When the Same Hike Lands on Fragile Ground
This chapter focuses on the conditions that exist before and alongside a policy-rate increase. The question is why the same rate-to-borrower mechanism produces a larger credit-loss response when leverage, income, inflation, or policy support is already strained. The value of this state-dependent view is practical: it tells us where to look before treating an average rate elasticity as a plausible scenario for a particular economy.
The vulnerability is built before the hike
A rate hike often receives the blame for losses that were made more likely earlier. During a long period of relatively cheap financing, households and firms may take on more debt, lenders may compete more aggressively for borrowers, and asset prices may support the appearance of strong collateral. None of those developments guarantees a later default. They do, however, leave more obligations whose cash flows are sensitive to a change in financing conditions.
The BIS study tests this idea by comparing rate hikes with the policy rate that prevailed beforehand. In one measure, the prior-year rate is classified as low or high relative to its five-year moving average. In another, the authors use the residual from a backward-looking Taylor-rule-type regression: a negative residual indicates that rates were relatively loose given the country’s inflation and growth conditions. These are practical approximations, not natural laws of what “easy” and “tight” mean.
The result is consistent across the two measures. When the policy rate was already high relative to its recent history, a further hike had only a small and temporary effect on loan-loss rates. When the prior stance was relatively loose, the response was more persistent. Using the Taylor-rule residual, a 1 percentage-point increase in the policy rate after loose conditions raised the loan-loss rate by about 0.2 percentage points, whereas tightening after a relatively tight stance was associated with little or no increase in the reported response.
The mechanism is cumulative. The rate hike does not need to create every vulnerability in real time. It can expose a debt structure built when low rates made repayment, refinancing, or leverage look manageable. If loans were originated with thin cash buffers, a reset in interest expense leaves fewer options: cut spending, reduce investment, draw liquidity, sell assets, or refinance under less favourable terms. A later default is then partly the result of the new rate and partly the result of the balance sheet that met it.
Private debt is a useful illustration of why evidence needs to be read rather than merely quoted. The BIS headline finding identifies high private debt as an amplifier, but a simple split at the sample median—approximately a 50% credit-to-GDP ratio—does not by itself produce a consistently larger loss response at every horizon. The authors then let debt compete with the initial interest-rate state in a multivariate specification. In countries with both relatively high private debt and lower initial interest rates, a 1 percentage-point hike increases loan-loss rates by about 0.2 percentage points, nearly twice the average effect.
Those results are not inconsistent. A single debt threshold can mix very different financial systems: a high credit-to-GDP ratio may represent stable long-term mortgages in one country and short-duration corporate or foreign-currency borrowing in another. Debt also tends to coexist with other conditions, such as low rates, rapid credit growth, or weak collateral. The multivariate result says that the combination contains more information than the debt ratio alone. It does not justify applying a high-debt multiplier to every economy with the same headline ratio.
The business cycle changes the borrower’s ability to absorb the same interest bill. The BIS finds that tightening during a downturn, defined through a negative output gap, raises loan-loss rates more than tightening during an expansion. After an expansion, higher rates alone are not associated with a significant increase in losses in that comparison. The difference is intuitive but important: when production and income are already weakening, an additional financing cost arrives when firms have less revenue and households have less secure employment with which to meet it.
The paper’s rule-of-thumb summary expresses the difference in basis points. One hundred basis points equal one percentage point. A moderate downturn—described as a roughly one-in-ten to one-in-fifteen-year scenario—adds about 16 basis points to the loss response to a 1 percentage-point hike. In a more adverse one-in-twenty-year downturn, the added effect is about 28 basis points; in a severely adverse one-in-fifty to one-in-one-hundred-year downturn, it is about 47 basis points. These figures are conditional scenario outputs from the estimated model, not promises that a recession of a certain label will produce an identical number.
Inflation can make the same sequence more difficult. Higher prices reduce the purchasing power of household income and can leave less of it available for debt service before wages catch up. For firms, input costs and weaker demand can compress cash flow at the same time that borrowing costs rise. The BIS finds that the positive rate-to-loss response is concentrated in states where inflation in the year before tightening was above its five-year moving average. The paper’s comparison assigns roughly 11 basis points of additional loss response to the higher-inflation state at the relevant horizon.
Inflation also complicates interpretation because its effect is not one-directional. A bout of inflation can reduce the real value of existing nominal debt, which may provide short-run relief to borrowers. But that relief can be outweighed when living costs rise, real income adjusts slowly, or the central bank tightens to prevent inflation from becoming persistent. A nominal policy rate is therefore not a sufficient description of the real burden facing a borrower.
The state before tightening is best understood as a set of interacting balances, not a scorecard. Low rates can coexist with high debt; high debt can coexist with fixed-rate contracts; a downturn can follow an earlier boom; inflation can cause both debt erosion and income stress. The relevant question is not whether any one variable is “high,” but whether the vulnerabilities that variable represents are connected to the borrowers and banks exposed to the hike.
The state transition has two important features. First, the hike is common to both paths; what changes is the condition from which the system enters it. Second, the adverse states overlap. A downturn may be related to earlier leverage and low rates, so their estimated effects cannot simply be added as if they were independent ingredients. The diagram is a causal map for interpretation, not a calculator.
Buffers change the path, not the existence of the shock
Fiscal policy is one way the surrounding state can either reinforce or cushion a rate increase. In the BIS analysis, restrictive fiscal policy is identified by an improvement in the primary balance over the three years before tightening, while expansionary policy is a deterioration. The terminology can be counterintuitive: an improving balance may mean that fiscal support is being withdrawn, whereas a deteriorating balance may reflect a government providing more demand or income support.
Rate hikes that followed fiscal restriction produced larger loan-loss increases. The rule-of-thumb difference is about 19 basis points relative to hikes following fiscal expansion. When fiscal stance and the business cycle are estimated together, expansionary fiscal policy still reduces the loss response while a downturn continues to amplify it. That suggests fiscal support is not merely standing in for a better cycle, although the primary balance is itself cyclical and cannot be treated as a clean experimental policy switch.
The central bank’s own balance sheet adds another policy dimension. A policy rate can rise at the same time that the central bank is allowing its asset holdings to shrink. In the BIS estimate, a decrease of one-half of a standard deviation in the balance sheet relative to its sample mean roughly doubles the increase in loan losses compared with the baseline. A concurrent expansion appears to limit the increase, plausibly because more liquidity remains available to the financial system while the policy rate is doing restrictive work.
That finding needs a careful qualifier. In many emerging economies, foreign-exchange reserves make up much of the central bank’s assets. A balance-sheet increase may therefore be associated with foreign-exchange intervention or a change in the exchange rate rather than with a pure domestic liquidity operation. The authors re-estimate the relationship while conditioning on FX intervention and the nominal exchange rate. Those factors matter, but their influence is less pronounced than the balance-sheet result. The evidence supports balance-sheet conditions as relevant to transmission, while leaving room for confounding in how that condition is measured.
The exchange rate can itself provide a modest buffer or amplifier. If a rate hike is followed by currency appreciation, the domestic-currency cost of foreign borrowing may fall and exporters may face a different adjustment than they would after depreciation. The BIS associates a half-standard-deviation appreciation with a roughly 0.02 percentage-point increase in loss rates five years after the shock, compared with about 0.1 percentage points after a comparable depreciation. FX interventions against the domestic currency are associated with smaller loss increases mainly in the first two years, after which the effect no longer matters in the same way. These estimates describe one part of the transmission and do not turn exchange-rate movements into a universal remedy.
Loan contracts form another buffer, and they often determine whether relief arrives before or after the stress. The BIS classifies countries with more than half of household loans at floating rates as having a predominantly floating structure. In those economies, credit losses rise more and faster after tightening, with an additional effect of about 11 basis points in the first year relative to countries with lower floating-rate prevalence. Fixed-rate loans delay the direct increase in scheduled payments. They do not remove unemployment, collateral, refinancing, or funding risks, but they can prevent the policy move from reaching a large stock of borrowers all at once.
The contrast between fiscal support, balance-sheet policy, exchange rates, and loan contracts is not a list of unrelated protections. Each changes a different link in the chain. Fiscal policy supports or withdraws from borrower income and demand. A central-bank balance sheet affects liquidity conditions. FX movements alter competitiveness and foreign-currency burdens. Fixed-rate lending changes the speed at which the policy rate reaches debt service. Their effects can therefore overlap, reinforce one another, or offset one another depending on the country’s financial structure.
This is why the BIS limits its combined exercises. When the authors place several adverse states together—such as a downturn, high debt, high inflation, floating-rate lending, and a loose pre-tightening stance—a 1 percentage-point hike produces roughly a 0.2 percentage-point increase in many combinations and about 0.25–0.30 percentage points in some combinations involving high inflation or restrictive fiscal policy. When none of the identified adverse states is present, the increase is not statistically significant in the corresponding scenario. The result is a nonlinear pattern: context changes the response, but the individual additions are not independent multipliers.
The range also shows why a severe scenario can become materially different from the average without being a forecast. The paper estimates that a 2–4 percentage-point rate increase arriving with a strong recession—a GDP fall of more than 3 percentage points—and some additional adverse conditions could be associated with a 1–2 percentage-point increase in credit-loss rates. Such an outcome could endanger financial stability, but assessing that danger requires a further step: translating losses into capital ratios, liquidity, credit supply, and possible feedback to the wider economy. The country-level loss equation does not perform that entire solvency analysis.
The most useful state-dependent question is therefore not “Which one variable predicts losses?” It is “What vulnerabilities and buffers arrive together with the hike?” A low rate may be benign when borrowers have fixed payments and strong incomes; it may be dangerous when it has supported high leverage and risky lending. A balance-sheet expansion may help liquidity, but it may also reflect FX reserves. Fiscal support may cushion households, but its size and persistence affect public-sector risk. Every buffer has a mechanism and a boundary.
The macro state sets the pressure applied to the system, but it still does not determine which bank inside that system will feel the pressure first. The next step is to open the bank balance sheet: loan quality, repricing, funding, securities, and liquidity determine how a shared macro shock is distributed among institutions.