Legally a company. Fiscally, 90% of GDP of local debt.
An IMF departmental paper published on 25 September: official local-government debt plus the debt of local government financing vehicles reached 90% of GDP in 2025. The vehicles were built because local governments could not borrow directly — they are legally state-owned enterprises, and the IMF counts their debt in its augmented definition of general government.
The company label moved the borrowing off the budget. It did not move the repayment: the interest comes back to local budgets, and past rescues have already swapped RMB 14 trillion of vehicle debt into official debt, with another RMB 10 trillion of swap quota running to 2028.
The deficit fell 3.5 points. The one-off dividends were worth about 4.
The 2026 Article IV, published on 21 September: the overall deficit narrowed from 14% of GDP in 2024 to 10.5% in 2025, “supported by large one-off dividend payments from SOEs and the Bank of Algeria of about 4 percent of GDP.” The stabilisation fund was exhausted in 2024; central-government debt reached 52.1% of GDP.
And the loop closes through the banks: six state-owned banks hold close to 90% of loans and deposits, and their credit is concentrated on the government and the same state companies. A dividend paid up today can be a loan drawn down tomorrow — the money moves; the public sector’s position does not.
Ten years of unpaid electricity, now a sovereign-guaranteed bond.
The state bulk electricity trader, NBET, stopped paying the generators in full. The bills ran from February 2015 to March 2025. They are now being settled through a ₦4 trillion bond programme issued by an NBET vehicle and “guaranteed by the full faith and credit of the Federal Government”.
This is the programme working: generators get paid and the sector gets liquidity. It is also a decade of arrears becoming explicit sovereign debt — a liability that existed the whole time, but only as unpaid invoices inside a state company, where no debt statistic was looking.
The debt sits in the company. The interest is a line in the budget.
Chile’s 2026 Budget Law sets aside CLP 548,000 million for the state railway, EFE, and the Santiago Metro. The glosas say why, in the law’s own words: EFE’s transfer pays “the interest on the domestic and external debt that the Company cannot cover with its available cash”; Metro’s covers costs it “cannot meet from its operating cash flow”.
The debt stays on the companies’ books, outside the central-government anchor. Only part of it carries an explicit guarantee — about 0.7% of GDP; the rest is serviced exactly the same way, from the Treasury, year after year. The 2027 budget reaches Congress on Wednesday with the same lines.
Ratings read the stock. The flow barely registers.
A new IMF working paper by Olivier Blanchard, Daniel Leigh and Prachi Mishra asks how much weight sovereign ratings put on debt versus forecast primary balances. In their model, a point of primary balance should be worth about fourteen points of debt. In the actual ratings of emerging markets it is worth 0.7 at S&P, 0.5 at Fitch, 0.1 at Moody’s.
Read the four pages before this one against that finding. A dividend that flatters one year’s deficit moves the flow the market barely reads. The vehicle debt and the refinanced arrears move the stock — which is exactly the number that sits outside the headline.
The money moves both ways. The headline sees one.
Borrowed through a vehicle in China, paid up as dividends in Algeria, parked as unpaid bills in Nigeria, paid from the budget in Chile. Four versions of the same flow between a State and its companies — and in every case the headline caught the flow and missed the stock:
Thursday, in English: The Anchor and the Perimeter — what the same rule says when it is run on the whole balance sheet.