EVIDENCE FILE · SOVEREIGN · AS OF 2025-12-31
The evidence behind Hungary
Every figure below carries the authority that published it and the date it was observed. Nothing on this page is typed; it is read out of the kernel when you load it, which is why it cannot say something the instrument does not.
The assertion
What binds first, when it binds, and what happens at zero. A breach is stated as a breach rather than as a deadline of zero days: they are different facts and reading one as the other is how a country already past a threshold gets filed behind one approaching it.
BINDING CONSTRAINT
106days
LIQUIDITY · binds on 2027-01-28
Already above the 60% of GDP reference value and rising: fiscal room is exhausted, and consolidation is assessed rather than chosen.
EVIDENCE: STRUCTURAL — derived from published figures by an accounting identity
ALREADY PAST A THRESHOLD
- RESPONSE
Every figure it rests on
11 inputs across 4 primitives. The evidence tier is not a confidence score — it says how the figure came to exist, and the weakest tier in a chain sets the tier of everything downstream of it.
| PRIMITIVE | INPUT | VALUE | SOURCE | OBSERVED | TIER |
|---|---|---|---|---|---|
| LIQUIDITY | reserveImportCover | 3.5months | worldbank:FI.RES.TOTL.MO | 2025-12-31 | STRUCTURAL |
| REFINANCING | debt | 77.7pct_gdp | eurostat:gov_10q_ggdebt/GD | 2026-03-31 | STRUCTURAL |
| REFINANCING | shortTerm | 5.4pct_gdp | eurostat:F31+F41 | 2026-03-31 | STRUCTURAL |
| REFINANCING | balance | -4.7pct_gdp | eurostat:gov_10dd_edpt1/B9 | 2025-12-31 | STRUCTURAL |
| REFINANCING | interestPaid | 3.8pct_gdp | eurostat:gov_10dd_edpt1/D41PAY | 2025-12-31 | STRUCTURAL |
| REFINANCING | marketYield | 5.69pct | ecb:IRS/M.HU.L.L40.CI.0000.HUF.N.Z | 2026-09-30 | STRUCTURAL |
| DEPENDENCIES | foreignBankClaimsUsdMillions | 83,834.63USD mn | bis:WS_CBS_PUB/Q.S..4R.U.C.A.A.TO1.A.HU | 2026-03-31 | IDENTIFIED |
| DEPENDENCIES | effectiveCreditors | 3.44count | bis:WS_CBS_PUB/Q.S..4R.U.C.A.A.TO1.A.HU | 2026-03-31 | IDENTIFIED |
| DEPENDENCIES | largestCreditorShare | 44.11pct | bis:WS_CBS_PUB/Q.S..4R.U.C.A.A.TO1.A.HU | 2026-03-31 | IDENTIFIED |
| RESPONSE | debt | 77.7pct_gdp | eurostat:gov_10q_ggdebt/GD | 2026-03-31 | STRUCTURAL |
| RESPONSE | annualChange | 2.4pp_per_year | eurostat:gov_10q_ggdebt/GD 2025-Q1→2026-Q1 | 2026-03-31 | STRUCTURAL |
- STRUCTURAL — derived from published figures by an accounting identity
- IDENTIFIED — read directly from the publishing authority
What would prove it wrong
A claim that cannot be wrong cannot be evidence. This one names the date it settles, the source that settles it, and the observation that would refute it — fixed before the outcome exists, not chosen after.
Hungary exhausts its liquid resources on or before 2027-01-28 (106.4 days) unless new resources arrive. Consequence: Stress-adjusted reserve cover falls below the three-month adequacy floor. If the buffer keeps draining, external payments or the exchange rate must absorb the adjustment.
- SETTLES ON
- 2027-01-28
- SETTLED BY
- eurostat:gov_10dd_edpt1
- REFUTED IF
- the buffer behind LIQUIDITY is observed at or above its 2025-12-31 level on 2027-01-28, measured from eurostat:gov_10dd_edpt1
What is not measured, and why
Three different absences, kept apart. A primitive nobody has published, a primitive that was measured and does not count down, and a primitive comfortable enough that no date exists are three states — and reporting them as one is how a gap gets read as safety.
NOTHING PUBLISHED
none
No authority publishes the figure. Absent, not assumed comfortable.
MEASURED, NOT A DURATION
- DEPENDENCIES
Read and real, but not a countdown. There is no rate at which a concentration drains, and forcing one would need a rollover assumption nobody publishes.
MEASURED, NOT BINDING
none
Measured and comfortable on the current trend, so no date exists.
Who is positioned to reach it
Foreign bank claims on this government, from the BIS consolidated banking statistics. The share is of the published all-reporters total, so these do not add to one and are not meant to: the remainder is the systems that report only in aggregate.
- 35.7%Austrian banking system$18.7bn · 2026-03-31 · IDENTIFIED
- 26.6%Belgian banking system$13.9bn · 2026-03-31 · IDENTIFIED
- 5.2%United States banking system$2.7bn · 2026-03-31 · IDENTIFIED
- 2.4%German banking system$1.3bn · 2026-03-31 · IDENTIFIED
- 1.8%Japanese banking system$1.0bn · 2026-03-31 · IDENTIFIED
- 1.8%French banking system$1.0bn · 2026-03-31 · IDENTIFIED
Do not take the sources on trust — do the work again
Citing a source is not evidence. The number beside the citation is still ours, computed by our code, and checking it means trusting that code. The audit word for what actually settles this is re-performance: fetch the original record yourself, redo the arithmetic, and see whether you land on the same figure.
There is a script in the repository that does exactly that, and the discipline that makes it worth anything is that it shares nothing with the platform: it imports no code from the application, goes to each authority over its own public API with no key and no account, and is not told what to expect. If it reused our fetchers it would re-perform our bug rather than the figure, and would agree with us for the wrong reason.
npx tsx scripts/reperform.mts hungary
Or skip us entirely and open the authorities directly. These are the exact requests the figures in section 02 come from.
- worldbank:FI.RES.TOTL.MO
- eurostat:gov_10q_ggdebt
- eurostat:F31+F41
- eurostat:gov_10dd_edpt1
- ecb:IRS
- bis:WS_CBS_PUB
A disagreement inside the last printed digit is rounding; anything wider is a finding, and the script prints both values rather than telling you the difference is small.
Which code produced this
The last link, and the one that makes the others checkable. The service reports the commit it was built from and a digest computed over the files on disk; clone the repository at that commit, recompute, and the two match or they do not.
- /api/provenanceThe commit, the digest, and the recipe for recomputing it.
- /api/integrityThe internal audit, run every thirty minutes, published whatever it says.
- /api/prime/entity/hungaryThis page as JSON, from the same call that rendered it.
- /mapEverything this platform does, and which of it is load-bearing.
Other entities: romania, italy, germany, france, greece, spain, poland, czechia, bulgaria.