Why ECL Still Fails Audits

ECL fails on staging logic, stale PDs, and missing forward-looking weighting — not math.

Pair with IFRS Explained, IFRS vs GAAP, IFRS Consolidation, and IFRS S1/S2.

Scope and Model Choice

Portfolio Approach Staging
Trade receivables, no financing Simplified, provision matrix No stages, lifetime ECL
Loans, debt securities General 3-stage Stages 1-3
Intercompany loans General, even if repayable on demand Assess SICR carefully
Financial guarantees General Track risk change

Link portfolios to Financial Statement Analysis materiality and Enterprise Risk.

The 3 Stages

  • Stage 1: performing, 12-month ECL, interest on gross.
  • Stage 2: SICR, lifetime ECL, interest on gross.
  • Stage 3: impaired, lifetime ECL, interest on net.

Document SICR triggers: 30 DPD backstop, rating notch moves, watchlist, macro thresholds.

Worked Example: Provision Matrix

Gross AR 10m:

Bucket Balance Loss rate ECL
Current 6.0m 0.3% 18k
1-30 2.0m 1.5% 30k
31-60 1.0m 5% 50k
61-90 0.6m 15% 90k
90+ 0.4m 40% 160k
Total 10m   348k

Calibrate loss rates to history, adjust for forward scenarios (base 50%, upside 20%, downside 30%), discount where material.

For general model: ECL = PD × LGD × EAD, probability-weighted across scenarios.

Forward-Looking Discipline

  • Select 2-4 macro drivers tied to portfolio (unemployment, rates, sector spreads).
  • Weight scenarios, retain source and vintage.
  • Reconcile to Cash-Flow Analysis and Treasury assumptions.

Evidence Binder

  • Staging memo, SICR thresholds, scenario weights.
  • PD/LGD sources, back-testing, management review with precision.
  • Recalculation file with preparer/reviewer sign — see Month-End Close and SOX 404.

Tie AR data to Bookkeeping Controls and Receivables DSO.

Worked Example: Intercompany Loan ECL (The Trap)

Parent lends subsidiary $2m, repayable on demand. “Repayable on demand = no credit risk” is the most common — and wrong — conclusion. Under the general model you still assess: can the subsidiary repay if demanded today from its own resources? If the sub holds $200k cash and burns $150k/month, the shortfall is immediate — ECL equals the cash shortfall over the expected recovery period, and “on demand” shortens, not eliminates, the horizon. Document: sub cash position, burn, forecast support, and any parent support letters (which themselves need probability weighting — history of actual funding vs stated intent). Reviewers challenge demand-loan ECL first because preparers most often shortcut it.

SICR Triggers: Operationalize, Don’t Philosophize

Vague policies (“we consider all relevant information”) fail review. Codify per portfolio:

Trigger Stage 2 entry Evidence
30 days past due (rebuttable presumption) Automatic unless documented rebuttal Aging + rebuttal memo
Rating downgrade 2+ notches (or internal score equivalent) Automatic Rating history file
Watchlist placement / forbearance granted Automatic Committee minutes
Forward-looking threshold breach (e.g., sector spreads +200bps) Case-by-case with memo Market data snapshot
Related-party borrower misses internal plan by >25% Case-by-case Forecast vs actual

Back-test annually: of last year’s Stage 1 population, what share defaulted or migrated within 12 months? Persistent under-migration means thresholds are loose. Tie the back-test to Financial Statement Analysis disclosure (ECL sensitivity is a required disclosure, not a nice-to-have) and to the Audit Committee pack where ECL is material.

Intercompany and Guarantee Edge Cases

  • Cash-pooling arrangements. Netting rights don’t erase gross exposures for ECL — assess each participant’s position, and zero-balance sweeps need same-day evidence or breaks accumulate unnoticed.
  • Parent guarantees of sub debt. The guaranteed party’s ECL reflects the guarantee (lower LGD), but the guarantor recognizes a separate liability under the guarantee. Both sides booked, both disclosed — see Related Party.
  • Sovereign and bank counterparties. “Zero risk” needs support (rating, history), not assertion — 2023 taught that lesson permanently. Even low-PD exposures need documented PD sources.

Reconcile every ECL input to the Month-End Close binder: provision matrix to AR aging (see Collections), PD/LGD to source vintage, scenario weights to treasury memos — see Cash Forecast.

FAQ

How often to update ECL?

Each reporting date. Stage transfers and scenario weights must reflect current conditions, not prior quarter rollforward.

What do auditors challenge first?

SICR completeness, Stage 2 under-identification, and unsupported low loss rates for aged buckets.

Next Steps

  1. Freeze staging policy and provision matrix methodology.
  2. Run scenario-weighted recalculation with review sign-off.
  3. Retain binder for year-end audit.