The Hike Is the Same; the Exposure Is Not
This series examines how central-bank rate hikes become bank credit losses across economies and individual institutions. Its central question is not whether a higher policy rate is good or bad for banks in the abstract. It is why the same nominal increase can be absorbed with little visible damage in one setting yet produce a sharp deterioration in another. The practical payoff is a way to read rate-risk evidence: start with the conditions and balance sheets that receive the shock, not with the shock alone.
The average number is not the answer
The headline finding from the BIS study is useful precisely because it is limited. Across an annual panel of 113 countries from 1990 to 2022, a 1 percentage-point increase in the policy rate is associated with an increase of about 0.1 percentage points in the loan-loss rate over a three-to-five-year horizon. The paper’s policy-oriented summary puts the five-year figure at roughly 0.12 percentage points after a 1 percentage-point increase in a short-term interest rate. That is a meaningful average effect, but it is not a universal conversion rule.
The denominator matters. The BIS country-level outcome is an annual loan-loss rate: net loan-loss provisions reported in the profit-and-loss account as a percentage of net loans outstanding. The question is therefore about the change in a country-level loss ratio over time, not about the amount a particular borrower will default on or the capital loss of a named bank. The estimate also describes a dynamic response. A hike does not pass through every loan contract on the day it is announced, and losses may appear as borrowers refinance, incomes weaken, collateral values change, or loans migrate into non-performing status.
The same study finds a much smaller average response in its bank-level panel: approximately 0.02 percentage points for a 1 percentage-point increase in the central-bank policy rate, with significance emerging only several years after the shock. Yet that average conceals a sharp split. Banks that already had an above-country-average share of impaired loans before tightening saw an increase of almost 0.1 percentage points, and the increase appeared significantly in the first year. Banks with healthier loan portfolios showed no significant increase in the corresponding estimate.
This is not a contradiction. A country-level ratio can combine many banks, borrower types, and balance sheets; a bank-level average can dilute the effect concentrated in a weaker subset. The studies also use different outcomes and designs. The BIS bank panel covers 3,350 banks in 92 countries, while the IMF study examines 2,147 banks in 31 countries from 1995 to 2023. The IMF measures loan-loss provisions as a share of average total assets and estimates their relationship with the level of the three-month interbank rate in a dynamic bank panel. These are related questions, not interchangeable experiments.
The safest interpretation is therefore conditional: higher rates are associated with higher credit losses on average, but the average tells us little about where the losses will concentrate, how quickly they will appear, or whether the bank’s other income streams will offset them. A coefficient can describe a population response without becoming a forecast for any one institution.
Four ways a bank can experience one rate move
The first distinction is between a borrower’s repayment problem and a bank’s reported profit. Higher rates can raise the debt service of households and firms carrying variable-rate debt, seeking new financing, or rolling over existing borrowing. That pressure can increase the probability of default. A higher discount rate can also lower the value of income-generating collateral, increasing the lender’s loss if default occurs and reducing the borrower’s net worth at the same time.
The second distinction is between a direct contract effect and a wider economic effect. A borrower with a loan that reprices quickly may feel the rate increase in the next payment. A household with a fixed-rate mortgage may not face that immediate change, although refinancing risk, unemployment, weaker house prices, and reduced income can still affect repayment. For firms, higher financing costs can suppress investment and operating cash flow. In an open economy, currency appreciation may weaken exporters, while depreciation can increase the burden of foreign-currency debt. The route from policy rate to credit loss is thus a sequence with different speeds, not a single switch.
The third distinction is inside the bank’s income statement. The IMF expresses pre-tax return on assets as:
Return on assets = net interest margin + net non-interest income − loan-loss provisions.
The net interest margin is interest revenue minus interest expense. If loan yields reprice faster than deposits, a bank can initially earn more. At the same time, provisions can rise as borrowers struggle, and non-interest income can fall when higher rates reduce the market value of securities or weaken fee-generating activity. A bank can therefore report a stronger average margin while a vulnerable loan book is generating larger losses. Conversely, the credit-loss channel can outweigh the margin benefit for a bank with risky borrowers and flexible-rate loans.
The fourth distinction is between an accounting signal and the underlying economic event. A provision is an amount set aside for potential or expected loan losses; it is not a perfect one-period measure of defaults and recoveries. The IMF notes that forward-looking expected-credit-loss standards and managerial smoothing can affect the gap between provisions and actual losses. Securities losses also appear differently depending on whether assets are held for trading, available for sale, or held to maturity. Credit losses are an important route to financial stress, but they are not the whole solvency story.
This map explains why the phrase “the effect of higher rates on banks” is incomplete. The policy rate is the common input, but the borrower channel and the bank channel are filtered through the state that existed before tightening. The output may be a provision increase, a margin gain, a securities loss, a capital strain, or several of these at once.
A better question than “Are higher rates good or bad for banks?”
Once the channels are separated, four diagnostic lenses become more informative than a general verdict.
What was already true before the hike? The BIS finds a more persistent loss response when tightening follows a relatively loose monetary stance. A policy rate below its five-year moving average, or a negative residual from a simple Taylor-rule relationship, is used as an approximation for that stance. The interpretation is not that a low rate mechanically causes a crisis. It is that a prolonged period of easy conditions may allow leverage and risk-taking to accumulate, leaving more borrowers exposed when financing costs reset.
How does the loan contract transmit the move? Economies with a high share of floating-rate loans experience larger and faster loss increases in the BIS analysis. The IMF reaches a similar conclusion for banks classified as having substantial flexible-rate exposure. Fixed-rate lending can delay direct payment pressure, while floating-rate lending can pass the policy move into debt service quickly. But loan structure is not destiny: unemployment and collateral effects can reach fixed-rate borrowers, and flexible-rate loans can temporarily increase interest income for a healthy bank.
Which balance sheet receives the pressure? Pre-existing impaired loans are a stronger warning sign than the headline average. On the liability side, deposit-funded banks may enjoy a slower rise in funding costs than banks reliant on wholesale markets, creating a larger initial margin cushion. On the asset side, a weak loan book can turn that cushion into a temporary illusion as non-performing loans stop generating interest. Securities classification and liquid assets add another layer: a bank that can meet liquidity needs without selling securities may avoid turning a valuation decline into an immediate reported loss.
Which policies and macro conditions surround the move? The BIS finds larger loss responses when hikes arrive during downturns, high inflation, restrictive fiscal conditions, or a shrinking central-bank balance sheet. The IMF finds that borrower-based measures such as tighter loan-to-value or debt-service limits, when in place before rates rise, can reduce the sensitivity of provisions—especially for flexible-rate loans and unemployment shocks. These findings do not mean that fiscal or prudential policy can erase the cost of tightening. They mean that the same monetary action is embedded in a wider policy mix and meets vulnerabilities that authorities may have influenced earlier.
These lenses should not be converted into a pile of independent multipliers. A downturn often coincides with weak demand and prior rate cuts; high debt can coexist with a particular lending structure; a balance-sheet change can partly reflect foreign-exchange intervention. The BIS therefore estimates some states separately and limits combined exercises to a small number of competing conditions. Its scenario results reach roughly 0.2–0.3 percentage points for a 1 percentage-point hike in combinations of adverse states, but that range is a conditional model result, not a promise about every economy that checks several boxes.
The evidence supports a more precise answer to the title. A rate hike is more likely to produce sizeable credit losses when it arrives after vulnerabilities have accumulated, reaches borrowers through rapidly repricing debt, and lands on banks with weak asset quality or little capacity to absorb other losses. It is more likely to be contained when loan contracts delay repricing, portfolios begin in healthier condition, funding is stable, liquidity is available, and prudential safeguards have limited leverage in advance. The next task is to follow that logic through the actual path from a policy decision to a borrower’s cash flow and a bank’s reported provisions—because that path explains both the timing and the variation that the averages conceal.