Same Loan, Three Different Numbers: How Basel, CECL, and IFRS 9 Measure Credit Risk

Introduction

Your credit card limit did not come from a spreadsheet someone filled out by hand. Neither did the mortgage offer sitting on your kitchen table. Both came from models that estimated how much you might borrow, how likely you are to stop making payments, and how much the bank could realistically recover if that happened. The math behind those models is the same across every major bank in the world. What differs is the time frame each one is required to look into the future.

This is my attempt to break down three frameworks that answer that question differently. Basel, CECL, and IFRS 9 each start from the same formula and arrive at very different numbers. Understanding why is what this article is about.

The formula everyone uses

At the center of credit risk measurement sits a deceptively simple equation:

EL = PD × EAD × LGD

PD, or probability of default, estimates the likelihood that a borrower stops making payments. EAD, or exposure at default, estimates how much money is actually at risk when that happens. LGD, or loss given default, estimates how much of that exposure the bank cannot recover. Multiply the three together and you get expected loss, the amount a lender should anticipate losing on any given position over a defined period.

At a glance, the formula looks almost too simple. Three inputs, one output. But how each of those inputs is estimated, over what time period, and under what assumptions, is precisely where risk practitioners spend most of their time. And it is where Basel, CECL, and IFRS 9 part ways.

Basel: Capital as the first line of defense

The Basel framework was not designed to tell banks how much to set aside for accounting purposes. It was designed to tell them how much capital to hold against the risk of unexpected losses. Banks can anticipate and provision for average expected losses, but it is the losses that exceed those expectations that can threaten solvency. Developed by the Basel Committee on Banking Supervision and now in its third iteration, Basel III applies to banks globally and sits at the foundation of how regulators measure whether a bank is adequately capitalized. Under IRB, PD, EAD, and LGD estimates feed into risk-weighted capital requirements designed to absorb those unexpected losses. Where provisioning connects to capital is through the shortfall rule: if a bank’s allowances fall short of its regulatory EL, the difference is deducted directly from CET1. If allowances exceed regulatory EL, some of that excess may be eligible for recognition in Tier 2 capital, subject to applicable limits.

PD: Basel uses a one-year horizon. A bank estimates the probability that a borrower defaults within the next twelve months. Under IRB, this PD is typically a long-run average default estimate with through-the-cycle (TTC) characteristics, reflecting performance across a full economic cycle rather than current conditions. This is deliberately conservative and cycle-smoothing by design. Large banks using the Internal Ratings-Based (IRB) approach build their own PD models; smaller banks rely on standardized regulatory tables.

EAD: Because Basel looks only one year out, EAD is calculated using a Credit Conversion Factor (CCF) that estimates how much of an undrawn credit line a borrower is likely to draw before defaulting. For a credit card with a $5,000 limit and $1,000 outstanding, Basel does not assume the full $5,000 is at risk. It applies a CCF to the unused $4,000 and adds that to the drawn balance.

LGD: Basel requires banks to use a downturn LGD, meaning the recovery rate estimated under stressed economic conditions, not average conditions. Under Foundation IRB, supervisory LGD parameters apply, for example 45% for many senior unsecured exposures. Under Advanced IRB, banks may use internal estimates subject to regulatory constraints and validation. In either case, Basel pushes LGD toward conservatism through supervisory parameters, downturn assumptions, and model validation constraints, even if the exact mechanics differ by approach. The 75% used in the example below reflects a conservative downturn assumption for an unsecured credit card exposure, not a universal regulatory floor.

Where it is used: Regulatory capital calculations, internal RAROC models, and loan pricing at large commercial and investment banks.

CECL: Recognizing the full picture upfront

Before CECL, U.S. banks operated under the incurred loss model, which required them to recognize a loss only after there was evidence that something had already gone wrong. The problem became clear during the 2008 financial crisis, when banks were forced to recognize massive losses all at once because they had not been building reserves early enough. In response, the Financial Accounting Standards Board (FASB) issued ASC 326 in 2016, commonly known as CECL, which took effect for large public banks in 2020. The shift was fundamental: instead of waiting for a loss to occur, banks now had to estimate and reserve for expected losses over the entire life of a loan from the moment it was originated.

PD: CECL requires a lifetime PD estimate. Rather than asking “what is the probability of default over the next twelve months,” it asks “what is the probability of default at any point between now and the maturity of this loan.” This estimate must also incorporate forward-looking macroeconomic assumptions, meaning banks are required to condition their PD on scenarios such as GDP growth, unemployment rates, and interest rate paths. Once the forecast horizon extends beyond what is reasonably supportable, banks revert to long-run historical averages.

EAD: For term loans, EAD is relatively straightforward. For revolving products like credit cards, it becomes significantly more complex. Because a card has no fixed maturity, CECL requires banks to estimate how much a borrower will draw down over the expected life of the account, how long that account will remain open, and how balances will evolve over time. This makes credit card portfolios one of the most technically challenging applications of CECL.

LGD: Unlike Basel, CECL imposes no regulatory floor on LGD. Banks estimate recovery rates using their own historical loss data, adjusted for current and forecasted economic conditions. This gives institutions more flexibility but also places a heavier burden on data quality and model governance.

Where it is used: Quarterly allowance for credit loss disclosures at U.S. banks, credit card and mortgage portfolio reserving, and forward-looking stress testing inputs.

IFRS 9: A staged approach to an evolving risk

While CECL requires banks to recognize lifetime expected losses on every loan from day one, the International Accounting Standards Board (IASB) took a different view when it issued IFRS 9 in 2014, with mandatory adoption in 2018. The standard applies to banks and financial institutions outside the United States and was similarly designed to replace the old incurred loss model. But instead of a single lifetime estimate applied uniformly, IFRS 9 introduced a three-stage framework that ties the measurement of expected loss to how much a loan’s credit risk has deteriorated since origination.

The staging framework Stage 1 applies to loans where credit risk has not increased significantly since origination. Only a 12-month ECL is recognized. Stage 2 is triggered when there has been a significant increase in credit risk, at which point the bank must recognize a lifetime ECL. Stage 3 applies to loans that are already credit-impaired, where lifetime ECL continues to apply and interest income is calculated on the net carrying amount rather than the gross balance.

PD: In Stage 1, PD is estimated over a 12-month horizon, similar to Basel. Once a loan moves to Stage 2 or 3, the bank must estimate a lifetime PD, similar to CECL. The key judgment call is determining what constitutes a significant increase in credit risk, which IFRS 9 leaves largely to the institution to define, creating variation across banks.

EAD: EAD follows the same staging logic. Stage 1 exposures use a 12-month EAD, while Stage 2 and 3 exposures require a lifetime EAD estimate. For revolving products, this creates a similar complexity to CECL once a loan moves out of Stage 1.

LGD: Like CECL, IFRS 9 does not impose regulatory floors on LGD. Banks use their own historical recovery data adjusted for forward-looking conditions. The key difference is that LGD assumptions interact with the staging decision, since a loan moving from Stage 1 to Stage 2 will typically see both a longer horizon and a more stressed LGD applied simultaneously.

Where it is used: Financial statements of banks across Europe, Asia, and most of the world outside the United States, cross-border lending portfolios, and early warning systems built around Stage 2 migration triggers.

Same loan, three different numbers

To make the differences concrete, consider a single credit card account. The numbers below are illustrative and intentionally simplified to highlight how each framework approaches the same exposure differently. They are not meant to represent actual bank model outputs.

• Total credit limit: $5,000
• Current outstanding balance: $1,000
• Undrawn amount: $4,000
• Expected account life: 3 years

The starting point is the same across all three frameworks. What differs is how each one defines the amount actually at risk, how far into the future it looks, and what recovery rate it assumes.

Basel

• PD (2.0%): Based on a one-year horizon. Typical U.S. credit card portfolios have historically shown 1-year default rates in the 1.5% to 3.0% range under normal economic conditions. (Federal Reserve Board, Charge-Off and Delinquency Rates on Loans and Leases, Q4 2025; Moody’s Annual Default Study, 2024)
• EAD ($4,000): Basel does not limit EAD to the current drawn balance. It applies a CCF of 75% to the undrawn $4,000, reflecting the likelihood that a borrower draws down more before defaulting. EAD = $1,000 + (0.75 × $4,000) = $4,000. (Basel III: CRE20, BCBS, 2017)
• LGD (75%): Downturn LGD for unsecured consumer exposure. The regulatory minimum for senior unsecured exposures is 45%, but banks applying conservative downturn assumptions to credit card portfolios typically estimate well above that floor. (Basel III IRB framework, BCBS, 2017)
• EL = 0.020 × $4,000 × 0.75 = $60

CECL

• PD (12.0%): Lifetime horizon across three years. Cumulative default probabilities for credit card accounts are materially higher than any single-year estimate, particularly when conditioned on a baseline macroeconomic scenario. (FDIC Quarterly Banking Profile; Federal Reserve DFAST disclosures)
• EAD ($3,500): CECL requires estimating how much of the credit line will be drawn over the full life of the account. This borrower is currently using $1,000 of a $5,000 limit. Over three years, utilization is assumed to rise toward the limit, with a lifetime average estimated at $3,500. (ASC 326–20, FASB, 2016)
• LGD (80%): No regulatory floor. Bank’s own historical recovery data on unsecured card balances, adjusted for forward-looking conditions. (ASC 326–20, FASB, 2016; CFPB Consumer Credit Card Market Report, 2025)
• EL = 0.120 × $3,500 × 0.80 = $336

IFRS 9

• Stage 1 PD (2.0%): 12-month horizon. Borrower has not shown significant credit deterioration since origination. (IFRS 9 para. 5.5.5, IASB, 2014)
• Stage 1 EAD ($1,500): 12-month expected balance. Higher than the current $1,000 to reflect potential drawdown within the year, but well below the lifetime estimate. (IFRS 9 para. 5.5.15, IASB, 2014)
• Stage 1 LGD (75%): Consistent with Basel assumption for a performing unsecured exposure. (IFRS 9 para. B5.5.28, IASB, 2014)
• Stage 1 EL = 0.020 × $1,500 × 0.75 = $22.50
• Stage 2 PD (12.0%): Significant increase in credit risk triggers a lifetime PD, identical to CECL assumption. (IFRS 9 para. 5.5.3, IASB, 2014)
• Stage 2 EAD ($3,500): Once a loan moves to Stage 2, lifetime EAD applies, consistent with the lifetime horizon now required. (IFRS 9 para. 5.5.15, IASB, 2014)
• Stage 2 LGD (80%): Stressed recovery assumption applied once loan migrates out of Stage 1. (IFRS 9 para. B5.5.28, IASB, 2014)
• Stage 2 EL = 0.120 × $3,500 × 0.80 = $336

Note that Basel produces a higher EAD than CECL here not because it looks further into the future, but because the CCF mechanism captures the full undrawn limit in a single year. CECL’s $3,500 reflects a lifetime average balance, which accounts for periods of lower utilization across the full three-year horizon. For this simplified example, Stage 2 converges numerically with the CECL case because both apply a lifetime horizon. In practice, the two can still differ materially due to staging judgment, scenario weighting, behavioral assumptions, and portfolio-specific modeling choices.

Why it matters beyond accounting

The differences between Basel, CECL, and IFRS 9 are not just technical. They have direct consequences for how banks manage capital, extend credit, and absorb losses during a downturn. The connection runs through one ratio that regulators watch above all others:

CET1 Ratio = CET1 Capital / Risk-Weighted Assets (RWA)

Where:

CET1 Capital = Common Equity — Regulatory Deductions

CECL

When a bank increases its allowance for credit losses, the charge flows through the income statement:

ACL↑ → Net Income↓ → Retained Earnings↓ → Common Equity↓ → CET1↓

Under CECL, this happens on day one. The full lifetime expected loss must be recognized at origination, meaning a bank building a large credit card portfolio takes an immediate and significant capital hit before a single borrower misses a payment. Broadly, higher allowances reduce retained earnings and therefore pressure CET1, though the exact impact can be moderated by transition arrangements, tax effects, and jurisdiction-specific regulatory capital filters.

Basel and CECL: the shortfall rule

Basel does not dictate provisioning levels directly. But the two frameworks are connected through the following:

If ACL < Regulatory EL: Shortfall = Regulatory EL — ACL → Direct CET1 Deduction

In practice, large banks monitor CECL reserve levels against Basel regulatory EL to assess whether a capital-impacting shortfall may emerge.

IFRS 9: the cliff effect

Stage 1: ECL = PD(12M) × EAD(12M) × LGD

Stage 2: ECL = PD(Lifetime) × EAD(Lifetime) × LGD

When a loan migrates from Stage 1 to Stage 2:

ΔECL = ECL(Lifetime) — ECL(12M) → ACL↑ → CET1↓ in a single reporting period

This jump can happen across an entire portfolio simultaneously during periods of broad credit stress, before actual defaults materialize. This cliff effect has been a persistent criticism of the staging framework since its introduction.

Closing Takeaway

All three frameworks start from the same equation. What separates them is the question each one is trying to answer.

Basel asks how much capital a bank needs to remain solvent when losses exceed expectations.

CECL asks how much a bank should reserve today for losses it expects to incur over the entire life of a loan.

IFRS 9 asks how provisioning should respond as credit quality changes over time.

Three questions, three answers, three different numbers from the same formula.

• Basel uses a one-year PD, a CCF-based EAD, and a regulatory LGD floor. Designed for capital adequacy, not loss forecasting.
• CECL front-loads the full lifetime expected loss at origination. The capital hit is immediate and does not wait for deterioration.
• IFRS 9 stages provisioning by credit deterioration. Stage 1 to Stage 2 migration triggers a cliff effect, where lifetime ECL replaces 12-month ECL across an entire portfolio simultaneously.
• All three converge on the same point: ACL increases reduce retained earnings, retained earnings reduce common equity, and common equity determines CET1. Provisioning is a capital decision, not just an accounting one.

Credit risk practitioners need to understand all three. Basel governs how much capital a bank holds. CECL governs how U.S. banks report losses. IFRS 9 governs how the rest of the world does the same. The same loan sits at the intersection of all three frameworks, and knowing only one means operating with an incomplete picture of the risk you are actually managing.

📩 Join Investor’s Handbook Digest — get the best investing, markets, and wealth-building insights each week.