What is Actually Happening in the Microfinance Sector?

This is Part 1 of a 3-part series on the recently released NCAER-MFIN report, “Assessing the Effectiveness of Regulated Small Borrowing in India” (March 2026). Across the three parts, we aim to make sense of the report’s findings against the broader microfinance context. In this first blog, we focus on a set of foundational aspects—namely, the study’s timeline, the representativeness of the sample, expenditure patterns, and the reported end use of microfinance loans among borrowers. The second blog in the series will explore claims relating to the purported benefits of microfinance, while the final blog will examine issues related to measurement and empirical strategy.

The recently released NCAER-MFIN report presents a picture of a healthy and robust borrower base in India’s microfinance sector. It suggests that microfinance has enhanced access to formal credit for low-income households by providing them relatively affordable loans and by reducing their dependence on informal sources. It further claims that borrowers are primarily using these loans for income-generating activities and are managing their indebtedness prudently, remaining well within established regulatory thresholds.

However, this portrayal sits at odds with the sector’s performance over the last 24 months, marked by rising NPAs, frequent write-offs, and, in some cases, distress-driven outcomes such as borrower suicides. Indeed, these are the reasons that a government scheme, namely CGSMFI 2.0, has been recently introduced. The scheme is intended to ease liquidity constraints that the sector has been facing, especially with regard to bank funding, because of its poor performance over the last 2 years.

Given this incongruity between recent events in the sector and the report’s positive claims for it, we believe that the report’s findings merit closer examination. In this first blog, we draw the reader’s attention to specific areas of concern for us in making sense of the report.

1. Survey Timeline and Sampling Method Should Be Made Explicit

A key limitation of the report lies in its silence on the timing of data collection. The findings could reflect conditions that either predate the crisis or capture only its early phase, thereby offering an incomplete view of borrower conditions. Microfinance is inherently cyclical, where periods of stability can endogenously give rise to behaviours that eventually create instability and downturns[1]. In this context, the survey’s timeframe becomes particularly important. And so, it is surprising that the exact survey period is not explicitly stated. Reporting the timeline is not a mere formality; it is a critical piece of context that allows readers to assess relevance, ensure comparability, and evaluate the credibility of the findings.

In the absence of an explicit timeline, the reader is left to piece one together via some triangulation. Two sources are helpful here: the group formation timelines mentioned in the FGD summaries and references in the executive summary. Evidence from the annexure suggests that fieldwork likely took place in mid- to late-2024. Additionally, the report compares survey-reported interest rates with industry data across two quarters and notes that one of these immediately preceded the survey. This places the survey timeline in Q2 FY 2024–25, i.e., between July and September 2024.

Data from CRIF High Mark show that the sector’s outstanding loan book peaked in March 2024 and has since declined. If the survey followed this peak, it may reflect a period of temporary stability. If it preceded the crisis, underlying stress may have been obscured by practices such as loan churning. Even if conducted during the crisis, the risk of sample distortion remains, depending on the precise timing of the survey.

Further, borrowers can broadly be categorised as performing or non-performing. The report does not clarify the method of composition of its sample. Given the absence of any discussion of recovery or post-default experiences among the borrowers that were part of the sample, it appears that defaulting borrowers were not meaningfully represented. A purposive sampling that favoured performing borrowers would clearly understate systemic stress.

The glaring omissions in the report with regard to both aspects that we have highlighted here – the precise timing of sampling, and the method of sampling – pose problems of interpretation and credibility and should be addressed forthwith. In addition, an explicit acknowledgment should be made about the potential drawbacks of the choices made in these two respects, because no sampling method is ever perfect and free of the evident dangers of mis-representation.

2. Loan Use and Repayment Patterns Should be More Adequately Studied

The report notes that credit was primarily used for productive purposes[2], citing that roughly 75% of borrowers reported using loans for business activities. However, this claim is puzzling when viewed alongside broader repayment patterns.

The majority of households rely for repayment on the income of other household members (71.8%) and on non-financed activities (36.1%), while only 51.8% repay from financed activities. There is also significant state variation: reliance on loan-funded activity is high in Madhya Pradesh (~82.4%) and Assam (~81.8%), but very low in more mature markets such as Telangana (~7.0%) and Tamil Nadu (~15.9%). Taken together, these patterns suggest that repayment is often supported by broader household cash flows rather than the financed activity alone, which may reflect either insufficient returns from the activity or a misalignment between loan structures and business gestation periods.

Furthermore, near-universal reporting of “productive use” (~94%) may obscure underlying issues. These may include partial deployment of loans toward the stated activity or their diversion to other purposes. Survey responses on loan use can also be sensitive to context. The report notes that borrowers were reached with the assistance of REs and their staff members but does not clarify whether interviews were conducted in the presence or absence of those staff members. This raises the possibility of response bias, particularly in centre-based interviews where staff presence may have influenced the interviewee’s responses.

The report’s findings on loan use also stand in contrast to observations from CMIE CPHS panel data. Drawing on 2021–24 data, our recent paper, “India’s Most Recent Microfinance Crisis: Theory, Empirics & Learnings,” suggests that a significant share of microfinance borrowing has been used for loan repayment (churning) and for expenditure on consumer durables and other non-productive purposes[3]. The differences between our findings and the NCAER-MFIN report’s findings may partly reflect the timing of data collection within the credit cycle or variations in sample design – but that is all the more reason for not interpreting the NCAER-MFIN’s report’s findings as representing a steady-state pattern of loan use, as the report’s authors appear to be doing.

To be fair, some of the underlying survey’s questions do attempt to capture such dynamics, but the report does not engage with them. For example, the lender choice question[4] was explicitly multi-response and unprompted, allowing borrowers to more truthfully reveal their decision drivers. In this context, the finding that 11.6% of borrowers report “adjusting an existing loan” may be indicative of loan churning, where new borrowing is used, at least in part, to manage existing debt. This is a non-trivial signal that merits closer attention than it receives in the report.

3. The Estimates of Consumption and FOIR Do Not Survive Careful Scrutiny

The report highlights that the sample’s average Fixed Obligation to Income Ratio (FOIR) stands at 18.7%, well below the Reserve Bank of India’s prescribed threshold of 50%. It presents this as evidence of manageable indebtedness. Yet, this sits uneasily alongside the report’s own acknowledgement that FOIR is an imperfect measure of repayment capacity, as it does not account for essential household expenditures. If FOIR is inadequate, then compliance with such a threshold provides only limited insight into households’ underlying financial stress.

The report partially addresses this limitation by constructing detailed household cash flow statements to estimate surplus income available for loan servicing. Without access to the underlying data and the associated cleaning procedures, it would be difficult to properly evaluate the estimates of surplus income so arrived. However, comparisons with external benchmarks can still help identify potential problems with those estimates.

The reported average MPCE of ₹3,742 corresponds broadly to the middle of the consumption distribution, around the 5th–6th decile, when benchmarked against both CMIE CPHS and the Household Consumption Expenditure Survey (HCES). The consistency of this positioning across two independent datasets strengthens the comparability of NCAER-MFIN estimates with those of HCES. However, the overall average may also obscure important variation within the sample, particularly across states, making it important to examine how state-level MPCE estimates compare with corresponding HCES benchmarks.

In Table 1, we compare NCAER-MFIN’s reported MPCE with that of HCES 2023–24 (conducted from August 2023 to July 2024) to assess potential underreporting of household expenditure. The comparison shows that HCES rural MPCE is consistently higher than NCAER-MFIN estimates—by up to 40% across states—with smaller gaps in West Bengal and Tamil Nadu and larger ones in Telangana, Punjab, and Madhya Pradesh. As the report notes, some divergence is to be expected given differences in sample design. However, the divergence in food expenditure is substantially larger, ranging from 10% to as high as 70% (notably in Bihar, Maharashtra, Madhya Pradesh, Punjab, and Telangana). Such differences, particularly in food expenditure, are difficult to explain solely by time period or by the report’s 68% rural sample composition.

Table 1: Benchmarking HCES MPCE with NCAER-MFIN Survey Data

Two concerns arise in this context. First, in several cases, the reported levels appear insufficient to meet even basic calorie requirements. Consider the case of Bihar, where food MPCE is reported at ₹1,130 per month (~₹38 per day). This appears difficult to reconcile with established poverty benchmarks and the survey’s internal MPCE-to-food-expenditure ratio. Second, there is considerable variation in reported food expenditure, ranging from ₹1,130 in Bihar to ₹2,015 in Tamil Nadu. While some variation is expected, a near doubling of per-capita food expenditure is difficult to justify, particularly given the relatively low income-elasticity of food consumption among low-income households, where expenditure levels would typically lie within a narrower band.

Taken together, these patterns raise concerns about potential underreporting of household expenditure. If expenditure is understated, repayment capacity is correspondingly overstated, which, in turn, would render even the cashflow-based assessments more reassuring than warranted. This, in turn, raises an important question: if borrower indebtedness is indeed as manageable as suggested, what explains the sharp rise in the sector’s gross non-performing asset (NPA) ratio to 16% by the end of FY25, nearly doubling within a year?

Conclusion

In this blog, we highlighted concerns related to survey timing, sample representativeness, interpretation of loan use, and inconsistencies in expenditure data. Taken together, these issues suggest that the report presents a more favourable picture of borrower conditions than is warranted by the underlying evidence.

———————————-

Footnotes

[1] Dwijaraj, Bhattacharya, Indradeep Ghosh, Madhu Srinivas, Navaneeth M S, and Shree Harini V. India’s Most Recent Microfinance Crisis: Theory, Empirics & Learnings. 2025. https://dvararesearch.com/indias-most-recent-microfinance-crisis-theory-empirics-learnings-dvara-research/

[2] By “productive use” we include loan uses showcased in report as Construction of shop/structure for business purpose, Farm/Crop Activity, Investment in existing Business, Investment in new Business

[3] Dwijaraj, Bhattacharya, Indradeep Ghosh, Madhu Srinivas, Navaneeth M S, and Shree Harini V. India’s Most Recent Microfinance Crisis: Theory, Empirics & Learnings. 2025. https://dvararesearch.com/indias-most-recent-microfinance-crisis-theory-empirics-learnings-dvara-research/

[4] The question on lender choice was “What are the things that you consider as important for choosing a loan provider? (Multiple response possible, no prompts by the interviewer)”

———————————-

Part 2 of the series is available here

Part 3 of the series is available here