Provision for Credit Losses Explained: Formula, CECL, and What It Means for Bank Earnings

May 9, 2026 · guides · 11 min read

Provision for Credit Losses Explained: Formula, CECL, and What It Means for Bank Earnings

When a bank reports earnings, one line item can swing net income by hundreds of millions of dollars from one quarter to the next. That line is the provision for credit losses. Understanding what it is, how it works, and what it signals about a bank's health is essential for anyone analyzing financial institutions.

This guide covers everything: the definition, the income statement mechanics, the balance sheet reserve, the shift to CECL accounting, and how investors use provision trends to judge earnings quality.

What Is Provision for Credit Losses

The provision for credit losses (PCL) is a non-cash expense that a bank records on its income statement to account for the expected cost of loans that will not be repaid. It is sometimes called the loan loss provision, the provision for loan losses, or simply the credit loss provision.

Think of it as a quarterly estimate. Management looks at its loan portfolio and asks: based on current conditions and what we expect going forward, how much of this portfolio will turn into losses? The answer gets charged to the income statement as an expense, reducing pre-tax income.

The provision is not a cash payment. No money leaves the bank when the provision is recorded. It is an accounting adjustment that reduces reported earnings and builds up a reserve on the balance sheet called the allowance for credit losses.

How the Provision Flows Through the Income Statement

The Basic Income Statement Mechanic

A bank's income statement follows a predictable structure:

Net interest income Plus: Non-interest income Equals: Net revenue (or total revenue) Minus: Provision for credit losses Equals: Net revenue after provision Minus: Non-interest expense (salaries, technology, occupancy) Equals: Pre-tax income Minus: Income taxes Equals: Net income

The provision sits directly between revenue and operating expenses. A large provision wipes out a significant portion of revenue before the bank even gets to its cost base. This is why provision changes drive so much quarter-to-quarter earnings volatility at banks.

A Simple Example

Suppose a regional bank generates 1 billion dollars in net interest income and 200 million in non-interest income, for total revenue of 1.2 billion. Operating expenses are 650 million. If the provision is 100 million, pre-tax income is 450 million. If the provision jumps to 400 million in a stressed quarter, pre-tax income collapses to 150 million, a 67 percent decline, even though the underlying loan business did not change.

The Allowance for Credit Losses on the Balance Sheet

What the Reserve Is

Every quarter, the provision expense flows into a contra-asset account on the balance sheet called the allowance for credit losses (ACL), also historically called the allowance for loan and lease losses (ALLL). A contra-asset account reduces the gross carrying value of the loan portfolio to a net figure.

The relationship is:

Gross loans (face value of all outstanding loans) Minus: Allowance for credit losses Equals: Net loans (the amount shown as an asset on the balance sheet)

The allowance represents management's best estimate of probable future losses embedded in the existing portfolio at the balance sheet date.

How the Allowance Moves

The allowance increases when the provision expense is recorded. It decreases when actual loan losses are written off, which are called charge-offs. When a loan is deemed uncollectible and officially written off, the bank debits the allowance and credits the loan asset. The allowance absorbs the hit rather than flowing through the income statement again.

If some portion of a previously charged-off loan is later recovered, the recovery is credited back to the allowance (a partial reversal) and also shows up in earnings.

The Coverage Ratio

Investors track the coverage ratio to assess whether a bank's reserve is adequate:

Coverage ratio = Allowance for credit losses divided by non-performing loans

A ratio above 100 percent means the bank has reserved more than enough to cover all currently non-performing loans. A ratio below 100 percent suggests the reserve may be thin relative to visible problem credits. Coverage ratios vary significantly across the credit cycle. During benign periods, ratios of 150 to 200 percent are common. During stress, ratios can compress toward or below 100 percent as NPLs surge faster than reserves can be built.

Net Charge-Offs vs. Provision

These two concepts are frequently confused.

Net Charge-Offs

Net charge-offs (NCOs) are the actual losses realized in a period: gross charge-offs minus recoveries. Charge-offs happen when a bank formally writes off a loan it has concluded will not be collected. This is the realized, actual loss event.

Provision

The provision is a forward-looking estimate of future losses. It is the expense booked today to prepare for losses the bank expects to realize in coming quarters.

Why They Diverge

Provision and net charge-offs can diverge significantly:

During an economic downturn, provisions run ahead of charge-offs. Banks rapidly increase their reserves in anticipation of rising defaults, booking large provision expenses before the actual charge-offs materialize. Earnings take the hit early.

During recovery, the reverse happens. As economic conditions improve, banks may reduce their allowance by booking a provision release, which is a negative provision expense that actually boosts earnings. Charge-offs may still be occurring, but the reserve is more than sufficient, so management releases some of it.

The gap between provision and NCOs tells a story about management's view of where credit quality is heading.

CECL: Current Expected Credit Loss Model

The Old Incurred Loss Model

Before 2020, US banks used the incurred loss model under US GAAP. Under that framework, a bank could only recognize a credit loss when it was 'probable' that a loss had been incurred based on observable evidence. This meant reserves were typically backward-looking: a bank waited until it saw concrete signs of deterioration before booking a provision.

Critics argued this approach made bank earnings pro-cyclical. In good times, reserves were thin because losses had not yet been 'incurred.' When the cycle turned, banks were forced to rapidly build reserves at exactly the moment their capital was most stressed, amplifying the downturn.

This dynamic was painfully visible during the 2008 financial crisis, when banks that had thin reserves going in were forced to take enormous provision hits as losses mounted.

CECL Under ASC 326

The Financial Accounting Standards Board (FASB) responded by issuing ASC Topic 326, known as CECL (Current Expected Credit Loss). Large public US banks adopted CECL beginning January 1, 2020. Smaller banks followed on January 1, 2023.

The core change: instead of waiting for losses to be 'incurred,' banks must now estimate and reserve for expected losses over the entire remaining contractual life of each loan from the moment it is originated.

CECL requires banks to use historical loss experience, current conditions, and reasonable and supportable forecasts of future economic conditions. This makes the allowance explicitly forward-looking in a way the old model was not.

What CECL Changed for Investors

Under CECL, allowance levels are structurally higher than they were under the incurred loss model. Banks hold reserves not just for loans showing signs of distress, but for expected lifetime losses across the entire portfolio including loans that are currently performing perfectly.

This means two things for investors. First, the allowance-to-loans ratio is a less useful cross-period comparison for banks that adopted CECL versus their pre-CECL history. Second, economic forecast changes now drive provision volatility even when no loans have actually deteriorated. If a bank revises its macro forecast downward, expected lifetime losses increase across the portfolio, triggering a provision build without any observable credit deterioration.

Day-One CECL Hit Explained

When a bank adopted CECL for the first time, it had to immediately true up its allowance from the old (incurred loss) level to the new (lifetime expected loss) level. This one-time adjustment bypassed the income statement and went directly to retained earnings as a cumulative effect of accounting change.

Large banks saw retained earnings reductions of hundreds of millions to several billion dollars on day one. JPMorgan Chase estimated a 4 to 6 billion dollar day-one hit when it adopted CECL. The hit did not flow through reported earnings, but it did reduce book value and tangible common equity.

Investors analyzing bank book values for periods immediately surrounding CECL adoption need to adjust for this one-time retained earnings reduction to make meaningful comparisons.

CECL's Impact on Provision Volatility

Because CECL requires forward-looking macro assumptions, provision expense is now more sensitive to economic forecast changes than it was historically. During the early COVID-19 period in Q1 and Q2 2020, large US banks collectively added hundreds of billions in reserves as their economic forecasts incorporated severe recession scenarios.

As those recession scenarios improved in late 2020 and 2021, the same banks released billions in reserves, booking negative provision expenses that significantly boosted reported earnings. Critics noted that earnings release looked unusually strong in 2021 partly because provision tailwinds were masking the underlying run-rate of the business.

CECL's forward-looking nature means reported bank earnings are now partly a function of macroeconomic forecasting rather than purely current loan performance. This adds a layer of subjectivity that investors must scrutinize.

Provision as an Earnings Management Tool

Discretion Within GAAP

The provision calculation involves significant management judgment. Banks must estimate default probabilities, loss given default, and expected future economic conditions. Reasonable people can reach different conclusions using the same data.

This creates opportunity for earnings management within GAAP boundaries.

Big Bath Provisioning

'Big bath' provisioning refers to the strategy of taking an unusually large provision in a period when earnings are already poor, often during a leadership transition or a period of restructuring. The logic: since the quarter is already bad, take the maximum defensible provision and start the next period with an over-built reserve that can be released in future quarters to boost earnings.

Regulators and auditors scrutinize big bath provisions. A provision that appears disproportionately large relative to observable credit deterioration can attract SEC comment letters or auditor pushback. But within the range of defensible estimates, management has latitude.

Cookie Jar Reserves

The inverse of the big bath is the 'cookie jar' reserve. During strong economic periods, a bank builds its allowance more aggressively than strictly necessary based on current expected losses. The excess reserve sits on the balance sheet as a buffer. In a future weaker period, the bank releases the excess reserve as a provision credit, smoothing reported earnings and masking the true credit deterioration.

Analysts watch for allowance levels that seem inconsistent with observable portfolio quality. An unusually high coverage ratio during benign credit conditions can signal a cookie jar build.

Credit Cycle and Provision Trends

Early Cycle

In the early stages of an economic expansion following a downturn, banks often have elevated allowances built during the recession. As credit conditions improve and losses decline, they release reserves. Provision expense is low or negative. Net income benefits from both improving revenues and provision tailwinds.

Mid Cycle

As expansion matures, loan growth accelerates and new loan vintages are underwritten under competitive terms. Provision runs roughly in line with charge-offs (neutral to earnings). Coverage ratios gradually compress as reserves grow more slowly than the loan portfolio.

Late Cycle

Credit concerns begin to emerge. Non-performing loans start rising. Banks build reserves ahead of expected losses. Provision expense increases, pressuring earnings. Investors who missed the early signals of deteriorating credit quality are caught off guard.

Downturn

Provisions surge. Charge-offs spike. Reserve builds absorb much of reported earnings. Capital ratios are pressured. Banks that entered the cycle with thin reserves face the most severe earnings impact.

Understanding where a bank sits in the credit cycle is central to interpreting provision trends. A rising provision at a well-capitalized bank early in a cycle is very different from the same provision spike at a thinly capitalized bank in a deteriorating economy.

Non-Performing Loans and the Coverage Ratio in Practice

What Non-Performing Loans Are

Non-performing loans (NPLs) are loans where the borrower has stopped making payments, typically 90 days or more past due, or loans in non-accrual status where the bank has stopped recognizing interest income because collection is in doubt.

The NPL Ratio

NPL ratio = Non-performing loans divided by total loans

A rising NPL ratio is a leading indicator of future charge-offs. Banks provision ahead of NPL increases if management has good loss forecasting. Banks that lag on provisioning will face large catch-up expenses when NPLs materialize into actual charge-offs.

Coverage Ratio Worked Example

Suppose a bank has:

Allowance-to-total loans ratio: 1.2 billion divided by 50 billion = 2.4 percent Coverage ratio: 1.2 billion divided by 850 million = 141 percent

The 141 percent coverage ratio means the bank's reserve exceeds its visible NPL balance by 41 percent, suggesting adequate cushion if NPLs do not grow significantly. If NPLs rose to 1.5 billion while the allowance remained at 1.2 billion, coverage would compress to 80 percent, which would typically prompt investor concern about reserve adequacy.

How Investors Analyze Provisions

Provision as a Percentage of Average Loans

Normalizing provision by average loan balances allows comparison across bank sizes and over time:

Provision rate = Annual provision expense divided by average loans

A provision rate of 0.25 to 0.50 percent is typical in benign credit environments for diversified commercial banks. Consumer-heavy banks (auto, credit card, personal loans) run structurally higher provision rates, often 1.5 to 3 percent, reflecting higher expected loss rates on unsecured consumer credit.

Comparing Provision to Net Charge-Offs

If a bank's provision consistently runs well above NCOs, the allowance is building. If provision runs below NCOs, the allowance is being drawn down. A sustained draw-down below NCOs raises questions about whether the allowance is adequate.

Judging Reserve Adequacy

Investors often build their own view of reserve adequacy by:

Comparing the bank's allowance-to-loans ratio to peers in similar loan segments. Looking at the trajectory of the ACL ratio over several quarters. Assessing whether coverage ratios are expanding or contracting relative to NPL trends. Examining management's commentary on macro assumptions embedded in the CECL model.

Provision Release as an Earnings Quality Flag

When provision releases are driving earnings beats, investors should be cautious. A bank that is beating consensus earnings estimates primarily through reserve releases rather than revenue growth or expense management has lower earnings quality than one generating organic improvement. The reserve release is a one-time tailwind that cannot repeat indefinitely.

Worked Example: Analyzing a Bank's Provision

Suppose two banks both report earnings in the same quarter with the following data:

Bank A:

Bank B:

Bank A is building reserves ahead of rising NPLs, running provision above NCOs, taking a measured but proactive approach to deteriorating credit. Coverage compression warrants monitoring.

Bank B is releasing reserves faster than losses materialize, running provision well below NCOs, and depleting the allowance at a time when that buffer may be needed. The 80 million provision number makes earnings look better than the 150 million in actual loan losses suggest. An investor focused only on the headline earnings number misses this dynamic.

Conclusion

The provision for credit losses is one of the most judgment-intensive and consequential line items in bank financial statements. Under CECL, it reflects not just current observable credit stress but management's forward-looking view of expected lifetime losses shaped by economic forecasts. That makes it simultaneously more informative and more manipulable than the old incurred loss model.

Investors analyzing banks need to look beyond the headline provision number to the direction of the allowance, the relationship between provision and actual charge-offs, the coverage ratio trend, and the plausibility of management's macro assumptions. Those signals, taken together, give a far clearer picture of true credit quality and the reliability of reported earnings than the income statement line alone.