Microcredit Was Supposed to End Poverty. Six Randomized Trials Across Four Continents Found It Doesn't.

The Idea That Won a Nobel Prize

When Muhammad Yunus and his Grameen Bank received the Nobel Peace Prize in 2006 for "efforts to create economic and social development from below," the award ratified three decades of development orthodoxy: that tiny loans to the very poor could unlock entrepreneurship, raise incomes, and shatter the cycle of poverty that trapped billions of people across the Global South. By the time Yunus stood in Oslo, the microfinance sector was disbursing roughly $25 billion per year to more than 100 million borrowers, and governments, NGOs, and private investors all treated credit access as one of the most effective tools for lifting families out of destitution.

The evidence for that belief was almost entirely anecdotal — compelling stories, yes, but randomized evidence, no.

Six Countries, One Question

Between 2003 and 2012, six independent teams of economists ran randomized controlled trials across four continents to answer what seemed like a settled question. Abhijit Banerjee and Esther Duflo randomized access to Spandana microloans across 52 slums in Hyderabad, India; Bruno Crépon tracked rural Moroccan households randomly offered credit by Al Amana; Alessandro Tarozzi ran a trial in rural Ethiopia; Britta Augsburg evaluated microcredit in Bosnia-Herzegovina; Orazio Attanasio tested group lending through XacBank in Mongolia; and Manuela Angelucci partnered with Compartamos Banco, Mexico's largest for-profit microlender.

The designs varied deliberately: nonprofit and for-profit lenders, loan sizes from $125 in Ethiopia to $1,500 in Mexico, interest rates spanning 12% to 110% APR, borrowers ranging from urban slum residents to rural farmers, and both group and individual lending models. If microcredit worked in principle, at least some of these trials should have found transformative effects.

In 2015, the American Economic Journal: Applied Economics published all six studies together in a single special issue, with Banerjee, Karlan, and Zinman writing the synthesis. The patterns across six independent datasets were remarkably consistent.

What the Trials Actually Found

Take-up was modest everywhere, with only 13% to 31% of eligible households actually borrowing when new microcredit branches opened, which itself challenges the narrative, because if credit access were truly a binding constraint on entrepreneurship, demand should have been overwhelming. Business activity increased slightly: borrowers were somewhat more likely to start or expand a small business, and investment rose in specific categories like durable goods in Hyderabad and livestock in Morocco.

But the outcomes that mattered most for the poverty-reduction thesis showed nothing across any of the six trials. Household consumption did not rise significantly, income did not increase, education outcomes for children did not improve, health indicators did not budge, and women's empowerment — measured through decision-making power, labor force participation, and household bargaining — showed no consistent gains anywhere on any continent.

Banerjee, Karlan, and Zinman put it plainly in their synthesis: the effects were "modestly positive, but not transformative." Six research teams working independently across four continents with different lenders, different populations, and different methodologies converged on the same dispiriting answer.

A Bayesian Confirmation

Individual trials can always be dismissed as underpowered or context-specific, so Rachael Meager, then at MIT, addressed this head-on in a 2019 paper by applying Bayesian hierarchical models to pool data from seven microcredit RCTs and estimate both the average effect and the genuine heterogeneity between sites. Her pooled estimate for the impact on household business profits was centered near zero, with a 95% posterior interval spanning from slightly negative to modestly positive; the impact on consumption was similarly negligible, and roughly 60% of observed variation across studies turned out to be sampling noise rather than true differences in effectiveness.

One finding stood out: at the very top of the profit distribution, above the 75th percentile, households with prior business experience did see gains. Microcredit helps some already-entrepreneurial borrowers expand, but it does not create entrepreneurs from scratch, and it does not reach the poorest borrowers at all.

Why Did Everyone Believe It Worked?

Three mechanisms explain the gap between perception and evidence. First, selection bias: early microfinance research compared borrowers to non-borrowers, which is roughly equivalent to comparing gym members to non-members and concluding that gym memberships cause fitness, because borrowers who repay and thrive remain visible while those who default drop quietly out of the sample. Second, the anecdotes were genuinely powerful (a woman in Bangladesh takes a $50 loan, buys a sewing machine, starts a tailoring business, sends her daughter to school), but for every borrower featured in a Grameen Bank annual report, dozens experienced no measurable change. Third, institutional momentum: by the time the RCTs published, microfinance had become a $100-billion-per-year global industry, and questioning it meant questioning livelihoods, mandates, and a development paradigm that had attracted billions in philanthropic capital and a Nobel Peace Prize.

The Strongest Counterargument

Jonathan Morduch at NYU and Timothy Ogden have argued that the six RCTs measure the wrong outcome entirely: that microcredit's primary value is not income growth but financial management. Poor households face extreme income volatility, where a loan due on Thursday when the harvest comes on Monday can mean the difference between feeding children and not, and credit smooths consumption over time by providing a buffer against shocks. Financial diaries research from Portfolios of the Poor confirms that credit access genuinely reshapes how the very poor manage day-to-day cash flows.

This objection has real force, but it concedes the central point: microcredit is a financial management tool for people who remain poor, which is a useful service, but not the claim on which a Nobel Peace Prize and $100 billion per year in lending were predicated.

What We Didn't Prove

All six RCTs measured outcomes over 18-month to 3-year horizons, and poverty reduction might require a decade of continuous credit access before benefits compound, though no trial has run long enough to test this. Take-up rates of 13–31% mean the trials measured the intent-to-treat effect of offering microcredit to communities, not the effect on actual borrowers; complier-average effects were sometimes larger, though still not transformative. The trials tested microcredit specifically, while microsavings, microinsurance, and mobile money are distinct products with potentially different effects. Geographic coverage excluded China, most of sub-Saharan Africa, and South Asia outside India.

What You Can Do

If you donate to international development, consider redirecting funds from microlending programs to evidence-backed alternatives: GiveDirectly's unconditional cash transfers show consistent effects on consumption, assets, and well-being across multiple independent RCTs, while Graduation programs that bundle asset transfers with skills training and coaching have produced sustained income gains in a six-country trial published in Science. Before supporting any poverty intervention at scale, check the Innovations for Poverty Action evidence base. If you work in microfinance, reframe your value proposition: credit access as financial management, not poverty escape. The microfinance story remains a cautionary tale about what happens when compelling anecdotes substitute for rigorous measurement across an entire industry for decades.