Infrastructure: Lessons from India. BIS Papers No. 106, Bank for International Settlements, Basel.
Frischmann, B.M. (2012) Infrastructure: The Social Value of Shared Resources. Oxford University Press, New York.
Gabor, D. and Brooks, S. (2017) ‘The digital revolution in financial inclusion: International development in the fintech era’, New Political Economy, 22(4), pp. 423–436.
Gallego-Losada, M.-J., Montero-Navarro, A., García-Abajo, E. and Gallego-Losada, R. (2023) ‘Digital financial inclusion: Visualizing the academic literature’, Research in International Business and Finance, 64, 101862.
GPFI (2016) G20 High Level Principles for Digital Financial Inclusion. Global Partnership for Financial Inclusion, Beijing.
GSMA (2023) The State of the Industry Report on Mobile Money. GSMA, London.
Gupta, A. (2024) Building Platforms as Digital Public Infrastructure: Trust, Adoption and Ecosystem Value using India’s UPI as a Case Study. Executive Fellow Programme in Management Dissertation, Indian School of Business.
ITU (2022) Measuring Digital Development: Facts and Figures 2022. International Telecommunication Union, Geneva.
Jack, W. and Suri, T. (2014) ‘Risk sharing and transactions costs: Evidence from Kenya’s mobile money revolution’, American Economic Review, 104(1), pp. 183–223.
Kanu, D.H. (2025) ‘Digital public infrastructure, the building blocks for digital financial inclusion and taxation of the digital economy in Liberia’, International Journal of Economics, Business and Management Research, 9(11), pp. 76–124.
Khera, P., Ng, S., Ogawa, S. and Sahay, R. (2022) ‘Measuring digital financial inclusion in emerging market and developing economies: A new index’, Asian Economic Policy Review, 17(2), pp. 213–230.
King, R.G. and Levine, R. (1993) ‘Finance and growth: Schumpeter might be right’, Quarterly Journal of Economics, 108(3), pp. 717–737.
Kouladoum, J.-C., Wirajing, M.A.K. and Nchofoung, T.N. (2022) ‘Digital technologies and financial inclusion in Sub-Saharan Africa’, Telecommunications Policy, 46, 102387.
Malladi, C.M., Soni, R.K. and Srinivasan, S. (2021) ‘Digital financial inclusion: Next frontiers, challenges and opportunities’, CSI Transactions on ICT, 9(2), pp. 127–134.
Mazzucato, M. (2023) The Big Con: How the Consulting Industry Weakens our Businesses, Infantilizes our Governments and Warps our Economies. Allen Lane, London.
Mazzucato, M., Eaves, D. and Vasconcellos, B. (2024) Digital Public Infrastructure and Public Value: What is ‘Public’ About DPI? UCL Institute for Innovation and Public Purpose Working Paper Series, IIPP WP 2024/05.
Mpofu, F.Y. and Mhlanga, D. (2022) ‘Digital financial inclusion, digital financial services tax and financial inclusion in the Fourth Industrial Revolution era in Africa’, Economies, 10(8), 184.
Munyegera, G.K. and Matsumoto, T. (2016) ‘Mobile money, remittances, and household welfare: Panel evidence from rural Uganda’, World Development, 79, pp. 127–137.
Naumenkova, S., Mishchenko, S. and Dorofeiev, D. (2019) ‘Digital financial inclusion: Evidence from Ukraine’, Investment Management and Financial Innovations, 16(3), pp. 194–205.
North, D.C. (1990) Institutions, Institutional Change and Economic Performance. Cambridge University Press, Cambridge.
North, D.C. (1991) ‘Institutions’, Journal of Economic Perspectives, 5(1), pp. 97–112.
Ozili, P.K. (2018) ‘Impact of digital finance on financial inclusion and stability’, Borsa Istanbul Review, 18(4), pp. 329–340.
Rochet, J.-C. and Tirole, J. (2003) ‘Platform competition in two-sided markets’, Journal of the European Economic Association, 1(4), pp. 990–1029.
Sarma, M. and Pais, J. (2011) ‘Financial inclusion and development’, Journal of International Development, 23(5), pp. 613–628.
Serrano, W. (2018) ‘Digital systems in smart city and infrastructure: Digital as a service’, Smart Cities, 1(1), pp. 134–154.
Suhrab, M., Chen, P. and Ullah, A. (2024) ‘Digital financial inclusion and income inequality nexus: Can technology innovation and infrastructure development help in achieving sustainable development goals?’, Technology in Society, 76, 102411.
Tay, L.-Y., Tai, H.-T. and Tan, G.-S. (2022) ‘Digital financial inclusion: A gateway to sustainable development’, Heliyon, 8, e09766.
UNDP (2023) Digital Public Infrastructure for Sustainable Development: Opportunities and Governance Considerations. United Nations Development Programme, New York.
Van, H.N. and Quoc, D.L. (2025) ‘Assessing the impact of digital financial inclusion on Sustainable Development Goals: Analyzing differences by financial development levels across countries’, Journal of the Knowledge Economy, 16, pp. 15140–15163.
Zuckerman, E. (2020) What is Digital Public Infrastructure? Open Markets Institute / Center for Journalism and Liberty.
Zuckerman, E. (2020) The Case for Digital Public Infrastructure. Knight First Amendment Institute at Columbia University.