Where Medicine Is Poison: Anatomy of Recurring Crises in Inclusive Finance
– Dr. M S Sriram
Every few years, inclusive finance goes through a crisis with issues of excessive lending, coercive recovery practices, and high interest rates. The article explains the crisis by looking at product design and client vulnerability and arrives at four sources of stress.
The two models of inclusive finance, namely the self-help group (SHG) model as well as the joint liability group (JLG) model, which replicated the classic Grameen Bank model, have performed well and have matured over the two decades of their existence. They have established good repayments, good profi ts and superior performance. The JLG model also attracted mainstream capital, and many firms were listed on the stock exchanges and were performing well on the profitability parameters. Despite the success and the lessons learnt from the past downturns, every three to four years, some crisis or the other hits either a significant organisation, a state (or a sub-geography in the state), or a model. In this article, we analyse why these recurring crises happen and explore the ways of dealing with them. In general, the trouble in the inclusive finance space could come from four sources: (i) the cracks in the lending model, (ii) policy-induced trouble, (iii) politics and electoral cycle-induced trouble, and (iv) organisation designbased trouble.
These are not mutually exclusive and are usually overlapping. Therefore, it is important to be aware of not only the risks arising from the operations but also those arising from the perception of the outside world. However, part of the problem was in the innovation in inclusive finance—the JLG—where the excess of medicine became a poison. This model was borrowed from the Grameen Bank of Bangladesh. Grameen Bank itself has since shifted from the classic model to Grameen II, where the concept of joint liability was dispensed while the meetings and aggregation of transactions continued. In the Grameen II model, the defaulting customer was put into a separate track with intensive follow-up and a top-up loan plus support until they could come back on the main track. We discuss the potential risk of such crises, in this model as well as in the following sections.
Cracks in the Lending Model
In the inclusive finance space, poor borrowers have very little skin in the game, and they are generally more leveraged than the average borrower. The regulatory restrictions on lenders accessing savings (unless the entity had a banking licence) made it structurally difficult to build customer stakes in the formal sector. The way this model worked was mostly because the customer valued the relationship with the lender and appreciated the ongoing cycle. The cracks in the model manifest differently in each lending framework.
SHG model: Theoretically, no cracks should appear in the SHG model. Unlike the other models, the SHG model was based on member savings at the base. The customers usually belonged to a cohesive community and had intimate embedded knowledge about each other. Therefore, loaning decisions and dealing with default were expected to be done purely on merit, as the savings of the decision-makers were at stake. If the group cohesiveness was intact, the information available for making a loan decision was better than formal systems. They not only used obvious information but also had information on behavioural patterns and habits because of living in proximity to the other members of the group. Therefore, even defaults could be assumed to be genuine. The group tolerated default because they had good knowledge and could understand the difference between a wilful and a non-wilful default.
However, the moment the leverage of the group went up, and external finance started flowing in, the relative stakes of the saving members reduced. Therefore, there was an increased risk with leverage. The fact that the SHG model had been adopted into the development agenda of the state through programmes like the National Rural Livelihoods Mission did not help matters, because the state not only funded group formation and gave seed capital but also controlled interest rates through subvention. This reduced the incentive for the members to exercise responsible behaviour and resulted in delinquency and dormancy of the group. Since the net borrowing of the collective was greater than their savings, the members would not lose their savings in the case of a large-scale default. They just lost access to their continued relationships with the group. The banks that lent to the group and the state that pumped funds would be the ultimate losers. The costs of default eventually devolved to the taxpayer. A failure in the SHG did not have a covariance effect because each SHG was a micro-independent entity. A domino effect did not happen in the case of SHGS and therefore we only hear of dormancy and failure, never a crisis.
JLG model: The JLG model, on the other hand, has had a periodic crisis. While there were multiple external factors such as political interference and policy announcements that triggered crises, they manifested only when there was something brewing underneath. The JLG model had some inherent cracks. The model operated on a very low or zero tolerance for default and put pressure on the groups to bail each other out in the case of an individual default. Several times, the loan was repaid by someone else in the meeting and the transactions were squared off outside of the group process. The model was coercive in nature, with the group meeting held up if a single customer defaulted. It was made clear during the group formation and training that the others were liable to pay up for the default of a fellow member. This was enforced through the social shaming of defaulters. In this model, because of the nature of co-obligation (and the moral pressure due to the oath; and conducting transactions in the open), the defaults got covered by repayment from the other co-obligants. The field-level functionary would know that there was stress and an incipient default, but the database of the organisation did not capture it. If this is accumulated for multiple instalments or spread to multiple members, then it resulted in a collective domino default.
This happened because the lender was not doing an assessment of the borrowers’ capacity to repay. Instead, the lender was using a coercive model of collections, and the borrowers were expected to exercise due caution before they took a loan. Thus, the assessment was outsourced to the borrowers themselves. While this model worked for small loans, for larger loan amounts, the borrower may be blindsided to assuming that they will “somehow” meet the instalments without a solid basis.
Defaults in a branch in JLG model should have been a localised phenomenon and usually should not have blown up into a crisis. However, we found that when a crisis erupted, multiple organisations got affected. Why does this happen?
This happened due to the structure of operations. The field staff were professionals who had outgrown being an extension staff of not-for-profit organisations. Their core skill set was in organising the community into groups and mobilising them to convene regularly. These staff transacted as per the standard operating procedures. Their skill sets were not about finance, loaning, and understanding the implications of the transactions on the balance sheet. So, there was a client who was possibly financially illiterate and a field staff who did not understand finance. This worked if the conveyor belt of loaning and repayment continuously worked. The field officer was incentivised to build the book—organise more groups and deepen the engagement with the existing groups. Repayment happened due to the repayment discipline, unless there was some deep stress. In the process, the system did not get early warning signals.
Having recognised the nature of lending, most microfinance institutions (MFIs) built inherent risk mitigation practices: the loans were to be given only to women in active workforce; utilisation checks were to be made; the amounts of loans went up gradually with each cycle; and an organisation-level risk matrix limited the exposure to a single customer, a branch, a district, and a geography. What this did not fully factor in was that there was a similar MFI lending to the same client with similar parameters in an independent data silo. At the customer level, they had much more leverage than the acceptable risk appetite of the MFI.
A part of this problem was addressed by credit bureaus to which data was to be compulsorily uploaded. There were continuing challenges in establishing the uniqueness of the customer. Even when that was done, the system still did not know the level of indebtedness of the customer in the informal sector. Therefore, when a crisis hit the sector, it turned out that the same three sets of issues came up—multiple MFIs lending to the same client, high (or perceived to be usurious) interest rates and coercive recovery practices.
While three issues were articulated, in each of the crisis cases, the underlying reason was largely multiple lending and overleveraging of the clients to the extent that they were not serviceable. It could be argued that the clients were responsible for their borrowing, and they should have been careful. Nobody forced them to get into debt. But this argument did not cut ice. In such situations, the customers resorted to showing the organisation in a bad light and the practices that were generally accepted during normal times—high interest rates and coercive recovery practices—became a punching bag. It did not help the provider of credit that all the borrowers belonged to the poor and vulnerable sections of society.
When there was an issue with the vulnerable sections (particularly poor women), their concerns were articulated by the gatekeepers of the community. This could be a religious group (as in the Kolar crisis of 2009) or the state (as in the Krishna crisis of 2006) (Shylendra 2006) and the AP crisis of 2010 (Sriram 2012) or the political class through the instrumentality of the state (as in the Assam crisis of 2023).
Grameen II Model
Would the Grameen model help? This model had a tolerance for default, an early recognition of stress, a top-up loan and an intense follow-up. Theoretically, this should have been the solution where instead of coercive lending practices, the lender put the borrower on a debt restructuring track and closely worked with the client to get them back on regular track. Similarly, there was a widespread practice of topping up the loan after it had gone halfway through to enhance the effectiveness, particularly if the business was doing well. However, experience showed that an additional loan resulted in the evergreening of the old loan. The incentive structure of the field staff was structured to compensate for growing business and keeping the books clean. The best way to keep a defaulting book clean was to evergreen the book! The 2024 crisis of the inclusive finance sector recognised this issue, and the Reserve Bank of India (RBI) had issued an informal advisory on this phenomenon called netting off.
Policy-induced Trouble
In general, the expectation was that the policy would be supportive of the organisations and their operations in inclusive finance. Whenever the market-based institutions undertook the agenda of inclusion, the policy response was to passively or actively support the agenda. If the market forces were in any case undertaking activities that would have fallen in the agenda of the state, it made sense to support such activities with a favourable policy framework. This happened in the early days of microfinance
in India.
Proactive policy support that blunts regulatory sharpness: The earliest mention of microcredit was seen in April 1999 as a part of the monetary policy announcement of 1999–2000. The statement recognized the initiatives that were undertaken by early players around the late 1990s. It acknowledged their representation to the RBI and set up a special cell manned by a senior banker. The statement also indicated that there should be no interest rate ceiling on loans to microcredit clients. This was followed up by another mention in the review of the monetary policy later in October 19994 where the concern to remove procedural bottlenecks for mainstreaming microcredit was expressed. The policy indicated that the RBI would look at measures to accelerate the flow of credit to MFIs. In February 2000, the RBI issued a circular to the banks on the microfinance portfolio. The document briefly defined microcredit and made six significant points.
(i) No interest rate cap on loans to MFIs and their loans to clients.
(ii) Freedom for banks to formulate their own model/conduit/intermediary for extending microcredit.
(iii) No criteria prescribed for selecting MFIs.
(iv) Banks to formulate their own lending norms.
(v) Banks to formulate a simple system, minimum procedures and documentation for augmenting the flow of credit by removing all operational irritants; and
(vi) The banks were to include microcredit at the branch, block, district and state credit plans with quarterly progress to be reported to RBI.
The approach of the state was of forbearance and encouragement as it recognised that the activity carried out by the markets was advancing the developmental agenda of the state. We found a similar approach with the Federal Reserve of the United States about sub-prime housing loans. As the housing sector boomed and had rapid growth, homeownership grew to an unprecedented 69.4% by 2004 (Moss and Bolton 2011). The Fed did not undertake any actions to douse this growth but instead issued comforting statements indicating that there were no downsides, and the prices may not crash. The growth of the housing market was fuelled by liberal sub-prime loans—given to individuals who had low credit scores and uncertain incomes. While the market-based players were building up the sub-prime portfolio, it was refinanced through collateralised debt obligations and mortgage-backed securities. Most of the refinance was coming from two significant organisations—Freddie Mac and Fannie Mae—both government-sponsored enterprises. Freddie and Fannie together had 50% of the country’s stock of residential mortgages and 70% of the newly issued mortgages (Moss and Bolton 2011). While the sub-prime loan book resulted in a major financial crisis, the policy approach was quite soft. As the markets were furthering the agenda of the state, the policymakers failed to see any red flags in the practices.
The above examples were of policy-induced trouble, where the policy was actively promoting the cause of inclusion through market mechanism. The alternative policy-induced crisis came from a fully opposite stance, where the state believed that the players were exploitative and a strong clampdown on their practices was necessary.
Regulatory clampdown paralysing the operations: The classic example of a policy clampdown goes back to the AP microfinance crisis of 2010 (Mader 2013). The (then undivided) AP government issued a tough ordinance, followed by a bill in the legislative assembly imposing severe restrictions on the operations of MFIs in the state. This clampdown was in response to the reported cases of suicides of some borrowers because of coercive recovery practices. The underlying reason was cracks in the JLG model, where multiple MFIs were lending to the same client, without a realistic assessment of the extant indebtedness and an understanding of the customer cashflows.
Additionally, all clients were identifi ed as poor and vulnerable. This meant that the state would intervene and regulate the activities if there was a crisis. An additional dimension in the case of AP was that it had a strong and competing SHG movement that was backed by the state (Sriram 2012).
The policy-induced crisis cut both ways—it could happen because the state was supportive of the markets and was unable to see the excesses of the market; or, if the state looked at the underlying activities and their exploitation and clamped down on the players. Maintaining a calibrated policy stance was usually difficult for the state.
Politics and electoral cycle-induced trouble: The trouble induced by politics and electoral cycles was derived from the model of lending. All the models of inclusive finance were susceptible to this risk. The SHG model was susceptible to capture. Any model that was based on organising communities into empowered units ran on democratic principles and was susceptible to political capture.
This was natural because it is the political players who were the masters of decoding the democratic processes. Therefore, wherever there was a collective of poor vulnerable women, it was natural for the political players to approach the community and see them as a voting bloc. The SHG movement was broken not by oppressive legislation but by kindness. The kindness emanated as funding for the groups; interest rate subventions and waivers which broke the contract of the community, with an external intervention.
The same principles applied to mainstream banks that had portfolios of the poor. There were repeated attempts to write off both interest and debt. This intruded into the contract between the lender and the borrower which broke the natural rhythm of transactions.
In the case of the JLG model, the model-induced trouble—of over-indebtedness —invited the political class and even a small sign of stress was enough for the political class to go on a fishing expedition.
In all the models, the common element was that the clients or borrowers were identified as poor and vulnerable who could not articulate and negotiate a debt restructuring deal on an individual basis and therefore needed representation to articulate their distress. Politics naturally provided a voice to these sections of society.
Organisations having portfolios exclusively belonging to the poor, vulnerable, and women were hit harder because they did not have the buffer of an alternative portfolio. The effect on the banks that had diversified portfolios was not as significant as that of the MFIs.
Organisation design-based trouble: One form of organisation that got into a crisis more often than the rest of the forms was the cooperative, particularly if the cooperative was also a bank. The design of the cooperative institution was based on mutuality, where members came together to save and borrow among themselves. The SHGs also operate on similar principles, except that they are informal and restricted in size. In the case of cooperatives, because of the principle of mutuality, they were governed by elected members who were also seeking services from the institution. Therefore, there was no demutualisation between the governance and the business. While this worked well in commodity and service cooperatives, it was problematic in banking where the borrowers were sitting on the governance of the institution.
When the institution was small and operated on the principle of mutuality, the net savers were expected to exercise control on the net borrowers and therefore the member control would work effectively. They would also have embedded knowledge about the other members and were able to exercise moral suasion to make the other members behave responsibly.
The problem in cooperative institutions started with anonymity—with large numbers of members; and when there was leverage—which reduced the member stakes. Both large numbers and leverage happened if the cooperative had a banking licence. In a cooperative bank, there was a possibility of significant non-member deposits. This opened itself to a crisis by the sheer design of the organisation. That the cooperative bank was a part of the payments system would encourage the regulator to apply strict banking type of regulation. This meant that the cooperatives would lose the speciality of its design. Therefore, for cooperative banks, the two possible scenarios were—they would not behave like a cooperative and behave like a bank, or they were open to an impending crisis due to leverage.
Is There a Way Out?
There are no silver bullets, and a single organisation cannot protect itself from a covariant attack; therefore the measures to be taken are complex and interdependent.
Ecosystem-level initiatives: One of the solutions that was implemented was to have an intermediary mechanism between strict regulation by the central banker and softer control by a self-regulatory organisation (SRO). While the law did not provide for the architecture of an SRO, the RBI recognised two organisations in the MFI sector as SROs—Sa-Dhan as well as Microfinance Institutions Network (MFIN). What was to be noted was: First, the writ of the SROs could run only on their members. Excluded from this framework were non-member institutions like the commercial banks; second, the SROs were ultimately funded and governed by the member organisations that were to be regulated. While there could be insulatory mechanisms, it was not completely demutualised. The apparent conflict of interest needed to be overcome by either providing regulatory teeth to the SROs or ensuring that they remained completely autonomous with mandatory funding from the members based on the business size. A deep look at the SRO structure was needed.
There are four credit bureaus, and the regulation requires that the lenders mandatorily upload their data to the bureaus. The nature of data, the classification and the establishment of the uniqueness of the borrower must be streamlined and standardised for the credit bureau to work effectively. The data provided by the bureau should be complete, agnostic to the form of organisation and should clearly identify total client liability and the number of loans on the client.
Firm-level initiatives: At the firm level, it is obvious that the lending practices must be responsible. In a competitive world, it was difficult to ensure that each firm followed the best practices and did not suffer collateral damage due to the practices of a competing organisation. This was difficult to achieve but a good idea was to build the aspect of informal indebtedness in the information silos. The source for this information would be field agents and other members of the group. Systems had to be built that considered verifiable physical assets, to determine an income proxy, than to take the self-declaration of the members about their income. Most importantly, organisations had to build lending systems based on cash flow assessments rather than coercive practices. The maturing of the Grameen Bank moving to the Grameen II model happened two decades ago. This maturing vis-Ă -vis the portfolios of the poor in the case of Indian MFIs had to happen as well.
It is also important to send signals that the engagement with the clients was fair, non-exploitative and non-extractive. The best way to send the signals was to practise much more responsible lending. Pending some of these fundamental reforms, the sector will face one crisis or the other, every few years. The sector must learn to live with the poison.
