Overview

A defensible rate-hike stress scenario requires an explicit rate measure, size, path, horizon, outcome, denominator, sample, and method. This chapter then maps the receiving macro state and bank exposures: repricing, asset quality, funding, securities, liquidity, and policy buffers. BIS estimates around 0.12 percentage points over five years, 0.2–0.3 in some adverse combinations, and roughly 0.3 under an IV specification are reference points, not plug-in multipliers. The final translation must cover provisions, margin, capital, liquidity, feedbacks, data gaps, and uncertainty.

Learning outcomes

A rate change is not a usable shock until its measure, size, path, horizon, and outcome denominator are named.
Adverse state variables should be modeled jointly where evidence exists, not added as independent increments.
Stress results need translation into capital and liquidity, with accounting limits, data gaps, and uncertainty visible.

Keywords

Stress testingHorizonDenominatorCapitalUncertaintyMonitoring

How to Read a Rate-Hike Stress Scenario

This chapter turns the evidence on policy rates and bank credit losses into a way of reading a stress scenario. The focus is not to manufacture one universal forecast from a published coefficient. It is to specify the shock, identify the conditions that receive it, map the bank exposures that transmit it, and keep uncertainty visible before a loss estimate is translated into a decision about capital or financial stability.

Start with the shock, not the multiplier

A stress scenario begins with a rate path, but “rates rise” is not yet a usable input. The analyst must first state which rate is moving. The BIS studies changes in central-bank policy rates and, in related exercises, short-term rates and exogenous monetary-policy shocks. The IMF uses the three-month interbank rate. These measures can move together without being identical. A scenario based on a policy rate should not quietly be evaluated with a coefficient estimated for a market rate, and a rate level should not be substituted for a rate change without explaining the choice.

The size and timing of the shock come next. A 1 percentage-point increase is the common reference in the BIS rules of thumb, but it can mean a one-time step, a cumulative move, or a change that is reached over several meetings. The answer also depends on the horizon. The BIS finds an average effect that builds over several years; the IMF’s annual panel estimates the relationship between the level of the three-month rate and provisions or profits in a given year, with dynamics represented by lagged outcomes. A first-year stress and a five-year stress are asking different questions of the same bank.

The outcome needs an equally explicit label. The BIS country-level result measures annual net loan-loss provisions as a percentage of net loans. The IMF measures loan-loss provisions as a percentage of average total assets and also models net interest margin, non-interest income, and return on assets. A change of 0.1 percentage points in a ratio is not a 10% increase in the amount of losses, and a ratio scaled by net loans is not interchangeable with one scaled by assets. If the denominator is hidden, the apparent precision of the scenario is misleading.

The sample and method matter as well. The BIS country-level estimate comes from an unbalanced panel of 113 countries over 1990–2022 and uses local projections to estimate responses through five years. Its bank-level exercise covers 3,350 banks in 92 countries and produces a smaller average response—about 0.02 percentage points for a 1 percentage-point policy-rate increase—while the country-level rule of thumb is about 0.1–0.12 percentage points over the longer horizon. The IMF uses an unbalanced bank panel of 2,147 banks in 31 countries from 1995–2023 and estimates dynamic relationships with bank and time effects, country-specific trends, and bank and macroeconomic controls.

These estimates are useful reference points, not plug-compatible multipliers. The BIS’s alternative exogenous-shock and instrumental-variable exercises retain a positive rate-to-loss effect and produce a larger IV magnitude, roughly 0.3 percentage points rather than 0.1. That is evidence that identification and the type of rate movement matter. It is not a reason to select the largest estimate whenever a scenario needs a more dramatic result. A defensible analyst would report which estimate was chosen, why its outcome and horizon match the scenario, and how the conclusion changes under the other plausible specification.

The next task is to describe the state that receives the shock. The most informative state variables are not a decorative appendix to the rate path. They are the conditions that determine whether borrowers have room to adjust. Was monetary policy already loose relative to recent history? Is private debt high in a system where the relevant borrowers have thin cash buffers?

Published by DeepMox

FAQ

Why does the loss denominator matter in a stress scenario?

The loss denominator matters because a 0.1 percentage-point change in provisions over net loans is not the same as one over average total assets or a 10% change in loss amounts. Hiding it makes the scenario appear more precise than it is.

Can adverse rate-hike effects be added together mechanically?

No. Adverse effects should not be added mechanically because low rates, debt, downturns, and balance-sheet changes can overlap or share causes. Use a coherent joint state and a corresponding combined estimate where the evidence provides one.

What should a bank stress test measure beyond provisions?

It should measure margin, non-interest income, return on assets, capital, liquidity, funding outflows, securities sales, and credit supply. The BIS and IMF do not provide a universal capital or failure probability, so the translation remains bank-specific.

Previous13 minWhich Banks Face Rate-Hike Losses First

Next12 minWhy Rate-Hike Losses Are Conditional