From Policy Rate to Credit Loss
This chapter follows the transmission from a central-bank policy-rate change to a bank’s loan-loss provisions and profitability. The key question is timing: which contractual and economic links make the same move hit one borrower immediately, another only at refinancing, and a bank after risk has accumulated? Following that chain gives us a better way to interpret the empirical estimates before asking why the chain becomes stronger in some macro-financial states than in others.
Debt service does not reprice all at once
A policy rate is not the interest rate on every loan. It is an upstream price that influences many other rates, often with different lags. The first distinction is therefore contractual. A household with a variable-rate mortgage, or a firm borrowing at a floating benchmark plus a spread, may see its interest payment reset relatively soon. A borrower with a fixed-rate loan may not see the scheduled payment change at all during the fixed period. A third borrower may be protected for now but exposed when a loan matures and must be rolled over at the new market rate.
That difference in timing matters because repayment capacity is a flow. If income stays constant while the interest bill rises, less cash remains for consumption, investment, wages, or other debts. For a firm, a higher interest bill can reduce operating cash available for investment and make a previously manageable refinancing requirement difficult. For a household, the same change can compress the budget available for goods and services. The borrower does not need to default immediately for risk to rise; a thinner cash buffer makes a later income or asset-price shock more consequential.
The BIS and IMF papers describe two related routes from that pressure to bank credit losses. The direct route runs through borrowers whose debt service responds to the rate. Higher payments can increase the probability of default, especially where the loan was made to a borrower already close to a debt-service limit. The indirect route runs through incomes and employment. Tighter financial conditions can reduce consumer spending and business investment, weaken cyclical firms, and raise unemployment. That can impair borrowers even when their original loans were fixed-rate, because the denominator supporting repayment—income—has fallen.
Collateral adds a second dimension. A higher discount rate lowers the present value of assets that generate future income, such as property or other financed assets. If collateral values decline, the borrower’s net worth and ability to obtain additional credit can weaken. If the borrower later defaults, the lender may recover less from the collateral. In the language of credit risk, the rate change can affect both the probability of default and the loss given default. Those are different quantities: one concerns whether the borrower fails to meet the obligation; the other concerns how much the lender loses after recovery is taken into account.
The wider economy can reinforce or offset these effects. Higher rates generally restrain spending and investment, but the circumstances in which a central bank tightens matter. A hike responding to an overheating economy is not the same event, for borrowers, as a hike responding to a supply shock that raises inflation while growth stagnates. Inflation may reduce the real value of existing nominal debt in the short run, but it can also raise living costs before wages adjust. In an open economy, a rate increase may appreciate the currency and help reduce the domestic cost of foreign borrowing, or it may occur alongside capital outflows and currency stress. Exporters, foreign-currency borrowers, and banks with cross-border exposures will not experience those paths in the same way.
These effects arrive in layers. The payment on a floating-rate loan may change first. Investment and consumption may respond next. Unemployment, non-performing loans, recoveries, and provisions can follow later. That sequencing is why a rate hike can look harmless in the first few quarters and still be associated with higher credit losses several years later. It also explains why a current loan-loss number is not a pure reading of the latest policy decision: it reflects the accumulated condition of loans made under earlier rates and earlier underwriting standards.
There is an important causal distinction here. The policy rate can be a trigger, while debt structure and weak income are mediators, and loan-loss provisions are an observed bank outcome. A recession can also cause the central bank to lower rates, creating a reverse correlation between rates and losses if the timing is ignored. The BIS addresses this problem with local projections, macro-financial controls, fixed effects, leads of the policy rate, and alternative monetary-policy shock measures. The IMF uses a dynamic bank panel with bank and time effects, country-specific trends, and controls for macroeconomic and bank characteristics. Neither design observes every borrower’s contract or proves that each link is operating with identical force in every country.
The diagram shows why “the rate went up” is only the beginning of the explanation. The policy move enters several channels, and the channels interact. A borrower may be helped by a stronger currency but hurt by a job loss. A bank may gain from loan yields repricing while losing from higher provisions. A delayed outcome is not evidence that the policy had no effect; it may be evidence that the relevant contract or balance-sheet transition takes time.
The bank sees several ledgers, not one
Banks do not record “the effect of rates” as a single line. They record provisions against loans, interest revenue and expense, securities income, operating costs, and ultimately return on assets. The IMF’s organizing identity is simple:
Return on assets = net interest margin + net non-interest income − loan-loss provisions.
The identity is useful because it prevents a common error: inferring credit risk from the direction of bank profits. Net interest margin is interest revenue minus interest expense. When loan rates adjust faster than deposit rates, the margin can rise after a hike. That is a price effect. It may provide a temporary cushion even as the loan book becomes more fragile.
Loan losses create a countervailing path. Banks set aside provisions when loans are expected to generate losses or when impairment becomes evident. As borrowers miss payments or migrate into non-performing status, the bank may need to recognize more expense and may also lose the interest income that the loan previously generated. The IMF calls this latter mechanism a volume effect: a loan that stops performing no longer contributes to the interest-earning asset base in the same way. For a weak bank with flexible-rate loans, the volume effect can eventually outweigh the initial increase in loan yields.
There is a third path through non-interest income. Higher rates reduce the market value of many fixed-income securities. Whether that decline immediately reduces reported profit depends on classification. Securities held for trading are marked to market through the income statement. Available-for-sale securities are also measured at fair value, but changes generally enter other comprehensive income rather than current profit. Held-to-maturity securities remain at book value unless they are sold, so the loss may become visible in profit and loss only when liquidity pressure forces a sale. The economic exposure exists across categories; the accounting location and timing differ.
Liquidity determines whether that third path becomes acute. A bank holding sufficient liquid assets may be able to meet withdrawals or margin calls by borrowing against eligible securities rather than selling them into a falling market. That can reduce immediate non-interest-income losses and support reported profitability. But liquid assets can have a lower return than riskier or less liquid assets, so the buffer carries an opportunity cost. It changes the bank’s ability to absorb a shock; it does not eliminate the underlying duration or credit risk.
The two studies measure these processes at different levels. The BIS country-level analysis uses annual net loan-loss provisions as a percentage of net loans for 113 countries over 1990–2022. Its local projections estimate how that ratio changes over horizons of up to five years after a policy-rate increase. The baseline estimate is about a 0.1 percentage-point increase after a 1 percentage-point hike, with the largest response appearing over a multi-year window. When the researchers replace ordinary policy-rate changes with more exogenous monetary-policy shocks, the positive response remains; an instrumental-variable local projection produces a larger estimate, roughly 0.3 percentage points rather than 0.1. That difference is informative about identification and shock selection, not a reason to choose the larger number as a forecast.
The BIS also repeats the exercise with bank-level data for 3,350 banks in 92 countries. The average response to a 1 percentage-point policy-rate increase is approximately 0.02 percentage points and becomes statistically visible only several years after the shock. This smaller coefficient does not erase the country-level result. It reflects a different aggregation level, a different sample composition, and a distribution in which many banks have little measured response while a vulnerable group has a much larger one. The study finds that banks with poorer pre-existing loan portfolios respond much more strongly, with an increase of almost 0.1 percentage points and significance in the first year.
The IMF’s design answers a related but different question. It uses annual data for 2,147 banks in 31 countries from 1995–2023 and models provisions, net interest margin, non-interest income, and return on assets as shares of average total assets. The main rate variable is the three-month interbank rate, entered with a squared term and interacted with bank characteristics and borrower-based macroprudential measures. In the provisions regressions, higher rates are associated with more provisions; the interaction is stronger for banks with flexible-rate loans and for banks with low market share. In the NIM regressions, the average rate effect is positive, but the result varies with deposits, loan flexibility, and non-performing loans. In the ROA regressions, average profitability can hold up even when a high-NPL, flexible-rate group experiences a negative net effect.
The numbers should therefore be read with their labels attached. A percentage-point change in a loss ratio is not the same as a percent change in the amount of losses. A provision ratio scaled by average total assets is not the same denominator as net provisions scaled by net loans. A five-year impulse response is not the same estimand as a contemporaneous annual level relationship. The BIS’s country-level estimate, its bank-level average, and the IMF’s component results can tell a coherent story without being substituted for one another.
Nor are provisions identical to realised defaults. The IMF notes that expected-credit-loss standards introduced a more forward-looking recognition framework in some jurisdictions, and that provisions can also reflect smoothing or other managerial choices. Over time, provisions may track underlying losses more closely, but any short-period interpretation needs that qualification. The same discipline applies to the bank’s profit identity: an increase in ROA can coexist with rising credit risk if margin income or another component temporarily offsets provisions.
The mechanism, then, ends not with one number but with a set of delayed observations. The policy rate changes the price of financing; contracts determine who feels it first; income, collateral, and exchange rates shape repayment; the bank records provisions and offsets them against margin and securities effects. The average studies establish that the loss channel is real enough to measure, but they also show why its magnitude cannot be separated from the conditions present before the hike. That is the point at which the same transmission chain branches: a rate move arriving after years of loose conditions or during a downturn meets a very different borrower and banking system from one arriving on resilient ground.