When the Loan Meant to Liberate Became the Thing to Escape

Few ideas in development arrived with more moral authority than microfinance. The pitch was almost irresistible: tiny loans, mostly to women, in places where formal banks would not open a branch and the only alternative was a moneylender charging whatever the desperation of the moment would bear. Give a woman a few thousand rupees to buy a buffalo or a sewing machine, and you were not handing out charity — you were unlocking enterprise that had been there all along, waiting for capital. Some of that genuinely worked, and any honest account has to begin there. But by 2010 India was living through a full-blown microfinance crisis, and the industry that had promised liberation from debt had, in one state at least, become the debt people were trying to escape.

What actually worked, and why it was believable

Start with the part that was real, because the failure is only interesting if you take the success seriously. Before microfinance reached rural India at scale, a poor household with no collateral and no credit history had essentially two options for borrowing: relatives, or a local lender whose rates could run to several percent a month and whose collection methods were not regulated by anyone. Against that baseline, a structured loan at a published rate, repaid in small weekly instalments, with a receipt, was a genuine improvement. It was predictable. It was priced. You could plan around it.

And the lending went to women, deliberately. That mattered more than the accountants expected. Money that arrives in a woman's name, is repaid in her name, and is discussed weekly in a group of other women changes something about who gets to decide things in a household. Careful studies have found effects on exactly this — women's say in household decisions, the freedom to choose what work to do — even where they found little else. The mechanism was not mystical. It was that a poor woman with access to a lump sum is a different person in a negotiation than a poor woman without one.

So the movement had real evidence behind it, real institutional innovation, and a founding story good enough to win a Nobel Peace Prize. That combination is precisely what makes what followed worth studying. The problem was never that the idea was fake. The problem was what happened when a modest, useful tool was scaled by people who needed it to be a revolution — and then by people who needed it to be a growth stock.

The mechanism nobody explains carefully enough

The engineering trick at the heart of classic microfinance is group liability. You do not lend to an individual; you lend to a small group of women who know each other, and the group is collectively answerable for the repayments. No one pledges land, because no one has land. What is pledged instead is standing — your reputation among neighbours you will see at the well tomorrow, at a wedding next month, for the rest of your life.

This is an elegant solution to a hard problem. The group knows things a loan officer never will: who is reliable, whose husband drinks, whose shop is actually making money. It screens borrowers better than a credit bureau could and it monitors them continuously and for free. When a woman is one week short, the other four lean on her, and usually that is enough, and usually that is fine.

But look at what the collateral actually is. It is social pressure. And social pressure is not a fixed quantity that stops at the right place — it is a force that scales with how badly the group needs the repayment to happen. While loans are prudently sized against what a household can actually repay, the pressure stays inside the range we would call community accountability: a reminder, a bit of shame, a neighbour covering for you and being covered next month. Push the loan sizes past repayment capacity, and the same mechanism keeps working — that is the trap. It does not fail. It escalates. The neighbours now need you to pay because they will be cut off if you do not, and the loan officer is standing in the lane with a target to hit, and the pressure that used to be a nudge becomes a crowd outside your door at dusk. The system did not break. It did exactly what it was designed to do, with a bigger number plugged in.

Group liability does not collateralise your land. It collateralises your standing among people you cannot avoid — and that kind of collateral has no upper limit on what it can be made to extract.

How prudence gave way to volume

Through the 2000s, Indian microfinance went from an NGO activity to a commercial one. Lending became a business funded by equity investors and wholesale bank borrowing, with the culmination being a large microlender's stock market listing in 2010. None of that is automatically corrupting — capital markets are how you fund lending at scale, and there was no other way to reach tens of millions of borrowers. But it changed what the institutions were optimising for, and the change ran all the way down to the field officer.

Once growth is the metric, the loan officer's job quietly inverts. She is no longer primarily assessing whether this household can carry this debt; she is disbursing against a target, with her performance measured by portfolio growth and by a repayment rate that looks spotless as long as new money keeps arriving. Meanwhile, several lenders were working the same villages, and there was no shared credit view — no functioning bureau that let one lender see what a borrower already owed to the other four. Each institution could honestly believe it had extended one reasonable loan. The household was carrying five.

In Andhra Pradesh this stacking was layered on top of an enormous state-backed self-help group network, so the same women were reachable by multiple channels at once. Estimates of household indebtedness in the state ran far above the national average. And a borrower who cannot repay loan one by earning has an obvious option available: repay loan one with loan two. That is not enterprise finance any more. It is a rollover, and it is the same mechanic that turns a consumer loan into the EMI debt trap — the instalment is met, on time, every time, by borrowing the instalment.

Andhra Pradesh, 2010

What happened next is genuinely painful, and it needs to be described carefully rather than dramatically. Through 2010, reports accumulated from districts across Andhra Pradesh of aggressive and coercive recovery practices — collection agents camping at homes, public shaming in front of the group, pressure applied to families of borrowers who had fallen behind. Alongside these came reports of suicides among indebted borrowers. The state government compiled cases it associated with microfinance debt; press tallies at the time ran higher than the official compilations. An investigation commissioned by one of the largest lenders was later reported to have found staff conduct implicated in some deaths, while the company publicly rejected responsibility for suicides in the state.

I want to be careful here, because this is where the debate usually goes wrong in both directions. These were reported and investigated cases with contested attribution, not a settled body count. Suicide has many causes, rural distress in that region long predates microfinance, and treating every death as a straightforward arithmetic consequence of a loan is neither true nor fair. Equally, the reverse move — pointing at the attribution problem to suggest nothing much happened — is worse. Something real was being reported, from many places, at once, about how money was being collected from people who could not pay.

The state's response was immediate and blunt. In October 2010 Andhra Pradesh promulgated an ordinance regulating microfinance lenders: registration with the government, disclosure of areas and rates, restrictions on where and how recovery could be conducted, monthly rather than weekly collections. The practical effect was that repayment in the state collapsed almost overnight. Borrowers stopped paying, in some part because officials signalled they need not, and portfolios that had reported repayment rates above ninety-nine percent went to near zero. The national sector contracted sharply — total lending and borrower numbers both fell hard over the following year — and lenders with heavy Andhra exposure had to be restructured.

The rethink, and what it fixed

The Reserve Bank of India convened a committee, which reported in early 2011, and out of it came a dedicated regulatory category — the NBFC-MFI — with rules aimed squarely at what had gone wrong. Borrower income thresholds and loan-size limits. A cap on how many lenders could lend to the same borrower and on total indebtedness. Interest rate and margin caps, since loosened and replaced in the RBI's later harmonised framework. Repayment-frequency choice for the borrower. Credit bureau reporting made mandatory, which was arguably the single most important item on the list, because it gave the industry the shared view of household debt whose absence had made the stacking possible.

That framework worked in the narrow sense that Indian microfinance survived, grew again, and has not repeated 2010 at that scale. But notice what the fix actually was. Regulators did not decide the mechanism was wrong. They put a speed limit on it. The instrument that turns social relationships into collateral remains in service; it is now bounded by exposure limits and a bureau feed. This is the ordinary pattern by which how technology gets captured plays out — the tool is not repudiated, it is regulated just enough to keep running, and the commercial logic that bent it in the first place is left intact underneath.

What the evidence actually supports

The research since is unusually clear, and it is more deflating than either camp wanted. Randomised evaluations run independently across several countries — in India, Bosnia, Ethiopia, Mexico, Mongolia and Morocco — converged on a consistent finding: expanded access to microcredit produces modestly positive but non-transformative effects. Households borrow, some invest in existing businesses, many use the money to smooth consumption through bad months, and the evidence for substantial reductions in poverty or large improvements in living standards is thin to absent.

Read that honestly and it is not a debunking. Consumption smoothing is valuable. Being able to absorb a medical bill or a failed harvest without selling your only productive asset is a real welfare gain, and if we had marketed microcredit as insurance-by-other-means it would have been an unambiguous success. What failed was the framing: a useful financial tool sold as a strategy for ending poverty. Once that was the claim, the sector needed a growth curve to match it, and the growth curve is what broke it.

Over-indebtedness looks exactly like a superb repayment rate. The portfolio is never healthier than in the last quarter before it dies.

The number that lies

Here is the part I keep coming back to, because it generalises far beyond microfinance. Every dashboard in that industry showed repayment rates above ninety-nine percent right up to the collapse. Those numbers were not falsified. They were true and they were meaningless, because a repayment made from a new loan and a repayment made from earnings look identical in the ledger, and a repayment extracted by a crowd outside a door looks identical to one made willingly. The metric measured whether money arrived. It could not see where the money came from or what it cost the person who produced it.

So the very indicator that the sector cited as proof of its borrowers' dignity and creditworthiness was the indicator most contaminated by their distress. The worse things got, the better the number looked — until the state changed one variable, borrowers were told they could stop, and the whole apparent asset turned out to have been made of pressure rather than income.

That is the lesson I would take to any lending system built on the poor, including today's app-based instant-credit business. Ask what the repayment rate is made of. If it is made of earnings, it is an asset. If it is made of refinancing and social coercion, it is a countdown. And if the answer is unknowable because no one is measuring household capacity to repay, assume the worst, because nothing in the incentive structure will surface it in time.

There are lending models that hold up better under this test, and they tend to share a feature: the lender and the borrower are not on opposite sides of the transaction. The Kerala cooperative model is imperfect and has had its own governance failures, but a credit society whose members are its owners has a structurally harder time treating over-lending as growth. And at the other end, when debt has already outrun any possible repayment, the honest instrument is not a restructuring but a write-off — which is why the ancient practice behind the debt jubilee is less quaint than it sounds and was invented precisely because rulers discovered that unpayable household debt, left standing, eventually destroys the society carrying it.

Microfinance deserved better than the mythology it was given, and its borrowers deserved better than an industry that mistook their desperation for demand. The small loan was a decent tool. It was never a theory of how poverty ends, and the years it was sold as one were paid for by the people who could least afford the difference.