Chirag Patnaik

India's household balance sheet from 1951-52 to 2025-26: net financial and physical savings, where financial savings go, new borrowing, the household debt stock, the mix of bank credit to households, gold loans, and repayment stress in microfinance and bank retail loans.

Borrowed time: the Indian household balance sheet

In 2022-23 household net financial savings fell to their lowest level since 1979-80. By 2024-25 they had recovered, but because households borrowed less, not because they saved more. Over the same period borrowing shifted into gold, the debt stock kept rising, and microfinance stress cleared as borrowers were cut off.

4.95%
of GDP, 2022-23 trough
Net household financial savings on the 2011-12 base: the lowest since 1979-80 (4.8%) on our splice of RBI's series.
7.1%
of GDP, 2024-25
Recovered on the new 2022-23 base. New borrowing fell from 6.5% to 4.9% of GDP, while gross financial saving stayed flat at about 12%.
49.1%
of GDP, household debt, Mar 2026
BIS series, up from a 33.1% low in 2016. Its last two quarters include a step we can't explain yet (see chart 3).
₹5.14 lakh cr
bank loans against gold, May 2026
5.5x the March 2024 level. Adjusted for the gold price, that's about 2.1x more gold pledged.

The chart doing the rounds shows household net financial savings at their lowest since 1976-77. Some version of that was true for 2022-23 and 2023-24. It is no longer the latest reading. The National Statistics Office moved national accounts to a 2022-23 base in February 2026. On the new base, net financial savings recovered to 7.1% of GDP in 2024-25, which is back inside the 7–8% range of the 2010s.

The recovery came almost entirely from the liabilities side. Households took on new debt worth 6.5% of GDP in 2023-24 and 4.9% in 2024-25. The assets side hardly moved: gross financial saving went from 12.3% of GDP to 12.0%. The slowdown in borrowing followed the RBI's November 2023 increase in risk weights on unsecured consumer credit. The timing fits a regulator-imposed brake, but these data can't prove that was the cause. What they do show is that household demand for credit didn't go away. Bank loans against gold jewellery were ₹0.93 lakh crore in March 2024 and ₹5.14 lakh crore by May 2026.

Three GDP bases are spliced here: 2004-05 (up to 2011-12), 2011-12 (2011-12 to 2023-24) and 2022-23 (from 2022-23). They don't match where they overlap. For 2023-24, the new base puts net financial savings 0.7 points higher than the old one (5.9% against 5.2%), mainly because the new base has lower GDP and higher gross saving. The charts draw each base as its own segment. Compare levels across a break with care.
1 · The long view

Household net financial savings, % of GDP, 1951-52 to 2024-25

Gross financial saving (bank deposits, insurance, provident funds, shares, currency) minus new financial liabilities. Each GDP base is drawn separately, and the dashed rules mark where the base changes.

Sources: RBI Handbook of Statistics on the Indian Economy, 2012-13 edition, Tables 3(a) and 10 (2004-05 base); 2025-26 edition, Table 12 (NAS 2025, 2011-12 base), with GDP from Table 1 of the 2016-17, 2018-19, 2022-23 and 2024-25 editions; 2025-26 edition, Tables 1 and 11 (2022-23 base). Net savings are computed from rupee flows over nominal GDP at market prices. The 2004-05 base segment uses the NAS headline household financial saving, which differs slightly from assets minus liabilities in the RBI flow-of-funds table.

Other people have made the "lowest since 1976-77" chart with a different data vintage. On ours, the last reading lower than the 2022-23 trough of 4.95% is 1979-80 (4.84%), a year of drought and oil shock. The 2020-21 spike to 11.7% was lockdown saving: people couldn't spend, not an increase in thrift. The fall that followed was the unwinding of that forced saving plus a borrowing binge.

2 · The scissors

What households put in, and what they borrowed, % of GDP, 1972-73 to 2024-25

Gross financial saving (assets acquired in the year) against new financial liabilities (loans taken in the year). The gap between the two is net financial saving.

Sources: RBI Handbook 2012-13, Table 11 (2004-05 base, up to 2011-12); Handbook 2025-26, Table 12 (2011-12 base) and Table 11 (2022-23 base). Liabilities are bank advances plus loans from NBFCs, housing finance companies, insurers and government. Bank advances also include the quasi-corporate sector. The 2023-24 split between bank and non-bank lending changes because of the HDFC–HDFC Bank merger, but the total doesn't.

Households have borrowed this much once before. New liabilities reached 6.6% of GDP in 2006-07, at the top of the pre-2008 credit boom. What differs is the saving that sat alongside the borrowing. In 2006-07 households put 17.8% of GDP into financial assets. In 2023-24, when borrowing hit 6.2%, they put in 11.4%. Apart from the lockdown year, gross financial saving has stayed between 10% and 12.5% of GDP since 2011-12. The 2024-25 recovery came from the borrowing line falling back, and the saving line didn't move.

3 · Where the saving went instead

Physical savings against net financial savings, % of GDP, 1951-52 to 2024-25

Physical savings are mostly housing and construction by households, plus machinery and equipment of household enterprises. This chart tests the strongest counter-argument: households didn't get poorer, they bought houses.

Sources: RBI Handbook 2012-13, Table 10 (2004-05 base); Table 11 of the 2016-17, 2018-19, 2022-23, 2024-25 and 2025-26 editions (2011-12 base, from 2012-13; latest vintage used for each year); Handbook 2025-26, Table 11 (2022-23 base). Gold and silver held as savings are excluded. They are small on the old base (0.2–0.4% of GDP) and larger on the new base (0.6–0.7%), which values them differently.

The counter-argument works for part of the period. On the old base, physical savings rose from 10.8% of GDP in 2020-21 to 13.4% in 2022-23, as the housing cycle turned. So some of the fall in financial savings did turn into bricks. But the shift stops there. On the new base, physical savings are flat at 14.1%, 14.1% and 13.9% of GDP across 2022-23 to 2024-25. That means the 2024-25 recovery in financial savings wasn't money coming back out of property, and the post-2023 borrowing wasn't mostly buying houses either.

What households put their financial savings into, % of GDP, 1983-84 to 2023-24

Gross flows into each type of instrument, all drawn on the same scale. The dashed rule marks the 2011-12 base change.

Source: RBI Handbook 2025-26, Table 12 (NAS 2025; up to 2010-11 the values match the 2004-05 base series). GDP matches the base: the 2004-05 base up to 2010-11 and the 2011-12 base after. Deposits include non-bank deposits. Insurance and pension covers life insurance funds plus provident and pension funds. Shares, debentures and MFs includes mutual funds and UTI units, as NAS nets them. Instrument detail is not yet published on the 2022-23 base.

The drop in bank deposits drives the long decline. Deposits took in 10.1% of GDP in 2006-07 and 4.6% in 2023-24. Insurance and pension flows have stayed between 3.3% and 6% of GDP for two decades, and were 4.3% in 2023-24. The shift to equities that the SIP-inflow headlines describe barely shows up in net terms: shares, debentures and mutual funds took 0.9% of GDP in 2023-24, the same as in 2018-19 and below the 2.4% of 1991-92. Currency is its own story. It was −2.2% of GDP in 2016-17 (demonetisation) and 2.8% in 2017-18 (remonetisation).

4 · The stock

Household debt outstanding, % of GDP, quarterly, 2007 to Q1 2026

Credit to households and non-profits from all lenders, at market value, as a share of trailing four-quarter GDP. Hollow points mark the two latest quarters.

Source: BIS total credit statistics, series Q.IN.H.A.M.770.A (adjusted for breaks). Quarters are calendar quarters, so Q1 2026 ends in March 2026. The RBI's June 2026 Financial Stability Report puts the ratio at 45.5% in September 2025 against a five-year average of 42.9%. The BIS level series jumps 7.3% in Q4 2025 alone (₹150.6 to ₹161.6 lakh crore), and GDP was rebased that quarter. We flag both last two points and don't lean on them.

The flow slowed but the stock didn't fall. Debt grows faster than GDP whenever new borrowing is larger than the existing debt ratio times nominal growth, which is roughly 4.4% of GDP at today's numbers. At 4.9% in 2024-25, borrowing was still above that line. Even without the last two flagged quarters the ratio rose 12 points in nine years, from 33.1% in 2016 to 45.5% in September 2025. That is within a point of the pre-2008 peak of 46.2%, and the two flagged quarters take it past. Before 2008 the build-up came with 14–18% of GDP in gross financial saving. This one comes with 11–12%.

5 · Where the borrowing went

Bank credit to households by product, indexed to March 2023 = 100 (log scale)

This shows whether gold loans replaced unsecured credit or grew on top of it. The dashed rule marks the RBI's November 2023 increase in risk weights on consumer credit.

Source: RBI Handbook 2025-26, Tables 42 (end-March) and 168 (monthly), sectoral deployment of bank credit, one vintage throughout. Scheduled commercial banks only. From December 2025 the monthly figures use month-end data, not the last reporting fortnight. Housing is left out because the July 2023 HDFC–HDFC Bank merger moved a large mortgage book into bank credit. Vehicle loans are the secured benchmark.

The break in 2023-24 is visible. In the year to March 2024 credit-card balances grew 25.6% and other personal loans 22.8%. After the risk-weight increase that fell to 10.6% and 7.5% in 2024-25, and card growth fell to 3.6% in 2025-26. Gold loans went the other way: up 4.4% in 2023-24, then 121% and 124%. Unsecured credit did keep growing, but more slowly. Gold loans came on top of it.

Bank loans against gold jewellery

Left: outstanding at the end of March, in ₹ lakh crore, plus the latest month. Right: the same loans divided by that March's Mumbai gold price, giving the tonnes of gold the loan book represents at current prices. Rising gold prices account for part of the rise, but not all of it.

₹ lakh crore outstanding
Tonnes of gold at March prices
Sources: RBI Handbook 2022-23 edition, Table 46, and 2025-26 edition, Tables 42 and 168 (sectoral deployment of bank credit, personal loans, item 4.8); Handbook Tables 164/166 (Mumbai gold price, March monthly average). Scheduled commercial banks only. NBFC gold loans (₹3.41 lakh crore in June 2026, per RBI data reported by ANI) are extra. From May 2024 at least one bank reclassified agricultural gold loans into this line (Business Standard, February 2025), which inflates the 2024-25 jump. The tonnes figure is illustrative: loan-to-value ratios aren't 100%, so the physical gold pledged is larger.

Deflating by the gold price is the fair test. Over two years gold went up 2.35x (₹65,229 to ₹1,53,299 per 10g, March to March). The gold loan book went up nearly 5x over the same years. So the rupee value of gold loans rose about twice as fast as the price of the gold behind them. Part of that is reclassification. Part is borrowers moving to gold after unsecured credit got more expensive. Part is ordinary households turning their savings into cash through borrowing, which is the reading the savings chart alone can't show.

6 · Did anything break?

Repayment stress: microfinance, and banks' unsecured retail book

Left: microfinance loans 31–180 days overdue, % of portfolio, all lenders. Centre: microfinance borrowers with a live loan, crore. Right: bank gross NPA ratios for credit cards and for all unsecured retail loans, %.

Sources: CRIF High Mark MicroLend reports (May 2026 and August 2026 editions for the loan book and overdue share; June 2024 and September 2024 editions for borrowers before September 2024). CRIF revises its history, so we use the latest vintage for each figure. RBI Financial Stability Reports, December 2021 to June 2026 (scheduled commercial banks, supervisory returns). The RBI didn't publish an unsecured retail GNPA before September 2022, and we left out one reading whose date the report doesn't state. Delinquency data built from credit-bureau records (TransUnion CIBIL, as cited in the FSR) run lower. We don't mix the two sources in one line.

This is the section where the strain reading has to give ground. Repayment stress did spike in microfinance, where the share of the book 31–180 days overdue went from 2.1% in March 2024 to 6.4% in December 2024. It has since come down to 1.6% by June 2026. Bank credit-card NPAs have stayed around 2%, and unsecured retail NPAs at 1.5–2.0%. There has been no general default wave.

How the microfinance stress got resolved is the part that matters. The loan book shrank from ₹4.43 lakh crore to ₹3.33 lakh crore, and the number of borrowers fell from 8.7 crore to 6.6 crore. The overdue ratio improved partly because 2.1 crore people, about a quarter of the borrower base, no longer have a live microfinance loan. Write-offs also played a part. Among loans disbursed in the previous 36 months, those more than 180 days overdue or written off reached 17.3% in December 2025. Lenders tightened the rules on how many loans a borrower could hold, and demand moved to other lenders. Chart 5 shows where some of it went.

Two readings

The repair reading

  • Net financial savings recovered to 7.1% of GDP in 2024-25, which is inside the 2010s range. On the latest numbers the "50-year low" is over.
  • Borrowing slowed because the regulator wanted it to, and that is prudential policy working.
  • The RBI says credit quality is improving. Housing NPAs were 0.5% in March 2026, down from 1.2% in 2019, and the share of prime-and-above borrowers is rising.
  • At 45–49% of GDP, Indian household debt is below Thailand (87%), Malaysia (70%) and China (59%).
  • Gold loans are formal credit replacing moneylenders, and higher gold prices mean each loan carries more collateral.
  • There has been no default wave. Microfinance loans 31–180 days overdue fell from 6.4% to 1.6% of the portfolio, and bank card NPAs are around 2%.
  • Some of the lost financial saving went into houses: physical savings rose to about 14% of GDP.

The strain reading

  • The recovery came from borrowing less, not saving more. Gross financial saving is still around 12% of GDP, against 14–18% in the mid-2000s, when borrowing last ran this high.
  • Borrowing slowed because of a regulatory brake, not because incomes rose. Demand then moved to the collateral households already had, their gold.
  • The debt stock kept rising through the recovery. The RBI's June 2026 report says nearly half of household borrowing is for consumption, and non-housing retail loans are 58.4% of the total.
  • Healthy bank balance sheets and fragile household ones can both be true. The risk has moved to households rather than disappeared.
  • Even after adjusting for the gold price, the gold pledged to banks roughly doubled in two years.
  • Microfinance stress was resolved by cutting borrowers off. 2.1 crore fewer people have a live loan, and write-offs rose. The overdue ratio improved partly because those borrowers are no longer counted.
  • The shift into houses stopped in 2022-23. Physical savings have been flat since then, so they don't explain the post-2023 borrowing.
What this asset doesn't claim: a household debt crisis, or a banking one. Bank asset quality is good. What it does claim is narrower: the savings recovery is a recovery in flows, and it happened while the debt stock and the amount of gold pledged kept rising. That is a household balance sheet under strain, and it is not the same thing as a household balance sheet that has healed.
Notes on method
  1. All ratios are computed from rupee values in RBI Handbook tables, with nothing typed in by hand. The 2004-05 base segment (1951-52 to 2011-12) comes from the 2012-13 Handbook. The 2011-12 base segment (2011-12 to 2023-24) uses the NAS 2025 household flows in the 2025-26 Handbook. For GDP, each year uses the latest vintage available in the Handbook editions we read.
  2. The 2004-05 and 2011-12 bases overlap at 2011-12 (8.0% against 7.4%). The 2011-12 and 2022-23 bases overlap at 2022-23 (4.95% against 5.3%) and 2023-24 (5.2% against 5.9%).
  3. Household here means the NAS household sector, which includes quasi-corporates and unincorporated enterprises. It is not only wage-earning families.
  4. The BIS debt ratio uses its own GDP and lender coverage and won't match the RBI's Financial Stability Report exactly. We use BIS for the quarterly history and cite the FSR where it adds something.