Supply-Side Expansion, Demand-Side Retreat

Credit is generally seen as one of the most important needs of businesses. This is perhaps true for corporates in India. But we often transfer this expectation to businesses of all kinds, including smaller ones and those that are primarily self-employment activities. The assumption does not hold up to closer scrutiny, and we see some important trends when it comes to tiny businesses in India, or what are increasingly called nano-enterprises (nano businesses operating at subsistence or self-employment scale).

We look at two sources of information in this article. First is the ‘Small Business Spotlight – December 2025’ report published by SIDBI and CRIF High Mark. Second is the surveys of unincorporated enterprises conducted by the National Sample Survey (NSS). These two sources do not coincide much in terms of coverage and approach. The SIDBI report is based on information from the lending system — a supply-side view. The NSS surveys provide a demand-side view of business credit, capturing self-reported outstanding loans of unincorporated enterprises.

What emerges is a picture of two distinct worlds. One is the universe of nano-enterprises captured by household surveys — largely home-based, own-account, and increasingly distant from credit. The other is the universe tracked by lending institutions — urban-concentrated, NBFC-driven, and expanding. The two overlap, but imperfectly.

The Supply Side: What SIDBI-CRIF Tells Us

As per the SIDBI-CRIF report, the credit portfolio of small businesses stood at ₹39.6 lakh crore outstanding as of September 2024, with 503 lakh active borrowers. This gives an average loan outstanding per borrower of around ₹7.9 lakh. By September 2025, the average had grown modestly to ₹8.2 lakh.

The aggregate picture has considerable nuance when we look at three groups: enterprises (registered entities), sole proprietors with entity presence (SPEP), and sole proprietors.

The landscape has been largely stable over the last two years in terms of structural composition, though it has grown in magnitude. Sole proprietors form about 70% of borrowers, and their average outstanding credit is largely unchanged at around ₹3.5 lakh. The average outstanding loan of enterprises has grown from ₹9.5 lakh in September 2023 to approximately ₹12 lakh by September 2025.

What is striking is that the average outstanding credit of SPEP is considerably more than the sum of the average outstanding credit for enterprises and sole proprietors. This suggests that SPEP borrowers are structurally different — likely a subset of bigger enterprises where sole proprietors also have better individual asset bases and familiarity with credit institutions.

The SIDBI report shows credit to small businesses expanding between 2023 and 2025, in terms of both outstanding credit and borrower counts. The growth is driven by sole proprietors, with or without entity presence. New-to-credit borrowers continue entering the formal system — a sign that supply-side expansion is reaching previously unbanked segments. The rising role of NBFCs in serving sole proprietors is notable; they now account for a significant share of small business lending.

What the report does not tell us: distribution of loans by size and interest rates faced by borrowers, rural-urban breakdowns beyond city-tier classification, and borrower counts by loan type (it only provides portfolio distribution by loan type). The terms on which new-to-credit borrowers are being served remain unclear.

The Demand Side: What NSS Surveys Show

The demand-side view from NSS surveys shows a different picture, albeit with qualifying limitations. We compare NSS 73rd Round (2015-16) with ASUSE 2023-24 — an eight-year gap containing significant structural disruptions whose differential effects on enterprise types deserve separate analysis.

Compared to 2015-16, in 2023-24, there is a decline in the proportion of non-agricultural non-construction unincorporated enterprises having outstanding loans exclusively borrowed and used for enterprise purposes — from 13.0% to 10.2%. The decline is also absolute: from 82 lakh to 78 lakh borrowing enterprises. Yet total outstanding loans have more than doubled, from ₹1.74 lakh crore to ₹3.98 lakh crore, and average loan per borrowing enterprise has grown from ₹2.1 lakh to ₹5.1 lakh.

If we look only at bank credit (commercial and cooperative banks), the picture is somewhat different. The proportion of enterprises with bank loans remained nearly unchanged — 4.2% in 2015-16 and 4.5% in 2023-24. In absolute terms, this is an increase from 26 lakh to 35 lakh enterprises. Total bank loan outstanding has more than doubled, from ₹1.12 lakh crore to ₹2.76 lakh crore.

There is another way to understand credit to unincorporated enterprises and how it has changed. Bank credit to unincorporated enterprises as a share of total non-agricultural enterprise credit from banks rose from 3.3% to 4.7% during the period 2015-16 to 2023-24. In the same period, the share of unincorporated enterprises in non-agricultural GVA (from national accounts) declined from about 11% to 8%. More credit is flowing to a sector contributing less.

The striking shift is in informal sources. The proportion of enterprises borrowing from informal sources has nearly halved — from 7.6% to 4.1%, and in absolute terms from 48 lakh to 32 lakh enterprises. Money lenders saw a sharp decline from 17 lakh to 5 lakh enterprises. Friends and relatives declined from 24 lakh to 15 lakh. Yet total informal credit outstanding has doubled, and the average loan per informal borrower has tripled.

The broad story: formal sources (banks, government) are reaching more enterprises with larger loans; informal sources (especially moneylenders and friends/relatives) are retreating. But even as fewer enterprises borrow informally, those who do are borrowing more. Credit is formalising and concentrating.

Who Is Borrowing Less?

The aggregate decline in borrowing enterprises — from 82 lakh to 78 lakh — masks sharp variation across segments. When we disaggregate by location (rural/urban), enterprise structure (own-account vs hired-worker), and gender of owner, a clearer picture emerges.

Figure 1: Enterprises vs Borrowers by Segment — Blue bars (total enterprises) grew; red bars (borrowers) shrank in most segments

The biggest segment is rural male own-account enterprises: 243 lakh units, nearly a third of all unincorporated enterprises. Their borrowing rate fell from 12.4% to 8.7%. In absolute terms, borrowers in this segment declined from 27 lakh to 21 lakh — despite the segment itself growing by 27 lakh enterprises.

Rural female own-account enterprises grew even faster — from 68 lakh to 106 lakh, a 57% expansion. Yet only 3.4% have any loan outstanding. That is 3.6 lakh borrowers among a crore enterprises.

The pattern holds across most segments. Urban male own-account: borrowing fell from 11.8% to 7.2%. Rural male hired-worker enterprises — the closest to ‘formal’ businesses in this universe — saw the sharpest decline: from 25.3% to 17.9%.

Figure 2: Formal Borrowing Rate by Segment — Only rural hired-worker female enterprises showed improvement

The sole exception is rural female hired-worker enterprises. A tiny segment — 1.3 lakh units, likely a few hundred sample observations — but the only one where borrowing rates held steady and formal credit access improved. Bank loan access jumped from 7.4% to 10.6%. This pattern is consistent with targeted policy effects, though we cannot attribute causation. The segment remains too small to move aggregate numbers.

Fewer Borrowers, Larger Loans

If fewer enterprises are borrowing, where is the credit going? The answer: to fewer borrowers taking larger loans.

Figure 3: Average Loan per Borrowing Enterprise — Hired-worker enterprises borrow 10-30x more than own-account enterprises

Average loan outstanding per borrowing enterprise has grown across all segments. For hired-worker enterprises, average loans now range from ₹7 lakh (rural male) to ₹15 lakh (urban female). For own-account enterprises, the amounts are far smaller — ₹0.4 lakh to ₹2 lakh — but even these have increased.

The gap between hired-worker and own-account enterprises is stark: a 10-30x difference in average loan size. A somewhat out-of-focus observation: OAE vs HWE is a more defining distinction than rural vs urban. The intent and organisation of the enterprise matters more than administrative geography — a point anyone interested in nano-enterprises needs to take into account.

Figure 4: Average Formal Loan per Borrowing Enterprise — Formal sources show even sharper concentration

The implication: credit is concentrating among a smaller number of larger borrowers. The enterprises that multiply — overwhelmingly own-account, often women-owned, operating from homes — are not the enterprises that borrow.

An Uneasy Discrepancy

The demand-side picture from NSS surveys sits uneasily with supply-side data. The SIDBI-CRIF report (September 2024) shows 361 lakh sole proprietors with active business credit — average outstanding of ₹3.6 lakh each. Adding sole proprietors with entity presence brings the total to over 420 lakh borrowers. ASUSE 2023-24 shows only 78 lakh enterprises with any loan outstanding.

Where are the other borrowers?

Part of the answer lies in coverage differences. SIDBI includes commercial vehicle loans, construction equipment loans, and loans against property taken by sole proprietors for business purposes; ASUSE excludes construction enterprises entirely and captures only loans explicitly borrowed and used for the enterprise. Agricultural loans are excluded from both. But these factors alone cannot explain a 5x gap.

A deeper issue may be the sampling frame. ASUSE samples enterprises based on the geographical distribution of the population. But enterprise distribution differs from population distribution — formal enterprises cluster in urban commercial zones. GST registrations, for instance, are heavily urban-concentrated.

We argue, mildly, that ASUSE may underrepresent non-household-premise enterprises. The gap between ASUSE borrowers (78 lakh) and SIDBI sole proprietors (360+ lakh) is too large to be fully explained by construction, vehicle loans, and loans against property — but it is also not enormous once we account for these differences.

Consider the registration data from ASUSE itself:

Only 0.44% of enterprises report Udyam registration, while the Udyam portal claims over 5 crore registered MSMEs (as of 2024). If ASUSE captured formal enterprises well, we would expect Udyam registration rates closer to 5-10%. The 0.44% rate suggests systematic undercoverage of the formal segment.

Bank account penetration offers a bridge. In ASUSE 2023-24, 76% of enterprises — 585 lakh — report having a bank account, up from 57% in 2015-16 (though much of this reflects Jan Dhan-driven account opening rather than enterprise banking). This is closer to SIDBI’s universe of credit-active sole proprietors. The surveys may be capturing overlapping but distinct populations: ASUSE finds the vast ocean of nano-enterprises (home-based, own-account, often invisible to formal systems), while SIDBI tracks enterprises that have crossed a threshold of formality (registered, banked, credit-active).

The Two Worlds of Small Business Credit for Nano Enterprises

What emerges from this analysis is a picture of two distinct worlds, each captured imperfectly by different data sources.

The first world: nano-enterprises in the ASUSE universe.

These are the 7.7 crore unincorporated enterprises captured by household surveys. The vast majority are own-account (87%), home-based, and operate at scales where credit is simply not part of how they function. Among those who do borrow, credit is increasingly formal — bank credit as a share of their borrowing has risen. But the base of borrowers is shrinking even as the enterprise count grows.

The second world: credit-active businesses in the SIDBI universe.

These are the 5 crore+ borrowers tracked by lending institutions — sole proprietors, SPEPs, and registered enterprises with active loans. This universe is urban-concentrated, NBFC-driven, and expanding. New-to-credit borrowers are entering the system. But the terms on which they borrow, and the distribution of loan sizes, remain opaque.

The two worlds overlap — the 78 lakh ASUSE borrowers are presumably a subset of the SIDBI universe — but they tell different stories. The supply side shows expansion; the demand side shows retreat. Both can be true if credit is flowing to a segment of enterprises that household surveys undercount (urban, non-household-premise, formally registered) while bypassing the nano-enterprises that surveys capture well.

Credit-Distant Enterprises: A Closer Look

To understand why most nano-enterprises do not borrow, we need to move beyond the binary of ‘has loan’ vs ‘no loan’. The NSS surveys allow us to do this by asking enterprises whether they faced problems in operation, including non-availability or high cost of credit.

How do we classify enterprises by credit state

ASUSE asks enterprises whether they faced problems in operation during the last 365 days, and if so, the most severe and next most severe problems. One option is ‘non-availability or high cost of credit’. Separately, the survey records outstanding loans by source at the end of the reference period.

We combine these to create four states:

• Satisfied borrowers: Have an outstanding loan, do not report credit as a problem
• Frustrated borrowers: Have an outstanding loan, but still report credit as a problem
• Discouraged enterprises: No loan, but report credit as a problem
• Credit-distant enterprises: No loan, do not report credit as a problem

This framework separates credit access from credit need. An enterprise may not borrow because credit is inaccessible (discouraged), or because it operates at a scale where credit is not central to how it functions (credit-distant). We cannot distinguish between enterprises for whom credit is genuinely irrelevant and those who have stopped seeking it after repeated discouragement — ‘credit-distant’ is our interpretation, not a self-evident fact.

Applying this framework to both surveys reveals a striking shift over eight years:

The most dramatic shift is in the ‘discouraged’ category — enterprises that do not borrow but report credit as a problem. This fell from 42 lakh to just 4 lakh enterprises. Frustrated borrowers (those who borrowed but still faced credit problems) also collapsed from 11 lakh to 1 lakh.

Where did these enterprises go? Three interpretations are possible. First, credit access improved dramatically and those who needed credit got it — the satisfied borrower count did rise modestly from 72 to 77 lakh. Second, enterprises that once aspired to grow (and thus needed credit) have settled into subsistence operations where credit is no longer relevant. Third, enterprise activity is one element of the livelihood portfolio for many households, and when finance is needed, it is accessed through borrowing for other purposes (housing, personal loans) rather than enterprise-specific credit.

We lean toward a combination of the latter two interpretations. The massive shift toward credit-distance — a 9 percentage point increase — cannot be explained by the modest increase in satisfied borrowers. The enterprise base itself seems to have shifted toward activities where credit is less central.

Who Became Credit-Distant?

The shift toward credit-distance was not uniform across segments:

The largest shifts occurred among male-owned enterprises, both own-account and hired-worker, in both rural and urban areas. Rural male OAEs saw the biggest change — from 78% to 91% credit-distant. These are the enterprises that form the backbone of India’s non-farm economy.

The exception, once again, is rural female hired-worker enterprises. This segment shows almost no change in credit-distance — 78% in 2015-16, 78% in 2023-24. This is consistent with our earlier finding that this is the only segment where borrowing rates held steady and formal credit access improved.

By Location and Enterprise Type

Credit-distance varies sharply by where and how enterprises operate:

The pattern is clear: credit-distance is highest among household-based, own-account enterprises — the quintessential nano enterprises. It falls as enterprises move to fixed premises and hire workers. But even among hired-worker enterprises with permanent structures, over 80% remain credit-distant.

Implications

Several implications follow from this analysis.

First, the universe of small businesses is dichotomous.

One group consists of nano-enterprises for whom credit is not salient. They operate from homes, do not hire workers, and function at scales where loans are neither sought nor missed. This group is growing in absolute numbers but retreating from credit. The other group — more visible to formal institutions — consists of enterprises that have crossed thresholds of scale, formality, and credit-relevance. This group is expanding its borrowing, driven by NBFCs and new-to-credit lending. A policy designed for one group may miss the other entirely, particularly the vast universe of nano enterprises.

We need to know more about the latter group — beyond its credit consumption — particularly its employment generation and value addition. The first group, numerous and visible through annual surveys, is declining in its contribution to value addition and yet consuming a larger share of bank credit, which tends to be the cheapest credit available. Meanwhile, delinquency trends in the SIDBI report, though at manageable levels overall, show PSBs with notably higher stress: PAR 181-720 (loans past due 6 months to 2 years) stands at 5.0% for PSBs versus 1.4% for private banks and 2.5% for NBFCs as of September 2025.  A similar pattern, but at lower levels, is seen for PAR 91-180. It would be important to examine whether public sector banks are lending more to relatively less productive enterprises and thus facing more delinquencies, and whether this pattern needs policy attention.

Second, the decline in ‘discouraged’ enterprises is significant but ambiguous.

In 2015-16, 42 lakh enterprises wanted credit but reported it as a problem. By 2023-24, this had fallen to 4 lakh. This could mean credit access improved dramatically. Or it could mean enterprises that once aspired to grow have settled into subsistence. Or — and this may be most plausible — enterprise activity is part of a broader livelihood portfolio, and credit needs are met through other channels. The data cannot definitively distinguish these interpretations, but the massive shift toward credit-distance suggests the problem is less about credit supply than about the nature of these activities.

Third, not every entity selling goods or services is an enterprise in the economically meaningful sense.

The period between surveys saw significant structural disruptions — GST implementation, demonetization, and the COVID-19 pandemic. Whatever their individual effects, the net result has been a deepening of the enterprise organisation dichotomy. Some fraction of enterprises from 2015-16 did not survive these changes; they have been replaced by a new wave of individuals whose approach to self-employment is more cautious, driven by desire to diversify the household livelihood portfolio, and credit-distant. At the same time, credit supply has improved — not solely through enterprise or business loans, but through secured loans and a changing mix of credit institutions. Improved credit supply, including cheaper bank loans and more expensive NBFC loans, is reaching those who want to borrow.

It may be time to redefine ‘enterprise’, recognising this dichotomy. We are perhaps counting self-employment as entrepreneurial activity and misdirecting resources. The needs of livelihood-safety-driven individuals and truly entrepreneurial individuals are different, and so are the actions they are likely to take.

A proposal: define enterprises by labour, not revenue.

One way to sharpen the definition is to use labour rather than turnover. Under this approach, an enterprise would require at least one hired worker — the owner-manager alone would not suffice. This is clearly a restrictive assumption that would misclassify some genuine solo entrepreneurs, particularly in knowledge-intensive services like software development or consulting. But for the vast majority of nano-enterprises in manufacturing and traditional services, the absence of hired labour is a reasonable marker of subsistence-orientation rather than entrepreneurial intent.

We need to destigmatise and simplify registration of employer-employee relationships. The E-Shram portal is a step in this direction, but only from the employee side. An employer-side equivalent — where hiring a worker triggers automatic, low-burden registration — could create a real-time census of enterprises that cross the subsistence threshold. This would leverage our digital public infrastructure without subjecting small employers to the full weight of labour law compliance that might otherwise discourage formalisation.

A labour-hiring-based definition could make enterprise-targeting policies more effective while improving the efficiency of livelihood support policies. As of now, a generous definition clubs livelihood activities and growth aspirants together, creating an opaque reality that data cannot capture and we cannot measure credibly. As the saying goes, to improve something we must measure it better; to measure it better, we must define it better.

Disclaimer: The views expressed are those of the author.

Data sources: SIDBI-CRIF High Mark Small Business Spotlight December 2025, NSS 73rd Round (2015-16), ASUSE 2023-24. Analysis uses weighted survey estimates for proprietary enterprises only.