Debt to GDP, 2014 to 2031
IMF WEO general government gross debt. Outturn solid, projection dashed.
General government gross debt
Statutory anchor, 55% present value
Projection, not outturn
Denominator instability
Three official nominal GDP figures are in circulation for overlapping periods: KES 18,993bn, 19,273bn and 20,817bn. Treasury's monthly ratio moves when the GDP base is revised, not only when debt changes. Between March and June 2026 the stock rose from KES 12,833bn to 13,013bn while the published ratio fell from 70.2% to 68.5%. Every ratio on this page stores its denominator.
Threshold monitor
MTDS 2026 sustainability benchmarksExternal debt is KES 5,684.69bn, 43.7% of the total stock. Multilateral creditors hold the majority. Over the year to June 2026 the composition moved in one direction: bilateral and Chinese exposure down, Eurobonds up. That is the arithmetic signature of the liability management programme.
External debt by creditor class
30 June 2026 · KES 5,684.69bnCurrency composition of external debt
Bulletin FX rates: USD 129.50, EUR 147.51, GBP 171.33, CNY 19.06
June 2026
March 2026
June 2025
Anomaly flagged, unresolved
The CNY share moves from 5.0% to 11.6% across three quarters. No portfolio shift of that size occurred: China's share of external debt by creditor was 11.9% in June 2025 and 10.9% in June 2026, stable throughout. The June 2025 currency row is almost certainly a reclassification or transcription error, not an economic event. The platform holds both values and refuses to plot a trend through them. A human scorer must resolve this with the Public Debt Management Office before any currency-risk claim is released.
Three liability management operations in nineteen months rebuilt Kenya's external redemption profile. The cliff that dominated commentary through 2024 is largely gone. What replaced it is a higher coupon and a heavier domestic redemption load.
Eurobond principal falling due
USD m · reconstructed from issue and tender reporting
Reconciliation failure — do not publish per-series figures
This ladder is built bottom-up from issue prospectuses and tender results because the Treasury does not publish a security-level Eurobond register in its monthly bulletins. It totals USD 10.23bn against the June 2026 bulletin's international sovereign bond line of USD 10.51bn. The gap is USD 286m, or 2.7%. Candidate explanations: the bulletin line may include the sukuk or accrued interest, or one original-size figure is wrong. Unreconciled. Evidence grade capped at 3.
Outstanding Eurobond series
KENINT · after the February 2026 tenderWhere the pressure went
The finding a redemption chart alone would hide
Domestic interest alone, KES 862.65bn, now exceeds total external debt service, principal and interest, of KES 716.10bn. Those two are not on the same definition, and total domestic debt service is larger still: FY2026/27 budgets a further KES 648.8bn of domestic redemptions. Stated on matched definitions the gap is wider, not narrower. External service came in 22% under budget as the buybacks reshaped redemptions. The binding constraint has migrated from refinancing risk to interest cost and domestic market absorption. A debt strategy still written around the Eurobond wall is answering last year's question.
A public-data replication of the standard debt dynamics decomposition. It is not the official Debt Sustainability Framework and does not reproduce it. Its purpose is to show what a live, auditable scenario surface looks like when the assumptions are visible and the back-test is published.
Scenario controls
Shocks apply from 2027 and persist
Preset scenarios
Debt to GDP under scenario
Model output, machine-indicative, unsigned
Baseline
Scenario
55% anchor
Reconciliation finding — the most important output on this page
Run the engine on Kenya's own stated assumptions and no residual, and debt falls to somewhere between 53% and 58% of GDP by 2032. The IMF's World Economic Outlook projects it rising to 75.1% by 2031. Both cannot be true.
Closing the gap needs a stock-flow residual of between 3.1% and 4.2% of GDP every year, KES 640bn to 880bn at the FY2026/27 base. That is debt arriving without passing through the deficit: arrears, below-the-line operations, valuation effects, guarantees called.
The range is the second finding. It moves on two choices nobody documents: whether the projection opens on the Treasury's publicly guaranteed ratio of 68.5% or the IMF's general government ratio of 71.6%, and whether the domestic and external effective interest rates are drawn from the same fiscal year. The slider defaults to 4.2%, the widest of the four specifications. Either the medium-term framework will not be delivered as printed, or a sum of this order lands on the stock each year from outside the budget. No published Kenyan document states which, and none states which specification it is using.
Closing the gap needs a stock-flow residual of between 3.1% and 4.2% of GDP every year, KES 640bn to 880bn at the FY2026/27 base. That is debt arriving without passing through the deficit: arrears, below-the-line operations, valuation effects, guarantees called.
The range is the second finding. It moves on two choices nobody documents: whether the projection opens on the Treasury's publicly guaranteed ratio of 68.5% or the IMF's general government ratio of 71.6%, and whether the domestic and external effective interest rates are drawn from the same fiscal year. The slider defaults to 4.2%, the widest of the four specifications. Either the medium-term framework will not be delivered as printed, or a sum of this order lands on the stock each year from outside the budget. No published Kenyan document states which, and none states which specification it is using.
What moves the ratio
Contribution to the 2027 change, percentage points of GDPEngine plausibility check
Not a back-test. Read the note below before citing it.
What this is, and what it is not
The engine reproduces interest at 41.6% of ordinary revenue against a published Treasury figure of 40.8%. That is not a back-test. The model input is the June 2026 debt stock with an FY2025/26 domestic interest outturn, an FY2026/27 budgeted foreign interest figure and an FY2026/27 revenue share; the published comparator is the FY2024/25 outturn. Nothing in the calculation is on FY2024/25 vintage, so the 0.8 point agreement is coincidental and carries no validation weight. It is a plausibility check on the order of magnitude, nothing more. A real back-test needs the FY2024/25 stock, interest and revenue, and is a task for the pilot.
Generate a decision brief
Question in, evidence-linked brief out
Policy question
Uses the current stress laboratory scenario
Decision brief
Every figure in this prototype traces to a named document with a publication date and a data vintage. Nothing is asserted from model memory. Where sources disagree, both values are carried and neither is averaged.
Source register
ESI grade 1 strongest, 5 weakestKnown gaps and unresolved conflicts
What a production build must obtain directlyGovernance state
SSII AI Use Protocol D.1
What this prototype is not
No output here has been scored, reviewed or signed. Under the AI Use Protocol, machine analysis may assemble evidence, triangulate sources and set out two-sided rationales. It may not assign a final rating, set a coverage flag, or release into a public product. Two named human scorers, reconciliation, right of reply and an eighteen-item quality check stand between this surface and anything a Cabinet paper may cite.