Ghana PAR Covenant & Loanbook Forecast

PAR30+ is projected to end May at 14.95%, sitting 1.05pp below the 16.0% covenant threshold. PAR90+ projects to 7.05%, 4.95pp clear of the 12.0% covenant. Both covenants are safe. The total loanbook grows modestly from GHS 460.9M to GHS 466.2M as new disbursements build Current, while every delinquent bucket from PAR60 downward contracts. Provision stock releases GHS 2.6M over the same window.

Over the May 16→31 forecast window (15 days), PAR30+ declines from 16.15% to 14.95%, ending 1.05pp below the 16.0% covenant. New disbursements (GHS 186.3M estimated, 92.3% retention) add GHS 11.9M to Current. Every delinquent bucket from PAR60 downward contracts: PAR60 −GHS 1.7M, PAR90 −GHS 2.4M, PAR120 −GHS 816K. PAR30 is essentially flat (+GHS 101K). GHS 4.2M in additional write-offs over the remainder clears the residual PAR150+ balance. Total book ends at GHS 466.2M (+GHS 5.3M, +1.1% from the May 16 opening). Projection is calibrated on a 6-month learning window (Nov 2025–Apr 2026) matching this period's 15-day day-of-month shape.

Layer	Opening (GHS M)	Opening %	Projected (GHS M)	Projected %	Δ (GHS M)	Prov Rate	Opening Prov	Projected Prov
Current	357.8	77.62%	369.7	79.30%	+11.9	1%	3.6	3.7
PAR0	28.7	6.23%	26.8	5.75%	-1.9	1%	0.3	0.3
PAR30	20.7	4.48%	20.8	4.45%	+0.1	20%	4.1	4.2
PAR60	17.8	3.85%	16.1	3.45%	-1.7	40%	7.1	6.4
PAR90	19.2	4.16%	16.8	3.60%	-2.4	60%	11.5	10.1
PAR120	16.9	3.66%	16.1	3.45%	-0.8	80%	13.5	12.9
PAR150+	0.0	0.00%	—	0.00%	-0.0	100%	0.0	—
TOTAL	460.9	100.00%	466.2	100.00%	+5.3		40.1	37.5
PAR30+	74.5	16.15%	69.7	14.95%	-4.8		36.2	33.5
PAR90+	36.0	7.82%	32.8	7.05%	-3.2		25.0	22.9

Methodology: 6-month trailing transition matrix (2025-11-01 to 2026-04-30), bias-corrected via backtest. Disbursement estimate seasonally adjusted with limit-change correction. Retention rate: 92.3% (median of stable months). Write-offs: deterministic via LASTSETTOARREARSDATE. Terminal bucket (PAR150+ / PAR180) projected deterministically — opening + upstream inflows − writeoffs; learned rates not applied to its own rollforward.

Period	Balance (GHS M)
Already executed (MTD)	16.0
Expected additional (to forecast date)	4.2
Total month	20.1

Already executed from DELETED_PRINCIPAL (LAH). Expected additional via LASTSETTOARREARSDATE + threshold. Write-off balance uses PRINCIPAL_BALANCE.

Full-month window (May 1→31). The BOP provision stock at May 1 was GHS 42.1M. As the delinquent stack contracts, the EOP stock falls to GHS 37.5M — a release of GHS 4.7M. Write-offs for the month total GHS 20.1M: GHS 16.0M was executed in the first half (May 1→16, confirmed from LAH actuals) and a further GHS 4.2M is expected May 16→31 based on loans whose arrears date crossed the 150-day write-off threshold. Bad debt recovery on written-off loans is GHS 1.0M for the month. The net Credit Loss P&L charge for May is GHS 14.4M.

Component	Forecast (GHS M GHS)
BOP Provisions	42.1
EOP Provisions	37.5
Δ Provision (EOP − BOP)	-4.7
+ Write-offs	20.1
− Bad Debt Recovery (WO loans)	−1.0
= Credit Loss P&L	14.4

Formula: CLP = EOP − BOP + Write-offs − Bad Debt Recovery. Full-month window. Math close enforced by clp-table.md.

Bridge from MTD actuals (May 1→16, GHS 12.9M) to the full-month forecast (May 1→31, GHS 14.4M). The EOP movement bar (−GHS 4.4M) reflects provision release as the existing book migrates toward performing buckets over May 16→31. New disbursement impact (+GHS 1.7M) is the net provision charge on new loans entering during the remainder, provisioned at 1% for Current. Write-offs remainder (+GHS 4.2M) is the additional credit charge from loans written off after May 16. Recovery remainder is less than GHS 1K. The net step from MTD to full-month is +GHS 1.5M, driven by remaining write-offs and new disbursement provisioning, partially offset by the provision release from improved book composition.

26 modified loans (refi or proactive restructure) are on the book at May 16, with a combined balance of GHS 50,150 — 0.01% of total. All sit in performing or near-performing buckets: 13 loans (GHS 22K) in Current, 6 loans (GHS 19K) in PAR0, 6 loans (GHS 5K) in PAR60, and 1 loan (GHS 3K) in PAR90. The pool is too small to constitute an intervention candidate under any scenario. No early write-off action is warranted, and the modified cohort has no material influence on the PAR30+ projection or the covenant distance.

PAR Bucket	Loans	Balance (GHS M)
Current	13	0.0
PAR0	6	0.0
PAR60	6	0.0
PAR90	1	0.0
Total	26	0.1

Modified = refinanced (ORIGINAL_LOAN IS NOT NULL) OR proactive restructure (offer ≤ original due date). Single metric — no sub-categories. Point-in-time classification at the snapshot date. Excludes postponements and reactive restructures.

4-period backtest (Jan–Apr 2026): total book median error −0.5%, range −4.9% to +2.9% — strong calibration. PAR30+ median +0.82% — model marginally overstates delinquency, the preferred direction for covenants. Mid-delinquency layers have the highest dispersion: PAR60 median +9.1% (−0.1% to +19.8%), PAR30 median +4.8%. Current and PAR120 track at −1.5%. PAR150+ shows −100% error across all periods — terminal bucket is near-zero at each backtest snapshot while actuals averaged GHS 890K; residual arrivals from upstream are not fully captured. Immaterial for the current forecast (PAR150+ opening: GHS 1,790).

Layer	Avg Actual (GHS M)	Months	Median Error %	Range (min / max)
Current	358.5	4	-1.5%	-6.3% / +4.8%
PAR0	30.0	4	-4.7%	-11.9% / +29.0%
PAR30	20.3	4	+4.8%	-1.2% / +23.7%
PAR60	17.2	4	+9.1%	-0.1% / +19.8%
PAR90	17.8	4	-4.0%	-11.4% / +1.9%
PAR120	18.3	4	-1.5%	-14.1% / +5.5%
PAR150+	0.9	4	-100.0%	-100.0% / -100.0%
PAR30+	74.5	4	+0.8%	-2.4% / +3.7%
TOTAL	463.0	4	-0.5%	-4.9% / +2.9%

Direction-aware coloring: aqua = conservative (safe), cherry = wrong direction or excessive magnitude, neutral = small/mixed. Terminal layer (PAR150+ / PAR180) is forecast deterministically — its error is expected to be near zero; large deviations indicate a modeling bug.

Ghana PAR Covenant & Loanbook Forecast — May 31, 2026

Structural Diagnosis