30Who Owes Whom
Global debt is roughly $325 trillion, about three times world GDP. But the total is almost meaningless. What matters is who owes, in what currency, to whom, and for how long.
Start with a clarification that removes most of the confusion in public debate. "National debt" mixes together four completely different things:
Table 30.1 — Four kinds of debt, and which ones are dangerous
| Type | Owed by | Owed to | Danger level |
|---|---|---|---|
| Domestic government debt in own currency | Government | Its own citizens and banks | Low. Cannot be forced to default; worst case is inflation. Japan at 250% of GDP. |
| External government debt in own currency | Government | Foreigners | Low-medium. Only the US, and to a lesser degree Japan and the eurozone, can do this at scale. |
| External government debt in foreign currency | Government | Foreigners | HIGH. This is what kills countries. Sri Lanka, Argentina, Zambia, Ghana, Pakistan. |
| Private external debt | Companies and banks | Foreigners | High. Becomes public in a crisis, because governments rescue banks. Ireland 2008, Korea 1997. |
Ask one thing: can the borrower create the currency it owes?
If yes, it can always technically pay. Japan owes yen and prints yen. The United States owes dollars and prints dollars. India's domestic debt is in rupees and India prints rupees. Default is a political choice, not an arithmetic necessity, and the real risk is inflation, not bankruptcy.
If no, the borrower must earn the currency through exports, tourism, remittances or new borrowing. When those dry up, it defaults — regardless of how large or small the debt looks relative to GDP. Sri Lanka defaulted with external debt around 60 per cent of GDP. Japan has not defaulted at 250 per cent. The ratio was never the point.
Economists call the inability of most countries to borrow abroad in their own currency "original sin". It is the single most important structural inequality in the world financial system, and it is why the dollar question matters far more to Colombo and Accra than to Tokyo.
The global picture
Table 30.2 — Government debt-to-GDP, selected countries, 2026
| Country | Gross govt debt / GDP | Share held by foreigners | Currency of debt | Risk |
|---|---|---|---|---|
| Japan | ~250% | ~14% | Yen | Low — owes its own money to its own people |
| United States | ~124% | ~24% | Dollar | Low — issues the world's reserve currency |
| Italy | ~135% | ~30% | Euro — cannot print it | Medium-high — the eurozone's structural weak point |
| France | ~115% | ~50% | Euro — cannot print it | Medium |
| China | ~90% (plus large local-government liabilities) | ~3% | Yuan | Low externally, high domestically |
| India | ~81% | ~5% | Mostly rupees | Low — see Chapter 37 |
| Brazil | ~88% | ~10% | Mostly real | Medium |
| Pakistan | ~70% | ~35% | Substantially foreign currency | High |
| Sri Lanka | ~100% | ~40% | Substantially foreign currency | Defaulted 2022 |
| Argentina | ~85% | high | Substantially dollars | Defaulted 9 times since 1816 |
Notice that debt-to-GDP does not predict crisis. Japan is triple Sri Lanka's ratio and safe. The currency and the ownership are what matter. Notice that debt-to-GDP does not predict crisis. Japan is triple Sri Lanka's ratio and safe. The currency and the ownership are what matter.
India's public debt is around 81 per cent of GDP — higher than most emerging peers, and a genuine constraint on fiscal space. But roughly 95 per cent of it is denominated in rupees and held domestically, largely by Indian banks (compelled by the SLR), insurance companies and provident funds.
This means India is structurally insulated from the crisis that destroyed Sri Lanka in 2022. India cannot be forced into external default by its own budget deficit. The 1991 crisis was a balance-of-payments crisis — not enough dollars to pay for imports — and India's $703 billion of reserves now covers roughly ten to eleven months of imports against three weeks in 1991.
This is one of the genuinely strong cards in India's hand, and it is rarely mentioned. Chapter 38 returns to it.
31Who Lends to the United States
America owes about $39.9 trillion. Roughly a quarter of it is owed to foreigners. The list of who holds it, and what they can actually do about it, is one of the most misunderstood subjects in geopolitics.
Figure 7. Foreign holders of US Treasury securities, 2026. Japan remains the largest at about $1.2 trillion, the UK second at roughly $897 billion — much of it custodial holdings for other owners — and China third at around $712 billion, down from a peak above $1.3 trillion in 2013. Note the presence of the Cayman Islands, Luxembourg, Belgium and Ireland: these are booking centres for hedge funds and custodians, not sovereign wealth. Source: US Treasury TIC data.
The composition of the $39.9 trillion
Table 31.1 — Who owns US federal debt, 2026 (approximate)
| Holder | Approx. share | Notes |
|---|---|---|
| US government trust funds (Social Security, Medicare) | ~18% | "Intragovernmental" — the government owing itself |
| Federal Reserve | ~11% | Acquired through quantitative easing |
| US mutual funds, pensions, banks, insurers | ~32% | American savers |
| US households and state/local government | ~15% | |
| Foreign official and private holders | ~24% | ~$9.35 trillion, March 2026 |
Claim: "China could destroy the US economy by dumping its Treasury holdings."
Reality: Four problems with this argument.
1. The size is wrong. China holds roughly $712 billion — about 1.8 per cent of total US federal debt and about 7.4 per cent of foreign-held debt. Japan holds substantially more. The Federal Reserve alone holds around six times as much.
2. Selling hurts the seller. Dumping $700 billion of bonds would crash their price, so China would take a large loss on the bonds it had not yet sold. It would also have to buy something with the proceeds — and the only markets deep enough are euros and yen, whose currencies would soar, damaging China's export competitiveness far more than it would damage America.
3. The Fed can absorb it. The Federal Reserve bought over $4 trillion of bonds during 2020–22 without difficulty. It could buy China's entire holding.
4. China has been reducing its holdings for a decade anyway — from over $1.3 trillion in 2013 to around $712 billion — and nothing happened.
The correct framing is the reverse. A creditor holding an asset it cannot sell without harming itself is not a creditor with leverage. It is a hostage. This is what Keynes meant: "If you owe your bank a hundred pounds, you have a problem. If you owe your bank a million, the bank has a problem." China holds American IOUs; America holds Chinese exposure. On the evidence of 2022, the more relevant risk to China is that the assets could be frozen, not that selling them would be effective.
India's position
India holds roughly $240 billion in US Treasuries — part of the $703 billion in reserves, with the remainder in other currencies, deposits with the BIS and other central banks, IMF positions, and gold. India's gold holdings have risen above 880 tonnes, with over 100 tonnes repatriated from London vaults since 2022.
India is therefore in the same structural position as every other reserve-holding country: it lends to the United States at low yield in exchange for safety and liquidity, and in doing so accepts exposure to American political decisions. There is currently no alternative that offers the same combination of depth and liquidity, which is the whole point of Part III.
32The IMF, the World Bank and the Price of Rescue
When a country runs out of foreign currency, there is one institution it can turn to. The terms it offers have shaped the development path of most of the world, and the argument about whether they help or harm is eighty years old and unresolved.
What the IMF actually does
The Fund lends foreign currency to countries facing balance-of-payments crises — that is, countries that cannot pay for imports or service foreign debt. It is not a development bank; the World Bank does long-term project lending. The IMF does emergency liquidity.
The lending comes with conditionality: a programme of policy changes the borrower must implement. The standard package, sometimes called the Washington Consensus:
- Cut the fiscal deficit — usually reducing subsidies and public sector wages
- Raise interest rates to defend the currency and curb inflation
- Devalue the currency to restore competitiveness
- Liberalise trade and remove import restrictions
- Privatise state-owned enterprises
- Deregulate markets and open to foreign investment
The IMF's case: a country in this position has run out of money. Someone must lend, and no private lender will. The conditions exist because the policies that caused the crisis — usually spending more foreign currency than the country earns — must change, or the loan simply funds the same behaviour and the crisis returns. A doctor who prescribes rest to a patient with a broken leg is not the cause of the break.
The critics' case: the conditions are contractionary at exactly the wrong moment. Cutting spending and raising rates in a crisis deepens the recession, throws people out of work, and cuts health and education budgets that harm a generation. The IMF's own Independent Evaluation Office has conceded that fiscal targets in several programmes were too tight. And the governance is indefensible: voting power is proportional to financial contribution, the United States holds an effective veto through its ~16.5 per cent share, and by convention the Managing Director is always European while the World Bank President is always American.
The honest position: both are substantially right. The IMF is usually called in far too late, when the options are genuinely terrible, and is then blamed for the fact that the options are terrible. It has also, repeatedly, imposed austerity beyond what was necessary and been slow to admit it — Greece being the most documented case. Neither observation cancels the other.
Table 32.1 — Selected IMF programmes and outcomes
| Country | Year | Size | Outcome |
|---|---|---|---|
| South Korea | 1997 | $57bn | Harsh terms, deep recession, then rapid recovery. Repaid early in 2001. Korea has never returned — and now holds huge precautionary reserves |
| India | 1991 | $2.2bn | Triggered liberalisation. Repaid by 1993. Widely seen in India as the crisis that forced overdue reform |
| Argentina | 2018 | $57bn | Largest programme in IMF history. Failed. Argentina defaulted again in 2020 |
| Greece | 2010–18 €289bn total | Economy shrank ~25%. The IMF later acknowledged it had underestimated the contractionary effects | |
| Sri Lanka | 2023 | $2.9bn | Followed the 2022 default. Stabilised, at the cost of a severe contraction |
| Pakistan | 2019, 2023, $6bn, $3bn, 2024 | $7bn | 24th programme since 1958. The pattern is the diagnosis |
India's 1991 programme is unusual in being remembered domestically as a turning point rather than a humiliation — though it was both.
The facts: reserves fell to about $1.2 billion, enough for roughly three weeks of imports. India airlifted 47 tonnes of gold to the Bank of England and Union Bank of Switzerland as collateral. The IMF lent $2.2 billion. The rupee was devalued twice in three days by a combined 18–19 per cent. Licensing was dismantled, tariffs slashed from an average above 80 per cent, and foreign investment opened.
What made the difference from Argentina or Pakistan was that India used the crisis to change direction and then kept going after the money was repaid. Growth averaged over 6 per cent for the next two decades. Reserves went from $1.2 billion to $703 billion.
The lesson is not "IMF programmes work". It is narrower: conditionality is only ever as useful as the borrower's own decision to reform. Countries that reform because they have decided to, recover. Countries that reform because they were made to, return.
33China as the World's New Creditor
In twenty years China went from a borrower to the largest bilateral creditor on earth, lending over $1 trillion to more than 100 countries. It is now the world's biggest debt collector, and it is discovering what that involves.
The Belt and Road Initiative, launched in 2013, financed ports, railways, power stations, highways and pipelines across Asia, Africa, Latin America and parts of Europe. The scale is without precedent for a single national lender: cumulative BRI-related lending is estimated above $1 trillion, exceeding the World Bank's development lending over the same period.
How Chinese lending differs
Table 33.1 — Chinese loans versus multilateral loans
| IMF / World Bank | Chinese policy banks |
|---|---|
| Typical interest ~2% or concessional rate | ~5% (rescue lending), often 4–6% for projects |
| Maturity Long, often 20–40 years | Shorter, typically 10–20 years |
| Conditions Explicit policy conditionality | No policy conditions — but commercial terms, Chinese contractors, sometimes Chinese labour |
| Transparency Published | Frequently confidential; non-disclosure clauses are common |
| Collateral Rare | Common — resource-backed, escrow accounts, revenue assignments |
| Restructuring Through the Paris Club, collectively | Bilateral, case by case; China is not a Paris Club member |
A developing country wants a port. The World Bank offers cheap money but demands governance reforms, environmental studies and competitive tendering, and takes seven years. China offers money in eighteen months, no questions about your politics, and builds it.
If you are a government facing an election in three years, the choice is not difficult. The problem arrives later, when the port does not generate the revenue projected and the repayment is due in dollars.
Claim: "China deliberately lends unpayable sums in order to seize strategic assets when countries default."
The evidence is more mixed than either side admits.
Against the thesis: The most-cited example, Hambantota port in Sri Lanka, is more complicated than usually reported. Detailed research by Deborah Brautigam and others found that Sri Lanka requested the project, that Sri Lanka's debt distress was driven mainly by international sovereign bonds rather than Chinese loans, and that the 2017 lease was a decision by Colombo to raise dollars, not a seizure. China has also written down or restructured a substantial number of loans quietly. And China has now become a large rescue lender — over $240 billion in bailouts to 22 countries between 2008 and 2021 — which is not the behaviour of a creditor hoping for default.
For the thesis, or at least for serious concern: the opacity is real and deliberate; non-disclosure clauses prevent other creditors from assessing risk. Collateralisation and escrow arrangements do give China seniority over other lenders. The World Bank and IMF have identified eight BRI recipients — Djibouti, Kyrgyzstan, Laos, Maldives, Mongolia, Montenegro, Pakistan and Tajikistan — at high risk of distress substantially connected to BRI borrowing. And China's reluctance to take losses alongside other creditors has materially slowed restructurings in Zambia, Sri Lanka and Ghana.
A fair summary: the evidence supports "commercially aggressive, strategically opportunistic, and institutionally immature as a creditor" more than it supports "deliberate trap". The distinction matters, because the policy responses are different.
What this means for the dollar
Here is the twist that is usually missed. Most Chinese lending is denominated in US dollars. China lent dollars because that is what borrowers needed and what commodity revenues were priced in. So a Zambian copper mine financed by a Chinese bank generates dollar revenue to repay a dollar loan to a Chinese lender.
China has therefore spent a decade extending the dollar's reach while trying to reduce its own exposure to it. Beijing is now actively pushing to redenominate new BRI lending in renminbi, and the New Development Bank has targeted 30 per cent local-currency lending by 2026. That shift, if it succeeds at scale, would be one of the more consequential changes in the system — because it would create genuine offshore demand for yuan that is not driven by sanctions avoidance.
34Four Debt Crises, One Pattern
Sri Lanka, Pakistan, Zambia and Argentina. Different continents, different politics, different creditors — and a sequence of events so similar it can be written as a single script.
The script. (1) Cheap foreign money becomes available, usually when global interest rates are low. (2) The country borrows in dollars for projects, or to cover a budget gap. (3) Something reduces foreign earnings — a commodity price fall, a tourism collapse, a policy error. (4) Global rates rise, the dollar strengthens, and the debt becomes more expensive in local currency at exactly the moment earnings fall. (5) Reserves drain defending the currency. (6) Fuel, food and medicine imports stop. (7) Default. (8) IMF programme, austerity, devaluation, contraction. (9) Slow recovery, and the beginning of the next cycle.
Sri Lanka, 2022
The clearest recent case and the most instructive for India, because it happened next door.
Sri Lanka borrowed heavily through international sovereign bonds during the low-rate 2010s — these, not Chinese loans, were the largest single component of its external debt. It then made three compounding errors: deep tax cuts in 2019 that removed roughly a third of government revenue; a sudden overnight ban on chemical fertiliser in 2021 that cut rice yields by around 20 per cent and destroyed tea export earnings; and a defence of an overvalued rupee that burned through reserves. COVID had already eliminated tourism, worth about $4.4 billion a year.
By April 2022 usable reserves were under $50 million. The country could not buy fuel. Power cuts ran to thirteen hours a day. Inflation exceeded 70 per cent. In July, protesters occupied the presidential palace and the president fled the country. Sri Lanka defaulted on about $51 billion of external debt — its first default since independence in 1948.
India extended roughly $4 billion in credit lines, currency swaps and deferred payments to Sri Lanka during 2022 — more than the IMF's initial programme and more than any other country. Part of this was settled in rupees.
Read that carefully, because it is a template. A financially stronger neighbour provided emergency liquidity, some of it in its own currency, at a moment when no one else would. That is precisely how a currency starts becoming regionally useful — not through summits, but through being the money that was actually available when it was needed. Chapter 40 argues this is India's single most realistic path.
Pakistan
Twenty-four IMF programmes since 1958 — among the highest in the world. The structural problem is chronic: a tax-to-GDP ratio persistently around 9–10 per cent, one of the lowest for any economy of its size; a large defence budget; energy sector circular debt; and export earnings concentrated in low-value textiles. External debt is around $130 billion, with substantial obligations to China, Saudi Arabia and the UAE alongside multilateral creditors. In 2023 reserves fell to about three weeks of imports before a $3 billion standby arrangement, followed by a $7 billion programme in 2024.
Pakistan's case demonstrates the limit of the IMF model: repeated liquidity support cannot fix a solvency problem rooted in an inability to collect tax.
Zambia
Africa's first pandemic-era default, in November 2020, on about $17 billion of external debt, of which roughly $6 billion was owed to Chinese lenders. Restructuring took until 2023–24 — more than three years — largely because bondholders, the Paris Club and Chinese lenders could not agree on comparable treatment, with each unwilling to accept a larger loss than the others. Zambian citizens spent three years in an economy with no access to international capital while the creditors argued.
Zambia is the clearest evidence that the international debt restructuring architecture, designed when creditors were a small club of Western governments, does not function now that China, private bondholders and multilaterals must all agree.
Argentina
Nine sovereign defaults since 1816, three since 2001. The 2001 collapse ended a currency board that had pegged the peso 1:1 to the dollar for a decade: when the peg broke, the peso lost about 70 per cent, dollar-denominated savings were forcibly converted at unfavourable rates ("corralito"), and poverty exceeded 50 per cent. The 2018 IMF programme was the largest in the Fund's history at $57 billion, and Argentina defaulted again in 2020.
Argentina is the standing demonstration that no external programme can substitute for domestic political consensus about fiscal discipline. It is also why Argentines hold an estimated $200–250 billion in physical US dollars — more per capita than almost any nation on earth. When a state repeatedly destroys its own money, its citizens adopt someone else's. That is dollarisation from below, and no policy reverses it quickly.
1. Never borrow long in a currency you cannot earn or create. This single rule would have prevented
most sovereign debt crises in history. 2. Reserves are not a luxury. The difference between India in 2026 and Sri Lanka in 2022 is substantially a
reserves buffer measured in months of imports. 3. Defending an overvalued currency destroys reserves and postpones nothing. Every country that
has tried has ended up with the devaluation anyway, minus its reserves. 4. Tax capacity is the foundation of sovereignty. Pakistan's problem is not that it borrows; it is that it
cannot collect. India's tax-to-GDP ratio of roughly 18 per cent (all levels) is adequate but not high for its
stage of development. 5. Crises are political before they are financial. Sri Lanka's tax cut and fertiliser ban were political
decisions, taken quickly, with catastrophic financial consequences.