Opening the Black Box on the Impact of Inclusive Credit

Contents

• Executive Summary
  • Main Findings
  • Implications
• Chapter 1: Why an Evidence Approach Matters More than Ever
• Chapter 2: How the Evidence Has Evolved: From Transformative Claims to Conditional Impact
  • 2.1 Early Promise and the Selection Bias Problem
  • 2.2 The RCT Revolution: Testing Transformative Claims
  • 2.3 From Averages to Distributions: The Heterogeneity Breakthrough
  • 2.4 When Credit Goes Wrong: Learning from Market Failures
• Chapter 3: Five Factors that Shape Outcomes
  • 3.1 Reading the Evidence: The Focus Note's Approach and Scope
  • 3.2 Who Borrows: User Demographics and Other Characteristics
  • 3.3 How Credit Is Structured and Delivered
  • 3.4 What Credit Finances: Use of Funds
  • 3.5 Where Credit Operates: Context
  • 3.6 When Credit Helps: Time Horizon
  • 3.7 Which Factors Matter Most?
  • 3.8 A Diagnostic Framework
• Chapter 4: Making It Work: From Evidence to Precision-led Impact
  • 4.1 Recommendations by Stakeholder Group
  • 4.2 Evidence Gaps and Priorities
  • 4.3 Credit as a Lever, Not a Ladder
• References
• Annex A: The Impact Pathfinder and Precision Causal Modeling
• Annex B: Summary Tables for the Five High-Level Factors
• Annex C: Influencing Factors
• Annex D: Impact of Credit on Women's Economic Empowerment
• Annex E: Impact of Credit on Jobs and Entrepreneurship
• Annex F: Impact of Credit on Poverty Reduction
• Annex G: Impact of Credit on Climate Adaptation and Resilience
• Annex H: Impact of Credit on Health
• Annex I: Impact of Credit on Energy Access

Executive Summary

Inclusive credit — credit products designed or adapted to reach low-income borrowers who are underserved by conventional financial institutions — remains one of the most celebrated and also most contested instruments for promoting development outcomes. Early proponents heralded it as transformative for poverty reduction. Rigorous evaluations have since revealed more modest or conditional average effects; and case evidence has documented outright harm in overextended markets. But these findings conceal as much as they reveal. Behind these aggregate findings lies a wide distribution of outcomes, positive for some borrowers, neutral or damaging for others, that depend on identifiable factors.

This Focus Note does not claim to have discovered heterogeneity in credit impacts. The randomized controlled trials (RCT) tradition and the Meager (2019, 2022) meta-analyses have documented it for several years. The note's contribution is to organize and operationalize what that accumulated body of dispersed conditional findings implies for the financial services providers (FSPs), funders, investors, policymakers, regulators, and other sector stakeholders who cannot conveniently and directly navigate the underlying literature.

The note moves the sector's understanding beyond whether credit "works," on average, toward the more productive question: under what conditions, for which borrowers, and through what product designs and delivery modes does inclusive credit generate meaningful benefit rather than harm? The five-factor framework presented here — WHO borrows, HOW credit is designed and delivered, WHAT it finances (is used for), WHERE (in what context) and WHEN (over what time horizon) — is a structured, evidence-grounded synthesis designed to make heterogeneous impact findings actionable for FSPs, funders, investors, policymakers, and regulators operating in low- and middle-income markets.

The note draws on findings from published research — 405 studies in CGAP's Impact Pathfinder, aka "Pathfinder" (CGAP 2026a) — and also new analysis of institutional-level borrower data through Precision Causal Modeling (PCM): a machine learning-enabled framework that identifies which borrower and product characteristics predict positive or negative outcomes within a given lender's portfolio. Annex A provides a detailed description of the methodology underpinning this note.

Main Findings

Research shows that credit is a high-variance tool. Unlike savings, some forms of digital payments, and several insurance products where evidence tends to be more positive on average, its effects are fundamentally conditional on five intersecting factors: who borrows, how it is designed and delivered, what it finances (is used for), where (in what context) and when (over what time horizon).

The importance of borrower characteristics stands out in the evidence base. Where conditions align — prior business experience, adequate financial capability, genuine control over loan proceeds — the gains are real and meaningful: increased enterprise profits, household income, and asset accumulation; more effective management of health emergencies, seasonal shortfalls, and livelihood disruptions; and stronger investment in children's education and household welfare. These gains compound across successive cycles for borrowers who systematically progress. Risk rises where those conditions are absent. Harm concentrates among very low-income groups repeatedly borrowing for consumption without a productive base, first-time entrepreneurs taking large loans without complementary support, and borrowers facing recurrent shocks under rigid repayment terms. Whether borrowers retain genuine control over how loan proceeds are used is a decisive moderator throughout. The combination of repayment liability without corresponding control, most documented among women borrowers in restrictive social contexts, is one of the starkest risk-without-benefit patterns in the literature.

How credit is structured and delivered can be decisive. It gives FSPs considerable agency over outcomes. Product design shapes results independently of who borrows. Repayment schedules aligned with actual cash flow cycles, loan sizes calibrated to the productive opportunity (under-lending is as consequential as over-lending), and digital disbursement into borrower-controlled accounts are each associated with stronger outcomes. When credit is structured with complementary support — training, insurance, direct-payment mechanisms — it can build borrower capability over time and extend meaningful access to groups that would not benefit from credit alone. Institutional conduct, including incentive structures, pricing transparency, and collection practices, shapes outcomes independently of product features. In fact, the worst market outcomes on record were conduct failures before they were anything else.

What exactly credit finances adds a third dimension. Credit serves borrowers through three distinct functions: (i) financing productive investment that generates returns above borrowing costs; (ii) smoothing consumption across income gaps, health shocks, and seasonal variation; and (iii) building resilience: reducing vulnerability to future shocks before they strike. Productive and consumption credit each carry well-documented evidence and distinct risk profiles. Specifically, productive credit generates the most consistently documented gains where absorptive capacity exists, while consumption credit becomes harmful when repeatedly used to bridge a structural income gap that borrowing cannot close. The third function, resilience-building, is the least measured because its returns show up in events that do not happen: crises avoided, damage that is less severe, recoveries that are faster, rather than in income gains that are easy to observe and attribute to a single loan event.

Where credit operates modulates the effects of even well-designed products directed at the right borrowers. Strong regulatory and supervisory frameworks are associated with safer and more beneficial credit markets, while rapid expansion without adequate oversight is the most consistently documented precursor to market-level harm. Market structure is a further contextual variable. In well-regulated markets, competitive dynamics can drive beneficial product innovation. Without adequate consumer protection, those same dynamics instead incentivize aggressive over-lending and multiple-borrower exposure. Social norms, particularly around gender roles, household decision-making, and community attitudes toward debt, act as contextual filters that amplify or negate the effects of otherwise well-designed products. In addition, climate conditions introduce covariate shock risk, where drought, flood, or cyclone can simultaneously convert viable loans into distress events across entire borrower populations.

When credit is deployed, and over what time horizon, shapes what is visible and what accumulates unseen. Compounding gains from productive investment credit are often invisible at standard evaluation windows of 18 months to three years, while harm from repeated consumption borrowing accumulates gradually across cycles — dynamics that point-in-time assessments miss. Two further temporal distinctions matter. Episodic borrowing to manage discrete shocks can be protective, yet repeated borrowing to bridge a persistent income shortfall compounds obligations without resolving the underlying gap. And long-run welfare is shaped as much by how borrowers progress across successive cycles as by any single loan event, with gradual, step-by-step increases in loan size outperforming both static limits and aggressive early escalation.

Implications

For FSPs, the priority is screening clients for absorptive capacity: whether a viable productive opportunity exists that can generate returns above the cost of borrowing, not repayment eligibility alone. This means drawing on business history, income regularity, transaction records, and household composition alongside conventional credit information, and explicitly assessing whether clients, particularly women, retain genuine control over loan proceeds. Products should be calibrated to borrower cash flows and capability, with repayment flexibility matched to experience, for example, structured contracts to protect first-time borrowers and adaptable terms that benefit experienced ones.

Training, insurance, and gender norms programming should be treated as integral to the lending proposition where they address the binding constraint. FSPs should build deliberate climate risk management capacity, as pausing lending and intensifying collections after shocks can compound client harm and portfolio deterioration where recovery protocols are not already in place. Tracking debt-service ratios, rollover rates, and loan size progression across cycles gives a more accurate picture of welfare than repayment rates alone.

For funders and impact investors, the shift is from funding access to funding precision: prioritizing providers with strong client screening, appropriate product design, and credible outcome measurement beyond volume metrics. Evaluation horizons should align with the productive logic of the credit being financed, and reporting disaggregated by borrower gender, loan size band, and extent and type of prior business experience to reveal the heterogeneity that averages conceal.

For policymakers and regulators, responsible lending standards should treat vulnerability factors such as irregular income and high shock exposure as signals calling for adjusted product design rather than exclusion. Active supervision should track borrower trajectories and rollover patterns across cycles, not only point-in-time portfolio quality. Open finance infrastructure and credit information systems serve two distinct purposes: precision lending on one hand and systemic over-indebtedness monitoring on the other. Both are necessary.

Chapter 1: Why an Evidence Approach Matters More than Ever

A decade ago, the central question in microcredit was whether it worked. A generation of randomized controlled trials (RCTs) had delivered a sobering answer: on average and in the short term, not much. These findings punctured microcredit's most ambitious claims but risked a different error: dismissing credit as a development tool altogether.

Asking whether credit "works" is somewhat like asking whether a particular medication works. The useful question is which formulation, for which patient, at which dose, and under what circumstances.

The problem was not the evidence but the question. Asking whether credit "works" is somewhat like asking whether a particular medication works. The useful question is which formulation, for which patient, at which dose, and under what circumstances. The analogy holds at the level of the question rather than the mechanism. Unlike a medication, whose effects operate through broadly consistent biological pathways across patients, credit's outcomes are inherently context-dependent: the same product can generate opposite results for different borrowers in different settings. Credit is a high-variance instrument. The same loan product can accelerate a capable entrepreneur's growth, leave a subsistence household unchanged, or push a vulnerable borrower into a debt spiral. That variance is not random; it follows systematic patterns that the evidence base can now help identify.

This Focus Note is organized around that reframing: under what conditions, for which borrowers, and through what product configurations does inclusive credit generate meaningful benefit rather than harm? Its contribution is to organize and operationalize what a decade of dispersed, conditional findings implies for financial services providers (FSPs), funders, investors, policymakers, and regulators who cannot conveniently navigate the underlying literature directly. The five-factor framework is a structured, evidence-grounded synthesis designed to make heterogeneous impact findings actionable. It does not attempt to resolve all contested questions in the credit literature or assess aggregate welfare effects across the sector. Its focus is narrower and more tractable: identifying the conditions under which credit generates benefit or harm and how those conditions can be improved.

The note is primarily addressed to FSPs and funders, where the five-factor evidence is most directly actionable. Policymakers, regulators, and supervisors will also find it relevant to consumer protection frameworks and credit market regulation, as will researchers to the structured identification of evidence gaps.

Why Now

Three developments make a synthesis both urgent and timely.

First, credit operates at a scale where imprecision is costly. The ATLAS financial inclusion data platform estimates total outstanding portfolios across the broader inclusive finance sector at approximately $1.5 trillion (ATLAS 2024). At this scale, even modest design or governance failures affect hundreds of millions of households. Debt crises in Andhra Pradesh, India (most notably in 2010), Cambodia (from the 2010s onward), and other markets demonstrated that aggressive, poorly governed expansion generates widespread over-indebtedness, forced asset sales, and erosion of trust — predictable consequences of deploying credit without adequate attention to context and borrower circumstances. Capital scarcity compounds the problem: in most low- and middle-income economies, credit is limited relative to demand, so imprecise deployment risks harm to unsuitable borrowers while forgoing the developmental returns that better-matched credit would generate.

Second, the financial inclusion agenda has shifted its emphasis toward outcomes alongside access. According to the 2025 Global Findex database (Klapper et al. 2025), 24 percent of adults in low- and middle-income economies borrowed formally in 2024 yet large shares of households remain unable to manage short-term shocks or pursue longer-term goals. Global policy frameworks, including the G20's financial well-being agenda (GPFI 2025), increasingly treat improved welfare and resilience, not outreach alone, as the ultimate objective of financial sector policy and investment. This shift is particularly consequential for credit, the financial service that most demands evidence-backed precision in deployment.

Third, measurement and analytics are catching up with policy ambition. CGAP's Impact Pathfinder synthesizes high-quality evidence from over 860 financial inclusion studies, including 405 studies on credit. Financial health frameworks, large-scale administrative data, and increasingly sophisticated causal methods now allow practitioners to move beyond proxy indicators toward outcome-oriented decision-making. CGAP's Precision Causal Modeling (PCM) research demonstrates that good-quality longitudinal administrative data, the kind most providers already hold, can generate operationally meaningful findings about what works for whom, substantially lowering the feasibility barrier for outcome-oriented analytics (Nielsen and Sirtaine 2025).

Macroeconomic and climate volatility, rapid digital credit expansion, and tighter opportunity costs on investor and donor capital intensify the stakes. These three developments together create both the imperative and the opportunity for a more evidence-informed approach to how inclusive credit is deployed.

Chapter 2 traces how the evidence on inclusive credit has developed — from early optimism through the RCT revolution to the current understanding of heterogeneous, context-dependent impact — and what that trajectory means for how the evidence should now be read.

Chapter 2: How the Evidence Has Evolved: From Transformative Claims to Conditional Impact

The microcredit evidence base has evolved through distinct phases, each building on the limitations of the last and progressively sharpening the questions the sector can ask and answer.

2.1 Early Promise and the Selection Bias Problem

When Muhammad Yunus and the Grameen Bank received the Nobel Peace Prize in 2006, microcredit's promise seemed clear: small loans would enable poor households to start businesses, generate income, and lift themselves out of poverty (Norwegian Nobel Committee 2006). From its origins in Bangladesh in the 1980s, microcredit had expanded to over 100 countries.

The early evidence largely relied on observational studies comparing borrowers to non-borrowers, and the results looked dramatic: borrowers appeared healthier, wealthier, and more empowered. But critics identified a fundamental problem: selection bias. People who choose to borrow may systematically differ from non-borrowers in ways that independently shape their outcomes, making it impossible to attribute observed differences to the credit itself.

2.2 The RCT Revolution: Testing Transformative Claims

Recognition that rigorous evidence was lacking led to the first generation of microcredit RCTs (Banerjee 2013). Between 2003 and 2012, the Abdul Latif Jameel Poverty Action Lab (J-PAL) and Innovations for Poverty Action (IPA) launched experimental evaluations in seven countries (Bosnia, Ethiopia, India, Mexico, Mongolia, Morocco, and the Philippines), randomly assigning access to microcredit to isolate causal effects from confounding factors.

Despite wide variation in program design, specifically group vs. individual liability, interest rates from 12 percent to 110 percent, and different loan sizes and schedules, the findings were consistent. The American Economic Journal: Applied Economics special issue summarizing six trials concluded that the studies found no clear evidence of poverty reduction or substantial improvements in living standards, and no robust evidence of improvements in social indicators (Banerjee, Karlan, and Zinman 2015).

The landmark Hyderabad study illustrates the pattern. Evaluating in the study the group lending model that had become microcredit's global template, Banerjee, Duflo, Glennerster, et al. (2015) found increased investment and profits for preexisting businesses but no significant changes in health, education, or women's empowerment. It also concluded that microcredit did not deliver the transformative social change its proponents had claimed. These findings did not mean microcredit was harmful, but they definitively challenged the transformative narrative.

2.3 From Averages to Distributions: The Heterogeneity Breakthrough

Meta-analysis of the accumulated RCT evidence revealed a crucial insight: heterogeneity matters more than averages. Rachel Meager's 2019 Bayesian analysis, pooling seven microcredit RCTs, concluded that average effects were "unlikely to be transformative and may be negligible." Meager's 2022 follow-up in the American Economic Review found something more actionable. While most borrowers saw no profit effect, gains were concentrated at the top of the distribution, particularly among those with prior business experience. This fundamentally reframed the policy question — from "Does microcredit work?" to "For whom does it work, and under what conditions?" Heterogeneous impacts were not a discovery of the Meager meta-analyses alone. Subgroup analysis had been common practice in development economics RCTs for at least a decade. However, they had not previously been organized into practitioner-accessible guidance at this level of granularity.

When a study reports a 10 percent average improvement, this might mean everyone experiences modest gains — or that 30 percent of borrowers substantially benefit while 70 percent see no change.

The reframing exposed the limitations of first-generation evaluation designs. When a study reports a 10 percent average improvement, this might mean everyone experiences modest gains or that 30 percent of borrowers substantially benefit while 70 percent see no change (Nielsen and Sirtaine 2025). Most early RCTs compared access to credit against no access, a design appropriate for testing whether access mattered but one that abstracted away from the variation in loan terms, repayment structures, and bundled services the heterogeneity findings now identify as crucial. Standard evaluation windows of 18 months to three years compound this problem, as they are often too short to detect compounding returns from productive investment credit and too narrow to observe the gradual accumulation of harm from repeated consumption borrowing. A single standardized indicator applied uniformly across heterogeneous borrowers will, by construction, obscure precisely the variation practitioners most need to understand (Lahaye and Clarke 2025).

A further structural feature of the evidence base warrants acknowledgment: randomized evaluations are feasible at the margin of program expansion when credit is newly introduced to a population but not in mature markets where credit is already embedded in local economic and social contexts. The accumulated RCT evidence therefore captures expansion-phase dynamics; both the long-run benefits that accrue over sustained exposure and the risks of market saturation and over-indebtedness that develop in mature markets may be systematically underestimated.

What the heterogeneity findings established was an agenda rather than a conclusion — the right questions without providing tools to answer them at the granularity practitioners require. A new generation of RCTs is now directly addressing this issue, testing graduated loan sizes, flexible repayment schedules, performance-linked contracts, and digital delivery. These trials are generating more actionable findings than earlier designs could resolve (Graney et al. 2024).

2.4 When Credit Goes Wrong: Learning from Market Failures

While the RCTs suggested microcredit was rarely actively harmful on average, a body of observational, qualitative, and case study research documented serious harms in specific markets where rapid commercialization outpaced consumer protection (Mader 2013). These studies face their own methodological limitations, but the patterns they identify are consistent: aggressive expansion, competitive pressure to over-lend, coercive collection, and vulnerable borrowers trapped in debt spirals.

In Andhra Pradesh, India, the 2010 over-indebtedness crisis following SKS Microfinance's IPO led to emergency regulations shutting down operations (Saxena 2014). In Cambodia, loan sizes increased sevenfold between 2012 and 2020, to nearly three times the average household income, with Human Rights Watch documenting forced land sales, child labor, and debt-driven suicides (Human Rights Watch 2025; Finch and Kocieniewski 2022). Similar patterns emerged in Sri Lanka and Jordan (Human Rights Watch 2021).

Muhammad Yunus famously reflected that he never imagined that one day microcredit would give rise to its own breed of loan sharks (Yunus 2011). The critique is more precisely directed at governance and incentive failures than at the commercial model itself. Commercial capital was instrumental in globally scaling financial inclusion, particularly in Latin America, and many commercial microfinance institutions (MFIs) have maintained strong client welfare standards. The over-indebtedness crises reflect what happens when aggressive growth targets, investor return expectations, and weak governance converge without adequate consumer protection or regulatory oversight.

These cases are not representative of inclusive credit overall, but they establish a critical point: serious harm follows identifiable and predictable patterns — aggressive expansion, weak protection, poor governance — that can, in principle, be detected and addressed before they become crises.

Chapter 3 draws on the full evidence base to identify the five factors that most consistently determine whether credit generates benefit or harm — and how they interact with one another across different borrower profiles and contexts.

Chapter 3: Five Factors that Shape Outcomes

3.1 Reading the Evidence: The Focus Note's Approach and Scope

This Focus Note primarily synthesizes evidence from CGAP's Impact Pathfinder, supplemented by insights from its PCM research with five financial sector institutions. These two sources do not constitute a unified evidence base. The Pathfinder draws on published research while PCM draws on proprietary operational data from an institutional sample. Both streams are used across the five-factor framework presented in detail in this chapter, and the note is explicit about which stream each claim draws on. Where findings from both streams point in the same direction, the convergence adds weight. Where they differ, PCM findings should be read with appropriate caution. Where individual studies are cited in the text, they serve as illustrations of broader patterns rather than as the sole evidentiary basis for the findings they accompany.

The Impact Pathfinder

The analysis in this chapter draws on 405 credit-focused studies in the Impact Pathfinder. The Pathfinder is an evidence synthesis platform covering over 860 high-quality studies of inclusive financial services. Studies span RCTs, quasi-experimental designs, systematic reviews, and other rigorous empirical research, and are assessed on methodology quality, relevance, and actionability to produce a confidence rating rather than a simple vote count. The evidence base covers publications from 1994 to November 2025, concentrated in South Asia and Sub-Saharan Africa. Annex A provides a full description of the Pathfinder's methodology, inclusion criteria, and the scope boundaries relevant to this note's enabling environment claims.

The Pathfinder is organized around the uptake and use of specific financial products. Enabling environment factors, including regulation, supervisory capacity, data infrastructure, credit information systems, market structure and conditions, and social norm regimes, appear in those studies as contextual moderators of product-level outcomes, not as independently evaluated interventions. The note's enabling environment claims therefore rest on a different evidential foundation from its product-design and borrower-characteristic claims. The enabling environment recommendations in Chapter 4 draw on practitioner consensus, CGAP's research, regulatory literature, and case evidence alongside Pathfinder synthesis, and are offered as indicative rather than evidence-based in the same sense as the product-level findings. Saying so does not undermine those recommendations; it correctly calibrates the confidence with which they should be held.

The Pathfinder synthesizes evidence from studies designed by researchers to measure outcomes they identified as relevant. This means the evidence base reflects what external observers measured, not what borrowers reported experiencing. How borrowers themselves understand credit decisions, assess risk, navigate repayment, and weigh the trade-offs involved is a distinct body of evidence encompassing financial diary research, qualitative debt stress studies, and work on borrower coping strategies, all of which falls outside the scope of this analysis. Findings on agency, harm, and the limits of product design should be read with that boundary in view.

Precision Causal Modeling

The Pathfinder evidence is supplemented by findings from CGAP's PCM research with five financial institutions across 19 countries. PCM is a causal discovery framework that inverts the conventional logic of evaluation. Rather than evaluating a pre-specified treatment, it begins with a defined measure of client progress and works backward to identify which lending practices are associated with stronger outcomes among otherwise comparable borrowers. Its credibility rests on population-level selection-on-observables diagnostics rather than experimental assignment. Three limitations apply wherever PCM findings are cited: (i) it cannot rule out unobservable confounding; (ii) borrowing pathway findings are subject to survivorship bias; and (iii) PCM findings have not yet appeared in peer-reviewed journals. Within these boundaries, PCM contributes an operationally grounded layer of evidence that is directionally informative and actionable. Annex A provides a full methodological account.

A note on terminology and evidence scope. This Focus Note uses "inclusive credit" as its governing term: credit extended to low-income individuals, households, and MSEs, including microcredit, agricultural credit, and digital credit, but not credit to large firms or the general consumer credit market. The evidence base is concentrated in microcredit and digital credit studies. Findings may not uniformly transfer to other inclusive credit forms, and the note flags where this caveat is relevant. Around 40 percent of the 405 Pathfinder credit studies explicitly used the term "microcredit," while many others examined microcredit without using that label. Microcredit is used where the evidence or narrative specifically concerns that segment, notably in the historical account in Chapter 2. Where the note refers to "credit" for brevity, inclusive credit is assumed.

The Five-Factor Framework

Analysis of the Pathfinder evidence reveals five high-level factors that consistently shape credit outcomes:

• WHO borrows: User demographics and characteristics
• HOW credit is structured and delivered: Product design, delivery channels, and bundling
• WHAT it finances: Use of funds
• WHERE it operates: Market, regulatory, and other ecosystem contexts
• WHEN is it helpful: Time horizon

These factors do not operate in isolation. They function as part of a broader system where their intersections determine the extent to which credit supports or undermines positive outcomes. For example, a woman's ability to benefit from credit depends not only on her own capabilities but on whether the product protects her control over funds and whether the social context supports her economic participation. Similarly, a low-income borrower's outcomes may depend on whether credit is structured to match their cash flows and whether complementary support is available. This chapter examines each factor in turn, keeping this intersectionality in view throughout. Annex B contains summary tables for the five high-level factors.

The concept of precision that runs through this note should be understood in probabilistic rather than deterministic terms. The five-factor framework identifies conditions associated with higher or lower probabilities of positive outcomes. A borrower with prior business experience, a flexible product, and a functioning market is more likely to benefit, but outcomes will still vary within such groups. Precision means more evidence-informed and better contextualized decisions. It does not guarantee individual outcomes.

Annexes D through I provide outcome-by-outcome summaries of the Pathfinder evidence base, covering women's economic empowerment, entrepreneurship and jobs, poverty reduction, climate adaptation and resilience, health, and energy access. Annex C provides a consolidated overview of the influencing factors themselves.

3.2 Who Borrows: User Demographics and Other Characteristics

Who borrows (i.e., gender, wealth, education, life stage, household composition) shapes how credit is used and whether it translates into sustained gains or increased risk. The evidence points to three key mechanisms through which borrower characteristics shape outcomes: (i) who controls resources; (ii) whether there is a viable income or asset base to absorb the loan; and (iii) whether the borrower has the capability and timing to convert credit into productive use. These characteristics do not operate in isolation. They consistently interact with other factors to shape outcomes, especially context and product design and delivery. A note of caution applies throughout: observed outcome differences across borrower groups partly reflect the systematically different products each group accesses rather than borrower characteristics alone. The evidence identifies who benefits under current conditions rather than who could benefit from well-designed credit.

These characteristics change slowly, if at all. Some, like wealth, education, and household structure, may shift over a lifetime. Others, like sex at birth, remain constant. What they share is relative stability compared to the more dynamic factors explored in later sections. Understanding these characteristics is essential for targeting lending, product design, and safeguards. Importantly, identifying these characteristics is not about sorting borrowers into "good" or "bad" candidates for exclusion, but about understanding how to better support different segments, particularly women and low-income households, through more precise design and complementary measures.

Gender

Women remain structurally underserved in inclusive credit markets. They receive fewer and smaller loans than men, face higher collateral requirements, and access credit on less favorable terms, despite evidence of comparable or lower credit risk (Alonso and Dezso 2024).

When loans are nominally taken out by women but appropriated by male household members, welfare gains disappear while repayment liability remains.

Gender is one of the most consistently important modifiers of credit impact. The central finding is that credit directed to women is associated with stronger household welfare gains than credit directed to men, but only where women genuinely retain control over loan use. Control is shaped by both product design (whether disbursement protects women's autonomy) and context (whether social norms support women's economic roles). Women borrowers are more likely to allocate resources toward household welfare. For example, research on migrant households in India found that directing credit to female members generated substantially larger gains in household consumption, income, and savings than directing equivalent loans to male household members (Bahure 2023). When loans are nominally taken out by women but appropriated by male household members, welfare gains disappear while repayment liability remains: a risk-without-benefit pattern that is among the most harmful combinations in the literature.

A qualification is warranted. Outright capture — where men direct loan proceeds to their own uses while women bear repayment liability — differs from voluntary reallocation toward a household member with greater absorptive capacity. The welfare-relevant concern is the liability-benefit mismatch: whether the woman responsible for repayment shares in the income the loan generates.

Credit directed to women also generates stronger health and human capital outcomes. Evidence from South Asia shows that women's participation in credit programs is associated with higher spending on education and improved maternal and child health, whereas comparable positive effects for men's borrowing is less consistently documented. In Bangladesh, wives' credit increased education spending while husbands' credit reduced it (Porter 2016). Women-owned enterprises with larger firms, higher education, and stronger entrepreneurial identity respond particularly well to appropriately sized credit. IPA synthesis work finds sizeable long-run gains in profits, employment, and assets when product design fits borrower characteristics (Graney et al. 2024).

The downside risks are equally gendered. Evidence from Cambodia reveals that women in debt-stressed rural households experienced compounding psychosocial and health deterioration when servicing loans taken for basic needs and health shocks, particularly where women were nominal borrowers with limited control over proceeds (Iskander et al. 2025).

A structural dimension beyond individual transactions should be noted. In rural Tamil Nadu, India, women increasingly function as their households' primary debt managers, bearing reputational consequences of default even where proceeds have been shared or appropriated by others (Guérin, Kumar, and Venkatasubramanian 2023). Longitudinal Observatory of Rural Dynamics and Inequalities in South India (ODRIIS) data find households devoting around 30 percent of income to debt servicing across all sources, a systemic burden invisible in studies measuring individual loan outcomes in isolation.

Restrictive gender norms do not preclude women from benefiting from credit. Rather, they call for product design and delivery approaches — direct disbursement, commitment devices, group programming, and engagement with male partners — that can work around or shift those constraints (Koning, Ledgerwood, and Singh 2021).

Economic Status

A borrower's economic status shapes whether credit supports income generation or primarily serves as short-term liquidity support. Evidence suggests that households with a viable asset base and stable earning capacity are more likely to productively deploy credit, while those facing extreme poverty or chronic income instability face greater risk of debt accumulation. However, outcomes for lower-income borrowers are not uniformly negative. They depend significantly on how credit is designed, what complementary support is available, and whether the broader context enables productive use. The critical question is not whether a household is "poor enough" to be excluded from credit, but whether the necessary conditions are in place for credit to generate benefit rather than harm, including appropriate product design, delivery mechanisms, and accompanying services.

Evidence also shows that the distribution of income gains from credit is uneven. Research in Vietnam and Bangladesh revealed that income gains from microcredit are concentrated among higher-income or better-off rural households while impact for the poorest is often statistically insignificant (Cuong 2008; Roodman and Morduch 2014). Similarly, household-level panel data from China show that a one percent increase in debt reduced wealth accumulation for poor households while increasing it for rich households (Wu, Yue, and Zuo 2023). These findings should be interpreted with caution. Economic status categories are not standardized across studies and may not be comparable across contexts, and there is meaningful variance within income segments.

Evidence from Ethiopia suggests that very low-income households with few assets and limited labor use a large share of credit for consumption and often repay by selling livestock or taking new loans (Siyoum, Hilhorst, and Pankhurst 2012). In contrast, better-off households are more likely to use loans for productive purposes. Similar patterns are reported in multiple countries; without a viable productive opportunity and minimum asset base, credit tends to be used to address immediate needs and can, in some cases, contribute to problematic debt over time.

For households facing extreme poverty, repeated shocks, or chronic income instability, the evidence suggests that other interventions, such as grants, social protection, insurance, or income stabilization, may need to precede or complement credit before it can play a productive role.

Productive Capacity: Business Size and Experience, Financial Capability, and Education

Business size and experience, financial capability, and education are strong moderators of credit impact, tilting it from a coping instrument toward a productive investment tool. RCT evidence shows that while microcredit has mostly positive or mixed effects on business outcomes overall, gains concentrate among borrowers who already have businesses or larger, more profitable firms. Effects for inexperienced or very small entrepreneurs are typically negligible or not positive, including in agricultural settings (Banerjee, Karlan, and Zinman 2015; Meager 2019).

Differences in entrepreneurial experience are among the strongest predictors of who can absorb and benefit from credit. An Egyptian RCT offering borrowers loans four times larger than they had previously taken found negligible average impacts, but machine learning applied to psychometric data revealed striking heterogeneity. "Top performer" borrowers substantially increased profits, while poor performers saw profits fall and were more likely to exit (Bryan, Karlan, and Osman 2024). Conventional lender criteria (e.g., firm size, track record) led to significant misallocation. Psychometric assessment of entrepreneurial type was a more reliable predictor of absorptive capacity, suggesting this has a dispositional dimension that experience and firm characteristics alone do not capture. Similar patterns appeared in Pakistan, where gains from flexible repayment were concentrated among more experienced or disciplined entrepreneurs (Bari et al. 2021). More experienced operators can identify profitable investments and effectively manage working capital while borrowers with little or no prior experience often redirect credit to household consumption, meaning credit rarely kick-starts successful new firms.

Educational level further differentiates who benefits from credit. The mechanism appears to be that borrowers with more years of education are better equipped to evaluate investment opportunities, manage financial contracts, and navigate market conditions. For example, one study in Vietnam found significant income gains only for borrowers with higher levels of education, while borrowers with less education showed no clear income effect (Cuong 2008). In Bangladesh, years of schooling and credit utilization were significant predictors of post-credit income for women (Islam 2015). Similarly, a study in Nigeria found that farmers with higher education levels were both more likely to access credit and more likely to use it to adopt climate-smart agricultural practices (Olutumise 2023).

Further, limited financial capability can constrain or even erode the benefits of credit. Pay-as-you-go (PayGo) solar and microcredit evidence indicates that some low-income households underestimate total repayment costs, increase spending more than expected, or struggle with contract terms, particularly where literacy and numeracy are low. These findings highlight the interaction between borrower characteristics and product design. Complex products with opaque pricing carry greater risk for borrowers with limited financial capability while simpler, more transparent products and complementary financial education can mitigate these risks.

Life Cycle: Age, Life Stage, and Household Composition

Age, life stage, and household composition moderate credit impact, although findings are often context-specific. What appears to matter is the timing of credit relative to the household life cycle. Households in phases of asset accumulation and income growth tend to seek investment-oriented credit and are more likely to generate sustained gains from it. Those facing high dependency burdens or income volatility tend to seek short-term liquidity support, which is more likely to be absorbed into immediate needs without building repayment capacity. The same household at different life stages will typically be in the market for a different product, and outcome differences across life stages reflect both underlying circumstances and the credit those circumstances call forth. The causal logic here runs from circumstances to credit demand, not the reverse. Life stage shapes what credit is sought and used for; it does not follow that different credit types determine life stage outcomes. Within this broader dynamic, responsiveness to credit-enabled innovation peaks in mid-life and declines among older household heads, as shown in rural Ghana (Twumasi et al. 2020).

The timing of borrowing relative to children's developmental stages also shapes long-run outcomes. Evidence from China shows that early parental access to credit, before schooling gaps emerge, can reduce long-term education disparities, while borrowing later in the household life cycle yields no detectable schooling benefits (You and Annim 2014).

Household composition introduces more consistent constraints. Larger households with many school-age children and few earners face tighter budget and labor trade-offs. Studies from Bangladesh, Vietnam, and China report associations between microcredit use and weaker education outcomes in larger households, particularly where children are drawn into household enterprises to support loan repayment (Islam and Choe 2013; Kondratjeva and Chen 2018; Doan, Gibson, and Holmes 2014). Similar pressures appear in energy transitions, where larger households face competing demands that can limit adoption of cleaner fuels even when credit is available.

Who Benefits from Credit: Conclusions

Credit is primarily a lever; it amplifies existing capacity rather than creating it from nothing.

Credit most consistently generates gains among borrowers with an existing productive base — assets, stable earning capacity, prior business experience, higher education — and women who retain meaningful control over loan use. Credit is primarily a lever; it amplifies existing capacity rather than creating it from nothing.

This does not mean credit is appropriate only for borrowers who are better off. Structured with complementary support — business skills, financial education, insurance, direct-payment mechanisms, gender-sensitive design — credit can build borrower capability over time and extend benefits to groups that would otherwise face higher risk.

For households facing extreme poverty, repeated shocks, or chronic income instability, credit alone rarely generates sustained gains and may cause harm. Grants, social protection, or income stabilization may need to precede or complement credit. Graduation programs (schemes that sequence asset transfers, savings, training, and psychosocial support before introducing credit) — most prominently the Bangladesh Rural Advancement Committee (BRAC) model — provide the strongest evidence for this, with rigorous long-run results from the Bandhan program in India (Banerjee, Duflo, Goldberg, et al. 2015; Bandiera et al. 2017).

Outcomes depend not on borrower characteristics alone but on how those characteristics interact with product design, context, and enabling conditions.

3.3 How Credit Is Structured and Delivered

Product design and delivery are among the most potent and actionable factors determining credit outcomes. The evidence consistently shows that how credit is designed matters at least as much as whether it is provided at all. FSPs have considerable agency over these features, making this a domain where evidence can directly translate into improved practice. Agency is necessary but not sufficient; incentive structures, competitive dynamics, and regulatory environments shape whether providers exercise that agency in borrower-beneficial ways.

Design features do not operate in isolation. Their effects depend on who borrows, what the credit is used for, and what broader context it operates in. A product that works well for experienced entrepreneurs in a well-regulated market may generate harm for first-time borrowers in saturated markets with weak supervision.

Repayment Flexibility

Grace periods and more flexible repayment schedules often improve investment and profit outcomes, particularly for borrowers pursuing higher-return, more illiquid investments who face short-term liquidity pressure. At the same time, grace periods can raise default rates and require careful targeting and risk management. Matching repayment to borrowers' actual cash flows, such as seasonal schedules for farmers and monthly cycles for borrowers with regular income, is associated with higher investment, higher profits, and reduced default rates. In Pakistan and Bangladesh, flexible repayment substantially increased business capital (by up to 80 percent relative to control groups) and profits (by 25 to 55 percent) (Bari et al. 2024; Battaglia, Gulesci, and Madestam 2024). Rigid weekly schedules have been linked in multiple studies to higher repayment stress and, in some contexts, to lower sales, profits, and working capital (Field et al. 2013; Banerjee, Karlan, and Zinman 2015).

"Debt juggling" (i.e., taking new loans to service existing ones) is a related mechanism. When repayment schedules do not align with income flows, borrowers take new loans to service existing ones, amplifying the real cost of borrowing in ways rarely traced by official interest rates and short-term evaluations (Guérin, Kumar, and Venkatasubramanian 2023).

Flexibility is especially critical for women and rural households. Women's income patterns are often more irregular and integrated with household consumption needs, and agricultural households face seasonal cash flow mismatches with standard repayment structures. However, flexibility most often benefits more experienced, capable borrowers. First-time or lower-capability clients sometimes perform better under structured, fixed contracts that limit overborrowing and establish repayment discipline. The optimal repayment structure therefore depends on who is borrowing, not only on product design itself.

How an institution responds to missed payments also shapes outcomes. CGAP's PCM evidence from five diverse financial institutions found that delinquency trajectory — whether a borrower is recovering or deteriorating across successive loan cycles — is a stronger predictor of downstream outcomes than any individual delinquency event. It also found that rigid zero-tolerance delinquency policies were consistently associated with worse client outcomes. Context matters for interpretation. For agricultural borrowers with seasonal income, for example, repayment variability is inherent. For workers with predictable income, a missed payment carries a different signal. In one group lending program, clients with access to late payment recalculation mechanisms achieved progression targets at rates 31 to 72 percentage points higher than those on rigid schedules, although as observational PCM evidence this cannot rule out underlying differences between the two groups. Repayment monitoring and institutional response should be calibrated to income profile and livelihood type, distinguishing distress-driven delinquency from cash flow variability rather than enforced through uniform rules.

Digital Channels and Digital Credit

"Digital" lending covers a wide spectrum, from fully algorithmic lending with no human touchpoints to the use of digital channels within an otherwise conventional model. For clarity, digital delivery refers to the use of electronic accounts (e.g., mobile money, bank accounts, digital wallets the borrower controls) for disbursement and repayment. Digital credit as a product category refers to algorithmically underwritten, typically short-tenor loans disbursed via mobile platforms, often without loan officer involvement. The two have different implications for borrowers.

The same loan, routed differently, produced materially different outcomes.

As a delivery channel, digital disbursement reduces transaction costs, expands geographic reach, and, where funds flow into borrower-controlled accounts, enhances the privacy and security of loan proceeds. PayGo financing further extends these benefits, combining digital payment collection with remote asset control to convert prohibitive upfront costs into manageable periodic payments. These models have been especially effective for solar home systems. A Ugandan RCT found that disbursing loans into women-controlled digital accounts increased business profits by 15 percent and capital by 11 percent relative to cash disbursement to the same women, with the largest gains for women who reported strong sharing pressure at baseline (Riley 2024). This is a finding about channel design since the same loan, routed differently, produced materially different outcomes. Evidence from women-focused fintechs similarly showed that tailoring repayment schedules to irregular revenue cycles can expand responsible access (Dalberg 2024). Digital accounts are not universally accessible, however; exclusion of borrowers without devices or connectivity is a real risk.

Digital credit as a product category introduces a distinct set of risks. Short-tenor, high-cost products can trap borrowers in refinancing cycles. The speed and automation of digital disbursement, combined with limited human touchpoints, compounds these risks for borrowers with limited financial capability or where consumer protection is weak. Borrowers are frequently unaware of terms, fees, and penalty structures, and algorithmic decisioning reduces opportunities to contest or seek recourse against lending decisions. Together, these features can allow exploitation and over-indebtedness to rapidly scale where regulatory and supervisory safeguards are absent (Duflos et al. 2024; Cassara, Zapanta, and Garz 2024).

The evidence is not uniformly negative. For example, well-governed algorithmic underwriting in a Chinese state bank simultaneously increased approval rates and reduced default rates by drawing on nontraditional signals (Li et al. 2024). The key distinction is between well-governed systems that can more fairly extend credit to previously excluded borrowers and opaque, under-regulated systems — often non-bank fintechs — where the absence of transparency and oversight creates conditions for exploitation and denial of recourse. Digital credit works best when accompanied by transparent pricing, robust consumer protection, adequate supervisory enforcement, and sufficient borrower financial capability.

Bundled Services and Complementary Support

Bundling credit with training, insurance, or advisory support has repeatedly been shown to improve both who benefits and how credit is used because credit alone is often insufficient to address the capability gaps, risk exposure, and information asymmetries that prevent borrowers from converting loans into productive gains. Syntheses of randomized and quasi-experimental evaluations find that standard microcredit contracts produce modest average effects on profits and virtually no impacts on consumption or poverty. However, impacts are substantially larger when credit is combined with business management, financial capability, or life skills training that is recurrent, market-oriented, and adapted to clients' time and mobility constraints (Jayachandran 2021; Banerjee, Karlan, and Zinman 2015).

In agriculture, bundling credit with climate insurance and advisory support can change both who borrows and how loans are used. Risk-contingent credit designs that link repayment obligations to weather or yield indices allow insurance payouts to cover part of loan obligations in bad years, reducing default risk and making credit accessible to risk-rationed farmers who would otherwise avoid borrowing (Ndegwa et al. 2024; Shee, Turvey, and You 2019). A related design — pre-committed emergency credit lines that automatically activate after a verifiable shock — achieves similar risk-reduction effects without requiring up-front insurance purchase. A large-scale RCT with BRAC in Bangladesh found that access to a guaranteed, post-flood emergency loan led treated farmers to expand cultivated land by 18 percent and crop production by 19 percent, with the product proving profitable for the lender — a rare combination of client benefit and institutional viability (Lane 2024). However, experience shows that added complexity can hinder uptake and claims as farmers often do not know bundled insurance is included or how to trigger payouts. Training and advisory support delivered alongside credit increase adoption of climate-smart practices by helping smallholders plan investments, interpret climate information, and integrate new technologies.

For women in restrictive social contexts, credit alone often cannot overcome entrenched norms around mobility, decision-making, and asset control, and in some settings has triggered backlash. Stronger and more durable gains in agency occur when economic services are bundled with components that explicitly work on norms, such as engaging male partners and community leaders, building women-only support groups, and creating opportunities for public participation (Chang et al. 2020; Jayachandran 2021).

The history of bundling is, however, mixed. Bundling has been used to force acquisition of add-on products, mis-sell insurance, and exploit information asymmetries. There are documented cases where expensive and opaque credit-linked insurance was effectively mandatory, and where digital credit apps layered fees and "protection" products onto captive users (Cassara, Zapanta, and Garz 2024). Patterns like these arise when provider incentives are misaligned and conduct rules are weak. Effective bundling therefore depends not only on thoughtful product design but on regulatory and supervisory safeguards that prohibit conditioning access to credit on unrelated products and that require clear consent and stand-alone value for any bundled element.

Credit Parameters: Loan Sizing, Loan Structure, and Secured vs. Unsecured Credit

The risk of lending too little may be at least as consequential as the risk of lending too much.

Loan size must match both opportunity and capacity. Larger loans can generate sizeable gains for screened, higher-ability, or more experienced borrowers but are risky or harmful when broadly given to clients without such profiles (Bryan, Karlan, and Osman 2024). "Large but patient" loans tailored to business cycles work very differently from large, rigid, short-tenor loans. PCM evidence adds a complementary insight: the risk of lending too little may be at least as consequential as the risk of lending too much. Insufficient capital relative to the borrower's productive opportunity was the single most consistent predictor of poor outcomes across the five partnerships. Progressive scaling of around 30 to 50 percent per loan cycle outperformed both flat loan sizes and rapid jumps of 100 percent or more, and minimum loan floors needed to exceed survival-level amounts for credit to serve a developmental rather than merely coping function.

Whether credit is secured or unsecured has significant implications for outcomes. Secured lending, including asset-collateralized loans, leasing, and PayGo finance, can support larger loan sizes and strong repayment among borrowers without traditional collateral. For example, an asset-collateralized water tank loan in Kenya requiring only a 4 percent cash deposit achieved over 40 percent uptake and virtually no repossessions while remaining commercially viable (Jack et al. 2023). Unsecured credit is essential for including borrowers without assets but carries higher credit risk and is particularly prone to predatory pricing and over-indebtedness when consumer protection is weak, a well-documented pattern in algorithmic mobile lending markets.

Interest Rates and Pricing

Interest rate levels have received substantial attention in the microcredit debate, but the evidence on their consequences for borrower outcomes is more nuanced than typically acknowledged by either advocates of rate caps or their opponents. Rate caps risk reducing access. If a ceiling falls below the rate at which lenders can profitably serve higher-risk borrowers, those borrowers may be excluded from formal credit markets (Alper et al. 2019; Maimbo and Gallegos 2014). Whether that exclusion harms or protects them is genuinely contested and varies by context. Barboni and Agarwal (2023) find that loan flexibility — specifically the ability to defer repayment — matters as much as rate level for borrower outcomes, suggesting that structural design features interact with price in ways that make rate-only analysis incomplete.

Rates clearly matter when they are predatory. APRs above 200 percent documented in some digital credit markets are harmful regardless of other product features. Within the range typically charged by responsible institutions, however, rate differences appear less important than structural design, borrower capability, and context. What matters more is debt service capacity relative to income; rate trajectory over time, including whether declining rates signal institutional recognition of progression; and whether the borrower's overall credit mix is becoming less expensive as the relationship matures. PCM evidence across all five partnerships is consistent: interest rate level consistently ranked below loan structure, sizing, and growth trajectory as a predictor of client success. These findings should be interpreted with caution as they reflect the rate ranges observed among five partners and do not address predatory pricing.

A further interpretive caveat: if these partner institutions were already implementing risk-based pricing at the time of data collection, the finding that rate level is a weak predictor of client outcomes may partly reflect that rates already incorporated borrower risk, leaving limited independent predictive variance for rate level alone to explain. Policy attention narrowly focused on rate reduction may therefore be less effective than a broader agenda simultaneously addressing product structure, transparency, and the enabling environment.

Loan Maturity and Tenor

Matching tenor to the expected return profile of the user is a core design principle. For MSEs with short order-to-cash cycles, short tenors may be entirely appropriate. For productive investments with longer gestation periods — business equipment, climate-smart agricultural inputs, education — short tenor creates mismatches that force premature repayment from income not yet generated. Evidence from India shows that when repayment structures better accommodate irregular business cash flows through grace periods or flexible schedules, borrowers are more likely to undertake higher-return illiquid investments without raising default rates (Barboni and Agarwal 2023). Extending tenor is not a free choice on the provider side. Many non-deposit-taking MFIs depend on short-term wholesale funding, which creates asset-liability mismatches, and many MSEs genuinely need short-term working capital rather than long-dated debt. The appropriate response is to match product to purpose, designing maturity and repayment to fit the underlying cash flow pattern rather than assuming longer is always better.

Institutional Conduct and Responsible Provision

Product design analysis captures only part of the "how" dimension. FSP institutional conduct — how providers behave toward clients, and the incentive structures behind that behavior — shapes outcomes independently of product features. Institutional conduct is addressed last in this section not because it is less important than product design but because it is more fundamental. The best designed product in the hands of a provider with misaligned incentives will reliably underdeliver — and the worst market outcomes on record were conduct failures before they were anything else.

In Ghana, a transparency intervention in mobile financial services reduced misconduct by 72 percent, with the largest welfare gains accruing to female clients.

Inadequate transparency about credit pricing and terms is associated with higher effective prices and poor product choices. The burden disproportionately falls on women, who face systematic overcharging where pricing is opaque (Bertrand et al. 2010; Bill & Melinda Gates Foundation 2024). In Ghana, a transparency intervention in mobile financial services reduced misconduct by 72 percent, with the largest welfare gains accruing to female clients (Annan 2025). Studies in East Africa found that digital credit products with opaque fee structures were associated with high rates of late repayment and default, while simple improvements to disclosure design materially reduced default rates (Kaffenberger and Totolo 2018; Mazer and McKee 2017). Aggressive expansion targets, commission-based loan officer incentives, and coercive collection practices can undermine even well-designed products. In fact, the market failures documented in Chapter 2 were primarily conduct failures, not product design failures.

A related and underappreciated dimension concerns FSP conduct during climate shocks. The standard response — intensifying collections and withdrawing from affected areas — is understandable as a short-term balance sheet measure but tends to compound client harm and deepen portfolio deterioration. Following the 2022 floods in Pakistan, 40 percent of MFIs reduced or stopped lending to climate-affected sectors at the exact moment clients needed capital to recover (Notta and Zetterli 2025). PCM evidence points toward the positive. In one group lending partner, loan officer tenure of nine or more years was associated with improvements in outcomes of up to 44 percentage points for the most vulnerable clients. In a second partner, customer care quality involving staff competence, rapport, and fair treatment emerged as a leading driver of outcomes in nine out of 11 countries. How services are delivered matters as much as what is delivered.

How Credit Is Structured and Delivered: Conclusions

The evidence reinforces a finding that runs throughout this Focus Note: outcomes are shaped by the interaction of multiple factors rather than by any single design feature. Repayment flexibility raises investment and profits for experienced borrowers but can encourage overborrowing among first-time clients. Digital delivery protects women's autonomy and expands access in some contexts while generating refinancing traps in others. Bundled services can multiply credit's effectiveness but add complexity that may depress uptake among borrowers who most need risk mitigation. Larger loans generate sizeable gains for screened, higher-ability borrowers and near-equivalent losses for those without the profile to absorb them. In each case, a feature that is beneficial under one set of conditions becomes neutral or harmful under another.

The challenge for providers is calibration: matching the full package of product characteristics — repayment structure, delivery channel, bundled support, loan size, pricing transparency — to the borrower's profile, intended use, and operating context. This requires ongoing monitoring to learn what works for whom. It also requires an active reading of contextual factors (e.g., regulation, market conditions, social norms around women's use of credit) as variables that shape what any given product configuration can achieve, rather than stable "givens" that can be assumed away.

3.4 What Credit Finances: Use of Funds

What credit is used for may be the most consistent cross-cutting predictor of outcomes. Across contexts, credit tends to generate the strongest positive impacts when it finances investments that expand productive capacity, strengthen human capital, increase mobility, and enable the adoption of climate-smart practices. A third category deserves explicit recognition: credit used to build resilience — financing investments that reduce vulnerability to shocks and improve recovery capacity over time, such as climate-smart inputs, water storage, diversification assets, or the restoration of productive equipment after a crisis. Resilience-building credit is conceptually distinct from both productive investment, where the return is measured in income, and consumption smoothing, where the logic is emergency coping. Its returns are partly counterfactual, measured in shocks that do not occur or faster recoveries. Credit for occasional consumption smoothing or emergency coping can also serve a legitimate protective function, helping households to avoid distress asset sales or maintain children's schooling during isolated shocks. However, when repeated borrowing for consumption or shock-coping substitutes for income growth or social protection over time, credit can entrench vulnerability.

The critical distinction is not between "productive" and "consumption" use as binary categories, but between uses that build capacity over time and repeated uses that service immediate needs without generating returns to cover borrowing costs. Use of funds also interacts with borrower characteristics, product design, and context. A productive investment loan to a capable borrower in a supportive market context generates very different outcomes than the same loan in a depressed economy or a loan to a borrower without the capability to deploy the capital.

Lenders can exert greater control over how credit is used when loans are delivered in kind or through structured value chains. In cash-based microlending, however, control is much weaker. Although most microcredit is formally designated for business or productive purposes, fungibility of money means borrowers can, and often do, redirect funds to consumption, household expenses, school fees, or debt repayment. An alternative is to ensure the availability of a sufficiently diverse mix of instruments so borrowers can self-select the type of credit that best matches their circumstances rather than use business credit for consumption by default.

Productive Investment

When directed toward equipment, inventory, technology, and climate-smart agricultural practices, credit often produces positive returns on business outcomes. A randomized evaluation of microcredit for small businesses across six countries showed that expanded credit access increased business investment, scale, and profits (Banerjee, Karlan, and Zinman 2015). Wider evidence from the Impact Pathfinder shows that asset-collateralized loans for equipment have dramatically increased durable asset ownership with strong repayment patterns. These types of productive investments often require larger loan sizes because many capital expenditures are invisible and must be paired with sufficient working capital.

Credit can enable purchase of improved inputs and technologies that both raise productivity and strengthen resilience to climate variability. Pathfinder evidence from Ethiopia, for example, suggests that farmers with access to credit employ significantly more climate adaptation strategies than their credit-constrained peers (Negera et al. 2025). The practices supported include improved drought-tolerant seed varieties, micro-irrigation systems that enable dry season cultivation, soil conservation measures, and climate-informed planting schedules. Evidence suggests that use of credit for climate-smart practices works best when it is timely, appropriately sized, and flexible; paired with climate information, agronomic advice, and/or insurance; and delivered through trusted group-based or community structures such as village savings and loan associations (VSLAs). A methodological caveat applies, however. Not all studies clarify whether credit was specifically directed toward climate-smart practices or whether adopting farmers simply had access to general-purpose credit alongside other resources, a distinction that matters for replicability.

Returns from productive investment are strongest for borrowers with the experience and capability to effectively deploy capital, and loan sizing is critical. Favorable market conditions, such as demand, stable prices, and strong supply chains and market linkages, are clearly important. Even well-structured investment credit can fail to generate returns in depressed markets with limited demand or where infrastructure is weak.

Human Capital and Mobility

Credit used for education and mobility also reflects the gains resulting from forward-looking investments. A loan program in Colombia enabled credit-constrained youth to complete tertiary education and enter better-paid formal jobs, with beneficiaries earning substantially higher starting salaries and having lower dropout rates than similar nonrecipients (Sánchez Torres and Velasco 2014). Education loans that directly pay tuition fees significantly raise women's enrollment and narrow gender gaps in tertiary education access, with particularly strong effects among lower-income women. This direct payment mechanism shows how the manner in which credit is delivered can enable forward-looking investments by ensuring credit finances its intended purpose while reducing diversion risk.

Credit for mobility, particularly vehicular, unlocks new economic and social opportunities for women by enabling them to travel outside the home (e.g., to access otherwise unreachable formal labor markets). While these investments may not produce immediate income gains, they can generate substantial long-term benefits through the empowerment they support.

Consumption Smoothing and Shock Response

The relationship between credit and consumption is nuanced. When households face isolated shocks — a single health emergency, a short-term flood, temporary unemployment, business income shortfall — borrowing to maintain consumption can be an appropriate coping strategy that prevents more damaging responses such as distress asset sales, reduced nutrition, or pulling children from school. For example, evidence from rural Vietnam demonstrates that during illness episodes, households with credit access maintained stable non-healthcare consumption by using borrowing to cover immediate medical expenses (Thanh and Duong 2017).

However, risks arise when credit is repeatedly used for consumption or recurrent shock-coping without an underlying pathway to income growth or asset accumulation. In Cambodia, evidence shows that many smallholders using microcredit as a coping strategy in the face of recurrent shocks experience high over-indebtedness, reduced consumption, forced asset sales, and, in some cases, loss of land (Green and Bylander 2021; Bylander et al. 2019). The key variable is not the use of funds at any single point but the pattern over time. Episodic credit can be protective for a genuine one-off shock with a plausible path back to normal income. Structural reliance on credit to bridge a persistent shortfall simply accumulates debt while the underlying problem remains unresolved. The Cambodian evidence cited is primarily observational and qualitative, documenting the mechanisms through which repeated coping credit becomes harmful rather than establishing causal estimates.

The key distinction is not between productive and consumption uses nor between occasional and repeated borrowing, but between credit that is well matched to the borrower's cash flow cycle and capacity to repay and credit that chronically substitutes for income or safety nets. Purpose-linked short-tenor credit — instalment financing, revolving working capital, emergency facilities with defined repayment paths — is not inherently problematic even when repeatedly used, provided it is sized to capacity and does not accumulate net debt. The risk profile deteriorates when credit repeatedly bridges a persistent income deficit rather than a bounded shortfall where obligations compound while the underlying problem remains unresolved. Risk is even greater for very low-income households lacking safety nets or facing high-frequency shocks.

Recovery Credit: A Distinct Challenge

Credit deployed to rebuild productive assets or working capital after a shock presents distinct design challenges. Without careful underwriting, recovery lending compounds harm when borrowers at maximum income constraint take on debt they cannot repay — the pattern that ad hoc FSP emergency lending tends to produce. Well-designed programs tell a different story. They establish underwriting criteria calibrated to the post-shock context: borrower character, track record, and opportunities in the new economic landscape rather than past income. Productive use requirements and active monitoring complete the package. Crucially, all of this is established before a shock occurs. Evidence from more than a decade of structured recovery programs shows repayment rates of 95 to 99 percent, often above pre-disaster benchmarks. For example, Haiti's largest microfinance institution, Fonkoze, achieved 97 percent repayment on 11,000 recovery loans after the country's 2010 earthquake. Similarly, recovery loan recipients in Nepal reported incomes 115 percent higher than nonrecipients following the 2015 earthquake (Zetterli et al. 2026). Reactive, ad hoc post-disaster lending is the risky response while deliberate recovery products with pre-established criteria can be both commercially viable and developmentally effective.

What Credit Finances: Conclusions

The evidence shows that what credit is used for is a powerful predictor of outcomes, however not an independent one. Productive investment credit generates the strongest and most consistent returns but only when borrower capability, loan size, and market conditions align. Human capital and mobility credit can transform long-term trajectories, yet its returns depend on functioning labor markets and delivery mechanisms that ensure funds reach their intended purpose. Resilience-building credit occupies a third position: its returns are protective and partly counterfactual, requiring evaluation frameworks that capture avoided shocks and faster recoveries alongside income gains. Consumption credit can serve a legitimate protective function when it bridges isolated shocks but becomes harmful when it chronically substitutes for income growth or social protection.

Providers, funders, and regulators should avoid treating the use of funds as a binary productive/consumption distinction. The diagnostic question is whether use builds capacity over time or primarily services immediate needs without generating returns to cover borrowing costs.

3.5 Where Credit Operates: Context

Wider contexts, such as regulation, institutional infrastructure, market characteristics, climate risks, and social norm regimes, influence whether credit expansion supports or undermines development outcomes. These factors vary by location and shape whether borrower capacity and product design can translate into positive gains.

Evidence across all outcomes examined shows that even well-structured credit may fail in weak regulatory environments or high-shock contexts, or when market conditions and social norms create barriers for borrowers. Conversely, strong enabling environments can partially compensate for weaker borrower profiles by providing safeguards, information, and complementary support. Context is a primary determinant of impact.

Regulation, Supervision, and Credit Infrastructure

Consumer protection measures — transparency requirements, responsible lending standards, effective supervision, accessible redress mechanisms — can substantially enhance credit's contribution to welfare. Evidence from Ghana shows that mandatory price disclosure reduced transaction fees by around 40 percent (Annan 2025). A multi-country review of mobile instant credit finds widespread problems of hidden prices, post-contract exploitation, and data-protection risks, concluding that robust rules and active supervision are essential as access expands (Cassara, Zapanta, and Garz 2024). Price caps are one response to predatory pricing but they carry risks. Caps set too low can exclude higher-risk borrowers and displace rather than prevent harm, making disclosure requirements and active supervision more reliably effective instruments.

At the system level, strong prudential regulation amplifies credit's positive effects on growth and stability. In Sub-Saharan Africa, credit uptake improves financial system stability primarily where supervisory enforcement is active (Damane and Ho 2024; Ofoeda, Mawutor, and Ohenebeng 2024). Where enforcement is weak, rapid expansion enables predatory lending, generates defaults and portfolio deterioration, and ultimately raises borrowing costs for households the inclusive finance policies intend to serve (Hua, Bi, and Shi 2023; Siddiki and Bala-Keffi 2024).

A responsible credit environment also requires clear licensing standards, capital and governance requirements, and, critically, gender-disaggregated loan-level data, which many supervisory authorities currently lack. Financial literacy is a further component, although standalone programs show mixed results. Effects are most consistent when integrated with product delivery.

Two infrastructure priorities serve distinct functions. First, open finance frameworks — account aggregation, interoperability, access to transaction histories — enable cash flow-based underwriting for borrowers without formal credit histories. Early evidence from Brazil and India suggest meaningful gains in access and cost (Medine and Plaitakis 2023). Second, credit information systems, including bureaus, registries, and cross-lender reporting, serve systemic stability, preventing unsustainable cross-borrowing from accumulating undetected until market-level crises emerge (McIntosh and Wydick 2005; Schicks 2010). Both types of infrastructure are necessary; neither substitutes for the other.

The findings in this subsection are directional rather than definitive. The Impact Pathfinder treats regulatory and enabling environment conditions as contextual moderators rather than independently evaluated interventions, so the evidence on which specific regulatory choices improve borrower outcomes is substantially thinner than the evidence on product design and borrower characteristics. Policymakers should hold these findings with appropriate caution.

Market Characteristics

Market characteristics operate on two distinct levels: (i) as structural infrastructure that is largely fixed and beyond the reach of individual lenders or borrowers to alter (e.g., output markets, transport links, agro-ecological conditions); and (ii) as market conditions that are more fluid and time-varying but equally capable of determining whether credit generates returns (e.g., competitive dynamics, macroeconomic shocks, climate events). Where structural market infrastructure is absent, credit frequently fails to generate gains regardless of product design. In Uganda, microcredit for women agricultural borrowers did not increase profits, yields, or commercialization. The binding constraint was absent output markets, not credit access; borrowers thus faced surplus crops they could not sell and redirected loans to consumption (Namayengo et al. 2023).

Indonesia's Kredit Usaha Rakyat (KUR) program illustrates the same dynamic at scale. Households in Kalimantan, Sulawesi, and Maluku-Papua showed significant increases in energy use and economic activity while those in parts of Java and Bali-Nusa Tenggara showed no or negative effects, reflecting differences in infrastructure, economic structures, and market access (Setyawati and Hartono 2025). As KUR is a heavily subsidized government guarantee scheme, results should be cautiously applied to commercial contexts. In Vietnam and India, returns to credit similarly varied with agro-ecology and proximity to markets — structural features individual lenders cannot alter (Luan, Bauer, and Kühl 2016; Tran, Wang, and Nguyen 2015).

More fluid and time-varying, market conditions affect credit impact through competitive dynamics and climate or macroeconomic shocks. In highly competitive markets, rapid credit expansion weakens lender screening and is associated with lower systemic stability (Feghali, Mora, and Nassif 2021).

Climate Stress and Shock Exposure

Climate shocks do not simply create demand for credit; they alter what credit does. The same instrument that reliably supports productive investment under moderate climate variability can become a source of harm under severe or repeated stress. How climate context shapes credit's impact is best understood across three phases.

First, before a shock, credit strongly amplifies adaptation. Farmers with access to formal credit in Kenya were approximately three times more likely than non-borrowers to adopt multiple climate-smart practices such as intercropping, rainwater harvesting, and terracing (Waaswa et al. 2024). FSPs that develop deliberate climate risk assessment and product adaptation approaches are commercially and developmentally better positioned than those that withdraw when conditions deteriorate. CGAP research in Pakistan found that 40 percent of MFIs reduced or stopped lending to climate-affected sectors, a reactive response that compounds client hardship while eroding the institution's own addressable market (Notta and Zetterli 2025).

Second, during a climate shock, credit can help households maintain production and avoid negative coping strategies such as distress asset sales, but as shocks become more covariate and severe, debt accumulates faster than income recovers. Contingent credit instruments, where repayment obligations are tied to a climate or price index, can protect borrowers and lender portfolios in bad years (Shee, Turvey, and You 2019), although they typically require some subsidy to remain affordable.

Third, after a shock, recovery credit deployed without adequate underwriting risks compounding harm. Structured programs with preestablished criteria and productive use requirements show repayment rates of 95 to 99 percent in the evidence (Zetterli et al. 2026).

In high-frequency shock environments, credit alone is insufficient. Grants, safety nets, community risk pooling, and risk layering with index insurance all become increasingly necessary as climate severity rises. The practical implication is that climate context warrants the same analytical attention as borrower characteristics and product design.

Social Norms

Social norms heavily impact women. Credit produces its strongest gains for women where these norms already support women's economic roles and is weakest or potentially harmful where norms are most restrictive. Evidence from rural India illustrates the stakes: women who were able to safeguard control over loan proceeds saw significant improvements in assets and income while those unable to shield funds experienced limited or negative impacts. The difference was driven not by individual characteristics but by the normative and household environment (Garikipati 2013). Product design can partially offset restrictive norms. Direct disbursement into women's own accounts, commitment devices, and group lending that expands mobility and public participation have each been shown to improve outcomes where norms would otherwise redirect the benefit (Garikipati 2013; Vaessen et al. 2014). Well-designed credit can also contribute to shifting norms over time, particularly when combined with gender-norms programming and community engagement (Ashraf, Karlan, and Yin 2010).

Gender norms are the most extensively documented social dimension, but caste, ethnicity, religion, and loan officer attitudes and implicit bias also shape who accesses credit and on what terms. The evidence on these factors is limited, reflecting a gap in the literature rather than evidence of their irrelevance.

Where Credit Operates: Conclusions

Context is not a residual but one of the primary determinants of impact, as evidenced in Uganda, where credit failed women agricultural borrowers due to absent output markets; in Cambodia, where repeated climate shocks converted productive credit into a debt trap; and in South Asia, where restrictive gender norms determined whether women could capture any benefit. PCM evidence adds a further dimension, as context does not merely shift the level of outcomes but reshapes which factors carry the greatest marginal impact for a given population. The same institution applying the same product suite can find that delinquency tolerance is the critical lever for one client segment while loan sizing is the critical lever for another. The five factors cannot be applied as a universal checklist; their relative importance must be diagnosed for each client segment and context.

A strong enabling environment can partially compensate for weaker borrower profiles — through disclosure regimes, safety nets, group-based approaches, and norms-shifting programs — shaping not only whether credit generates returns, but who captures them and who bears the risks. Where shocks are frequent and credit repeatedly used for coping rather than investment, grants, safety nets, and community risk pooling may need to precede credit.

Where complementary interventions are absent and the enabling environment cannot reasonably be improved in the near term, deploying credit to vulnerable borrowers is, by definition, not precision-led. The responsible course may be to sequence other instruments first or to invest in building enabling conditions before scaling credit access.

3.6 When Credit Helps: Time Horizon

Time horizon is one of the most important and most commonly neglected dimensions of credit impact assessment. Time operates through three distinct mechanisms. The first is accumulation of outcomes: the same credit product that appears neutral at 18 months can reveal benefit or harm at three years or six years. The second is the time profile of credit use: whether credit finances a purpose that generates returns within or beyond the loan period. Here, contrast matters. Productive investment gradually builds returns, rewarding longer horizons; repeated coping credit compounds obligations without resolving the underlying gap. The third mechanism is the borrower's multicycle trajectory: long-run welfare shaped by how borrowers progress or deteriorate across repeated cycles, not by any single loan event, a dimension invisible in cross-sectional studies. Each mechanism has distinct implications for product design, evaluation, and supervision.

The Temporal Asymmetry

Unlike insurance or social protection, credit generates a financial obligation that must be serviced regardless of whether the purpose for which it was taken succeeds. Time therefore operates asymmetrically. When credit finances productive investment, returns gradually accumulate and time works in the borrower's favor. When repeatedly used to cover a persistent income shortfall or recurrent shocks, obligations accumulate while the underlying problem remains unresolved (Duvendack et al. 2011; Duvendack and Mader 2019).

The key distinction is between episodic and structural use. The relevant question is not short vs. long tenor but whether credit is matched to a purpose that generates the capacity to repay. Credit used for a one-off shock can be protective, preventing distress responses that would otherwise deplete productive assets. The same product repeatedly used to bridge a persistent income shortfall creates a different dynamic. Repayment obligations compound while the underlying gap remains, and the borrower moves from managing liquidity to servicing debt. In rural India, households using credit for health shocks had significantly higher debt a year later and were five times more likely to have taken on additional debt (Dhanaraj 2016). Systematic reviews find that the longer clients remain in debt purely for coping, the more likely benefits diminish and eventually turn negative (Stewart et al. 2010; Duvendack and Mader 2019).

The Hyderabad microcredit experiment illustrates the positive side of the asymmetry. At 18 months and at three years, treatment effects were near zero for most households. At the six-year follow-up, households with preexisting businesses owned firms with 35 percent more assets and twice the revenues of the control group (Banerjee et al. 2019). Meta-analytical work confirms the pattern: average effects on income and consumption are modest, but gains are concentrated among borrowers at the upper end of the performance distribution and among borrowers with an existing productive base, emerging only over longer horizons (Meager 2022; Hou 2023). Apparent neutrality at 18 months should not be treated as evidence of no effect.

Borrowing Trajectories

CGAP's PCM pilots find that clients who achieved positive outcomes, particularly improved financial health or increased income, were not those who received a single optimal loan but those who progressed along structured borrowing pathways over time. Progressive loan size increases of roughly 30 to 50 percent per cycle outperform both flat loan sizes and rapid scaling across capacity segments. Larger follow-on loans can generate strong returns for higher-ability borrowers, while over-rapid scaling for others increases risk without commensurate gains (Bari et al. 2024; Bryan, Karlan, and Osman 2024).

Most MFIs fund their portfolios through short-term wholesale credit lines, limiting their capacity for longer-tenor products. The design imperative is therefore not necessarily longer individual loans but better-sequenced products: structured renewal and graduation mechanisms that support progression within standard tenors. The BRAC graduation model is one template that sequences asset transfers, savings, and livelihood support before introducing credit. A six-country evaluation found lasting income and asset gains among ultra-poor households who completed the full sequence (Banerjee, Duflo, Goldberg, et al. 2015).

When Credit Helps: Conclusions

The "when" analysis yields three operationally distinct findings.

1. On evaluation windows: The six-year Hyderabad follow-up, revealing compounding gains invisible at 18 months, illustrates a structural problem in standard evaluation practice. Apparent neutrality at short follow-up windows is not evidence of no effect but of a window too narrow to observe outcomes whose logic plays out over years.
2. On product design and borrower progression: Positive long-run outcomes are associated not with a single optimally designed loan but with structured progression across cycles: loan sizes scaling at roughly 30 to 50 percent per cycle and product features graduating with evolving borrower capacity. Since most MFIs rely on short-term wholesale funding, the design implication is better-sequenced products within standard tenors rather than longer individual loans.
3. On supervision: Point-in-time monitoring is structurally mismatched to how credit outcomes unfold. Repayment rates at any single point capture neither the accumulation of harm from repeated consumption borrowing nor the compounding of gains from productive investment. Trajectory-based indicators — debt-service ratios across cycles, repeat and rollover borrowing rates, portfolio delinquency patterns — are more analytically appropriate.

3.7 Which Factors Matter Most?

The evidence reviewed in this chapter identifies a considerable number of influencing factors (e.g., borrower economic status, market conditions, regulation, social norms) that bear on credit outcomes. However, no published studies systematically compare their relative strength across contexts. A precise ranking is not currently possible.

Although current evidence does not allow a precise ranking, the frequency with which factors appear across studies provides a rough indication of relative influence. By that measure, Annex C shows that product design and delivery features are most frequently cited, closely followed by borrower and community socioeconomic characteristics, reflecting where evaluation designs have concentrated attention.

Product design and delivery is meaningful precisely because it is adjustable in the relatively short term. A caveat applies to financial capability where evidence on financial education programs is mixed, as effects are typically modest, short-lived, and difficult to sustain at scale. The more robust pathway is indirect: designing simpler, more transparent products that reduce the capability demands placed on borrowers rather than attempting to raise capability to meet complex product requirements. Ecosystem factors (e.g., economic conditions, climate exposure, gender norms, regulatory frameworks) are equally important but evolve slowly and require coordinated action across multiple stakeholders.

These findings should be interpreted with caution. The Impact Pathfinder can only analyze factors identified in published studies, and frequency in the literature reflects research attention as much as true influence. This caveat applies with particular force to regulatory and enabling environment factors, which the Pathfinder codes as contextual moderators rather than primary influencing variables. Annex C also provides a directional indication, not a definitive ranking.

No single actor controls all five factors. FSPs have the most leverage over product design and delivery; funders and impact investors over incentives and analytics; and governments and regulators over the regulatory frameworks and market infrastructure that make precision lending possible at scale. Civil society and development organizations have the most leverage over social norms through gender and community programming.

3.8 A Diagnostic Framework

Figure 1 translates the five-factor framework into a set of structured diagnostic questions. For each factor, it poses the questions a practitioner would need to answer to assess whether the conditions associated with positive outcomes are present for a given borrower and credit situation 

The five sets of questions converge on a single operational test at the center of the figure: does this credit match the conditions that generate benefit rather than harm?

[Figure 1: Five-Factor Diagnostic Framework — see original publication for the full figure. The framework poses structured diagnostic questions for each factor: WHO (borrower characteristics and control over loan proceeds), HOW (product design, repayment flexibility, disbursement channel, bundled support), WHAT (use of funds and productive vs. coping function), WHERE (regulatory environment, market conditions, climate exposure, social norms), and WHEN (time horizon and evaluation window). These five sets of questions converge on a central operational test: does this credit match the conditions that generate benefit rather than harm?]

The diagnostic questions in Figure 1 have practical implications. Answering them well and acting on the answers requires different inputs from different actors. FSPs control product design and underwriting. Funders and investors set incentives and performance expectations. Policymakers and regulators shape the enabling conditions that determine whether precision lending is even possible at scale. Chapter 4 takes up what each of these stakeholder groups can do — and how far different institutions can realistically go given their current data infrastructure and institutional capacity.

Chapter 4: Making It Work: From Evidence to Precision-led Impact

Whether a loan improves livelihoods or deepens financial distress depends on who receives it, how it is structured, what it finances, where it is deployed, and when it is deployed relative to borrowers' capacity and context. The evidence in this Focus Note consistently points toward what might be called precision credit: the deliberate matching of product design, complementary support, and evaluation approach to the specific circumstances of the borrower and the market. Access barriers remain significant in many markets and this note does not treat the access problem as resolved. But expanding access without attending to these conditions is as likely to cause harm as to generate impact. This chapter translates that evidence into recommendations for the stakeholders best placed to act on it. FSPs have the most direct control over HOW credit is designed and delivered — the factor most frequently cited in the Pathfinder evidence — and over the way they assess WHO is likely to benefit from a given product. Funders and impact investors shape the incentives and evaluation horizons that determine whether providers exercise that agency in borrower-beneficial ways. Policymakers and regulators are primarily responsible for the WHERE conditions under which lending occurs, in particular, the regulatory frameworks, market infrastructure, and data systems that determine whether precision lending is even possible at scale.

A Note on the Evidentiary Basis of the Recommendations

The Impact Pathfinder synthesizes studies of the impact of financial services on clients. Regulatory and enabling environment factors appear in those studies as contextual moderators, not as independently evaluated interventions. The Pathfinder therefore provides strong evidence on what FSPs and funders can do to improve outcomes and considerably thinner evidence on which specific regulatory choices produce better borrower outcomes. The recommendations for policymakers and regulators consequently rest on a more mixed evidentiary foundation and should be read with that calibration in mind.

4.1 Recommendations by Stakeholder Group

The recommendations that follow are organized by stakeholder group. They are not of equal weight or feasibility. Some can be acted on within existing institutional capacity by adjusting repayment schedules, improving pricing disclosure, or incorporating trajectory monitoring into existing data systems. Others require meaningful investment in new capabilities, whether in data infrastructure, analytical tools, or complementary service provision. A few depend on coordinated action across institutions and cannot be meaningfully advanced by any single actor. Some recommendations codify principles already familiar to well-run institutions. The contribution of the evidence is in specifying how recommendations should be applied, for which borrower segments, and with what calibration. Readers should treat the list as a menu to be prioritized against current capacity rather than a checklist to be adopted wholesale.

Financial Services Providers

Several of the actions below reflect practices that many FSPs already apply in some form (e.g., flexible repayment, screening for absorptive capacity, appropriate bundling). The evidence's contribution is in the specificity of application. Which actions matter most depends on borrower segment, product type, and operating context. A substantial gap exists between institutions that acknowledge these principles and those that consistently apply them at the right level of granularity. FSPs can:

• Screen for absorptive capacity, not just repayment eligibility. Conventional credit scoring measures whether income or assets exist to service a loan, not whether a viable productive opportunity exists that will generate returns above the cost of borrowing. The WHO analysis in Chapter 3 shows prior business experience to be one of the strongest predictors of positive outcomes while inexperience and over-optimism are consistently associated with higher risk. FSPs can use business history, income regularity, transaction history, existing debt exposure, and household composition alongside conventional credit information (Fernandez Vidal and Caire 2024). They can explicitly assess whether clients, particularly women, retain genuine control over loan proceeds, as the WHO evidence shows that welfare gains disappear while repayment liability remains when loans are nominally taken out by women but appropriated by male household members.
• Design products around borrower cash flows and capabilities. The HOW analysis identifies repayment flexibility, appropriate loan sizing, disbursement mechanism, and complementary services as the most actionable design variables. FSPs can align repayment schedules with income patterns, using grace periods for investment loans and harvest-linked loan structures for smallholders. They can right-size loan amounts through graduated approaches (PCM evidence shows progressive scaling of around 30 to 50 percent per cycle outperforms both flat loan sizes and rapid jumps). Digitized disbursement into borrower-controlled accounts, particularly for women, also protects control over loan proceeds. FSPs can calibrate flexibility to borrower experience, as structured contracts protect first-time or lower-capability borrowers while more experienced borrowers benefit from adaptable terms.
• Strategically bundle and ensure transparency. The HOW evidence shows that credit alone is often insufficient, with business training, insurance, and gender norms programming each addressing capability gaps and social constraints that credit cannot overcome on its own. The appropriate complement depends on the binding constraint for the relevant borrower segment: training for first-time entrepreneurs, insurance for climate-exposed borrowers, plus gender norms and privacy-enhancing features in restrictive contexts. Transparency is a precondition; the evidence shows that how loan terms are framed shapes borrowing decisions. FSPs should ensure clear disclosure of effective interest rates and all fees. Bundling also carries conduct risks, and explicit consumer consent and regulatory safeguards against forced product acquisition are essential accompaniments to any paired services strategy.
• Graduate based on demonstrated capacity and build institutional pathways. Rapid graduation without adequate screening is a consistent risk factor. PCM piloting shows clients regularly outgrowing products without clear transition pathways, which leads to stagnation or exit. FSPs can build structured progression pathways with loan sizes, product features, and complementary support calibrated to the client's evolving capacity. Most financial institutions rely on short-term wholesale funding, limiting capacity for longer-tenor products. The answer is smarter sequencing, with successive loan cycles that together achieve what a single longer loan would, incrementally building borrower capacity within tenors the institution can fund.
• Make the internal commercial case for precision investment. Building the data systems, screening capacity, and analytical infrastructure that precision lending requires involves real upfront cost. The commercial logic runs in its favor, as product configurations associated with better client trajectories consistently coincide with lower default rates and higher retention. The costs of imprecision in portfolio write-offs, regulatory intervention, and reputational damage also repeatedly exceed the costs of better screening. FSPs should internally make this case on commercial grounds, not only on impact grounds, and accordingly plan the investment horizon.
• Measure financial health and borrower trajectories, not just repayment. High repayment rates can coexist with deteriorating household welfare as a borrower may continue to pay on time while their debt-service ratio rises, liquid buffers deplete, and reliance on rollover borrowing deepens. FSPs should track metrics that reveal trajectory, such as debt-service ratios, repeat borrowing rates, rollover frequency, and indicators of growing liquidity strain. PCM evidence shows that progression velocity — the rate at which borrowers move through loan cycles with improving trajectories — is a more meaningful indicator of long-term welfare than repayment rates alone.
• Build institutional capacity to maintain lending through climate shocks. The standard FSP response to climate shocks — pausing lending, intensifying collections, withdrawing from climate-affected areas — can compound client harm and portfolio deterioration where recovery protocols are not already in place. FSPs should instead build deliberate climate risk management, including physical climate risk assessments of the lending portfolio; contingent funding facilities triggerable by climate events; and recovery lending programs with pre-established underwriting criteria, productive use requirements, and deployment protocols — all designed before a shock occurs rather than in response to one.

Funders and Impact Investors

Several of the actions below can be incorporated into existing portfolio assessment and reporting processes with modest adaptation. A number of leading impact investors already apply elements of them. Others require investment in evaluation infrastructure or longer-term capital commitments. Funders and impact investors can:

• Fund precision, not volume. Portfolio growth is a weak proxy for impact. The five-factor framework shows that the same product generates strongly positive outcomes for some borrowers and mixed outcomes for others, depending on borrower characteristics, product calibration, and context. Funders and impact investors can prioritize providers that demonstrate strong client screening, appropriate product design, and credible outcome measurement beyond outreach metrics. They can treat repayment rates as necessary but insufficient evidence, ask what the institution knows about which clients benefit, and ask how it uses that knowledge.
• Recognize that precision and sustainability are aligned, and help FSPs make the investment. Precision in credit deployment is not in tension with institutional sustainability; it is a condition for it. The market failures documented in several places in the note were commercial as well as developmental. Episodes of over-indebtedness in markets such as Cambodia and Bosnia each triggered portfolio deterioration, regulatory tightening, and investor withdrawal. FSPs face a real short-run barrier. The cost of building data and analytics capacity falls before returns arrive, and many institutions, particularly smaller ones, lack the balance sheet to bridge that gap unaided. Funders are well-placed to help bridge the gap by providing patient capital and technical assistance for precision infrastructure, and by not penalizing the investment period with return expectations calibrated to a volume growth model.
• Assess context before financing expansion. The WHERE analysis shows context to be one of the primary determinants of impact. Before supporting scale, funders should assess the enabling environment. Where enabling conditions are weak, the appropriate response depends on the funder's mandate. DFIs and donors with technical assistance capacity should consider directing resources toward strengthening enabling conditions through regulatory support, credit information infrastructure, or consumer protection frameworks — before or alongside portfolio investment. Commercial impact investors should, at minimum, price enabling environment risk into their underwriting and set explicit thresholds below which expansion is not supported.
• Support complementary services as integral to the financing proposition. The HOW and WHAT findings show that credit alone is often insufficient, and this is predictable from borrower characteristics and context. Business training, insurance, gender norms programming, and advisory support are not peripheral enhancements. Funders and impact investors can finance bundled approaches where borrower characteristics and context indicate credit alone is unlikely to generate positive outcomes, and treat the cost of complementary services as part of the cost of responsible deployment.
• Clarify use cases before specifying metrics. Investment impact measurement and management (IMM) design often begins with indicator selection rather than with a clear articulation of the decision it is meant to inform. Investors and fund managers should establish the specific purpose that outcome evidence will serve: informing portfolio construction, triggering a follow-on commitment, managing impact risk, or enabling learning and product iteration. Shared clarity on use cases — between DFIs, fund managers, and FSPs — is a precondition for right-fit IMM and for avoiding measurement that is either too burdensome to sustain or too shallow to inform decisions (Nierakkal, Lahaye, and Clarke 2025).
• Align evaluation horizons with the productive logic of the credit being financed. The WHEN analysis shows that evaluation windows of 18 months to three years are often structurally too short to detect compounding gains from productive investment credit. The practical response is to improve what shorter windows can reveal: borrowing trajectory indicators, such as debt-service ratios across successive cycles, repeat and rollover borrowing rates, and loan size progression, can signal whether a portfolio is building or eroding borrower capacity without requiring multi-year outcome studies.
• Require distributional reporting. Average effects conceal the heterogeneity that Chapter 3 identifies as the central analytical finding of the accumulated evidence. Funders and impact investors should require reporting disaggregated by borrower gender, loan size band, and where available, prior business experience and household income tier. They should monitor early warning indicators of deteriorating trajectories, including rollover rates, repeat borrowing frequency, loan size trends within a cohort, and rising delinquency.

Policymakers and Regulators

The recommendations below draw on consumer protection evidence (strong in the Pathfinder), documented consequences of rapid expansion without supervisory capacity, and logical inference from product-level findings alongside CGAP's research. They should be applied with appropriate recognition that the evidence base on which specific regulatory choices generate better borrower outcomes is thinner than the evidence on product design and borrower characteristics. Policymakers and regulators can:

• Build the enabling infrastructure for precision lending and credit stability. The most important contribution policymakers can make is to build infrastructure that makes better inclusive finance possible. Three priorities serve distinct functions. First, fully interoperable instant payment systems (IPS), with open participation rules, third-party initiation rights, and real-time settlement, generate the transaction data trails that enable cash flow-based precision lending (Cook, Lennox, and Sbeih 2021). Second, open finance and data-sharing frameworks extend the IPS data layer to credit underwriting, enabling cash flow-based lending over collateral or bureau history (Jeník, Mazer, and Fernandez Vidal 2024). These frameworks must be built with active oversight from the outset (Dias, Mazer, and Jeník 2026). Third, credit information systems (bureaus, registries, reporting mandates) remain essential for systemic stability, monitoring aggregate exposure, detecting multiple borrowing, and preventing market-wide over-indebtedness. Cutting across all three, the most consequential regulatory reform is creating licensing avenues and proportional frameworks for non-bank financial institutions, including fintechs.
• Enforce transparency and consumer protection as a foundation. Policymakers and regulators should require clear disclosure of effective interest rates and all associated fees, ensure accessible grievance and dispute resolution mechanisms, and enforce penalties for abusive practices. Effective consumer protection requires a broader ecosystem approach that combines transparent pricing with product governance requirements, suitability standards, and enforcement capacity, specifically, the framework set out in the G20/OECD High-Level Principles on Financial Consumer Protection (OECD 2022) and elaborated upon for digital credit contexts by Izaguirre et al. (2025).
• Actively monitor market conditions and product risk and intervene before harm accumulates. The WHEN and WHERE analyses show that the costs of repeated consumption credit and the benefits of productive investment credit accumulate gradually, dynamics missed by cross-sectional snapshots. Policymakers and regulators should track credit penetration and borrowing intensity at the market level, rollover rates and refinancing patterns at the product level, and debt-service ratio trends at the borrower segment level (Kumaraswamy and Beer Kremnitzer 2025).
• Establish responsible lending standards that frame vulnerability as a product design challenge. Responsible lending standards should support a more demanding standard: assessing not only whether income exists to service a loan but whether the product design is appropriate to the borrower's circumstances and risk exposure. When it is appropriate, assess what adapted design (e.g., flexible repayment, smaller initial loan sizes, bundled risk mitigation) is needed for responsible provision (Zetterli 2023).
• Integrate financial health measurement into supervisory frameworks. Current supervisory frameworks are organized around financial stability and portfolio quality, which is necessary but insufficient for assessing whether credit is generating positive outcomes for borrowers. As financial health measurement frameworks develop, policymakers and regulators should incorporate outcome indicators into monitoring requirements, including measurement of over-indebtedness across formal and informal debt sources.
• Publish system-level disaggregated data to enable segmented analysis. The heterogeneity finding that runs through Chapter 3 cannot be acted on without the data to see it. Policymakers and regulators should require and publish system-level data, disaggregated at minimum by borrower gender, loan size band, collateral type, sector, and lender type.
• Coordinate credit policy with social protection and establish debt relief mechanisms. In climate- and shock-prone contexts, credit is not a substitute for safety nets. The poorest households often require a sequenced package of support before credit can play a productive role. Policymakers should also create regulatory space for structured recovery lending. Evidence shows that well-designed post-shock programs outperform reactive emergency disbursement on both repayment and welfare outcomes (Zetterli et al. 2026). In many LMICs, borrowers facing over-indebtedness have no formal exit short of asset loss. Accessible debt relief mechanisms are a necessary complement to prevention-oriented regulation.

4.2 Evidence Gaps and Priorities

Borrowers in low-income contexts rarely engage with a single credit relationship — yet debt juggling across formal credit, informal borrowing, rotating savings groups, and in-kind obligations is largely unobserved in impact studies.

The evidence base synthesized in this Focus Note reveals significant gaps in understanding the heterogeneity of credit's effects — the conditions under which the same product generates gains for some borrowers and harm for others. This is the nexus addressed by the five-factor framework and is where the gaps are most consequential for practice.

The most pressing methodological gap is the near invisibility of the borrower's full financial life in standard credit evaluations. Rigorous studies observe the formal loan under study, but borrowers in low-income contexts rarely engage with a single credit relationship. Debt juggling — balancing formal credit against informal borrowing, rotating savings groups, in-kind obligations, and informal credit from employers, landlords, and moneylenders — all shape whether a formal loan generates benefit or harm yet are largely unobserved in impact studies. Financial diary research and longitudinal ethnographic work documents informal debt constituting the majority of household credit and debt-service ratios that would be alarming if standard metrics captured them.

Understanding heterogeneity requires understanding the full portfolio. CGAP's Financial Inclusion 2.0 initiative is actively developing financial health as a holistic measurement construct, capturing day-to-day financial management, resilience, and opportunity-taking as dimensions of financial wellbeing (CGAP 2026b). Applied to credit evaluation, the approach could reveal the cumulative borrower experience that single-product impact studies miss (Nielsen, Sirtaine, and Spaven 2026).

A second gap concerns how findings on the five factors translate across credit types. Most of the evidence base comes from microcredit. Whether the framework's findings transfer to digital credit, PayGo financing, education lending, or supply chain finance — and where they need to be adapted — remains largely untested. This is particularly consequential as digital credit has rapidly expanded into contexts where supervisory capacity is limited and product features substantially differ from traditional microcredit.

A third gap: most evaluations follow borrowers for 18 months to three years, systematically underestimating compounding gains from productive investment credit and missing the slow accumulation of harm from repeated consumption borrowing. Longer follow-up and trajectory-based measurement that tracks successive borrowing cycles rather than single loan events would substantially sharpen the evidence on heterogeneity over time.

There are significant evidence gaps beyond this heterogeneity nexus. Three further gaps are well-acknowledged but lie outside the scope of this note. The enabling environment evidence base is almost entirely absent, particularly the causal effects of specific regulatory interventions on borrowers. Natural experiments around policy changes represent an underexploited opportunity. Community-level spillovers, intergenerational effects, and market-level dynamics from credit expansion are similarly thin. And global household debt measurement is constrained by the near-total invisibility of informal and multi-source debt in official statistics.

4.3 Credit as a Lever, Not a Ladder

Credit appears to function more as a lever than a ladder. It amplifies existing capacity among borrowers who already have viable opportunities and capabilities, but on its own does relatively little to create capacity where it is absent. Credit can accelerate progress where households have business experience, stable or predictive cash flows, and genuine investment opportunities. Where these foundations are missing, credit alone may add financial burden without benefits.

However, this framing should not be seen as fixed. When credit is paired with complementary support (e.g., training, insurance, gender-sensitive design, appropriate product structure), it can help build capability over time, expanding the group of borrowers for whom it functions as an effective lever.

The sector has long recognized that access alone is insufficient. What has been slower to follow is the analytical infrastructure to act on that recognition at the level of individual borrowers and products.

The evidence now helps clarify when credit functions as a lever and when it risks becoming a liability. The five factors highlighted throughout — who borrows, how the credit is designed and delivered, what it finances (is used for), where (in what context) and when (over what time horizon) — provide a structured diagnostic basis for more evidence-informed deployment decisions. The sector has long recognized that access alone is insufficient, a fact made clear by the over-indebtedness crises of the 2010s. What has been slower to follow is the analytical infrastructure to act on that recognition at the level of individual borrowers and products. An evidence-informed precision approach now makes that possible.

Doing so requires several important transitions: from volume to outcomes, from access as an end in itself to precision in how credit reaches and serves those who can benefit from it, from average effects to understanding heterogeneity, and from credit in isolation to credit embedded in protective and opportunity-building ecosystems with complementary services. This is not a case against expanding access. Significant barriers remain in many markets, from documentation requirements and infrastructure gaps to regulatory constraints that entirely exclude large populations from formal credit. It is a case for ensuring that where access is extended, it is extended with attention to the conditions that determine whether it generates benefit or harm. The evidence shows what works for whom and under what circumstances. The remaining challenge is not only analytical but operational: whether providers, funders, regulators, and policymakers are willing and able to translate the evidence into how credit is designed, delivered, and governed.